Op-Aide: An intelligent operator decision support system for diagnosis and assessment of abnormal situations in process plants.

摘要

Abnormal situation is a general term used for any significant departure of the process from an acceptable normal range of operation. During abnormal situations, the operators are confronted with information overload. They are also required to take complex decisions in a short time during such situations. Due to the high stress involved in such situations, the ability of the operator to effectively recall process knowledge is significantly hampered. This leads to significant safety and economic impact. Automated systems to aid the operator in effective decision making during abnormal situations can improve the ability of the operators to perform effective abnormal situation management (ASM). Existing systems for operator decision support do not provide quantitative information about the abnormal situations. However this information is required for effective ASM. Hence, in this work, an intelligent system for operator decision support, Op-Aide, has been developed to assist the operator in quantitative diagnosis and assessment of current and future consequences of abnormal situations. The blackboard architecture of the system is open and allows for modular development. The current implementation of Op-Aide consists of modules for data acquisition, process monitoring and diagnosis, parameter estimation, and situation assessment using process simulation.; Abnormal Situation Management (ASM) involves timely identification and mitigation of any departure of the process from its normal range of operation. Ineffective ASM has significant safety and economic impact on the chemical industry. In this work, an intelligent operator decision support system, called Op-Aide, has been developed to assist the operator in quantitative diagnosis and assessment of abnormal situations. A set of characteristics desirable in any ASM decision support system has been identified. Op-Aide has been developed based on an open, modular, blackboard-based architecture. In this approach, independent modules provide data acquisition, process monitoring, fault diagnosis, and situation assessment capabilities. Novel algorithms and techniques that provide these different functionalities have also been developed.; An original B-Splines based data compression algorithm has been developed for the data acquisition module. This technique achieves high data compression while providing computationally efficient, on-line retrieval of historical data. The process monitoring module uses Principal Components Analysis (PCA) to detect process abnormalities. Fault diagnosis is performed by two different experts. One of these is a novel multiple fault diagnosis algorithm that uses signed digraphs to interpret faults identified by PCA. A new B-Splines based adaptive system for trend analysis (ASTRA) has also been developed to identify the root causes for abnormal situations. Conflicts between the two diagnosis experts are resolved using a rule-based approach. Once root causes have been diagnosed, their magnitude and rate of change are estimated by the fault parameter magnitude estimation module using a computationally efficient, nonlinear, dynamic optimization technique. The immediate and future consequences of root causes can also be estimated in Op-Aide using the process simulation module.; Op-Aide has been implemented in Gensym's expert system shell G2, MATLAB and C. The application of the system and its various components have been successfully evaluated on a simulation of Model IV fluidized catalytic cracking unit. The B-Splines based data compression algorithm and ASTRA have been successfully demonstrated on real-time data from a crude distillation unit and an industrial scale ethylene plant.