متدولوژی بهبود فرآیند بر اساس روش تجزیه و تحلیل آماری چند متغیره
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|16847||2004||17 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Control Engineering Practice, Volume 12, Issue 8, August 2004, Pages 945–961
A systematic procedure for process improvement methodology is proposed based on multivariate statistical process control methods. To take advantage of a large amount of historical data, the procedure employs a combination of hierarchical clustering method and statistical process control methods to detect and analyze the key factors that significantly affect the performance of processes. This methodology consists of four sequential steps: (1) Data collection and multivariate statistical analysis; (2) hierarchical clustering and operation mode detection; (3) selection of dominant variables; (4) a new operational guideline and its validation. The proposed procedure was applied to improve the heat efficiency of an industrial hot stove system located at Pohang Iron & Steel Co. (POSCO) in Korea. The implementation results show that the proposed methodology helps us systematically improve the operating conditions of the hot stove system.
Process improvement is an essential activity in most of the industries in order to survive in the increasingly competitive global market. Therefore, there has been increasing attention on the methodologies for process improvement. Basically, process improvement involves two complementary steps: the reduction of product variations within pre-specified limits and continuous process improvement through finding the effects of common (or chance) causes upon the performance of processes. The key idea of the first step is to achieve process improvement through reducing the variations of product to a pre-specified limit based on Statistical Process Control (SPC) methods. Achieving this goal requires three sub-steps: (1) detection of the occurrence of abnormal situations based on control charts such as Shewart, CUSUM, and EWMA charts, which are extensively used for process monitoring; (2) diagnosis for the assignable (special) causes that are generating the abnormal situations; and (3) elimination of those causes. In the second major step, the goal is to provide an opportunity to improve process performance through finding the relationship between product quality and common causes, presented within the process itself. In order to analyze this relationship, there are typically two methods according to the methods of data gathering: designed experimental methods and non-experimental methods. Designed experimental methods have been developed in several related ways such as a fractional factorial design, an orthogonal array, and evolutionary operation (Taguchi, 1986; Box & Draper, 1998). These methods have been developed for providing the optimum operating conditions. However, these methods have limits in the scope, the frequency, and the number of situations in which experiments have been conducted. Furthermore, these methods have possibilities of unexpected risk and require high experimental cost for implementation. On the other hand, non-experimental methods aim at extracting new and important information to improve process performance based on the process data obtained directly from process information system (PIS). Although these non-experimental methods require no cost and risk in gathering process data, there are limits in the range of possible process improvement. However, at most of modern industrial sites, it is believed that the collected historical data would be a gold mine of information, if only the “important” and relevant information could be extracted painlessly and quickly (Piovoso, Kosanovich, & James, 1992). Therefore, there has been increasing attention on the methods to collect historical data for process improvement such as higher productivity, better product quality, increased safety, and lower environmental pollution. Thanks to the deployment of PIS in the industrial world, engineers can get a huge amount of process data representing operating conditions. The industrial data shows that variables are highly correlated and often collinear, signal-to-noise ratio (SNR) is low, and even there are missing data. These complex characteristics of the data make it difficult to analyze the data and extract the information using conventional univariate SPC charts. Recently multivariate statistical techniques such as the principal component analysis (PCA) and the partial least squares or projection latent structures (PLS) have received increasing attention as alternative data analysis tools (Jackson, 1991; Geladi & Kowalski, 1986). This is due to their ability to efficiently handle a large number of highly correlated variables, measurement errors, and missing data. More important is their ability to provide an operator with useful information on process improvement by projecting the high-dimensional process data into the low-dimensional space defined by a few latent variables. Based on these multivariate statistical techniques, several researchers have proposed various methodologies for monitoring the performance of a continuous process over time, which aims at detecting any abnormal situations due to the assignable (special) causes (Kresta, MacGregor, & Marlin, 1991; MacGregor, Jaeckle, Kiparissides, & Koutoudi, 1994; Wise & Gallagher, 1996). Then, after identifying assignable (special) causes for abnormalities, process improvements can be achieved through eliminating the causes by changing the process operating conditions. Nomikos and MacGregor (1994) used the multi-way principal component analysis (MPCA) as an extension of PCA, for treating batch or semi-batch processes. Monitoring charts based on MPCA are capable of tracking the progress of new batch runs and detecting the occurrence of observable upsets. Kourti, Nomikos, and MacGregor (1995) discussed the methodologies based on PCA and PLS for diagnosing the assignable (special) causes. The methods mentioned above have mainly focused on the detection and diagnosis of the assignable (special) causes through comparison with the recipe data, which represent normal operation trajectories. However, there are limitations for process improvement because such methods do not provide any information about the common causes, which can provide significant opportunities for process improvement. The purpose of this paper is to provide a methodology to uncover the chances to improve based on multivariate statistical analysis methods, which can lead to the process improvement in view of its performance. The proposed methodology has the characteristics of hierarchical analysis procedure, which aims at providing a new operational guideline for process improvement through detecting and analyzing the common causes that significantly affect process performance. This paper especially focuses on the case in which there are several operation modes depending upon changes of operating conditions. The proposed analysis procedure, combined with an operator's experience, can lead to the operational guideline for evolutionary process improvement. The proposed methodology is validated by applying it to the industrial hot stove system, located at Pohang Iron & Steel Co. (POSCO) in Korea. The implementation results showed that the proposed methodology could be used as an efficient tool for detecting and analyzing the common causes, which are differentiating operation modes. And it could also provide significant opportunities to improve the operating conditions of the hot stove system. In order to explain the proposed methodology, this paper is organized as follows: First, the proposed methodology is explained according to its procedure. Second, a brief description on the hot stove process is presented. Third, the application results to the hot stove process are discussed and the implementation results given. Finally, conclusions are provided.
نتیجه گیری انگلیسی
In this work, the process improvement methodology based on multivariate statistical analysis methods was proposed. It provides a systematic and efficient process analysis method for process improvement according to the hierarchical analysis procedure. The hierarchical analysis procedure successively uncovers the common causes that are significantly affecting the process performance. The proposed methodology was applied for improving the heat efficiency of the hot stove system, located at Pohang Iron & Steel Co. (POSCO) in Korea. The implementation results show that it leads to about 3% higher than that of the previous year. Consequently the proposed methodology provides engineers with significant opportunities to improve the operating condition of the process.