Machining of large dies based on the prediction of the press/die deformation
Abstract
In this work a methodological scheme for a reduction of both the try-out and lead-time of complex dies is presented. The finite element simulation of the tool behaviour along the stamping process results in criteria for the best design of high-cost dies/punches. Using it, modifications of the main geometry, components and functional parameters are recommended. Examples are deeply explained as experimental evidences.
Keywords: Stamping, Try-out, Bending, Dies, Simulation
By: D. del Pozo (a), L.N.López de Lacalle(b), J.M. López a, A.Hernández(b)
(a) Robotiker Technology Center, Parque Tecnológico, Edificio 202, 48170 Zamudio, Spain, dioni@robotiker.es
(b) University of the Basque Country, UPV/EHU. Faculty of Engineering, Department of Mechanical Engineering, Alameda de Urquijo s/n, C.P. 48013, Bilbao, Spain.
login or register to download the paper.
Dear Author,
I will like to know if you have machined a real die with your system, and the final behaviour of it.
Is your example a real part?.
Yes I have though about it, and in recent studies the measurements are in the order of magnitude of the paper. For example, for a holding force of 800 Tons, the applied adjust to the machining of the die was 0.3 mm, and a considerable saving in time.
Hi Dionisio, I found the paper interesting, especially the contact FEA topic. I got two first questions for you:
Question No.1:
Since the final deformation of the FEA models is normally magnified graphically for visualization purposes in FEA packages. Therefore the deformation map of the contact surface only shows the general trend of deformation of the contact surface; it does not correctly show the true geometry of the deformed die. In addition, the FEA results depend a lot on the boundary conditions of the FEA models such as material properties, meshing types, number of elements in the analysis model; especially in non-linear contact analysis problems, these factors dramatically influence to the final analysis result. That leads to the correction of the CAD die model based on the resulted FEA map of the contact surface is not accurate. The questions, therefore, are:
1.1 How to compensate this error in the reality in term of CAD re-modeling of a die and modeling approaches?
1.2 Do you think that the experiment database should be constructed for practically compensation purposes when correcting the CAD die model based on the FEA map of the contact surface?
Question No.2:
As mentioned in the Question No.1, the FEA results depend very much on the boundary conditions of the FEA models, including material properties, meshing types, number of elements in the analysis models.
2.1 Have you considered the influence of these factors to the final FEA results?
2.2 When these factors are changed, are these differences among the FEA results less then 5%?
Thank you very much,
Le
Hi Le:
I reply to your two questions:
- Question No 1:
As it is indicated in section 2 of the paper, they have been carried out multiple studies until getting to the expressed simplified model, considering the material properties, the number of elements, the types of elements. The objective: to have a FEA map with the appropriate accuracy. So:
1.1- There is no need to re-modeling the CAD of a die. The deformation map is in z-direction of the die, and it is applied only to one of the workpiece: die or punch.
1.2.- Can you explain again this second question?
- Question No 2:
As I say in the previous answer, I have considerer the differents material properties, meshing types, number of elements in the analysis, so:
2.1.- Yes, I have considerer influence of these factos to the final FEA results.
2.2.- For example:
a) If you change the material properties, your are changing the material of the die, and in this case the results are significanty diferents.
b) In the case of the meshing type, the difficulty, more than the difference, is in the complexity of the die geometry.
c) The number and dimension of elements of the analysis depend on the dimensions of the die.
Dear Mr del Pozo,
Were you been using Catia or UGS. What do you think about the FEM codes of this sofware packages, have you used them?.
Actually very good work...
Hi Dionisio
Many thanks for your answers; this is really interesting; and here are more discussions and comments about this topic:
Comment No.1:
- Dionisio said: 1.1- There is no need to re-modeling the CAD of a die. The deformation map is in z-direction of the die, and it is applied only to one of the workpiece: die or punch.
Le : As stated in the paper at step 8: “By applying the deformation map of contacting surface to the CAD geometry of the die, the corrected geometry is obtained”, and at step 9: “this deformation map will be applied to the machining process of the initial surface of the die, based on the die-press deformation”.
If there is no-need to remodel the CAD of a die, it means that step 8 is unnecessary, isn’t it? Moreover, since a deformation map is applied to the machining process of a die (step 9), it means that the Z-value of the die surface must be readjusted, and how can you readjust the Z-direction of the die based on the deformation map in CAM/CNC software? Especially if the surface of the die is a free-formed one.
Related to the question 1.2: “Do you think that the experiment database should be constructed for practically compensation purposes when correcting the CAD die model based on the FEA map of the contact surface?”
Since the FEA results depend on many factors, and different FEA solvers (Marc, Ansys, Patran, etc.) also normally give different results; therefore the experiments are usually needed to correct the analysis models. I think that if the experiment database is constructed, then it would be beneficial for the similar projects, especially for compensation of the errors between the FEA and the actual results.
Comment No.2:
2.1 Related to the influence of the boundary conditions applied to the FEA models. If you considered these influences to the final FEA results, then how much different among the results are? , It would be good for the readers if these different numbers are included to the paper, since this is a very important part of the FEA, and it shows the analysis results are reliable or not .
We had used different solvers from Marc, Patran, Ansys for contact analysis problems, and we found that the changes in boundary conditions, sizes and types of element are significant and need to be taken into account, especially in non-linear FEA.
2.2. Related to the meshing types and elements: Dionisio said; “In the case of the meshing types, the difficulty, more than the difference, is in the complexity of the die geometry”.
LE: Today with the advances in Computer Graphics, meshing problems in FEA are easily done with many options of meshing as well as types of elements. Therefore the complexity of the geometry is not a big deal in meshing anymore. The pure-FEA packages such as Msc. Patran and Marc. Mentat are very strong; and even the embedded FEA modules in Pro.Eng, Solidworks (Cosmos) as well as UG also provide powerful tools for meshing. For our experiences, Patran is the best one for meshing:
You may CLICK HERE to view a few meshing photos with different types of meshing and geometry.
2.3. Related to the number and size of elements: Dionisio said: “The number and dimension of elements of the analysis depend on the dimensions of the die”
LE: Of course if the dimensions of the die change, then the number of the elements would be changed in cases the size of elements is keep un-changed or changed. However, with the same geometry of a die (volume is not changed), there will be FEA models with different numbers and sizes of elements.
You may CLICK HERE to view the meshing with the same die and punch models, but the size of elements is changed, see Figure 1 and 2.
In FEA, the smaller the size of the element, the better accuracy of the analysis is resulted. In addition, there are many types of elements used in FEA, in 3D analysis problems, the typical ones are Tetrahedral elements with 4, 6, or 10 nodes, and Hexahedral elements with 8 or 16 nodes. The bigger number of nodes in one elements, the better accuracy of the analysis is resulted. Of course the bigger number of elements and nodes, will lead to more time-consuming in the analysis and require more powerful computers.
This is why I questioned about the changes in the size and number of elements. Normally, if the differences among the FEA results less then 5%, then the FEA results are reliable.
Many thanks for your reply!
Hello Hieu:
This is a great contribution to our IPROMS 2006 debate!
Many thanks to you and to our authors for your active involvement!
May Italy and France both win the World Cup!
Best wishes.
DTP.
Yes, I have used the 2 software packages CATIA (from Dassault) and I-deas (now from UGS), for the development of this work.
First I-deas with NASTRAN solver and second CATIA with ELFINI solver.
The reasons have been the following:
1.- I-deas with NASTRAN solver for the acceptance and reliability of the results.
2.- CATIA with ELFINI solver to check the results in this software and validate it from the point of view of this work.
Nowadays, Dassault has acquired ABAQUS as solver, but you can use NASTRAN, ABAQUS, MARC with CATIA V5.
On the other hand, the die-maker can use VAMOS CAA V5, a die design development tool with CATIA V5 software for the 3D geometry of the die.
Hi Le:
2.1 In the FEA model, there are some simplifications and hypothesis as for example, the press model, application form of the load (depends on the press), the form the die is attached to the press, and so on. They all are checked and validated.
2.2. You are right. Die-maker use generally software package for die design as CATIA from Dassault, I-deas-NX from UGS or Wildfire from PTC, and tools as VAMOS, CADCEUS, NX mold and die tooling, for the 3D design of dies.
2.3. I have used in the work Tetrahedral elements with 4, 6, or 10 nodes, and the differences among the FEA results were less than 5%, but the computation time was significantly important. The figures in the paper are tetrahedral elements with 4 nodes.
With respect with hexahedral elements there is no method that I know to mesh a complex die geometry in a simple manner.
Thank you very much for your comments.
Hi Dioni,
I see your paper is very worthly commented. I think it is a good way to improve the quality of further work.
I hope see you soon.
Hi Dionisio,
Many thanks for your exellent reply. If it is okay, could you please send me on of the CAD model of the die or punch in Parasolid format. I and my colleages will try to do the meshing with hexaheral elements. My email is Le-ChiH@cf.ac.uk.
I still prefer Tetrahedral elements since it is very convinient for meshing, no matter of the complexity of the 3D objects, especially in non-linear contact analysis problems. Another advatage of using the Tetrahedral elements is that we can flexibly change the number of elements or apply the fine-meshing in the areas where it is important to the analysis.
However, if we know different ways of controlling the meshing types of the FEA models, then it is good and privides us more options for validation of the FEA models.
Many thanks,
Le
Hi Dionisio,
There are two papers at IPROMS 2006 related to your research. Click to the following titles to read the papers:
Paper No.1:
Springback prediction with FEM analysis of advanced high strength steel stamping process
Paper No.2:
Le, Cardiff University, Wales, UK
Hi Le:
The two papers are related with the process design for obtaining the desired stamped part, the method-plan, and in this stage only the simulation of material flow is taking in account, and it is not considered the whole 3D geometry of the tools: die, punch, blankholder.
Thanks
Dioni
Hi Dioni,
You are definitely right!
However, it would be good and wonderful to see how FEA, especially non-linear contact analysis problems were solved theoretically (in term of problem modeling) and practically (in term of application areas).
Many thanks,
Le, Cardiff University, Wales, UK.
I have some news than in RBTK are working now with Batz in this interesting subject.
Have you though about to check if real errors are similar to those related in your work?
Sincerely,