A generic intelligent nonconformance tracking and recovery (GINTR) system.

摘要

This research proposed a generic distributed architecture for nonconformance diagnosis based on the multi-agent systems (MAS) technology. The system architecture consists of two layers: the system layer and the application layer. The system layer has a bus framework consisting of a set of unit diagnosis agents and one mobile agent. The application layer is the kernel of the proposed system architecture. It is at the bottom of the proposed system and implements all the elementary nonconformance diagnosis tasks. Under the proposed distributed system architecture, a generic intelligent nonconformance tracking and recovery (GINTR) system is designed and developed. In conjunction with the soft computing and case-based reasoning (CBR) technologies, the system can identify the root causes of a nonconformance and provide timely corrective actions. The application of the GINTR system follows a series of steps. First, a new format is proposed to represent the nonconformance's feature. The beauty of the new feature representation approach is that it can handle the situation where there are diverse feature attribute types and uncertain feature attribute values. Second, a distance measure is adopted to recover the missing attribute values based on the existing knowledge base. Then a RBFN is trained to locate the root causes of the nonconformance and the results are further tuned with the help of the Genetic Algorithm (GA) method. Finally, the identified root causes are used to retrieve the diagnostic recovery solutions from the database. The experimental results indicate that the GINTR system is capable of identifying a nonconformance's root causes and providing recovery solutions. The results also demonstrate that the distance measure is an effective approach to recovering the missing attribute values and the system's capability in tracking and recovering a nonconformance can be significantly improved by applying the feature weights.; To demonstrate the effectiveness of the proposed GINTR, a prototype of the GINTR system has been developed using the Microsoft.NET technology. Microsoft.NET is considered as the next generation distributed computing paradigm. Since it is built on open standards and embraces all programming languages, the integration barrier compounded by the myriad of competing proprietary software technologies that plague the industry has been removed. The prototype system developed is generic, fault tolerant, and agile. It can be adapted to diverse complex manufacturing environments with minimum difficulty. (Abstract shortened by UMI.)