Concurrent optimization of parameter and tolerance design based on the two-stage Bayesian sampling method

ABSTRACT

Researchers usually expect to reduce costs while improving product robustness in product quality design. The concurrent optimization of parameter and tolerance design assumes that the output response follows a normal distribution. However, non-normal responses are also common in product quality design. As for the concurrent optimization of parameter and tolerance design with a non-normal response, a novel total cost function based on the two-stage Bayesian sampling method is proposed in this paper. First, the hierarchical Bayesian model constructs the functional relationship between output response, input factors, and tolerance variables. Secondly, a two-stage Bayesian sampling method is used to obtain the simulated values of the output responses. The simulated response values are used to build the rejection cost and quality loss functions. Then, the genetic algorithm is used to optimize the constructed total cost model, including the tolerance cost, rejection cost, and quality loss. Finally, the effectiveness of the proposed method is demonstrated by two examples. The research results show that the proposed method in this paper can effectively improve product quality and reduce manufacturing costs when considering the uncertainty of model parameters and the variation of the output response.

KEYWORDS:

• Quality design
• Bayesian method
• parameter design
• tolerance design
• quality loss function

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.

Additional information

Funding

The work was supported by the National Natural Science Foundation of China [72171118, 71771121, 71931006]; Natural Sciences and Engineering Research Council of Canada [RGPIN-2018-03862]

Notes on contributors

Yan Ma

Yan Ma is a Ph.D. candidate in Quality Management and Quality Engineering from Nanjing University of Science and Technology, China. Her current research interests include Bayesian statistics, robust parameter design, and quality engineering. Email:[email protected]

Jianjun Wang

Jianjun Wang is a Professor at the Department of Management Science and Engineering at Nanjing University of Science and Technology. He is a senior member of the Chinese Society of Optimization, Overall Planning, and Economical Mathematics. He is a reviewer of several international journals, such as JQT, EJOR, IJPR, CAIE, and QTQM. His current research interests include quality engineering and quality management, robust parameter design, Bayesian modeling and optimization, and industrial statistics. Email: [email protected]

Yiliu Tu

Yiliu Tu is a professor at the Department of Mechanical and Manufacturing Engineering, University of Calgary, Canada, and Zi Jin Scholar Chair Professor at the School of Economics and Management, Nanjing University of Science and Technology, China. His current research interests are OKP (One-of-a-Kind Production) product design and manufacture, ultra-fast laser micro-machining technology, and quality engineering and management. He is a senior member of SME (Society of Manufacturing Engineers) and a professional engineer of APEGGA (The Association of Professional Engineers, Geologists, and Geophysicists of Alberta). Email: [email protected]