Title: Improved product quality through causality analysis in product engineering

Authors: Shichang Du, Lifeng Xi, Jun Lv

Addresses: Department of Industrial Engineering and Logistical Engineering, School of Mechanical Engineering, Shanghai Jiaotong University, Shanghai, 200240, China. ' Department of Industrial Engineering and Logistical Engineering, School of Mechanical Engineering, Shanghai Jiaotong University, Shanghai, 200240, China. ' Department of Business Administrative, School of Business, East China Normal University, Shanghai, 200240, China

Abstract: An integrated causality analysis method is developed to investigate how to identify the causal relationships between manufacturing process variables and product quality variables for improved product quality in product engineering based on data from a questionnaire. This method is especially advantageous in following two situations: a manufacturing process variables and conditions are inconvenient, or even infeasible to collect data through distributed sensing systems a large number of potential factors need to be studied simultaneously. Firstly, fuzzy logic is utilised to analyse the lingual, non-numerical, potentially vague or even somewhat inaccurate data from the questionnaire and transfer it to a pre-processed dataset. A new algorithm is developed to generate automatically membership functions. Secondly, a Bayesian network is built from the pre-processed dataset by integrating manufacturing process knowledge and generic learning algorithms. Thirdly, the causal relationships revealed by the Bayesian network are used for quality improvement. Finally, the developed integrated approach is demonstrated through one case study.

Keywords: causality analysis; product engineering; fuzzy logic; product quality; process variables; learning; Bayesian network; quality improvement.

DOI: 10.1504/IJMSI.2009.025295

International Journal of Materials and Structural Integrity, 2009 Vol.3 No.1, pp.47 - 65

Published online: 18 May 2009 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article