| Summary: | 博士 === 中原大學 === 機械工程研究所 === 93 === The present study attempts to develop an error detection and diagnosis mechanism (EDDM) of single version software for fault-tolerant system. The failures caused by software fault of a PC-based computer-controlled system will be focused on. Error classification is established by means of the knowledge-representation based error hierarchy design. The detection mechanism employs the hook process to capture the message in and between the various application programs and the operating system, and detects whether the monitored application program is failed. Since the hook technique uses the capture mode but not the interrupt mode. Therefore, it will not affect the executing application program or the operating system. The diagnosis mechanism can identify the failure type and location of the failure message and make predictable estimation on the executing application program. A fuzzy reasoning and verification Petri nets (FRVPNs) model is established to achieve the purpose of reasoning and decision-making the failure event of the monitored application program for the EDDM. According to the symptoms of failure event, the linguistic variables, the fuzzy sets and its membership functions, and the fuzzy reasoning rule by use of “IF…THEN…” proposition are designed for the FRVPNs model. Through the hierarchical design of the fuzzy rule tree decision (FRDT) associated with using the PN technique to transform the fuzzy reasoning rule into the PRVPNs model, the inference speed and accuracy can be improved. In addition, the inconsistent rules such as conflict, redundancy, circularity, and incompleteness etc. that likely exist in the fuzzy knowledge rule base will be verified and modified by the FRVPNs model.
|
|---|
