Politecnico di Milano School of Industrial and Information Engineering MSc in Mechanical Engineering Advanced Manufacturing Processes 2021-2022 Academic Year FEM simulation of tube bending processes Francesco Flamini 995700 Claudia Ghidoni Pastorelli 976445 Matteo Lamberti 995030 Massimiliano Moreni 977593 1 ABSTRACT Tube bending is a metal forming process (usually a cold one) highly used in lots of industrial fields, requiring precision and skilled knowledge. The following work is mainly focused on the application of Finite Element Method (through many different softwares, depending on the applications needed) in order to create a simplified model of a particular structure, to prevent the most common defects and observe the parts’ (or assemblies’) behaviour. All the most important bending processes will be introduced and, for each of them, a numerical model will be associated, based on the specific conditions we are taking into account. 2 INDEX LIST OF FIGURES…………………………………………………………………………………4 INTRODUCTION…………………………………………………………………………………..8 BENDING PROCESSES AND FEM SIMULATION…………………………………………...10 • • • • • • • • Rotary draw bending………………………………………………………………………..11 Roll bending………………………………………………………………………………...24 Heat induction bending ……………………………………………………………………..31 Laser bending……………………………………………………………………………….36 Shear bending……………………………………………………………………………….41 Push bending………………………………………………………………………………..49 Press bending………………………………………………………………………………..66 Hydro bending………………………………………………………………………………68 REFERENCES…………………………………………………………………………………….71 3 LIST OF FIGURES Fig. 1. Tube bending process [taken from Google images] Fig. 2. Tube bending process using heat [taken from Google images] Fig. 3. FEM analysis of bent tubes [taken from Google images] Fig. 4. FEM analysis of bent tubes [taken from Google images] Fig. 5. Rotary draw bending Fig. 6. Sketch of NC bending process of tube Fig. 7. Eight-node hexahedral isoparametric element in natural and Cartesian coordinate systems Fig. 8. A FEM model of tube bending Fig. 9. Deformed meshes when bending angles are (a) 0.0; (b) 0.3; (c) 0.6;(d) 0.9; (e) 1.2; (f) 1.57 rad Fig. 10. Stress distribution along the bending direction of tube with bending angle 0.3 rad Fig. 11. Relationship between maximal wall thickness changing ratio and bending angle Fig. 12. Elastic-plastic finite element model of springback of thin-walled tube NC bending Fig. 13. Schematic diagram of mandrel simplified reasonably Fig. 14. Curves of bending velocity and mandrel drawing velocity vs time Fig. 15. Constraint condition of springback computing Fig. 16. Simulation and experiment results of springback angle Fig. 17. Curve of bending angle vs springback angle of stainless steel tube bending Fig. 18. Curves of bending angle vs springback angle of stainless steel tube bending with different relative bending radius Fig. 19. Curves of bending angle vs springback angle of stainless steel tube bending with different relative tube diameter Fig. 20. Curves of ending angle vs springback angle of tube bending with different elastic modulus Fig. 21. Curves of bending angle vs springback angle of tube bending with different yield stress Fig. 22. Curves of bending angle vs springback angle of tube bending with different strength factor Fig. 23. Curves of bending angle vs springback angle of tube bending with different hardenability value Fig. 24. Forming principle of rotary draw bending method and sketch of standard mandrel Fig. 25. FEM model for NC bending process Fig. 26. Geometry model of mandrel with three balls Fig. 27. Comparison between the experimental and simulation results cencerning the cross-section distorsion vs bending angle Fig. 28. Tools and axis of the three-roll-push-bending machine Fig. 29. Three main steps of the three-roll-push-bending process Fig. 30. Tube section Fig. 31. Tool set-up for the Three-Roll-Push-Bending-Twisting with a description of the kinematic of the tools 4 Fig. 32. Final setup of the machine. The roller B is constrained to A by a spherical-joint element, which allows the centre of B to rotate about that of A, keeping the same relative distance Fig. 33. Tube mock-up ‘S’: Comparison between measured and numerical profiles. The red points represent the edges of the final meshed, whereas the measured points have been plotted with colour scaled in relation to their deviation (or distance) from the mesh-surfaces. 3D visual of the whole tube (a), view of portion undergoing TRPBTon the ZX- (b) and YZ-plane (c); curvatures obtained via RDB (d) (f) Fig. 34. S-shaped tube mock-up. (f) and (g) residual stresses distribution in terms of Von Misses equivalent stress while (c) and (d) accumulated plastic strain in the zones bent via RDB. (a), (c), (f) bottom view, (b), (d), (g) top view. In (b) the welding seam is highlighted Fig. 35. Variation of residual stresses and plastic strain along path-1 (f) and -2 (g) are plotted after the completion of the forming process. Path-2 follows the welding seam. e stress tensor orientation, S11 is aligned with the longitudinal direction Fig. 36. Compressive test with numerical results and experimental ones: (a) reaction force, (b) ux displacement, (c) uy displacement vs z-displacement along the loading-axis; in the model FEM-stress the residual stress and accumulated strain from the previous process phases are considered, on the contrary, these aspects are not taken in account in the model FEM-virgin Fig. 37. Induction bending machine Fig. 38. FEM meshed model Fig. 39. Pipe shape after loading Fig. 40. Pipe shape after unloading Fig. 41. Temperature distribution in local heating zone Fig. 42. Circumferential equivalent plastic strain Fig. 43. Wall thickness and thickening Fig. 44. Ovality/spring back angle with bending angle (a/b) Fig. 45. Ovality with load step Fig. 46. Circumferential scanning scheme Fig. 47. Factors of influence Fig. 48. Finite element model Fig. 49. Sketch of laser scanning process Fig. 50. Temperature with time Fig. 51. Axial stress with time Fig. 52. Axial strain with time Fig. 53. Displacement with time Fig. 54. Axial plastic strain/time Fig. 55. Bending angle with n of scans Fig. 56. Schematic sketch of shear bending equipment Fig. 57. Schematic illustration of shear bending process Fig. 58. Typical products by various pushing forces: (a) splitting; (b) successful; (c) wrinkling; (d) rupturing 5 Fig. 59. Finite element meshes after the deformation Fig. 60. Distribution of tube effective strain Fig. 61. Strain distributions across tube width in the steady state region (line TB)- maximum strains have occurred around π = 90° Fig. 62. Distribution of thickness strain along routes T and B Fig. 63. Cross section configurations of the deformed tubes with different initial thickness obtained by simulation and experiment: (a) simulation, (b) experiment and(c) experiment (t0=3.0 mm) Fig. 64. Effect of the axial pushing pressure on the cross section ovality of the deformed tube Fig. 65. Configuration of the deformed tube Fig. 66. Effect of the initial thickness on the thickness strain of the tube-sheared part obtained by simulation and experiment. Fig. 67. Relation between the pushing and shearing strokes Fig. 68. Effect of the tube initial thickness on the as a function of the tube initial thickness obtained by simulation shearing load obtained by simulation Fig. 69. Effect of the initial thickness and pushing pressure on the distribution of thickness strain along the hoop direction of the tube-sheared part Fig. 70. Tube bending stress and strain analysis Fig. 71. FE modelling of tube free bending Fig. 72. Model assembly position Fig. 73. Bending simulation under condition Fig. 74. Bending simulation under condition Fig. 75. Bending simulation under condition Fig. 76. Bending simulation under condition Fig. 77. Experiments of 3D free bending under different conditions Fig. 78. Schematic of an elbow granular-media-based push-bending process: (a) before bending, (b) bending Fig. 79. FEM model of tubular blank and DEM model of granular filler Fig. 80. Stress-strain curve of 1Cr18Ni9Ti Fig. 81. Distributions of stress in granular filler and tube: (a) granular filler, (b) tube Fig. 82. Forming force of the granular-media-based push-bending process: simulation and experimental results Fig. 83. Effects of particle sizes of granular fillers on wrinkling of thin-wall elbow tube: (a) particle size 0.98 mm, (b) particle size 1.58 mm, (c) particle size 2.08 mm Fig. 84. The contacts’ distribution on tube wall (unfolded into a plane): (a) particle size 0.98 mm, (b) particle size 1.58 mm, (c) particle size 2.08 mm Fig. 85. The tube bending process treated Fig. 86. Meshed geometries of die, tube and rod Fig. 87. Urethane rod penetrating the tube Fig. 88. Predicted tube geometry after bending 6 Fig. 89. Von Mises stress distribution in the formed tube Fig. 90. Von Mises strain distribution in the formed tube Fig. 91. MOS bending technique Fig. 92. Influence factors on free bending technology Fig. 93. Bending profile experimentally used Fig. 94. Measured geometry of EN AW-5182 tubes Fig. 95. Standard deviation and density function of experimental bending Fig. 96. Strong wrinkling at radius 1 Fig. 97. Weak wrinkling at radius 4 Fig. 98. Decrease of wrinkling by decrease of lubricant (a to c) Fig. 99. Simulation result of test geometry (max EPS): a) batch C4 b) batch C1, C3 Fig. 100. Technique used to measure radius and bending angle Fig. 101. Correlation between experiment and simulation varying the weld seam position Fig. 102. (a)ANOVA in terms of EPS inner. (b) Scatter plot Fig. 103. (a)ANOVA in terms of bending angle. (b) of diameter and thickness in relation to EPS inner ANOVA in terms of bending radii Fig. 104. Press bending process (1. upper roll; 2-3 lower rolls; 4 – pipe (Semi-manufactured); 2θ - bending angle; h – the distance covered by the upper roll during the bending; R – the bending radius) Fig. 105. FEM of press bending Fig. 106. The final position of press bending Fig. 107. Stress-strain relation reached through tube tensile test Fig. 108. Distribution of wall thickness of angle 2Ζ=25° Fig. 109. Distribution of wall thickness of angle 2Ζ=90° Fig. 110. Wall thickness trend Fig. 111. Diagram of double-layered tube specimen Fig. 112. Dimensions of ultra-thin-walled elbow (number 1-1 to 5-5 are the measuring locations of section flattening after bending) Fig. 113. Tooling of the double layered tube hydro bending Fig.114. Bending results under different levels of internal pressure: (a) defect-free double-layered elbow formed under Py; (b) section view of double-layered elbow under pressure Py; (c) wrinkling under pressure 0.1Py; (d) rupture under pressure 1.2Py; (e) ultra-thin-walled elbow formed under pressure 0.8Py; and (f) ultrathin-walled elbow formed under pressure Py Fig. 115. Finite element meshes for double-layered tube hydro bending: (a) shell type outer tube (b) solid type outer tube 7 INTRODUCTION As one kind of key lightweight components for “bleeding” transforming and loading carrying with enormous quantities and diversities, metallic bent tubular parts have attracted increasing applications in many hightechnological industries such as aviation, aerospace, shipbuilding, automobile, energy and health care since they satisfy the current needs for products with lightweight, high strength and high performance from both materials and structure aspects. To bend the tubular materials with certain bending radius, bending angle and shapes, the tube bending, one important branch of the tubular plastic forming fields, has been a vital manufacturing technology for lightweight products. Many tube bending approaches have been developed in response to the different demands of tube specification, shapes, materials and forming tolerance. Fig. 1. Tube bending process [taken from Google images] Fig. 2. Tube bending process using heat [taken from Google images] According to the forming conditions, there are cold bending at room temperature and hot (heat) bending with elevated temperature. From the viewpoint of loading conditions, there are pure bending, compression bending, stretch bending, roll bending, rotary draw bending and laser bending. There are stainless steel tube bending, aluminium alloy tube bending, copper tube bending, magnesium alloy tube bending and titanium alloy tube bending in materials point of views. From aspects of tubular shapes, there are round tube bending, rectangular tube bending and other irregular section tube bending. There are seamless tube bending and welded tube bending from the tube fabrication respect. For any bending process, upon bending deformation, the complex uneven tension and compression stress distributions are induced at the extrados and intrados of bending tube respectively, which may cause multiple defects or instabilities such as wrinkling, over thinning (cracking), 8 cross-section distortion and springback. The accurate prediction and efficient controlling of these physical phenomena are the ever existing knots in forming manufacturing .Up to now, regarding different tube bending processes with various loading conditions, great efforts have been conducted on investigations of the multiple defects/instabilities, selection/optimization of the forming parameters and tooling to promote the development of tube bending science and technology by using analytical, physical experimental and numerical methods. For example using a FEM analysis we’re able to optimize the process non only in terms of reducing the defects of the tubes but even from an economical point of view. Fig. 3 and Fig. 4. FEM analysis of bent tubes [taken from Google images] The optimization strategies associated to a particular type of process require enough knowledge about the basic physical stages determining the achievement of the desired results. Furthermore, to obtain a correct optimization, the quantitative results reached through a numerical analysis are of high importance too. Even if the analytical approach is limited (mathematical equations describing the process too much complicated), the finite element method (able to consider numerous dependencies on physical constants as well as to describe complicated geometries and kinematics) allows us to describe coupled simulations of plastic and thermal processes. 9 BENDING PROCESSES AND FEM SIMULATION 10 Rotary draw bending It is the most common process used in the bending industry. This type of bending has a great variety of tooling options and offers the best quality and accuracy in terms of bending results, particularly for tight radii and thin wall tubes. The characteristic of this bending method is that the centre bending die, which is used to form the angle of tube parts, rotates with the workpiece together, and the die set sometimes is equipped with a wiper and a mandrel, depending on the size and shape of workplace. A die set up for a rotary draw bending is shown in Fig. 5. Fig. 5 Rotary draw bending Rotary draw bending is called in this way because the tube is being drawn into the bending area past the tangent point. At one end, the tube is tightly pressed between a centre forming die and a clamp die that holds the tube. The pressure die restrains the free end of the workpiece and allows it to move in a straight line, avoiding the part to kick out. As the workpiece is being drawn and rotating around the centre die, the pressure die, which is either static or boosted, transfers the workpiece to the centre die at the tangent point, so as to get the desired angle and radius.[1] In the following pages some examples of rotary draw bending technique applied to different processes are presented in addition of FE simulations. Using a three-dimensional rigid-plastic FE simulation system called TBS-3D (tube bending simulation by 3D FEM) a Finite Element simulation system of an NC bending process of a thin-walled tube (D/t ≥ 20, where D is the external diameter and t is the wall thickness) has been simulated neglecting the effect of springback; moreover deformed meshes under different bending stages, the stress distribution along the bending direction, and the relationship between the maximal wall thickness changing ratio and the bending angle have been studied. Thin-walled tube forming processes are widely used in the aerospace and automobile fields since the products obtained are much lighter in weight and with enough strength. For this manufacturing process NC bending machine instead of a traditional bending methods could be used since the first can provide higher precision and efficiency; this also means that the bending process will be affected by an higher number of process parameters and the design phase will be more complicated with respect to the conventional trial-and-error approach. The NC tube bending process can be represented as follows: 11 Fig. 6 Sketch of NC bending process of tube. The final desired shape is obtained through the use of the pressure die, the clamp die and the bending die. Wrinkling and ovalizing of the cross-section are smoothed by the usage of a mandrel and wiper die, meanwhile the pressure die can also offer a pushing force along the longitudinal direction to avoid an extreme reduction of wall thickness in the outer tensile area. The traditional tube bending process has been rebuilt using a FE simulation through which it is possible to perform a “virtual bending” on a computer, helping not only in design phase but also in the analysis and optimization of the process and the die parameters. As a result numerous and sophisticated “trial-and-error” experiments can be avoided, reducing cost and time without losing in quality and efficiency. The eight-node hexahedral isoparametric element is adopted to simulate the tube bending process. The following figures show this element in natural and Cartesian coordinate systems. Fig. 7 Eight-node hexahedral isoparametric element in natural and Cartesian coordinate systems. 12 The velocity and displacement vector of a node inside the element is: where π₯π , π¦π , π§π are the coordinates, π’π₯π , π’π¦π , π’π§π are the velocities of some one-node of this element, ππ is the shape function of the eight-node hexahedral isoparametric element with ππ , ππ , ππ natural coordinates of the node (i is the number of the node of the element). An important parameter that needs to be set up during the NC bending process of tube is the friction, so an arctangent friction model has been introduced into the FE simulation using the following equation: Where m is the friction factor, k the yield shear stress, π’π is the velocity vector of the workpiece relative to the die, π’0 a verysmall positive number as compared to π’π , and t the unit vector in the direction of π’π . For what concerns the FE model triangular elements have been chosen to simulate the pressure die, the bending die and the wiper die, the tube has been discretized into some eight-node hexahedral isoparametric elements. The following picture is showing the element choice: Fig. 8. A FEM model of tube bending. Computation conditions: • • • • The material used is 304 stainless steel and the flow stress model is π = 1356(0.001 + π)0.549 πππ. Tube dimensions: outer diameter of 28 mm, length 228 mm, and wall thickness of 1 mm. Process parameters: the rotation speed of the bending die is 0.05 rad/s , the bending radius is 56 mm, and the pressure die remains stationary. Die parameters: the length of the wiper die, the pressure die and the mandrel are 140 mm. 13 • Friction condition: the friction factor between the pressure die and tube is 0.01 and the friction factor between the wiper die and tube is 0.0. As a conclusion to this bending process some pictures collecting the obtained results have been reported below: Fig. 9. Deformed meshes when bending angles are (a) 0.0; (b) 0.3; (c) 0.6;(d) 0.9; (e) 1.2; (f) 1.57 rad. Fig. 9 shows that mashes deform differently according to the different deformation stages when the bending angles are 0.0 ,0.3 ,0.6 ,0.9 ,1.2 and 1.57 rad. The deformation is mostly focused on the bending area, the higher is the distance from the bending area, the smaller is the deformation. In part (a), L presents the length to be bent into a 90°(1.57 rad) bent tube; (f) shows the shape of the tube after 90°bending and dL presents the elongation of the tube itself after the bending. Fig. 10. Stress distribution along the bending direction of tube with bending angle 0.3 rad. From this figure that shows the stress distribution along the bending direction with a bending angle of 0.3 rad, it can be understood that the outer area is undergoing tensile stress , while the inner area is undergoing compressive stress and the stress neutral layer moves closer to the inner area. 14 Fig. 11. Relationship between maximal wall thickness changing ratio and bending angle. Figure 11 shows the link between the maximal wall thickness changing ratio and the bending angle: the maximal wall thickness reducing ratio in the outer tensile area doesn’t change a lot with increase of the bending angle and the maximal wall thickness increasing ratio in the inner compression area increases linearly with the bending angle. As a conclusion of this FE simulation system for the NC precision bending process of thin-walled tube it can be stated that: this process makes the tube elongates to some degree and the outer area is undergoing tensile stress while the inner area is undergoing a compressive stress meanwhile the neutral layer moves closer to the inner area. As reported above the maximal wall thickness reduction ratio in the outer tensile area changes a bit with an increase of the bending angle and the maximal wall thickness increase ratio in the inner compression area increases linearly with bending angle. [2] Effect of the springback on the forming quality of rotary draw bending In the previous pages the springback effect has been neglected, but it is a key factor that affects the forming quality of tube bending since after the bending process the elastic deformation releases and the bent-tube is sprung back. [3] The elastic-plastic finite element method can be used to analyse the springback back process of thin-walled tube NC precision bending and the entire process is solved through the use of two algorithms: the dynamic explicit algorithm and the static implicit algorithm. This numerical simulation method can be useful also to study the effect of geometry and material parameters on the springback phenomenon to effectively control it and the factors that affect the forming quality. If the value of the springback of NC tube bending is beyond the error range, the accuracy of tube bending can’t satisfy the requirement of applying, so that in the past many researchers performed lot of studies on it making reference to the springback theory of beam bending and considering as parameters the action centre of the pressure die and the friction force between mandrel and tube. The results were improved using an approximate expression in trigonometric form for the displacement field and as constitutive relation the total deformation theory was employed; moreover without increasing the accuracy of the required hardware, the method of bend-rebend was used in addition of process control. Nowadays by numerical simulation method, combined with theoretical analysis and experiments, the springback rule of thin-walled tube NC precision bending and the effect of other parameters have been carried out. 15 The whole process includes three phases: tube bending, mandrel drawing and tube springback; the processes of tube bending and mandrel drawing are dynamic, this leads to the usage of the dynamic explicit algorithm, while the process of tube springback is static, so the static implicit algorithm is used to solve it. During simulation tube bending process can be divided into two stages: tube bending and mandrel drawing are performed at the first stage whose operations are controlled through the loading of dies, in the second stage the process of tube springback is carried out. Independelty from the fact that a dynamic explicit algorithm or a static implicit algorithm is used in FEM, it is required to solve the nonlinear equilibrium equation: π = ππ’Μ + πΉ − π = 0 Where M is the system mass matrix, π’Μ is the vector of accel-eration, F is the vector of internal force and P is the vector of external force [4]. Since the aim of the simulation is to predict with a certain accuracy the springback phenomenon, the process of retracting mandrel must be taken into account and, in this case, the method for predicting the springback angle available in literature could not be used. The springback process is the result of a stress self-balancing of tube which depends both on the bending process and on the retracting mandrel: the total springback angle considering retracting mandrel is much smaller than that not considering it and the maximum difference between them on average is 107%.[4] The software chosen to establish the finite element model is DYNAFORM (LS-DYNA) since it can do both dynamic explicit and static implicit computing and it can model a seamless transition between the two algorithms. Figure 12 shows the elastic-plastic FEM of spring back. Fig. 12 .Elastic-plastic finite element model of springback of thin-walled tube NC bending Establishment of finite element model To reduce the computational cost it has been taken advantage of the symmetrical structure so only half of the model has been established, then a symmetry constraint is assigned. Firstly the geometry models of dies are required, only after the ones of the contact surfaces between dies and tube; for the mandrel structure a simplified model shown below has been used: Fig. 13. Schematic diagram of mandrel simplified reasonably 16 The mandrel body and mandrel heads are linked to each other with CONSTRAINED-JOINT-SPHERICM. Firstly the mesh of the tube with a refinement in the field of big deformation has been computed, then the one of the dies is obtained by offsetting the mesh of the tube. For the simulation fully-integrated-shell-elements and five integral points along the thickness are used and for the material model a transversely anisotropic material model is chosen followed by some experimental results. The contact condition is treated by using a penalty function method [4] in which a contact force (Pc) proportional to the penetration deepness is assigned to the slave nodes to avoid the situation in which slave nodes could penetrate master surfaces; the mathematical expression for the contac force is ππ = πΎπ × πΏ, πΏ ≥ 0 where πΎπ = 0.1 is the contact interface stiffness. For what concerns the friction, Coulomb friction model in which the friction force fc is proportional to the contact force ππ ( ππ = π × ππ where π is the friction coefficient) is used. In the loading phase, during tube bending, the clamp and the bending dies rotate, the pressure die pushes ahead, and the mandrel keeps stationary. When the bending process is finished (π‘0 ), the mandrel begins to be drawn back till the process is finished (π‘1 ). Bending velocity and mandrel drawing velocity vs time are shown below: Fig.14. Curves of bending velocity and mandrel drawing velocity vs time For what concerns the constraint condition shown in figure 15 the translation and rotary of bent-tube must be constrained in order to unchange the energy of spring back computing; the constraints are imposed at the end of the bent-tube since it deforms hardly. Fig. 15. Constraint condition of springback computing The spring back time needs to be reasonably estimated, so that a comparison between different spring back times applied to the model and the computing results has been performed, when the effect of the increasing time on the computing precision is smaller than the requirement an optimal spring back time is found. To control the whole springback process it has been divided into small several steps (artificial stabilization with a scale factor =0.005 [4]), this is necessary to accelerate the convergence of springback computing and increase the accuracy. 17 At each iteration the matrix is updated. The experiment equipment W27YPC-63 full-automatic hydraulic pressure tube bender controlled by programmable logic controller is used to validate the finite element model of thin-walled tube NC bending; since the experimental results and the ones from the simulation are in agreement with an error ratio that doesn’t exceed 18.57%, the finite element model is considered reasonable. Fig.16. Simulation and experiment results of springback angle Following the validated FEM, the springback rules of 1Crl8Ni9Ti stainless steel tube and LF2M aluminium alloy tube NC bending with thickness 1 mm, diameter 38mm and bending radius 57mm have been studied. It is possible to see that when the bending angle is small, the trend of springback angle with bending angle is nonlinear, the reason is that the whole bent-tube process considers both the curved part and the transition part and the springback of the transition part is more significant. On the contrary, when the bending angle is big, the springback angle changes linearly with the bending angle and this is because the transition part kept stable after the bending process is developed to steady forming. The trend changing of bending angle vs springback angle is displaced in the picture below. Fig.17. Curve of bending angle vs springback angle of stainless steel tube bending The effects of relative bending radius and relative tube diameter on the springback have been studied. When the bending angle is big: the bigger the bending radius is, the bigger the springback is; the reason is that as the length increases more material took place in the springback. On the contrary, when the bending angle is small: the bigger the bending radius is, the smaller the springback is and the more the nonlinear changing trend of springback angle with bending angle is; this is because the transition part of benttube is shorter and the process of tube bending is developed to steady forming early. 18 Fig. 18. Curves of bending angle vs springback angle of stainless steel tube bending with different relative bending radius When the bending angle is big: the bigger the tube diameter is, the bigger the springback is. The reason of that is the same of the previous case. When the bending angle is small: the bigger the tube diameter is, the smaller the springback is and more the nonlinear changing trend of springback angle with bending angle is; this is because the rate of the elastic deformation in the same whole deformation is smaller Fig. 19. Curves of bending angle vs springback angle of stainless steel tube bending with different relative tube diameter If now the effect of the retracting mandrel is accounted for the total springback angle is much smaller than that not considering it and the maximum difference between them in the paper is 107.34% so to predict accurately the springback, the retracting of the mandrel needs to be considered. [4] Also the effect of material parameters on the springback rule has been carried out and they have been set on the values of lCrl8Ni9Ti stainless steel tube. The parameters changed are the elastic modulus E, yield stress ππ , strength factor K and the hardenability value n. The following pictures show the trends changing the parameters above: 19 Fig. 20. Curves of ending angle vs springback angle of tube bending with different elastic modulus Fig. 22. Curves of bending angle vs springback angle of tube bending with different strength factor fig. 21. Curves of bending angle vs springback angle of tube bending with different yield stress fig. 23. Curves of bending angle vs springback angle of tube bending with different hardenability value The significance of effect of these four parameters on the springback of thin-walled tube NC bending become weaker and weaker changing the rate of springback angle when the bending angle computed was 180°. Effect of the mandrel on the quality and bending limit in rotary draw bending A quite important element in the rotary draw bending process to increase the quality of the product and the bending limit is the mandrel, so that an analytical model of it including mandrel shank and balls has been performed; moreover some mandrel parameters such as mandrel diameter d, mandrel extension e, number of balls n, thickness of balls k, space length between balls p and nose radius r need to be selected preliminarly and so some reference formulas was required to do this. To validate the analytical model an experiment has been carried out and a 3D elastic–plastic FE model of the NC bending process is established using the dynamic explicit FEM code ABAQUS/Explicit. From this analysis becomes straightforward to see the influence of mandrel stress distribution during the bending process and its impotance in the NC precision bending process such as wrinkling prevention. The following figure shows the tools required in the thin-walled tube NC precision rotary draw beding process: 20 Fig. 24. Forming principle of rotary draw bending method and sketch of standard mandrel. During this bending process the tube is clamped against the bend die, it is drawn by the bend die and the clamp die then it rotates along the groove and after it has passed the tangent point it rotates along the groove of the bend die to obtain the desired degree of bending radius. Since the forming process has been completed the mandrel is withdrawn and the tube is unloaded. During this operation the tube needs to substain rigid contact forces from the tooling system made of the bend die, the clamp die, the pressure die, the wiper die and the mandrel, thus precison and strict control of parameters are required to obtain the wanted final shape. To provide a rigid support during the bending process the ball-and-socket-type flexible mandrel, including balls, is positioned inside the hollow tube; to select preliminarly the mandrel parametrs some formulas have been deduced after the stating of important assumpions: 1) No interference is present between the mandrel shank and the balls, this means an ideal contact condition. 2) The balls rotate about Z-axis and should point to the bending center O. 3) The bending deformation starts from the early stable stage to stable one, and only the materials in the local regions near the tangent line deform largely so that in these regions wrinkling, over-thinning and section ovalization may occur. The maximum extension length The maximum mandrel extension length ππππ₯ can be computed by the following equation: and it shows that the mandrel extension length largely depends on the ratio D/t , the bending radius R and the mandrel diameter d. If this maximum value is exceeded the tube will interfere with the mandrel shank and over thins or then cracks. There exists also a minimum value of e, ππππ , that it is equal to r, if it is further reduced the mandrel and wiper die would not cooperate well and the wrinkling may happen. The thickness of balls The thickness of the balls k should be a reasonable value, this means that if it too low the mechanical connection strength may be compromised, whereas if it is too high the clearance between tube and point C on balls is too big so that the contact is no more rigid. The next formula shows the link between some parameters and k: where ππππ₯ is the maximum value of clearance. 21 The number of flexible balls This vaule can be computed using the following formula: Where the cross-section degree with maximum ovalization is α and πΌπ is the effective arc of ovalization. The space length between balls The space length between balls p is obtained by looking at the geometry shown in the above picture and the formula: In order to verify the above proposed analytical modeling of mandrel, some experiments have been carried out and the resulting parametrs can reasonably justify the analytical model. FEM model If compared the static implicit algorithm, the dynamic explicit FE algorithm shows important advantages such as little solution costs, few difficulties in simulating complex contact and large deformation process and it is also able to predict imporant problems as wrinkling and cross-section distortion phenomena without performing several experiments. The following pictures illustrate the FE model of the process and the geometry model of the mandrel shank and flexible balls: Fig. 25. FEM model for NC bending process. Fig. 26. Geometry model of mandrel with three balls For what concerns the choice of element-type the four-node doubly curved thin shell S4R had been adopted to describe three-dimensional deformable tube, moreover to describe with sufficient accuracy the bending process five integration points have been selected across the thickness. Two parts can be identified axially in the tube: a clamped portion and a to be bent portion, this last is the one that will actually undergo the bending deformation. In addition to what already said the external and internal rigid tools are modeled as rigid bodies using 4-node 3-D bilinear quadrilateral rigid element R3D4 and total number of meshing element is 5889. For what concerns the material properties, the uniaxial tension test applied to stainless steel is used to obtain them and the traditional Coulomb model has been chosen to describe the interfaces’friction conditions; 4 values of friction coefficient have been chosen according to the different friction situatins present. Some boundary conditions need to be set in the FE model: 1) ‘‘Surface-to-surface contact’’ was chosen to represent the contact interfaces between the tube and the die so that sliding is allowed between the 2 surfaces. 2) ‘‘Kinematic constraints’’ is used to describe the mechanical constraints in addition with ‘‘Penalty method’’ to improve the computation efficiency and accuracy. 22 3) The ‘‘Displacement/rotates’’ and ‘‘Velocity/angular’’ were used to apply the same boundary constraints and loadings as in the true bending process. 4) Since the relative movement between mandrel shank and flexible balls is quite complex, the‘‘Connector element’’ is employed to define the contact conditions between mandrel shank and floating balls. At the end, following the same forming rule and dies structure as the NC bending machine, the previously described elastic–plastic FE model has been validated by using PLC controlled bender W27YPC-63NC. The following figure shows the comparison of FE cross-section distortion degrees with the experimental ones; it is easy to see that the maximum distortion degree increases linearly with the bending angle and the error between the simulation and the experimental results is lower than 5%, this means that the model is considered correct. Fig. 27. Comparison between the experimental and simulation results cencerning the cross-section distorsion vs bending angle The effect of mandrel parameters has been invistigated using the established explicit FE model and it shows the following results: 1) The larger is the mandrel size the more the neutral axial will be moved outward and this will increase the minimum wrinkling energy and anti-wrinkling ability, but it makes outside tube over-thinning. 2) When the mandrel extension length increases, the neutral axial will be moved outward but the ability in restraining the wrinkles is much less than the one of the mandrel size, moreover the excessive extension will cause the tube over-thin or even crack. 3) The significance of ball number’s effect on the neutral axial position is between the ones of the mandrel size and mandrel extension length: increasing the ball numbers improves the cross-section distortion degrees in an efficient way, while the role for controlling the distortion is limited when the ball number exceeds the one computed through the formulas and increasing the ball number will lead to the over-thinning of the tube and at an higher manufacturing cost. [5] 23 Roll bending Operating principle A set of tools for the three-roll-push-bending process is shown in figure 28. The process, based on two holding rolls, one bending roll and one setting roll, is driven by axes which are numerical controlled. Among these the C-axis designates the feed of the tube. The setting roll is moved in the bending plane by either rotating around the centre of the bending roll (Y-axis) or translating in radial direction (P-axis). Although three rolls are sufficient to perform the bending operation, a rear holding roll is used to stabilize the tube during the bending operation.[6] Fig. 28. Tools and axis of the three-roll-push-bending machine. The three-roll-push-bending process can be described by three steps as displayed in figure 29. The set-up stage includes the clamping of the tube and the closing of the tool. After this the starting step follows where the setting roll is moved to the end position with a simultaneous feeding of the tube, which is performed by the Caxis. When the desired position of the setting roll is reached, a constant radius R is bent by constant feeding of the tube. 1. Set up 2. Starting 3. Bending Fig. 29. Three main steps of the three-roll-push-bending process. An arc is formed depending on the position of the setting roll defined by P-axis and Y-axis and the feeding rate of the C-Axis. As the setting roll can be moved while the tube is being pushed forward and since the machine has an additional A-axis allowing for rotation of the tube around the C-axis during the bending process, arbitrary 3D geometries can be bent. In article [7] is shown a numerical simulation of three-roll push bending plus twisting (TRPBT) of rectangular tubes. For this purpose, tubes are subjected to different combinations of these forming processes and then measured using a Coordinate Measuring Machine (CMM). The results are used to validate the FE simulation of the forming methods. 24 Fig.30. Tube section. Set up of the process for the numerical simulation The scheme of the configuration used for this process is shown in Fig. 31, consists of a fixed main (A) and a mobile slave roller (B) constrained to the former. The roller B, whose rotation axis is parallel to those of A, and the roller C are fixed on an arm, rotating about the same axis of A. An angular actuator, operating under displacement control, produces the angular displacement θ of B. The rear side of the tube is constrained to a booster, which pushes the tube forwards. C C Fig. 31. Tool set-up for the Three-Roll-Push-Bending-Twisting with a description of the kinematic of the tools. The tube is forced to pass through a system of rollers that are free to rotate about their centres. In the setting phase, the tube is approaching the rollers. Two holding rollers support the upstream part of the tube preventing undesired deformations in the zone that has not been deformed yet. After the tube has passed through the two rollers, the roller B moves about the rotation axis of A according to an angular displacement θ, causing the curvature of the tube. Up to this point, the tube keeps going forward while the rollers remain fixed. The value of θ determines the curvature on the plane x-z, illustrated in Fig. 31. From the knowledge of θ, it is possible to evaluate the theoretical curvature Rc that the tool set-up should impose. The analytical procedure is based on finding the radius of the circle tangent to both rollers A and B having the centre lying on the x-axis. However, 25 the springback induces effectively a greater curvature Rc,eff with respect to the theoretical one Rc, as shown in Fig. 31. Successively, the tube is deformed by the twisting roller, inducing a curvature Rtws on the plane y-z. The twisting can be set up by adjusting the position HS of the twisting rollers. For HS = 0 the twisting roller is tangential to the tube profile and twisting is not produced. The shapes of the samples obtained by experimental tests are acquired using a CMM. The measured profiles of the tubes will be compared with those estimated by the numerical analysis. FE model of three-roll-push-bending plus twisting Below is explained a numerical model able to predict shape and mechanical behaviour of formed tubes with satisfactory accuracy, in a reasonable time and using computational resources within reach of most companies working in the tube-forming branch. Acceptable and realistic accuracy target was set at 10%. The workpiece is approximated as shell elements. This approximation is reasonable due to the small ratio between the thickness of the tube and the remaining dimensions. Thanks to this approximation computation time decreases. Infact, the exploitation of brick elements would imply an increment in computational heaviness, accompanied by more numerical instabilities. Obviously, shell elements cannot capture the stress concentration and local deformation occurring at the junctions between the tube walls. Nevertheless, they provide a satisfactory representation of the macroscopic geometry and the global deformation process of the tube which is the required output of the model. The optimal shell size, obtained experimentally, is equal to 1x5 mm and the longest side of each element is parallel to the tube axis. The shell element thickness is imposed to be entirely in the interior of the tube and, thus, the nodes are representative of the external surface. Material constitutive properties are implemented considering the results of the tensile tests. The material model is formulated based on the Hill yielding criterion with an isotropic hardening rule. The mechanical properties of the welding seam are inferred from microhardness tests and implemented in the FE model. The rollers as well are modelled with shell elements but are regarded as rigid bodies (with infinite stiffness in order to reduce the computational cost) because their compliance is very low in comparison with that of the workpiece. For computational lightness, it has assumed that the cylinders can freely rotate about their centre and their contact with the workpiece is assumed frictionless but in a real machine tool, the cylinders are not completely free to rotate so this is an assumption that can lead to oversimplify the model. Furthermore, the high contact pressure tool-workpiece may induce tangential frictions forces may affect the final shape. To account it a Coulombian model wherein the friction coefficient along the tangential direction is taken equal to 0.25 has been used. Fig. 32. Final setup of the machine. The roller B is constrained to A by a spherical-joint element, which allows the centre of B to rotate about that of A, keeping the same relative distance. Model of the S-shaped tubular mock-up The FE model of the above-described forming processes is now used to simulate the fabrication of the Sshaped mock-up. Thanks to the FE model we’re able to obtain not only the mesh but also all the nodal information in terms of hardening, plastic strain, residual stress, mass, displacement, velocity. Both the final 26 geometry and its material properties (hardening, plastic strain, residual stresses) are recovered in a successive simulation to reproduce the compression test. Fig. 33 compares numerically predicted and measured profiles of the S-shaped mock-up. The following table reports the curvature radii estimated both from measurements and numerical model. Measured and estimated curvature radii from mock-up. Fig. 33. Tube mock-up ‘S’: Comparison between measured and numerical profiles. The red points represent the edges of the final meshed, whereas the measured points have been plotted with colour scaled in relation to their deviation (or distance) from the mesh-surfaces. 3D visual of the whole tube (a), view of portion undergoing TRPBTon the ZX- (b) and YZ-plane (c); curvatures obtained via RDB (d) (f). 27 Once again, the validity of the FE model is confirmed, as 5 out 6 curvature radius values are well below the accuracy threshold (10% from the real value). Until now, the comparison has shown only the capability of the FE analysis to predict correctly the final geometry of the workpiece but as we will see the model is also able to compute the distribution of residual stresses, strain hardening and plastic deformations. Structural analysis and full-scale compression test In order to obtain a correct prediction of the structural response of the mock-up under in-service loading conditions we need the estimation of the strain hardening and residual stresses distribution after the manufacturing process. For this purpose, the residual stress and the plastic strain field are extracted from the FE model and shown in Figs. 34 and 35. Fig. 34. S-shaped tube mock-up. (f) and (g) residual stresses distribution in terms of Von Misses equivalent stress while (c) and (d) accumulated plastic strain in the zones bent via RDB. (a), (c), (f) bottom view, (b), (d), (g) top view. In (b) the welding seam is highlighted. 28 Fig. 35 illustrates residual stresses and plastic strain after the manufacturing process measured along two paths (path-1 and path-2) running along the two opposite tube walls of minor width. Fig. 35.Variation of residual stresses and plastic strain along path-1 (f) and -2 (g) are plotted after the completion of the forming process. Path-2 follows the welding seam. e stress tensor orientation, S11 is aligned with the longitudinal direction. Such walls have been selected because are expected to undergo larger strains than the lateral walls. Path 2 runs along the welding seam. 29 It is reasonable to argue that the in-service mechanical response of the component is influenced by the stressstrain history experienced during the forming process. To further investigate this issue, the mock up is subjected after manufacturing to an instrumented destructive compression test. The outcomes of the experimental test are interpreted by comparison with an FE simulation of the compression test (Fig. 36). For this purpose, the numerical model of the tube is recovered and exploited for the simulation, reproducing the same loading conditions. To highlight the effects of the loading history experienced by the mock-up during the forming process, two different FE analyses have been performed: the former (FEM stress) takes into account the residual stresses, plastic strain and hardening accumulated during the fabrication process, the latter (FEM virgin) assumes the material as virgin. Fig.36. Compressive test with numerical results and experimental ones: (a) reaction force, (b) ux displacement, (c) uy displacement vs z-displacement along the loading-axis; in the model FEM-stress the residual stress and accumulated strain from the previous process phases are considered, on the contrary, these aspects are not taken in account in the model FEM-virgin. Results It is interesting to observe that the FE simulation of the compression test is able to satisfactorily reproduce the buckling phenomenon and the overall displacement field if the actual material conditions are implemented in the numerical model. Conversely, if the numerical model does not incorporate the forming load history (FEM virgin in Fig. 36), the load-bearing capability of the structure is significantly underestimated, and the tube displacements differ to a larger extent from the experimental ones. Considering the load history experienced by the tube during forming improves significantly the accuracy of the FE structural analysis. In this way, the maximal magnitude of compressive load, which is crucial to assess the mechanical resistance, is accurately estimated with a relative error of 0.3%. However the FE model has difficulty in predict the evolution of the bucking phenomenon. 30 Heat induction bending Pipeline transmission is the dominant way for long distance transportation of oil and gas, because it allows high efficiency, high safety and low costs. So there is a persistent development of pipeline projects, in which bend pipes are important fittings for changing direction of oil and gas: about 30-40% of the pipes are bent during the construction.[8] Today, there is a very high demand for transporting oil and gas by pipelines with high pressure in order to increase the capacity: this service condition requires high strength. Since it can be performed in the field cold bending is preferred, but when the bend radius is small with a thickwall pipe it is necessary a hot bending process (Induction bending), which applies localized induction heating and after a subsequent water cooling, allowing the microstructure to be very changed and so the mechanical properties, such as the yield strength. So, through this thermal treatment an increase of hardness is obtained, but it is always followed by a higher brittleness: in order to guarantee the safety, subsequent tempering heat treatment is required to obtain a good toughness, relaxing also the microstructure from stresses.[9] Local induction heating is an advanced technique with important advantages: high efficiency, higher production quality and lower costs; however there some problems like springback and cross section ovality: during bending, the wall of the bending outside is thinned and the wall of the bending inside is thick-ened, the cross-section of the bend pipe becomes oval and bending angle suffers some spring-back caused by elastic deformation after unloading. In engineering design, the thinning of the pipe wall is not allowed to exceed 7% and the ovality is not allowed to exceed about 12%, so the bending radius is generally greater than 3.5 times the outside diameter of the pipe when using conventional pipe-bending techniques with local induction heating. Usually, the widely used materials are carbon steels, stainless steels (Duplex, Super), special steels such as Inconel and also Aluminium, Titanium, Monel, Copper, Clad Pipe; as we had already underlined the most classic application is for oil and gas transportation, but also important applications in petrol-chemical sector, power plants, industrial equipment, building constructions and also ship building. Induction Bending principle [8] [9] [10] Fig. 37. Induction bending machine The machine is composed of some mainly important components: pipe, pusher, coils for local induction heating, bending arm with a clamp, water cooling device and guide rollers, as shown in figure above. 31 During the pipe is bended, it is contemporary heated up by a local electro-induction process; the hot deformed zone is controlled then by water wall. Usually, induction heating coils and cooling coils are mounted to ensure uniform heating in the range of 850-1100 °C: the induction coil can be adjusted with a 3-plane movement. By adjusting the radius arm and front clamp, the required bend radius can be fixed. There is one pointer to display a correct degree of turning and arc lengths are marked on the pipe, which can be moved slowly, even if the bending force is applied by a fixed radius arm arrangement. Once everything is set as required, hydraulic pressure, water level, and switches are inspected and then the induction bending operation is started. Upon reaching the required temperature range, the pipe is pushed forward slowly at a speed of 10-40 mm/min, and the operation is stopped when the specified bend angle and arc length are reached. Just beyond the induction coil the heated material is quenched using a water spray. During the process of bending pipe, deformed metal is generally subjected to tension stress and the wall of pipe thins on the bending outside, while it is subjected to compression stress and the wall of pipe thickens on the bending inside. Moreover, the cross-section geometry of the bent pipe is oval. Process Parameters The important parameters that affect the induction bending process are: • • • • • • • Pipe Diameter Surface Contamination Process Parameters like Temperature, Speed, Cooling rate, etc Bend Radius Bend Angle Process Interruptions Hardenability of the Pipe Material, etc. Induction vs cold bending Induction Bending Must Slow Process Better control Highly efficient complex machineries are used for hot bending. Hot bending Process is efficient for larger pipe diameters High costs Requires lower force Cold Bending No Fast Process as no heating & cooling involved Less control Simple machines Not suitable for large pipe diameters. Cold bending is limited to smaller diameter piping only. Not so expensive Higher force FEM finite element model [8] As we have understood before it’s very important to control the thickness of the pipe to ensure its performance. Thickness variation is one of the most important design criteria and, also, quality control variables to ensure mechanical properties. To obtain a required uniform thickness, pipe pushing load is needed to control with revers moment. Large deformation elastic-plastic finite element method (FEM) is used to develop a coupling thermomechanical simulation system: the whole deformation process of bending pipe has been realized by the computer. By FEM simulation, stress variation, strain and temperature fields and cross-section geometry with bending pipe process can be obtained. In addition, the spring-back of bending pipe after unloading can be predicted. All of them are very important for guiding practically the process. 32 FEM involves multiple non-linear factors (material and geometry), and contacts during the de- formation process of bending pipe. In addition, local induction heating technique makes it more difficult to obtain the model and solution. For the simulation, elastic-plastic large deformation FEM is used to develop 3D coupling thermo-mechanical simulation system for the solution. In this particular investigation, we consider as the stock' s material the pipeline steel X70: it is assumed isotropic and the Von Mises yield criterion for yield behaviour. To model the material behaviour the following equation is used, given by a constitutive model: the flow stress at which the material starts to flow is function of the equivalent strain, the equivalent strain rate, and temperature. Induction bending is a thermo-mechanical process analysed through solving the heat conduction equation coupled with deformation analysis. During the bending process, the temperature distribution of the pipe can be calculated using the partial differential equation: (where k, is thermal conductivity; p is density; c is specific heat; qv is heat generation term representing the heat released due to plastic work; x, y are position coordinate; t is time) In this evaluation we have to consider the water-cooling process and so find also the related heat loss, by considering a convection heat transfer process. Coupling thermo and mechanical issues, a FE model for hot elbowing pipe has been built in this study. According to the symmetry of the pipe bending structure about the x, y plane, only a half-part of +z is analysed in order to reduce the number of element. Guide rollers are defined as rigid along with ignoring the influence of friction between rollers and pipe on bending pipe to reduce simulation time. Bending arm and clamp are replaced with large rigid elastic elements. The effect of these simplifications on bending pipe is weak and fits the practical process. The main calculating conditions are the following: pipe original diameter (Do), 1 000 mm; wall thickness (b), 20 mm; bending radius (R), 6 000 mm; pushing speed (v), 0. 5 mm/s and so on. The FEM meshed model is shown in the following Figure, in which 2858 3D solid elements are involved. Fig. 38. FEM meshed model 33 Analysis and discussion After using the simulation system previously developed and loaded, the whole forming process of pipe bending was simulated. The pipe shape after loading is shown in figure 39 with a bending angle 94", while 40 gives the pipe shape after unloading. It is evident that a greater springback of pipe-bending occurs. Fig. 39. Pipe shape after loading Fig. 40. Pipe shape after unloading The temperature distribution in the local heating zone is shown in the subsequent figure, during the bending process. First, the node temperature varies from room temperature 20 °C to 900 °C, and then drops quickly after a while due to the heat radiation and water cooling. The circumferential equivalent plastic strain at a certain cross-section of the deformation area is shown in 42, when bending angle is about 45°. Fig. 41. Temperature distribution in local heating zone Fig. 42. Circumferential equivalent plastic strain The equivalent plastic strain value decreases at the former half stage and increases at the later stage from bending outside to bending inside. Furthermore, strain of bending inside is more than outside. So, the distribution rule agrees with that of pipe-bending theory. The relationship between wall thinning or thickening ratio of the pipe wall thickness and bending angle for the outside, middle and inside of bending is shown in Fig. 43. Wall variation at the frontend of pipe is relatively greater than any other. The main reason is that the deformation is difficult at the beginning of bending. According to the engineering quality specification of the thin-wall and large-diameter pipe, the thinning ratio of the wall thickness on the bending outside is not allowed to exceed 7%. Generally, the change ratio of the wall thickness is expressed in the form of the following equation: ∅= π0 − π1 π0 34 where ∅ is change ratio of the wall thickness; b0, is initial wall thickness; and b1, is formed wall thickness. In addition, it can be seen from Fig. 43 that the wall thickness increases at the middle. The neutral layer of pipebending process leans to the bending outside. Based on the bending theory of material mechanics, this phenomenon will occur as deformation cross-section is subjected to axial force besides shear force and torque force. So, the simulation results are in accord with that of theory. Also we must consider quality requirements for engineering design for the ovality (10-12%) of the bent pipe, calculated by: πΏ= ππππ₯ − ππππ (ππππ₯ + ππππ ) β 2 Fig. 43. Wall thickness and thickening The relationship between the ovality and the bending angle is demonstrated in the figure 44. With an increase in bending angle the ovality firstly increases, and then decreases near to the end of bent pipe as shown in figure 44 (a). Moreover, it can be also seen that the ovality of the pipe after unloading is less than that prior to unloading. The relationship between spring-back angle and bending angle after unloading is illustrated in figure 44 (b). The spring-back angle curve shows a gradual increase with the bending angle. Figure 45, instead, presents the relationship between the ovality of certain cross-section near the leading end and load step. At the beginning of bending, the ovality increases gradually with an increase in load steps, but reaching a high value, it remains at a steady value. Fig. 44. Ovality/spring back angle with bending angle (a/b) Fig. 45. Ovality with load step 35 Laser bending Laser bending is a noncontact method that used laser energy in order to obtain 2D bending, 3D shaping, and precision alignment of metallic and non-metallic components: sheets and tubes can be formed depending upon the non-uniform thermal stresses induced by laser heating. Compared to laser bending of sheet metals, the laser tube bending is more complex because much more process parameters are involved. It is also become one of the basic parts of laser metal forming because it overcomes the troubles that characterize the tube bending technique. Moreover, tube bending is an attractive application in many high-technological industries such as aerospace, aviation, automobile, shipbuilding, energy and health care. The bending process is based on the non-uniform distribution of temperatures on specimen surface, which generate different thermal stress between the irradiated area and the cold side of the tube. The mechanism used in laser tube bending is Buckling mechanism where the diameter of the laser beam is chosen to be much bigger than the tube thickness as seen in Fig. 46. Hence the inner side of the tube which faced the laser beam is homogenized heated in thickness direction causing material expansion. The expansion is restricted by unheated material which leads to plastic compression. Due to the cooling, the material suffers shrinkage and shortening in axial direction forcing the tube to bend towards of laser beam. [11] [12] Laser Bending principle [11] Fig. 46. Circumferential scanning scheme Laser tube bending is similar to the heat treatment process and the surface temperature of the material has to be less than the melting point. During the process there is a lot of nonlinear phenomena like the temperature, microstructure, and stress field changes: all are significantly related to each other. In this case we can describe a tube with circumferential scanning shown in the figure above: the tube rotates on its axial axis around 180°, where the laser light is defocused on the tube surface with laser beam diameter much bigger than tube thickness. The heated area suffers wall thickening and compressive plastic deformation because of the thermal expansion limitation by the unheated material. When the laser source is turning off a rapid cooling occur, with material that shrinks on the heated surface. As a result, the shortening of the material in the axial direction forced the tube to bend towards of laser source. During bending process heat is generated because of strain energy, but it is very small compared to input laser beam energy so it can be neglected. There are many influencing factors on laser tube forming, such as forming process parameters, material’s properties, optic parameters, and geometrical parameters. Figure 47 gives factors of influence on laser tube bending. 36 Fig. 47. Factors of influence The melting point and yield strength are basic conditions to restrict to the forming process. For a given forming objective, material’s properties and geometrical parameters of the tube are already decided. The aim of the laser tube bending is to acquire the desired bending deformation by the reasonable match of the process parameters. As a high nonlinear forming process, it is hard to express the relationship of bending angle and process parameters by using a general formula; so the finite element method is a powerful tool to study the process. Using finite element method (FEM) process simulation, can be not only the final bending angle be acquired, but also kinds of fields, such as the stress, strain, displacement, and temperature fields. The FEM process simulation is an ideal researching method and very complex: a series of treatment techniques are necessary according to the characteristics of the laser tube bending. In this paper, it is adopted a thermomechanical FEM model considering the correlation between material’s properties and temperature. To some extent, the FEM process simulation can realize the optimization of forming processes, too. However, it is hard to obtain the optimum process parameters only through FEM simulation. FEM simulation It’s considered a typical coupled thermal-mechanical problem, where the temperature and displacement fields affect each other. The finite element method based on the 3D thermal elastic–plastic was established to realize the simulation of the laser bending of tubes and so to reduce analytical errors significantly. The fields of thermal stress, temperature, and displacement can be obtained simultaneously by using the coupled model. In laser bending process, the temperature and deformation of the heated zone are characterised by sharp variations during a very short time. Steep temperature and stress gradients occur. Higher precision can be acquired using brick element. Thus, brick elements are selected when the sheet is discretized during simulation. The brick element is divided into full integral element and reduced integral element according to numerical integral way. For bending problem of the thin sheet, we can obtain unreasonable results because of larger precision loss, while selecting full integral element. Using reduced integral way, not only the calculation time of integration can be decreased, but also analytical precision can be improved because of eliminating the influence of imperfect high-order term. Meanwhile, a fine mesh is necessary around the laser beam path due to steep temperature and stress gradients in the heated zone. It is essential to have at least eight integration points in the thickness direction; in order to cut the total number of elements in the simulation, a coarse mesh is used outside the heating zone. 37 1. Boundary and initial conditions The natural cooling-down method is adopted in the paper: in other word, the tube is cooled down naturally in the air. Heat convection and radiation exists between the tube and the surroundings, but therefore, the irradiation boundary condition is nonlinear. The displacements at one end of the tube are restricted, and the other surfaces are displacement free. It is assumed that the tube is initially stress free and strain free. Initial temperature of the tube is 20 °C. 2. Simulation A specimen of Φ 10×80×1 mm (diameter × length × thickness) is analysed. Twenty-node brick elements are used. The total element number is 10,400. The finite element mesh is shown in figure 48; in 49 instead the scanning process. Fig. 48. Finite element model Fig. 49. Sketch of laser scanning process The process parameters are: • Laser output power: P=1 kW • Scanning velocity: v=1.5 m/min • Laser spot diameter: d=5 mm • Initial temperature of sheet metal: T0=20 °C • Scanning wrap angle: 180° The right figure shows the sketch of the laser scanning process and cross-section of the tube in the heated zone. The central zones of laser scanning side and non-scanning side are selected to study the characteristics of the laser tube bending. 38 • Single scan process Fig. 50. Temperature with time Fig. 51. Axial stress with time The left figure shows the variation of temperature with time. The laser spot gets to the zone I at 0.294 s. The irradiation lasts for 0.196 s. The temperature reaches its peak value at 1,099.25 °C at 0.49 s. Subsequently, the temperature drops down sharply until it reaches to the room temperature. The stress and strain vary with the change of temperature greatly. Right figure, instead, gives the variation of axial stress with time. At the beginning, the stress state is tensile: when the laser spot irradiates the zone I, the thermal expansion of materials due to temperature rise goes up, but the surrounding materials restrict the expansion and then the stress of zone I become too compressive. At the beginning of the cooling stage, the stress in axial direction recovers from compressive to tensile because of the shrinkage. Subsequently, the tensile stress diminishes gradually. Finally, the stress state is in compressive due to the more shrinkage of material. Zone II is not irradiated directly by laser beam. The temperature rises slowly due to the heat conduction. The temperature reaches its peak value at 111.64 °C at 5.80 s. There is a great gradient between the laser scanning and non-scanning side. At the beginning, the stress state in zone II is compressive. The stress keeps a lower value all the time. Thus, the plastic deformation in zone II is not induced. Fig. 52. Axial strain with time Fig. 53. Displacement with time In figure 52 is showed the variations of the thermal strain, total strain, and plastic strain with time. The change of thermal strain is similar to that of temperature. The thermal strain increases from zero to its peak value 0.0143, and then decreases to zero finally. This corresponds with the movement of laser spot. The axial plastic strain is dependent upon the axial stress. At the beginning of the laser irradiation, the material in the heated zone expands due to the temperature rise. When the temperature continues to go up, the axial stress of in the heated zone becomes too compressive because of the restriction of 39 the surrounding materials. The axial compressive stress rises greatly with the increase of the temperature. In the meanwhile, the yield strength decreases with the temperature rise. The compressive plastic deformation is brought out in the heated zone. Obviously, the gradient and development of the temperature between the laser scanning side and the nonscanning side leads to the changing complexity of the stress and strain. The final strain of the laser scanning side resulting by the shrinkage of the heated zone is −0.0084 in axial direction. The length of the laser scanning side becomes shorter than that of non-scanning side after cooling. The length difference between both sides makes the tube produce the bending angle. • Multi-scan process Fig. 54. Axial plastic strain/time Fig. 55. Bending angle with n of scans A bending angle acquired in a scan is small. In order to obtain the desired bending angle in practical engineering, the tube has to be scanned repeatedly along the same scanning path. Figure 54 gives the variation of the axial plastic strain of the irradiated side with time. The axial plastic strain of the irradiated zone becomes larger and larger with the increase of the number of scans. Meanwhile, the axial plastic strain of the non-irradiated side is almost zero. The different step of the axial plastic strain difference is generated between two successive scans. Thus, the axial plastic strain difference between the irradiated side and non-irradiated side augments in the multi-scan laser bending process. Consequently, the bending angle of the tube increases. Figure 55 instead, gives the influence of the number of scans on bending angle, it is seen that the relationship between the number of scans and the bending angle keeps linear; the bending angle acquired in the first scan is largest. • Conclusions - The gradient and development of the temperature cause the changing complexity of the stress and strain. Consequently, the length of the laser scanning side is shorter than non-scanning after cooling. The length difference between both sides makes the tube produce the bending angle. The relationship between the number of scans and the bending angle is about in direct ratio. - 40 Shear bending Tube shear bending is a suitable bending technology to realize a considerable small bending radii in order to minimize the space of bent tubular parts. The forming process can also be studied through an elasto-plastic 3D finite element simulation and the results on the deformation behaviour and strain distribution demonstrate that the forming process is a combination of shearing and bending deformations. Nowadays light weights, low cost and compact products are the most requested on the market moreover tubular components, especially, with small bending radii need to fit into tighter spaces to be adopted in hydraulics, fuel piping and in many devices used in narrow spaces such as automotive engine parts and air conditioning appliances. Another important challenge in case of tubes is to produce one-piece tubular form to replace casting or multiple welding operations. With conventional bending methods the minimum bending radius that can be obtained is almost more than 1– 2 times of the tube original diameter, with shear bending of tubes on the other hand very small radii, compact and unified elbow tubular parts can be manufactured. A typical frame of a shear bending equipment and a schematic illustration of the process are shown below in the pictures 56 and 57: Fig. 56. Schematic sketch of shear bending equipment 41 Fig. 57. Schematic illustration of shear bending process The tube (#4, Fig.57) is inserted in the hole of dies (#5,#7) and the mandrels (#6, #8) are placed inside where the mandrel of moving die is fixed to the moving die itself. Then using a hydraulic cylinder (#11) the moving die slides and applies a shearing force on tube; at the same time, another hydraulic cylinder (#1) pushes the end face of tube and material is supplied into dies during the process. Thanks to this configuration the tube is undergoing a shearing deformation continuously. Two load cells (#2, #10) measure the values of pushing and shearing forces and the hydraulic cylinders are connected to a hydraulic power pack and controlled independently. The final shape of the tube is a crank shape product and ππ and ππ radii (die corner radius and mandrel head radius, respectively) determine the inside and outside bending radii of the bent tube. Experimental data Process parameters used: - A1050, aluminium tube diameter π·0 =30mm wall thickness π‘0 = 1.5 mm die corner radius 3 mm radial clearance of die-tube and mandrel-tube of 0.1 mm During the shear bending process, to overcome the friction force and to avoid the extreme thinning outside, a pushing force is applied on the tube-end side; the resulting product obtained by applying different values of the force is shown in the following picture: Fig. 58. Typical products by various pushing forces: (a) splitting; (b) successful; (c) wrinkling; (d) rupturing 42 Only in the case (b) a correct shear bending product is obtained, if the pushing pressure on the tube is not enough to send material into the dies a tube splitting happens (case(a)) and, on the contrary, if the pushing pressure exceeds a threshold a wrinkling in the tube outbreaks (case(c)), finally, in the worst case of too high pressure applied there is the rupture of the workpiece (case(d)); therefore the amount of pushing force should be accurately measured. Finite element simulation settings: - ELFEN finite element code By exploiting the symmetrical shape, only half of the model is considered Dies and mandrels are assumed to be rigid Tube is modelled using solid elements The number of tube elements in the thickness are 4 and in circumferential direction are 20 A mesh refinement has been done in the axial direction where there is a severe deformation The bending angle is 90° The bending die has been moved by 100 mm The used flow stress-strain relation of the tube material is π = π(π)π The isotropic hardening rule is assumed Von Mises yield criterion is used The following table collects the simulation parameters: During the experiment many values of pushing pressure have been used, only for P=30 MPa no failure mechanisms have occurred in tube. On the base of the deformation of the meshes around different routs (T=Top, L=lateral, B=bottom), the amount of deformation can be stated: the elements on the tube lateral side (route L) have sustained more shearing deformation if compared to the top and bottom sides, in fact approaching to routes T and B the degree of deformation reduces and elements in these regions have been subjected to significantly less shear distortion. The figure number 59 shows what said: 43 Fig. 59. Finite element meshes after the deformation In a conventional pure bending process the tube suffers only a pure bending moment since the inside and outside bending radii are concentric and there is a single bending centre; on the contrary in shear bending, on the base of the die geometry, each element of tube feels an additional shearing force as well as a bending moment. The amount of deformation shown by each element is related to its position in the peripheral direction: elements around routes T and B are subjected to a high bending deformation due to die geometry whereas the maximum shearing occurs on route L. FEM results The effective strain distribution after the bending process is shown below: Fig. 60. Distribution of tube effective strain It can be seen clearly that in the vertical part of the tube the effective strain along tube longitudinal direction is uniform whereas it varies in tube width direction. 44 When π = 90° the maximum effective strain (ε≈1) is reached in the region across tube width and the result is in good agreement with the corresponding theoretical value for a 2D ideal shearing process. It can be seen that moving towards the centre (route L), the bending component reduces as the shear components increase until the shear strain becomes uniform and reaches its maximum value. Moving to route B or T, since the shearing deformation decreases, the effective and normal strains reduce. Figure 61 shows the effective strain distribution of tube outer layer in the steady state region across tube width (lines TB) and the x and y strain components: Fig. 61. Strain distributions across tube width in the steady state region (line TB)- maximum strains have occurred around π = 90° For the value of the tube thickness strain along routes T and B a comparison between simulation and experimental results is shown in Fig. 62. Fig. 62. Distribution of thickness strain along routes T and B In this experiment a pushing pressure of 55 MPa has been applied on the tube. Point C on route B (Fig.62) and point F on route T are under compression in the longitudinal direction and thickness strain in these regions is positive; on the contrary point D on route T and point E on route B are under tension and therefore thickness strains in these points are negative. [13] 45 Effects of initial thickness in shear bending process In the previous pages the effects of initial thickness on the deformation characteristics during shear bending process wasn’t discussed, in the following paper this aspect has been investigated both by experiments and numerical simulation through the use of FEM (ELFEN code). The results show that, under the effect of same pushing pressure, increasing the initial thickness the tube thinning and cross section deformation increase. The ability of a tube to be bent is strictly correlated to geometrical factors such as the bending radius, the diameter and the thickness of the tube. Experiments and numerical simulations show that during shear bending processes no failure occurs only if the pushing pressure is inside a suitable range and performing some experiments it has been demonstrated that the range of pushing pressure for a successful forming enlarges if the tube initial thickness is increased. The following picture illustrates both the experimental and simulated results related to the cross section deformation using tubes with different thickness value but with the same amount of pushing pressure (P= 40 MPa) applied to understand the effect of the initial thickness on the deformation behaviour. Fig. 63. Cross section configurations of the deformed tubes with different initial thickness obtained by simulation and experiment: (a) simulation, (b) experiment and(c) experiment (t0=3.0 mm) The thicker the initial tube is, the greater its cross section deformation is, especially around the inner part of the bending and this result is completely against the traditional behaviour in conventional tube bending; the reason is that increasing the initial thickness the bending rigidity of the tube rises and the bending becomes harder as a direct consequence. Therefore, the tube material cannot follow the die shape well and the tube cross section needs to sustain more deformation. What stated is shown in Fig. 63 (c), in fact a gap is formed between the tube and die wall due to a shrinkage registered along the diameter. However, increasing the pushing pressure ovality of the cross section can be reduced; the next figure shows the experimental and simulation results of the relation between the pushing pressure and the cross section ovality of the deformed tube. 46 Fig. 64. Effect of the axial pushing pressure on the cross section ovality of the deformed tube The thinning of tube wall during shearing deformation increases the lower the pushing pressure is since during the process the tube sustains elongation and its cross section deforms; if the pushing pressure is enough high the ovality is prevented, always assuring not to have wrinkling as a drawback of an excessive pressure applied. π‘ π‘0 Another aspect is considered: the effect of the initial thickness on the thickness strain ππ‘ = ln ( ). Experiments and simulation state that thickness strain decreases with an increasing of the initial thickness. Fig. 65. Configuration of the deformed tube. Fig. 66. Effect of the initial thickness on the thickness strain of the tube-sheared part obtained by simulation and experiment. In order to explain the reason of this behaviour, the relation between the pushing and shearing strokes of the moving die (Figure 67) as a function of the tube initial thickness and the effect of the tube initial thickness on shearing load (Figure 68) are plotted. 47 Fig. 67. Relation between the pushing and shearing strokes as a function of the tube initial thickness obtained by simulation Fig. 68. Effect of the tube initial thickness on the shearing load obtained by simulation It can be seen that in the graph shown in figure 67 the slope of the curves decreases by increasing the tube initial thickness, this means that a lower amount of material is supplied into the dies increasing the tube initial thickness. So under the action of the same pressure, the tube shows more elongation and the shearing load increases by increasing the tube initial thickness, as displaced in figure 68. Moreover by applying an higher pushing force on the tube it is possible to understand that thinning can be reduced; this is confirmed by the results shown in the following figure: Fig. 69. Effect of the initial thickness and pushing pressure on the distribution of thickness strain along the hoop direction of the tube-sheared part These curves demonstrate that applying an higher pushing pressure on the thick tubes can prevent the tube wall from thinning. [14] In conclusion, to summarize what has been said previously: - Increasing the tube thickness, the bending rigidity increases too, as a consequence the tube cannot follow the die profile and the cross section ovality increases accordingly. Raising the applied pushing pressure the cross section high ovality can be prevented. A thick wall tube has more tendency to thinning and rupture with respect to a thin tube but to reduce these drawbacks the applied pushing pressure must be as high as possible within a range in which wrinkle does not occur. 48 Push bending 3D free bending process is an advanced metal forming technology, characterised by high flexibility and efficiency. It can be easily found that copper tubes filled with low melting point alloys are able to induce the most uniform stress distribution, associated to better formability and moderate thickness changing rates. Tube bending process is one of the main used manufacturing technologies for lightweight production and 3D free bending technology has the big advantage of being suitable for most tubes and profiles due to its geometrical flexibility and efficiency. It can be used to produce complex hollow tubes with asymmetric cross-sections and continuous varying radii. Also, the bending of different R values and angles can be achieved without changing the die, which greatly extends the flexibility of the bending machine. On the other hand, during the tube bending process, wall thickness and cross section changes are unavoidable. However, due to the axial force, wall thickness and cross section changes are restrained during 3D free bending technology. 3D free bending is a dieless forming technology, making it hard to keep stable and restricted springback during the bending process. However, several materials could be used inside the tubes for supporting roles. These materials are not only used for restraining thickness changes and springback of the tubes, but also to further improve formability. [15] Figure 70 shows 3D bending process, during which the internal side of the tube is subjected to compressive stress while the outside is associated to tensile stress. For this reason a thickening of the inner side of the tube and reduction of its outer portion is obtained. In the meantime, the tube was subjected to feeding force, so the neutral layer offset outwards to restrain reduction of the wall thickness and reduce the deformation of the cross section of the bent tube. The tube bending stress and strain analysis can be represented always in Fig.70, where ππ and ππ are bending radius of inside and outside wall of tube, respectively, t is the thickness of tube at any point and t0 is the initial thickness of tube, N is the interaction between filling materials and tube, and P is the feeding force. Fig.70. Tube bending stress and strain analysis FEM simulation: ABAQUS/explicit is applied to numerically simulate the forming process and the model used for this process is shown in Fig.71, including a tube, a bend die, a spherical bearing, a guider, a pressing and different filling materials. Each die model created during the geometry modelling was oriented in its own coordinate system and was independent of the other geometry models, then the dies and tube were assembled in the global coordinate system. These data were put in CATIA software and then imported to ABAQUS/explicit FE code. The tube was made of copper and filling materials were steel balls, PU rubber and low melting point alloys, respectively. The interaction among the parts is set to general contact and the analysis step is adjusted to dynamic explicit. Since the lubrication unit is designed near the bend die, the coefficient of friction between the dies and the tube is assumed as a very low value. Thus, based on a great deal of simulation results, the coefficient of friction between the dies and the copper tubes were identified as 0.02. Meanwhile, the pressing and guider are set to encase. The tube specifically built for a specified movement speed in the Z direction and the bearing is set having two movements respectively in the X and Y direction, while the remaining degrees of freedom are all fixed. 49 Fig.71 FE modelling of tube free bending To simulate the free bending process using different materials, simulations for four copper tubes are carried out based on the conditions summarized in the table. The filled materials included steel balls, PU rubbers, and low melting point alloys. The relative position sizes are shown in Fig. 72. Fig.72 Model assembly position Influences of different filling materials in 3D free bending can be investigated through a simulation without filling materials (1). The simulation results are represented in Fig.73. Fig.73. Bending simulation under condition (1) The simulation results under condition (2) are presented in Fig. 74. The simulation data of bending process using low melting point alloys are represented in Fig.75. The low melting point alloy establishes a complete contact with tube wall after solidification and the stress distribution is uniform during bending process. Ideal forming results are indicated in Fig.75. The simulation results under condition (4) are shown in Fig. 76. Due to the lowest elastic modulus, PU rubber was unable to provide internal pressure. If compared to condition (1) (Fig. 73), during the forming process, the tube bears less stress and the stress distribution is more uniform when the filling materials were replaced with PU rubber. 50 Fig.74. Bending simulation under condition (2) Fig.75. Bending simulation under condition (3) Fig.76. Bending simulation under condition (4) The experimental process and results under different conditions are shown in Fig. 77. Fig.77. Experiments of 3D free bending under different conditions 51 Results: • • • • The interaction N between filling materials and tube has a significant effect on formability. The more the uniform distribution of N is, the better the formability and moderate change in thickness attained along the radial direction. Compared to the flexible filling, copper tubes filled with solid filling (such as low melting point alloys and steel balls) could contribute to better formability and less change in thickness. On the other hand, filling tube with steel balls led to bad formability and great change in thickness than unfilled tubes. Meanwhile, large numbers of spherical indentation are developed on the inner surface results in degradation of the inner surface quality. The ellipticity decreased regardless of the filling material and filling with low melting point alloys reduced ellipticity to 1.467%. A 3D FEM-DEM coupling numerical model can be usually developed to investigate the interaction between thin-wall elbow and granular filler for a granular-media-based thin-wall elbow push-bending process. [16] This procedure is about filling a tube through granular media and pushing it into a die to bend a tubular blank into an elbow shape. Depending on the mechanical properties of granular filler an elbow tube having t/D<0.01 (wall thickness/outer diameter) and R/D<1.5 (ratio of bending radius/outer diameter) can be produced. To use in a correct way the 3D FEM-DEM model it should consider both the tubular blank deformation behaviour (since it is continuum, FEM is applied) and mechanical properties of granular filler (based on discrete media, so a discrete element method DEM will be applied). Referring to traditional normal-diameter normal-wall elbow tubes push-bending process, it consists of pushing a tube between a die and a mandrel to bend a tubular blank obtaining a final elbow shape. Unfortunately, there exist many problems using the traditional push-bending process to form large diameter thin-wall elbow tubes, especially wall thinning-fracture and wrinkling. For that reason, in recent years granular-media-based forming process has been developed and it’s now commonly used due to its flexibility, resistance to high forming temperature and good characteristics associated to pressure-transmission. While on one side, in traditional push-bending process, forming force of punch is immediately imposed on tubular blank with filler providing passive supporting in the tube, on the other side, in the new tube bending process, forming force of punch is applied to granular media which, with their flow in mould cavity, transform the tubular blank into a curvature. After a 3D FEM-DEM is developed, using this coupling model, some key forming parameters, such as forming force, wall thickness distribution, wrinkling are simulated and compared to experimental results. About the mechanical process, a tubular blank with predetermined dimensions and filled with granular media is inserted into the die and then deformed according to the shape of the die. The tubular blank is bended under forming force of punch, constraint force of die, and interaction of granular filler. The forming force of punch and constraint force of die can be analysed and studied using FEM simulation. On the contrary the granular filler cannot be simulated through FEM, since it’s a discrete system, not a continuum one. 52 Fig. 78. Schematic of an elbow granular-media-based push-bending process: (a) before bending, (b) bending In the granular-media-based push-bending process, the deformation behaviour of tubular blank can be determined by minimum principle of potential energy: (1) where {δ}^e is displacement vector of nodes, [k] e is element stiffness matrix, {F}e is equivalent nodal forces which can be expressed as: (2) where {P} collects forming force of punch and constraint force of die, {N} is interaction between tubular blank and granular filler. Considering Eq. (2), Eq. (1) can be rewritten as: (3) In the granular-media-based push-bending process, the mechanical characteristics and motions of granular filler can be determined by minimum principle of generalized complementary energy: (4) where {ζ}e is displacement vector of discrete elements in granular filler, [D(α)]e is tangent stiffness matrix of granular media, {N} is interaction between tubular blank and granular filler, {T} collects external forces except {N} and α ≥ 0 denotes granular media only produces compressive contact forces. The third and fourth equations can be numerically solved with FEM (finite element method) and DEM (discrete element method) to analyse the granular-media-based push-bending process. Talking about FEM-DEM coupling method, DEM simulation involves following the motion of every particle in the flow and modelling each collision or contact between particles and between particles and boundary. The time evolution of the particles is advanced using an explicit finite difference scheme. A damping is imposed on the dynamic equations of motion to dissipate kinetic energy and obtain static equilibrium. The granular material considered cast iron and all the properties are listed in the first table below. In addition, the friction coefficient is set to 0.1 in DEM. The main geometry conditions associated to tubular blank are reported in the 53 second table. The size considered for a single FEM element of tube is 1mm x 1mm to obtain the best compromise in terms of accuracy and stability. The die and punch were modelled in FEM to be rigid. Tubular blank is set to 1Cr18Ni9Ti stainless steel, the material parameters are listed in the third table, and the stressstrain curve of 1Cr18Ni9Ti is shown in Fig. 3. The friction coefficient is set to 0.02 in FEM. The forming process simulation is based on the following passages: at first a downward displacement is applied on the punch and it produces a forming force on the granular fillet. The fillet confined in tubular blank is compressed, applying contact forces on tube inner walls. The contact forces between granular filler and tube drive the tubular blank into a curvature. The contact forces between granular filler and tube are calculated in DEM, and the boundary determined by the geometry of deformed tube is provided by FEM. Simultaneously, at the end, the tube deformation is simulated through FEM while the forming force (due to contact forces between the granular filler and the tube) is given by the application of DEM. To couple FEM and DEM models, a particular module for data-exchanging needs to be set. Applying it contact forces on the rigid wall elements in DEM and corresponding deformable tube elements in FEM can be easily associated. The contact forces on a wall element in DEM are equally imposed on the nodes of the corresponding tube element in FEM in a format of concentrated force. The nodes’ coordinates on a tube element in FEM are used to create the corresponding wall element in DEM. At this point the total time consumption of 3D FEM-DEM analysis can be split in three parts: time-consuming of DEM simulation for granular media π‘π·πΈπ , time consuming of FEM simulation for tubular blank π‘πΉπΈπ , and time-consuming of data transmission π‘πππ‘π−π‘ππππ . The total consumption of time is finally expressed as: π‘πππ’πππππ = π‘π·πΈπ + π‘πΉπΈπ + π‘πππ‘π−π‘ππππ For the granular-media-based push-bending process, π‘πππ‘π−π‘ππππ mainly depends on read/write of files estimated by data volume, π‘π·πΈπ and π‘πΉπΈπ are mainly based on total number of elements, which can be determined knowing element size and dimension of tubular blank. Wrinkling and bending phenomena can be observed imposing the element size of FEM almost equal to the particle size used in DEM. That means that, in conclusion, the total time-consuming π‘πππ’πππππ depends mostly on particle size of granular media and tubular blank dimensions. Generally, in 3D coupling analysis for granular-media-based push-bending process, the most time-consuming is about 200 h. Fig. 79. FEM model of tubular blank and DEM model of granular filler 54 Fig. 80. Stress-strain curve of 1Cr18Ni9Ti In order to fully analyse the fundamental mechanism of the granular-media-based push-bending process the distributions of stress in granular filler tube can be represented in Figure 81. From there it can be seen how an important stress concentration occurs at the area A in granular filler corresponding to the major deformation area B in tube. The starting point of local buckling and cross-section deformation are positioned in the area B of tube. So, in conclusion, the stress concentration occurring at the area A in granular filler prevents the presence of wrinkling and cross-section deformation. Fig. 81. Distributions of stress in granular filler and tube: (a) granular filler, (b) tube Forming force is an important process parameter determining choice and design of forming equipment. It is hard to calculate forming force of granular-media-based push-bending process applying current traditional models and formulas, because of the deformation characteristics of granular media and interaction between tube and granular filler. At this point through 3D FEM-DEM approach, the forming force of the process can be calculated and compared to experimental data obtained, which are graphically represented in Figure 82. 55 Fig. 82. Forming force of the granular-media-based push-bending process: simulation and experimental results Experimental data shows slower and slower increase because lubricating oil between tube and die is gradually squeezed out leading to friction changes larger. The coupling model cannot reflect this working condition. Wrinkling is one of the most important defects in thin-wall elbow tube bending. Because of additional axial compressive force introduced by ‘push’-bending, the wrinkling at inner bending arc becomes more severe. As it can be noticed looking at Figure 81. stress concentration of granular filler present at the deformation area prevents the thin-wall tube wrinkling. It must be noticed that not all the granular filler can be correctly used to face wrinkling phenomenon. Through Figure 83 the effects of particle size on wrinkling of thin-wall elbow tube during granular-media-based push-bending process are represented. Contacts between granular filler and tubular blank provide supporting forces for inner bending arc of thin-wall elbow. The contacts’ distribution on tube wall is unfolded into a plane and shown in Figure 84. Fig. 83. Effects of particle sizes of granular fillers on wrinkling of thin-wall elbow tube: (a) particle size 0.98 mm, (b) particle size 1.58 mm, (c) particle size 2.08 mm 56 Fig. 84. The contacts’ distribution on tube wall (unfolded into a plane): (a) particle size 0.98 mm, (b) particle size 1.58 mm, (c) particle size 2.08 mm A novel bending method, based on the principles of a push bending process, has been developed to create an elbow section starting from a straight tube, making advantage of an inner pressure and a rigid die. [17] The die guides the initial tube into the elbow, then it is pushed by a punch, while internal pressure is used to avoid wrinkling and buckling phenomena. Going on a FEM simulation will be applied and two different cases will be compared and analysed: using internal hydraulic pressure and using a urethane rod to supply the internal pressure. Fig. 85. The tube bending process treated A great number of bending processes for transforming a straight tube into an elbow section already exist, but, at the same time, they present many disadvantages. The new process developed here uses a urethane rod inside the tube. This tube is then set inside the die and a plunger located at the edge end of the tube and rod exerts an axial loading pressure on both. In addition a counterpunch is used to control tube and rode at the other extremity. The final result will give the tube and rod pushed into the die, creating an internal pressure in the tube. At this point a finite element model can be developed to describe the tube deformation and the stress/strain distributions inside the tube. The FEM/FEA (or finite element analysis) method applied in this case is LSDYNA, a non-linear software. The method of solution used by LSDYNA 3D provides fast solutions for large deformation dynamics and complex contact/impact problems. Using this integrated product, the problem is pre-processed in ANSYS, the explicit dynamic solution is obtained using LSDYNA 3D and results are manipulated using the standard 57 ANSYS post-processing tools. Taking advantage of symmetry a half of the problem is modelled using 3D solid brick elements. Traditionally, shell elements have been favoured for use in simulating bulge forming and pipe bending operations, but in this case the aim was just to give a proper representation of the deformation of the urethane rod. Solid elements are actually used to model the tube and the die, to avoid difficulties related to the mixing of different element types. The tube material behaviour can be described only through a non-linear elastic-plastic material model, since it is plastically deformed. In LSDYNA there are two strain rate independent plasticity models available: the classical bilinear kinematic hardening model and the classical bilinear isotropic hardening model. Both models use two slopes, the elastic modulus and the tangent modulus to represent the stress–strain behaviour of the material and can be used satisfactorily for most engineering metals. The only difference between the two models is the hardening assumption, where the kinematic hardening model assumes secondary yield to occur at 2σy while isotropic hardening occurs at 2σmax. The urethane rod behaves in a non-linear elastic manner and the LSDYNA software provides three options for modelling non-linear elastic materials: the Blatz-Ko rubber model is used for compressible foam type materials such as polyurethane rubbers, the Mooney–Rivlin model is used to model incompressible rubber materials and the viscoelastic material model is used to model the behaviour of glass and glass-like materials. In this case the Blatz-Ko model was chosen as the most appropriate material model. Poisson’s ration is automatically set to 0.463 when using this model and material response is defined through the strain energy density function: with πΌ2 and πΌ3 strain invariants. The rod shear modulus was experimentally determined. Starting from the ANSYS pre-processing tools, simulations are processed to produce a 90° bend in a straight tube of inner diameter 37.6 mm, outer diameter 40 mm and length 140 mm. The die had an inner radius of 40 mm and a die bend radius of 70 mm. The outer diameter of the urethane rod was 37.6 mm and its length was 140 mm. Fig. 86. Meshed geometries of die, tube and rod During the analysis an algorithm is developed to guarantee the correct surface contact between die and tube and between tube and urethane (rod). The die is modelled as a rigid body and a loading curve defining displacement over time is applied to the nodes at the free end of the tube and rod to model the punch force. An elastic Coulomb friction law is used adopting a coefficient of friction equal to 0.1 between die and tube and 0.15 between tube and rod. 58 Problem: About the simulation results, the first thing noticed is that the meshes associated to tube and rod needed a redesign at the area nearest the die bend. This is because it is the first deformed location with significant contact problems. To solve those problems the mesh density was distorted to increase then nearer the die bend. Despite that approach the redesigned mesh is still subjected to problems related to the contact algorithm. A clear example is shown in figure 87, where the rod has penetrated the tube and the die. Fig.87. Urethane rod penetrating the tube Solutions: 1. Using a first analysis the problem is considered as associated to the fact that the end nodes of the rod, tube and die are coincident, causing problems for contact algorithm. When trying to establish contact the algorithm searches for probable contact nodes in an area nearby the node under examination. At the tube and rod end it is thought that there may be not enough nodes to properly establish contact. To investigate this case the rod’s length model is reduced so that the end nodes of the rod and tube are not at the same z-location. At the end the simulation with reduced rod length doesn’t show any significant improvement. 2. Here is provided a new analysis to find out if a different material model would more accurately predict the behaviour of the rod, applying a two parameter Mooney- Rivlin material model to represent the urethane material. Experimental tests are conducted on the urethane material and the Mooney–Rivlin constants are determined from experimental stress–strain curves. The constants used are C01 = 0.152655E − 03 and C10 = 0.133499E − 02. It’s found that this model would give better results, but a new issue is now observed in the urethane model. It is possible to remove the hourglass behaviour and improve the contact algorithm behaviour changing solution control parameters as advised in the LSDYNA user documentation. There are however still significant problems in the urethane rod mesh—it seems that the model is having a great deal of trouble accurately describing the compression of the urethane rod. To solve this problem a hole was introduced into the urethane rod. The idea of this hole is to allow space for the elements in the centre of the rod to compress and deform. After this initial deformation the rod deforms normally and as expected afterwards. It should be noted that this hole would not be present in the actual forming process but is simply a method adopted to allow the finite element solution of the problem. Results obtained using Mooney-Rivlin model at the end of the forming process are shown in Fig.88: 59 Fig.88. Predicted tube geometry after bending In the following figures (89 and 90 respectively) show the distribution of Von Mises stresses and strains in the deformed tube. It can be clearly seen how the stress distribution is non-homogeneous and is higher on the back of the tube than on the front portion. Fig. 89. Von Mises stress distribution in the formed tube Fig. 90. Von Mises strain distribution in the formed tube MOS bending process technique Among all the possible free bending techniques (and in particular push bending techniques) one of the latest and most significant one is MOS bending process, associated to 6 DOF and mainly used by BMW to produce semi-finished products for the tube hydroforming process. [18] For this particular application the bending process is analysed to have a better understanding of the whole process. In the first experimental bending series, the process precision is measured and evaluated by standard deviation and process capability. In the second bending series, the analysis focuses on the weld seam position and the wrinkling behaviour of the tube. The correlations shown between process parameters and the resulting bending geometry are the basis for the validation of a process simulation model developed in LS-Dyna. The finite element model is applied to provide a numerical based sensitivity analysis of the bending process, considering the process and the tube parameters. 60 Generally speaking, the previously mentioned MOS technique is used to create 3D bent tubes as the preforming step into the hydroforming process. In Fig. 91 it is shown the bending principle on which MOS technique is based: Fig. 91. MOS bending technique The main factors influencing MOS technique are schematized in figure 92, ordered in terms of their effects on the bending result: Fig. 92. Influence factors on free bending technology The process deviation is obtained bending tubes to a 3D part, as it is shown in figure 93. The represented profile is made of a union of 4 bents characterized by different radii and bending directions. Fig. 93. Bending profile experimentally used The tubes used for the bending series are obtained starting from a coil through a roll-forming process and welded longitudinally using the high-frequency method. The manufacturing process just mentioned is based on two parts welded together using an electromagnetic field, to have at the end a very strong join. The material used is an EN AW-5182 aluminium alloy. An inhomogeneity of tube dimensions can be easily noticed, looking at Figure 94, for three different batches. This inhomogeneity regards, in particular, diameter and sheet thickness. 61 Fig.94. Measured geometry of EN AW-5182 tubes The process deviation is rated from the bending line of 26 tubes from the batch C4. The bending line is reconstructed by means of iterative cylinder fitting through the measured tube surface. To calculate the deviation between the single bends, the bending line is oriented as follows: 1. The origin of the bending line is located at the start position (Figure 93) and the local x-axis points tangentially along curve. 2. The z axis is fixed at the tube end position. 3. After the coordinate transformation to global Cartesian coordinates, the bending line was interpolated equidistant. The measurement of standard deviation (shown in Figure 95) is performed along the bending line from start to endpoint. It shows that the deviation increases with increasing distance from the orientation point. Therefore, the development of the course is nearly linear and can be expressed as a deviation of 9.2 mm per 2,650 mm of bending line (or 3.54 mm/m bending line). Regarding the process capability indices, it is now possible to calculate the necessary specification limits for subsequent forming or joining operations: (1) A stable process chain is required, so the process capability index Cp must be greater than 2. Since the tube profile is considered symmetric from the middle of the bending line the maximum distance in both directions is 1,325 mm. Therefore, the standard deviation is calculated by π = 1.325 m x 3.54 mm/m = 4.64 mm. The minimum total specification limit at the end of the tube has to be at least (USL-LSL) = 55.68 mm (calculated by the previous relation defining Cp (1)). These bounds results in a preforming tool with movable segments to meet the demands of the bending deviation. Fig. 95. Standard deviation and density function of experimental bending FEM model is developed to study numerically the based parameters. The simulation is built in LS-Dyna and the tools (mandrel, guide and die) are defined as rigid shells, without considering thickness. The stress and strain distribution from the roll-forming process is simulated and mapped to the closed tube mesh before the beginning of bending simulation. During the bending process, the tube is pushed with a 10-fold increase in pushing velocity. Standard penalty-forming contacts were used for all contact zones. The coefficients of friction are 0.08 for the contact between guide/tube and die/tube and 0.15 for mandrel/tube. The weld line strength of the tube (e.g., increased yield strength) is constituted by a hardness increase during roll-forming simulation in the area of the strip edges and also mapped on the tube mesh. The definition of the weld line thickness differs from the overall sheet thickness (Figure 94 border of thickness distribution) with an increase to 2.3 mm. Therefore, a single element row is selected along the weld line, and a new thickness is defined for 62 this area. The influence of weld seam position on the bending result and tube wrinkling tendency is analysed in order to provide an overview of the precision of the simulation model. While testing 26 tubes of batch C4, wrinkling phenomenon is observed in 10 cases. Strong wrinkling occur on the profiles only in radius 1 and 4: wrinkling in radius 1 (Figure 96), weak or no wrinkling in radius 4 (figure 97). When the same profile was bent with tubes of batch C1 and C3 (10 tubes each), no wrinkling was observed. No significant differences are found analysing the mechanical tube properties, the tube surface roughness and the tube homogeneity. The maximum effective plastic strain EPS resulted by the simulation are represented in figure 99. It can be finally concluded that the tube dimensions have a significant influence on wrinkling behaviour during the free bending process. Fig. 96. Strong wrinkling at radius 1 Fig. 97. Weak wrinkling at radius 4 Fig. 98. Decrease of wrinkling by decrease of lubricant (a to c) Fig. 99. Simulation result of test geometry (max EPS): a)batch C4 b)batch C1, C3 Figure 98 shows the influence of the amount of lubrication on the wrinkling occurrence for the same bending profile. It follows, therefore, that friction behaviour also has a strong relation to wrinkling occurrence. With the assumption that a higher amount of lubricant results in a lower friction coefficient, the simulation model shows strong wrinkling for friction coefficients around 0.10, medium wrinkling around 0.16 and no wrinkling starting from 0.18. The distribution of the friction coefficient on the different tool couples also results in different wrinkling behaviour. In addition to the wrinkling phenomenon analysis, another bending series is performed to show the effect that weld seam position has on the bending result. A series of a single, two-dimensional bends is performed changing the weld seam position from 0° to 180° (0°: weld seam position at the inner portion of the bend). A good method related to radius measurement is developed and shown in figure 100. Fig. 100. Technique used to measure radius and bending angle 63 After reconstructing the bending line, the bending plane is transformed into the global Cartesian x-y plane. To determine the exact centre point of the radius on the bending line, an optimal bend was fitted to the bending line by means of nonlinear optimization (least-square distance). The bending angle was measured using the straight lines at the start and end of the fitting profile. For the radius measurement, a group of points around the determined centre of the bend was used to perform a circle regression on the bending line raw data. It is necessary to get a minimum number of points (in respect to the measurement interval) in the bending area with constant curvature, as plotted in Figure 100, to receive a robust radius measurement using the regression method. Though figure 101 it can be observed the weld seam influence on the bending result, plotting measured angles and radii. Fig. 101. Correlation between experiment and simulation varying the weld seam position At this point, after the simulation model has been defined and its performance demonstrated, a numerical based sensitivity analysis can be performed to gain information about the influence of the parameters shown above on the tube wrinkling behaviour and the bending result in terms of bending angle and bending radius. Therefore, the simulation model in LS-Dyna is integrated in the optimization tool LS-Opt to perform a Design of Experiments (DOE) analysis. For each design point, the run starts with the initialization of the tube mesh, followed by the bending simulation and a spring-back operation. In the DOE, the following parameters are varied: diameter (63.85-64.15 mm), sheet thickness (1.9-2.1 mm), coefficient of friction (COF) guide/die/mandrel (0.08-0.18), mandrel position (-12 – 0 mm in relation to guide position), weld line thickness (2.1-2.35 mm), weld line position (0°-180°) and scale factor for the yield curve (0.85-1.15). The statistical experimental design includes 400 sampling points, distributed by a space-filling algorithm. The DOE is studied as associated to the wrinkling behaviour of the process. Figure 102 (a) shows the plotted result of the analysis of variance (ANOVA) in terms of input parameters and EPS inner. Here, the tube dimension (diameter and thickness) shows the highest sensitivity in terms of the EPS inner. Wrinkling, as previously noticed, is caused by a change in tube diameter. In the scatter plot shown in Figure 102(b), the diameter and the thickness are plotted over the EPS inner. Data points are divided into two levels. The experiments with EPS below 0.4 have no wrinkle occurrence, whereas all points with EPS above 0.4 have wrinkles in the inner bend. When we observe the diameter-to-thickness relation of the tube, we can see those experiments with a big diameter and a small thickness (big diameter of neutral layer) have no wrinkles. Conversely, tubes with a small diameter in the neutral layer show wrinkling in the bending result. Analysing the DOE with refers to the bending result as bending angle and radius the main contributing factors are the mandrel position, the tube diameter and COF of the die, as represented in figure 103. 64 Fig. 102. (a)ANOVA in terms of EPS inner. (b) Scatter plot of diameter and thickness in relation to EPS inner Fig. 103. (a)ANOVA in terms of bending angle. (b) ANOVA in terms of bending radii As final step some points can be listed in order to improve the MOS bending technology performance, reducing the measured process deviation: • • • • Change bending profile and tools design to avoid any wrinkling by variation of tube dimension within the production tolerances Increase precision of inner and outer tube lubrication to lower process deviations Increase precision of tube dimension to limit process deviations Develop tools allowing an adjustment of the contact gap between tools and tube to adapt their position regarding to tube dimensions 65 Press bending Press bending technology is based on a device having 3 main rolls: rolls 2 and 3 are fixed on a table (on which a press is set), while roll 1 is fixed by the battering ram of the same press, as shown in Figure 104. [19] The process consists of the following passages: in order for the bending to verify, pipe 4 is placed on rolls 2 and 3, as roll 1 moves vertically on the pipe causing it to deform. The bending radius, as well as the bending angle, are controlled using quotes x and h. The distance between rolls 2 and 3 can be adjusted moving the rolls horizontally. The three rolls are placed on a surface with a channel in contact with the pipe. The channel’s dimensions are proportional to the pipe’s diameter. This procedure is highly advantageous because, through it, different bending radii can be produced without changing the rolls, because the bending radius depends on the position of the three rolls, not on the one of the rolls. On the opposite the negative point of that procedure is the fact that the bending radius is found indirectly using quotes h and x and it is more difficult to modify. Fig.104 Press bending process (1. upper roll; 2-3 lower rolls; 4 – pipe (Semi-manufactured); 2θ - bending angle; h – the distance covered by the upper roll during the bending; R – the bending radius) One of the most common problems associated to tubes production is the wall thickness variation, due to the fact that on the outside of the bending profile the wall is subjected to tensile stress and the wall becomes thin, while on the inner part of the bending profile compressive stress appears and the tube wall becomes thick. In this case, to study the wall thickness variation phenomenon for a press bent tube, a, ABAQUS FEM is applied. Finite element model: ABAQUS CAE is the main FEM software used to model the press bending process, as shown in figures 105 and 106. Fig.105 FEM of press bending Fig.106 The final position of press bending The tube is modelled as 3D deformable part of which material behaves elastic-plastic and the tools are modelled as 3D discrete rigid. Shell elements S4R are used to model the tube geometry. Contact between various pairs of surfaces: bend die-tube, pressure die-tube, wiper die-tube is defined using *CONTACT_SURFACE_TO SURFACE contact command, which allows the sliding phenomenon between these surfaces with a Coulomb friction model. The implement friction coefficient was 0,1. The material 66 considered is laminated steel OLT 35(STAS 8183-87). The mechanical properties of the tube material are obtained applying a tensile test on a straight tube. The stress is derived from load force and instantaneous geometry of the tube. The strains are determined from extensometer directly applied on the tube during the test. Figure 107 shows the Stress-strain relation used to describe the material behaviour during the simulation. In addition, the elastic modulus considered is E=2.1x105 [N/mm2 ] and Poisson’s ratio is 0,3. Fig.107 Stress-strain relation reached through tube tensile test If, at first, an angle 2Ζ = 25[°] and a diameter of 27 [mm] are chosen, a distribution of the wall thickness of the tube bending made using finite elements, is obtained, as shown in figure 108. On the other hand, in figure 109 the distribution of the wall thickness of the tube bending at angle 2θ = 90[°] and diameter 27 [mm] is represented. Fig.108 Distribution of wall thickness of angle 2Ζ=25° Fig.109 Distribution of wall thickness of angle 2Ζ=90° Fig.110 Wall thickness trend 67 Hydro bending The typical difficulties associated to manufacturing ultra-thin-walled elbow tubes made through traditional bending processes can be solved applying the double-layered tube hydro bending method. [20] Typically the hydroforming process provides a simple alternative to manufacture such thin-walled tube elbows. The specimen used for tests about the effect of internal pressure, the wall thickness distribution and the bending defects consists of an outer tube made of carbon steel and an inner tube made of stainless steel, with the thickness ratio of outer to inner layers equal to 10. The yield pressure of the double-layered tube is derived from theoretical analysis. In addition, the mechanism of the wrinkling and stabilizing of the inner tube is defined through the finite element analysis. It can be shown that the stability of the inner layer during bending improves with the increase of the internal pressure, while the wall thickness of two layers decreases with increasing internal pressure. The new double-layered tube hydro bending method for manufacturing the ultrathin-walled elbow components is based on many different procedures: (a) wrapping the ultra-thin inner tube with a thicker outer tube. (b) sealing and pressurizing the double-layered tube with a liquid medium. (c) double-layered tube bending with internal pressure support. (d) separating the outer layer from the elbow. The double-layered tube specimen is prepared by wrapping the ultra-thin-walled tube with a mild steel outer thicker tube, which will be then removed after the hydro bending process. The double-layered tube is sealed with end caps by welding at the two ends, as shown in figure 111. The inner tube is made of stainless steel, while the outer one is made of low carbon steel. The mechanical properties of the inner and outer tubes are shown in the following table, obtained from uniaxial tensile tests on samples cut along the axial direction. Through figure 112 it can be represented the dimensions of the ultra-thin-walled elbow. Fig.111. Diagram of double-layered tube specimen Fig.112. Dimensions of ultra-thin-walled elbow (number 1-1 to 5-5 are the measuring locations of section flattening after bending) 68 Experimental procedure: A hydraulic press is used to conduce this experimental method and is identified in figure 113. The internal pressure is provided by a pressure pump and servo controller to keep a constant value during the bending process. After the internal pressure is linearly increased to the target pressure, the upper die moves downward to bend the tube on the boundary of the die cavity. Four pressure values are chosen in the hydro bending experiments, as shown in the table below. The slide gage is used to measure the wall thickness of two layers whereby the cross-section dimensions are measured with an outside lock-joint transfer caliper. Fig.113. Tooling of the double layered tube hydro bending Changing the levels of internal pressure new experimental results of double-layered hydro-bending can be obtained and shown in figure 114. Fig.114 Bending results under different levels of internal pressure: (a) defect-free double-layered elbow formed under Py; (b) section view of double-layered elbow under pressure Py; (c) wrinkling under pressure 0.1Py; (d) rupture under pressure 1.2Py; (e) ultra-thin-walled elbow formed under pressure 0.8Py; and (f) ultra-thin-walled elbow formed under pressure Py. It is highly possible that the rupture occurred first at the outer layer, as the total elongation of the outer layer material is much smaller than the one of the inner layers, as mentioned a previous table. The deformation at the outside arc exceeds the plastic limit of the outer layer and leads to fracture. The increasing axial stress at the compression side during bending process causes wrinkling (or bifurcation instability) of the thin-walled tube if the compressive stress exceeds the critical value. The internal pressure has an important role in the successful forming of the double-layered elbow. On one side, the higher the internal pressure, the more a ‘‘tight fit’’ bond would be formed during bending. The bond between the two layers would increase the critical wrinkling stress of the inner layer greatly. On the other side, a larger additional 69 axial tensile stress could be introduced to the inner layer if a higher internal pressure is applied. The additional axial tensile stress would decrease the magnitude of the axial stress at the compression side. Finite Element Model: ABAQUS/Explicit is the finite element software used to study the wrinkling mechanism associated to lower pressure values. Just half of the numerical model in the view of geometry and loading symmetries is built up, as shown in figure 115: Fig.115 Finite element meshes for double-layered tube hydro bending: (a) shell type outer tube (b) solid type outer tube Upper and lower dies are considered rigid, while the inner tube is modelled using S4R (four-node quadrilateral shell with reduced integration) elements, with the mesh size of 6 mm and with five integration points along the thickness direction. The outer tube is separately modelled with shell (S4R) and solid elements (C3D8R, eight node three-dimensional (3D) brick element) for comparison purposes. For the lateral case, four-layer brick elements are created along the thickness direction, and the element dimension is around 2.5 mm 3 5 mm 35 mm. The material model applied for two layers is isotropic, homogeneous and elastic–plastic material following the Mises yield criterion. The ends of the two layers are tied together to simulate the tensile forces of the outer layer applied to the inner layer. The main properties of the material near the weld line are considered to be the same as the base material of the external tube. In addition to that, the internal surface of the inner tube and the end caps are subjected to the application of a constant pressure. The bending behaviour associated to the double-layered tube in presence of inner pressure is analysed through experimental and numerical simulations, focusing on the effects of internal pressure on the section flattening, on the wall thickness distribution and the wrinkling mechanism of the inner layer. The following results are found: 1. The section stiffness of the double-layered tube can be increased greatly with the increase of the internal pressure. 2. The thickness distributions of two layers are very similar, with the maximum thinning points located at an angle from the symmetrical section. The wall thicknesses of the two layers decrease as the internal pressure increases. 3. The internal pressure delays the onset of wrinkling and enhances the stability of the inner layer. It is shown from the simulation that the maximum magnitude of compressive axial stress at the inner arc decreases as the internal pressure increases. The inner tube is stabilized when the maximum magnitude of compressive axial stress is less than the critical buckling stress of the inner layer throughout the bending process. This is the reason for the prevention of the wrinkling under higher levels of internal pressure. 70 REFERENCES [1] Z. Jin, S. Luo, and X. Daniel Fang, “KBS-aided design of tube bending processes,” Engineering Applications of Artificial Intelligence, vol. 14, no. 5, pp. 599–606, Oct. 2001, doi: 10.1016/S09521976(01)00016-1. [2] M. Zhan, H. Yang, Z. Q. Jiang, Z. S. Zhao, and Y. Lin, “A study on a 3D FE simulation method of the NC bending process of thin-walled tube,” Journal of Materials Processing Technology, vol. 129, no. 1–3, pp. 273–276, Oct. 2002, doi: 10.1016/S0924-0136(02)00664-7. [3] R. J. Gu, H. Yang, M. Zhan, and H. Li, “Springback of thin-walled tube NC precision bending and its numerical simulation,” Transactions of Nonferrous Metals Society of China, vol. 16, no. SUPPL., pp. s631–s638, Jun. 2006, doi: 10.1016/S1003-6326(06)60268-9. [4] R. J. Gu, H. Yang, M. Zhan, H. Li, and H. W. Li, “Research on the springback of thin-walled tube NC bending based on the numerical simulation of the whole process,” Computational Materials Science, vol. 42, no. 4, pp. 537–549, Jun. 2008, doi: 10.1016/J.COMMATSCI.2007.09.001. [5] L. Heng, Y. He, Z. Mei, S. Zhichao, and G. Ruijie, “Role of mandrel in NC precision bending process of thin-walled tube,” International Journal of Machine Tools and Manufacture, vol. 47, no. 7–8, pp. 1164–1175, Jun. 2007, doi: 10.1016/J.IJMACHTOOLS.2006.09.001. [6] H. Hagenah, M. Merklein, H. Hagenah, D. Vipavc, R. Plettke, and M. Merklein, “Numerical Model of Tube Freeform Bending by Three-Roll-Push-Bending,” 2010. [Online]. Available: https://www.researchgate.net/publication/266357349 [7] S. Ancellotti, V. Fontanari, S. Slaghenaufi, E. Cortelletti, and M. Benedetti, “Forming rectangular tubes into complicated 3D shapes by combining three-roll push bending, twisting and rotary draw bending: the role of the fabrication loading history on the mechanical response,” International Journal of Material Forming, vol. 12, no. 6, pp. 907–926, Nov. 2019, doi: 10.1007/s12289-018-14530. [8] X. tong LI, M. ting WANG, F. shan DU, and Z. qiang XU, “FEM Simulation of Large Diameter Pipe Bending Using Local Heating,” Journal of Iron and Steel Research, International, vol. 13, no. 5, pp. 25–29, Sep. 2006, doi: 10.1016/S1006-706X(06)60090-3. [9] T. Zhou, H. Yu, J. Hu, and S. Wang, “Study of microstructural evolution and strength–toughness mechanism of heavy-wall induction bend pipe,” Materials Science and Engineering: A, vol. 615, pp. 436–446, Oct. 2014, doi: 10.1016/J.MSEA.2014.07.101. [10] W. Xun, Z. Jie, and L. Qiang, “Multi-objective Optimization of Medium Frequency Induction Heating Process for Large Diameter Pipe Bending,” Procedia Engineering, vol. 81, pp. 2255–2260, Jan. 2014, doi: 10.1016/J.PROENG.2014.10.317. [11] Y. Guan, G. Yuan, S. Sun, and G. Zhao, “Process simulation and optimization of laser tube bending,” International Journal of Advanced Manufacturing Technology, vol. 65, no. 1–4, pp. 333–342, Mar. 2013, doi: 10.1007/s00170-012-4172-6. [12] K. I. Imhan, B. T. H. T. Baharudin, A. Zakaria, M. I. S. B. Ismail, N. M. H. Alsabti, and A. K. Ahmad, “Investigation of material specifications changes during laser tube bending and its influence on the modification and optimization of analytical modeling,” Optics & Laser Technology, vol. 95, pp. 151–156, Oct. 2017, doi: 10.1016/J.OPTLASTEC.2017.04.030. [13] M. Goodarzi, T. Kuboki, and M. Murata, “Deformation analysis for the shear bending process of circular tubes,” Journal of Materials Processing Technology, vol. 162–163, no. SPEC. ISS., pp. 492– 497, May 2005, doi: 10.1016/J.JMATPROTEC.2005.02.090. 71 [14] M. Goodarzi, T. Kuboki, and M. Murata, “Effect of initial thickness on shear bending process of circular tubes,” Journal of Materials Processing Technology, vol. 191, no. 1–3, pp. 136–140, Aug. 2007, doi: 10.1016/J.JMATPROTEC.2007.03.007. [15] X. GUO, X. CHENG, Y. XU, J. TAO, A. ABD EL-ATY, and H. LIU, “Finite element modelling and experimental investigation of the impact of filling different materials in copper tubes during 3D free bending process,” Chinese Journal of Aeronautics, vol. 33, no. 2, pp. 721–729, Feb. 2020, doi: 10.1016/J.CJA.2019.02.016. [16] H. Liu, S.-H. Zhang, H.-W. Song, G.-L. Shi, and M. Cheng, “3D FEM-DEM coupling analysis for granular-media-based thin-wall elbow tube push-bending process”, doi: 10.1007/s12289-019-014738. [17] S. Baudin, P. Ray, B. J. mac Donald, and M. S. J. Hashmi, “Development of a novel method of tube bending using finite element simulation,” Journal of Materials Processing Technology, vol. 153–154, no. 1–3, pp. 128–133, Nov. 2004, doi: 10.1016/J.JMATPROTEC.2004.04.205. [18] N. Beulich, J. Spoerer, and W. Volk, “Sensitivity analysis of process and tube parameters in freebending processes,” in IOP Conference Series: Materials Science and Engineering, Nov. 2019, vol. 651, no. 1. doi: 10.1088/1757-899X/651/1/012031. [19] V. Adrian Ceclan, G. AchimaΕ, L. LΔzΔrescu, F. Mioara Groze, and A. Lect, “FINITE ELEMENT SIMULATION OF TUBES PRESS BENDING PROCESS.” [20] L. Hu, B. Teng, and S. Yuan, “Effect of internal pressure on hydro bending of double-layered tube,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 226, no. 10, pp. 1717–1726, Oct. 2012, doi: 10.1177/0954405412457517. 72