An effective strategy for improving the precision and computational efficiency of statistical tolerance optimization
Wenhui Zeng,Yunqing Rao,Peng Wang
DOI: https://doi.org/10.1007/s00170-017-0256-7
IF: 3.563
2017-03-24
The International Journal of Advanced Manufacturing Technology
Abstract:Statistical tolerance optimization is a more practical and economical method for assigning proper tolerances on components in product design. Monte Carlo simulation is the simplest method to simulate probability distributions of random variables (tolerances). However, the precision of this method is proportional to the root of sample size, and a large sample size leads to excessive computational time. Moreover, during the entire process of statistical tolerance optimization, the optimum value cannot be found when the precision of fitness in each iteration is not sufficiently high, and when these inaccurate values are used in many generations with large number of assembly analyses. In this paper, a strategy with multiple sample testing is proposed to solve this problem by improving the precision of fitness and penalty values and reducing the computational time cost. Based on this strategy, high precision of fitness could be obtained for the next generation using metaheuristic algorithms such as genetic algorithm, cuckoo search, and particle swarm optimization, and a feasible solution that satisfies all the constraints could be obtained. A practical industrial application is provided to demonstrate the effectiveness of the statistical tolerance optimization problem, and the experimental results confirm the efficiency of the proposed strategy.
engineering, manufacturing,automation & control systems