University of Twente Student Theses
Using optimisation techniques to solve a production problem, with application to the deep drawing process of an automotive part
Veldman, Else (2006) Using optimisation techniques to solve a production problem, with application to the deep drawing process of an automotive part.

PDF
8MB 
Abstract:  At the University of Twente an optimisation strategy for metal forming processes is being developed in cooperation with the metals industry. One of the industrial partners of the UT project is INPRO in Berlin, which is a subsidiary of amongst others Volkswagen. Volkswagen is interested in optimising the deep drawing process of an automotive part in which cracks occur. For the research described in this report, the optimisation strategy is applied to solve this problem. First, the cause of the cracks has been investigated and it was concluded that large deformation due to forming is causing the cracks. Subsequently, the optimisation strategy is applied. This strategy comprises a structured procedure to model the problem in such a way that a mathematical optimisation algorithmcan be applied to it. Besides themodelling part, the strategy comprises a screening procedure to reduce the number of variables and a mathematical algorithm to solve the optimisation model. This algorithm can be coupled to any Finite Element Method (FEM) programme. For the automotive part two processes are distinguished: the reference process and themodified process. Themodified process is the reference processmodified by INPRO. Thismodified process proved to reduce the number of cracks that occur in the part. The optimisation strategy is applied to two optimisations of the process of the automotive part: 1. Optimise the reference process to compare these results with the modified process 2. Optimise the modified process to further reduce scrap Initially, for the first optimisation of the reference process an optimisation model with six variables has been defined. After an analysis of the responses, it appeared that the responses suffer from numerical noise. To improve the results the optimisation model was adjusted. This led to a simplified optimisation problem with only one variable. This problem has been solved with two different sets of variable settings for the unimportant variables: one set is contains the original settings for the design variables and one set contains the optimal variables settings based on screening results. For screening, a number of calculations are executed to estimate the effect of the variables on the responses. After solving this problem an optimum was found that improved the process with respect to the reference process. The optimisation with the optimal variable settings resulted in a better optimum that the optimisation with the original settings. The modified process still outperforms the optimised reference process. Furthermore, a second optimisation of the modified process was performed. The optimisation model does not differmuch fromthe one for the reference process: the definitions of the responses are equal, only the variables are different. Screening was performed to eliminate the unimportant variables. The screening results showed that there are no dominant variables. Therefore, no variables are excluded from the optimisation model and the result is an optimisation model with eleven variables. Because this are many variables for the optimisation strategy, this strategy is no further applied. For optimisation the variables are set to their minimum or maximum values as defined in the optimisation model. The choice for these settings is based on the screening results. This simple optimisation led to a large improvement with respect to the modified process. For the optimisations the FEMprogramme AutoFormis used. Based on these simulations and the identified cause for the cracks, the optimisations delivered good results. However, the AutoForm simulations did not coincide with reality and simulations in other simulation programme and the cause of the cracks is still uncertain. This makes it difficult to draw strong conclusions on the effect of the optimisation on reducing the cracks in reality. 
Item Type:  Essay (Master) 
Clients:  INPRO 
Faculty:  ET: Engineering Technology 
Subject:  52 mechanical engineering 
Programme:  Mechanical Engineering MSc (60439) 
Link to this item:  http://purl.utwente.nl/essays/58318 
Export this item as:  BibTeX EndNote HTML Citation Reference Manager 
Repository Staff Only: item control page