Study On Identification Methodology Of Critical-to-quality Characteristics For Complex Products

Identification of critical-to-quality characteristics is helpful to control the product quality and make a stable manufacturing process. It can reduce the product variation, improve product quality. The complicated products with the complexity of structure, technology and management result in a sharp increase of the quality characteristics, and more difficulties with lower efficiency of quality control. In order to improve the efficiency of quality control and reduce the enterprise cost, it is necessary to study the quality characteristics of complex product, recognize and strictly monitor the important characteristics of product quality.There are four methods on identification of critical-to-quality characteristecs for complex products have been studied in the paper.(1) An identification method of critical-to-quality characteristics for complex products based on L1 norm is proposed. For the specialty of high dimension and small-simples, the concept of norm in functional analysis is adopted. And SVM classifier is used to test the classification prediction ability of critical-to-quality characteristecs subset, which can test identification performance of the presented method.(2) An identification model of critical-to-quality characterictics for complex products has been extablished based on rough sets and gravitational search algorithm. The method uses gravitational search algorithm to search in quality characteristics space. The adaptation function established by rough set theory is use to evaluate the characteristics combination searched by above algorithm. The effectiveness of reducing the dimension and identification reliability is verified by experimentation.(3) A two-stage attribute clustering analysis method has been established to identify the critical-to-quality characteristics. In the method, mutual information is used to measure the correlation between the characteristics. The example shows that because of the elimination of noise and redundant characteristics, the two-stage clustering method can identify critical-to-quality characteristics effectively.(4) A hybrid method of filter and wrapper based on heuristic algorithm for identifying the critical-to-quality characteristics is represented. Firstly, in order to reduce the dimensions, quality characteristics are filtered by the relevance and redundancy analysis. Then, the cricical-to-quality characteristics are searched by particle swarm optimization(PSO) in characteristics space. This method has advantages of filter and wrapper. Experimental results show that, the method is effective and practically applicable.