The higher the value, the more rapid the metal hardens during shaping operations. The n value of most sheet metals used for forming, falls in the range of 0.2 - 0.6. Strain hardening also affects the point of maximum force in the punch stroke. The higher the n value, the later it occurs. This could be beneficial. When n = 0, the material would possess insufficient ductility to bend round the punch-profile radius and so failure would always occur, regardless of the value of anisotropy. It has an important role in defining the formability of the sheet metal.
Metals and their properties in metal forming
Material properties play a very crucial role in defining product quality. Steel, copper, aluminium and their alloys are most frequently used materials for formed parts. Moreover, the metal that is chosen for a particular application depends on the requirements of the part, the cost of manufacturing, and the availability of the metal. Requirements of parts vary due to the fact that though the primary consideration for one part may be strength; it may be surface appearance for another. Thus, for each material type or class of material, there are large numbers of alloys and heat treatments, which affect material properties and formability.
Effect of process and tool parameters in product quality
The process parameters like blank holding force and friction coefficient have a profound influence in defining the product quality. In addition, a proper value of blank holding force can ensure a wrinkle-free product without excessive thinning. This is a very important feature with respect to aesthetics in several industries like automobile, furniture and household appliances. A high value of coefficient of friction results into galling wear as well as fracture. Similarly, very fine lubrication increases the cost of product. To maintain the surface finish of the product during drawing, the friction is one of the most important parameters.
Tool parameters like die and punch radii ensure a smooth flow of metal, which is the most important requirement in deep drawing operation. Improper selection of these parameters may create defects like tearing and wrinkling, which is directly related with surface finish of the product in most of the cases and in a few cases with functional aspects too.
Software programme
The software programme is mainly divided into three modules: input module, computation module and output module. The function of each of the modules include:
Input module:
The programme gets inputs from the file, which contains the training patterns, weights and biases. A neural network is initialized in this module. Various functions are prepared to read the data from these files in the programme. The data read from this input file is used in the computation module. Several functions are also incorporated to normalise as well as randomize the patterns before processing in computation module.
Computation module:
It is the core module of the programme, and consists of functions to carry out the tasks such as error calculations, back propagation algorithm and training. The parameters are adjusted according to the error calculated and finally the neural network gets trained up to predefined accuracy.
Output module:
The output module does the front-end tasks such as calculation of output for the testing data and prints the output on screen or store in a file. It consists of the inputs and the outputs corresponding to those inputs. It also has the facility to either store or print the whole input training data as a file for future reference.
Selection of parameters: After studying the importance of all parameters, these are compared based on their influences on strains involved during the drawing operation. Further, efforts are also taken to select all parameters that represent deep drawing as a whole.
To simulate the deep drawing in Pam Stamp 2G all input parameters are needed. Hence, same parameters are selected for neural network system design. These input parameters are as follows:
· Yield strength (Mpa)
· Strain hardening exponent
· Strength coefficient (Mpa)
· Plastic strain ratio (R value)
· Limiting drawing ratio
· Die radius (mm)
· Punch nose radius (mm)
· Clearance between die and punch (mm)
· Blank holding force (KN)
· Coefficient of friction
· Thickness (mm)
· Punch travel (mm)
Out of these parameters, a few are material properties, a few are tool design parameters and the remaining are process parameters.
The developed neural network also has three output parameters, which are as follows:
Major strain (e1)
Minor strain (e.2)
Check whether this part can be deep drawn or not (1 or 0); if it can be drawn, output should be 0, else output should be 1. Intermediate values are interpreted as the probability of the part being rejected.
Generation of Data
To generate the data a lot of simulation of cylindrical deep drawn cup with commercial sheet metal forming software Pam Stamp 2G are carried out. The materials taken into consideration are mild steel, copper and aluminum alloy.
The properties are established from ft available literature.
Characteristics of materials used for simulations
In this project, the following parameters including the tool design parameters, are kept constant during all simulations:
Die radius = 7mm
Punch radius = 9.5mm
Clearance = 1.25 mm
Punch travel = 0 - 90mm
The following parameters including the process parameters are varied during simulations so that the neural network gets trained within these two extreme limits of these two extreme limits of these parameters.
The simulation data:
Based on the materials and a range of other parameters several combinations of all parameters are formed to perform simulations.
Conclusion
It is possible to arrive at the workable material, tool and process parameters. The number of FEM simulations will be large, but finite in number, and these will encompass infinite forming conditions arising from infinite combinations of the parameters considered. If the material parameters cannot be changed, the consequences of design changes in the tools and the processing conditions can be examined. Thus, consequences of possibilities that were thus far not simulated can be examined. Moreover, the tool offers excellent possibilities for vendors who are neither able to afford an FEM code nor have skilled people to operate it. OEMs can use this tool to optimise the part and process design. Finally, with a reduction in wastage and savings on time, substantial cost reductions can be expected. |
