املاک و زمان مشاور فرآیند آماری برای کنترل کیفیت موثر

An advisory decision support system has been presented in this paper. This system helps in collecting statistical data and thereafter analyzes the enormous volume of data and aids in making quality related decisions. In contrary to conventional SPC applications where the analyzed results have to be interpreted by quality control specialists, MES based unmanned manufacturing environments require automation of the interpretation process. The developed advisory system helps in selecting and designing control charts based on various cost, rule or heuristics models. The system also provides interpretation expertise by configuring and applying various rule sets. On violation of these rules, signals are generated by the system and the expert system advices for appropriate remedial actions. Thus the system acts as an advisory support system.

The basic goal of using quality control techniques is to streamline the manufacturing system by minimizing the occurrence of quality related problems. Most of the time, problems related to quality of products have many controllable sources, be it the vendors of raw materials, equipment used to process such materials, methods used for processing, the personnel involved or any other specific source as identified by the organization. Such factors usually affect the quality at any stage of the production process and hence, there has been an ardent need to monitor those problems effectively and efficiently through proper design and deployment of an appropriate quality and process control system. In the present day context, the three basic functionalities of statistical process control system, i.e. specification, inspection and control [2] are to be expanded so as to make the manufacturing processes more competitive. From time to time, various methods have been suggested to design the appropriate control charts for specific processes [8] and [3] and interpret those charts for maintaining the quality of the materials or products manufactured by the processes [1], [6] and [7]. However, it is realized that in order to achieve the desired objective of capturing the process data on real time basis and for on-line interpretation of those data, an integrated approach is to be undertaken. This will ensure the maximum effectiveness of the quality control function in a manufacturing organization under the manufacturing execution system (MES) environment [5]. In this paper a modular and integrated software architecture is proposed and developed. It uses the generic methods to address various problems of quality control and facilitates mapping of the unique quality requirements of the manufacturing organization. The modules of the developed software primarily use statistical techniques to interpret the data captured during various stages of the manufacturing processes. Different objective methods are used to analyze the enormous volume of the generated data and aid in making quality related decisions. It will have the ability to respond to any request for quality clearance to be provided for either the raw materials procured or finished products produced. It also helps in systematizing the data collection procedures at various identified qualities related checkpoints and thus aids in effective troubleshooting. The quality control module addresses to the above-mentioned areas of concern in a very comprehensive manner and raises alarms whenever out-of-control situations occurred. It also ensures discipline in the data collection procedures and helps in breaking down and analyzing the complex quality control system by assigning costs at the various stages of the manufacturing process [4] and [6].

While designing the advisory system, a few alternative architectures are evaluated. It is identified that the overall performance of the system will depend on two critical components, that is, data integration component and advisory system component. The comparative performance of the integration component based on a 1000 KB of data payload is summarized below. These are the average elapsed times to complete 10 simultaneous request–reply communications. a) Socket based integration = 0.002 s b) EAI–File adapter based integration = 0.05 s c) EAI–Database adapter based integration = 0.02 s However, a like-to-like comparison of the performance of various modules of the advisory system cannot be done because of the disparity in the modeled variables. But it can be observed that the heuristics based models are performing best as compared to the cost based optimization models, whereas the rule based models perform moderately. Based on the real time cost data from the integrated ERP system, the optimization engine is able to determine the feasible solutions for 7 input and 3 output variables in approximately 0.02 s and after 75 to 80 iterative steps on the average. The expert system application is designed and developed to serve as a real time on-line adviser to the operators of the manufacturing processes so as to help them to design and apply the appropriate control chart techniques, detect the out-of-control situations and also suggest the possible causes behind those situations. A rule based expert decision support interpreter is also incorporated into the system. The interpretation logic of the expert system is such that if some of the predefined rules are violated, then the process will be out-of-control. On violation of any rule, exceptions will be generated by the system and the expert system will advice for the appropriate remedial actions and hence, acts as an ideal advisory support system. The developed advisory system is then implemented on a user acceptance testing environment using real time data collected from a multi-location manufacturing plant producing diverse products, such as industrial reduction gear boxes. These multi-product and multi-location testing data have proved the scalability of the system over a three-tier architecture. After the Beta implementation with only one gear hobbing machine attached to the SCADA server, it is ascertained that the advisory system is now able to detect a probable out-of-control situation within 5–6 h, whereas earlier it usually took more than a day to detect the problem.