FEA compliant parametric tolerance allocation of mechanical assembly using neural network and differential evolution algorithm

• Published in: International journal of computer integrated manufacturing Vol. 25; no. 7; pp. 636 - 654
• Main Authors: Jayaprakash, G., Sivakumar, K., Thilak, M.
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 01.07.2012
• Subjects: FEA; flexible parts; gravity; neural network; rigidity; tolerance allocation
• ISSN: 0951-192X; 1362-3052
• DOI: 10.1080/0951192X.2012.665184

The technological and financial limitations in the manufacturing process are the reason for non-achievability of nominal dimension. Therefore, tolerance allocation is of significant importance for assembly. The purpose of tolerance design in product components is to produce a product with the least manufacturing cost possible, while meeting all functional requirements of the product. Limitations of the conventional tolerance allocation methods are as follows. The cost tolerance model developed by regression analysis has fitting error. Only dimensional tolerances of components alone are considered for allocation while the effect of geometric tolerances on functional requirement of the product is not considered. It is based on an assumption that all parts of the assembly are rigid. But in reality, every mechanical assembly consists of at least one or more flexible parts which undergo significant deformation due to gravity, angular velocity, etc. In this article, a back propagation (BP) network is applied to fit the cost-tolerance relationship. A parametric CAD model is developed to determine assembly constraint equation (the functional requirement) based on geometric and dimensional tolerances. Finite element analysis is used to determine the deformation of components in an assembly. An optimisation method based on Differential Evolution (DE) is then used to locate the combination of controllable factors (tolerances) to optimise the output response (manufacturing cost plus quality loss) using the equations stemming from the trained network. Integration of statistical tolerance design with finite element analysis guarantees that the optimal tolerance values of various components of the assembly, obtained as end product of the tolerance design will remain within tolerance variation. Then the product can function as intended under a wide range of operating conditions for the duration of its life. An application problem (motor assembly) is used to investigate the effectiveness and efficiency of the proposed methodology.