The Fault Monitoring Of The Closed-loop Industrial System Based On The Time Series Model

With the development of the science and technology, the complexity of the modern industrial process is growing exponentially. Great casualties and property losses will be caused in the event of failure of these complex industrial processes. Therefore, the improvement of maintainability and reliability of the industrial process is an urgent thing. The fault monitoring technology can effectively improve the maintainability and reliability of the process, increase the economic benefit, and reduce the risk of accidents.Statistical process control (SPC) is a kind of important data driven fault monitoring approach, which is widely applied in all kinds of industrial processes for process monitoring.In order to guarantee stability, the modern industrial process added feedback control technology to reduce the influence of external factors. Since the feedback controller has weakened the influence of external factors, the statistical process monitoring technology’s ability to monitor feedback control system has weakened. Therefore, the development of effective statistical process monitoring technology has important research significance for the feedback control system. This paper presents a fault diagnosis method for feedback control system based on time sequence model. Compared with the traditional statistical process monitoring techniques, the proposed method has less miss rate and faster detection rate. The details of the paper are summarized as follows:The fault monitoring approach in batch processes has been proposed in the paper. Inaddition, the influence of the EWMA/DEWMA controller on batch processes with a time delay are analyzed. A time series model can be used to express the batch process with a time delay and a EWMA/DEWMA controller. An improved parameter-reset recursive least square (PRELS) method is used to identify the coefficients of the time series model in real time. Then, a dynamic principal component analysis (DPCA) method is used to monitor those identified coefficients, and an influence matrix method is used to isolate the fault.A general feedback control system has been considered in the paper. Inaddition, all kinds of problems of applying traditional SPC to monitor feedback control system are analyzed. The paper proposes a statistical real-time monitoring technique to solve those problems. A general feedback control system has been derived a general time-series model, and the stability of the identified coefficients of the time-series model by RELS is proved. The statistical process control method can be used to monitor the coefficients of the time-series model.A general feedback control system with a time delay has been studied in the paper. Inaddition. two situations of that the set point is constant and time-varying are analyzed. Considering the nonuniqueness of the time-series model, a new time delay identification method based on adaptive LASSO is proposed. The system controller performance index and stability index are computed based on the delay and AR term of the time series model, and the statistical process monitoring technology is used to monitor those indicators in real time.The modeling and fault monitoring approach of the two-dimensional (2D) dynamic batch process have been proposed in the paper. Inaddition, the2D dynamic of the batch process is analyzed. A2D-ARMA model can be used to express the2D batch process. An improved2D adaptive LASSO method is used to identify the two-dimensional dynamic batch process. The paper proposes two stability evaluation indics based on the inner-wise matrix and2D-ARMA model. The SPC control chart is adopted to monitor these stability indies.Finally, the induction and summary of the paper is given, and the unsolved problems and future research direction is introduced.