Production process optimization using feature selection methods

Abstract

A manufactured product is required to undertake an acceptance testing before delivery in order to ensure the quality of the product can meet industry standard. The acceptance testing usually consists of a large number of functional tests, which consumes considerable production time. In order to optimize the production process, acceptance testing strategies that could identify a subset from all available functional tests while still assure the inspection accuracy need to be identified. In this study, a statistical quality control method - Lot Acceptance Sampling Plan (LASP) - and strategies derived from feature selection techniques are presented. Feature selection is a process of selecting a segment consisting of the most significant variables from the original ones. The feature selection techniques outperform the LASP by means of choosing functional tests according to their importance instead of randomly and therefore largely reduces the inspection time. The derived strategies are implemented and their performance is tested on an acceptance testing dataset provided by Siemens. The results of the case study have shown that the feature selection based acceptance testing strategies can reduce as high as 81% inspection time while keeping the same accuracy with current industrial strategy to distinguish a qualified and non-qualified product.