عنوان مقاله :
پايش داده‌هاي سانسور شده از چپ در فرايندهاي چندمرحله‌يي مبتني بر مدل رگرسيون تحليل بقا
عنوان به زبان ديگر :
MONITORING LEFT-CENSORED DATA IN MULTISTAGE PROCESSES ON THE BASIS OF SURVIVAL ANALYSIS REGRESSION MODELS
پديد آورندگان :
اسدزاده، شروين اسدزاده، شروين دانشگاه آزاد اسلامي واحد تهران شمال - گروه مهندسي صنايع
كليدواژه :
مدل رگرسيون تحليل بقا , سانسورشدگي از چپ , مقدار ارزش انتظاري , فرايند چندمرحله‌يي , تابع درست‌نمايي
چكيده فارسي :
امروزه توليد اغلب محصولات حاصل فرايندهايي با چندين مرحله است. در هر بخش ازفرايندهاي چندمرحله‌يي متغيرهايي با ماهيت قابليت اطمينان، مي‌توانند عملكرد فرايند را تحت تأثير قرار دهند. اين متغيرها ويژگي‌هاي منحصر بفردي چون سانسورشدگي و پيروي از توزيع‌هاي خانواده مكان ـ مقياس و لگاريتم ـ مكان ـ مقياس دارند. در اين نوشتار به كمك راهكارهاي كنترل فرايند آماري پيشرفته و رويكردهاي مدل‌سازي، داده‌هاي قابليت اطمينان در حضور سانسورشدگي از چپ پايش شده است. بدين منظور، ابتدا مدل‌سازي داده‌ها به كمك مدل‌هاي رگرسيوني تحليل بقا مورد مطالعه قرار گرفت. سپس رويكردهاي پايش بر اساس نمودارهاي كنترل جمع تجمعي و ميانگين متحرك موزون نمايي به كمك تشكيل تابع درست‌نمايي و مقدار ارزش انتظاري ارائه
شده و متعاقباً توسط شاخص‌هاي متوسط طول دنباله مورد توجه قرار گرفته است. در نهايت عملكرد برترين نمودار كنترل به صورت مطالعه‌ي موردي در كارخانه‌ي تك‌تاب ظريف مورد بررسي قرار مي‌گيرد.
چكيده لاتين :
Nowadays, most products are the results of multistage processes with cascade property. In each step of multistage processes, several variables with reliability characteristic may affect the performance. These reliability variables have some unique properties such as censoring and following the family of location-scale and log-location-scale distributions including Weibull, Log-normal and Log-logistic. The purpose of this paper is to monitor left-censored reliability data with the aid of advanced statistical process control (SPC) techniques in line with statistical modeling approaches. To this end, data modeling has been studied using survival analysis regression models. In fact, the accelerated failure time (AFT) regression model has been employed to establish the relationship between quality variables. Then, two monitoring schemes including a cumulative sum (CUSUM) and an exponentially weighted moving average (EWMA) control charts have been proposed by constructing likelihood function and conditional expected values (CEVs) respectively. It should be noted that the transformation of Weibull distribution to standard smallest extreme value distribution has been done to effectively remove the cascade property in the discussed process. Subsequently, the performance of the proposed control schemes has been examined and investigated in terms of average run length (ARL) criterion under various censoring levels of low, medium and high (20%, 50% and 80%). Finally, the performance of the superior (the cumulative sum) control chart has been investigated in a real case study in Taktab Zarif company located in Kashan, Iran. The quality variable of interest is actually the tensile strength of thread which is affected by the proportion (weight) of raw materials. Moreover, the values corresponding to tensile strength of threads are recorded just in case that they are beyond the specified level. In other words, we encounter some data which are left-censored and thus remedial action should be considered to alleviate the mentioned obstacle for optimal multistage process monitoring.
عنوان نشريه :
مهندسي صنايع و مديريت شريف
