Theory And Methods Of Reducing And Controlling Variation Of Multivariate Quality Characteristics

Quality is a permanent problem. It is not only a lifeline of military products, but also a key problem of research and production in navigation, guidance and control system domain. This project is of important theory and practice values in the process of products design and production.Variation is the fundamental cause of poor quality. Reducing and controlling variation has become a core research domain in modern quality engineering science. In this paper, we systematically study some theories, methods and implementation techniques of reducing and controlling variation of a product with multivariate quality characteristics in its forming process by using positive analysis and simulation, which are based on variation theory and process conception. It is important for organizations to enhance product quality, minimize quality loss and strengthen market competitiveness. The main results are as follows.1. Variation theory and continuous quality improvement. First, we introduce the development of quality concept, the causes of resulting in variation, the statistical rule of process output and new quality loss principle. Then we point out three ways of implementing continue quality improvement and the relationship among them. Finally we provide two kinds of techniques for reducing and controlling variation in quality engineering domain and their basic principles.2. An assessment and measurement method for multivariate process capability. Process capability indices (PCIs) are singer-number measures for the capability of a process in meeting specification limits. These indices have been widely used in assessing the capability of manufacturing process and choosing appropriate suppliers. In this paper, three new procedures are provided to assess multivariate process capability. That is cumulative non-conformity rate method, principal component method and composite index method. It can be proved that our procedures are very useful in assessing process performance by using some positive analysis.3. Studies of measurement process. A measurement process is an important support process and it is necessary in the manufacturing process to assure and improve product quality. In the paper, we point out some limits of analysis of variance in measurement process study and propose methods of isolating the source of variation in gauge analysis and monitoring measurement process.4. Identification in multivariate quality control. In a multivariate quality control process, acommon statistic is Hotelling's T2. In order to detect small shift or trends sensitively, multivariate cumulative sum (MCUSUM) chart and multivariate exponentially weighted moving average control chart (MEWMA) are recommended. However, the common problem that still exits in these techniques is how to discriminate the variables that are really being out of control when the multivariate control charts are signaling. To solve his problem, we propose three kinds of diagnostic procedures. The first one is binary depth search technique, which divides the total variables into two partial variables and remove correlation of one partial variable to the other by residual analysis, then use residual vector to construct Hotelling T2 statistic and to seek which partial variables are out of control. Once out-of-control partial variables are determined, we utilize binary depth search on these variables further until variables being out of control are identified. The second one is signal discrimination technique. The third one is that when multivariate chart issues a signal, we utilize union diagnostic chart to identify the variables and/or the principal components that caused the signal. By using simulation and positive analysis, we have proved that these procedures can solve the multivariate process control problems well. In this field, some innovation results are obtained.5. Detection and isolation of the variation sources in a complex manufacturing system. In order to reduce variation and seek quality improvement opportunities, it i...