Fault Detection & Diagnosis And Fault Tolerant Control For Process Control System

The operation of process industry usually takes place in extreme conditions, such as high-temperature & high-pressure or low temperature & vaccum. The maloperation or system faults will cause the hazardous status, for example, production stoppage, equipment explosion or toxic gas leakage. To improve manufacturing safety, it is necessary for control system to effectively detect and diagnose fault. On the other side, many process industries are characteristic with low dynamic performance, and the control precisions are lower than the aerospace control precision. It is more possible for fault detection & diagnosis and fault-tolerant control to successfully apply to process industry. Some new satisfactory tolerant control methods with multi-objects and multi- constraint conditions need to be further explored. The fault detection & diagnosis and fault-tolerant control technique for process control system are investigated in this paper. Some main research findings and innovation are listed as following:1. Based on the industry practice, rational describing models for sensor faults and valve faults are founded, which can amend the existing two fault description models. Grounded on the above fault models, an effective plant active tolerant control method is proposed by incorporating the intelligent modeling strategy with an adaptive threshold scheme. The simulation results, including three-tank benchmark problem and DAMADICS benchmark problem validate the proposed method.2. A fault detection & diagnosis method based on feature sample extracting, Kernel PCA and the gradient arithmetic of kernel function is developed. The feature extraction method can solve the calculation problem of the kernel matrix K during monitoring process. The contribution degrees of each variable to Hotelling's T~2 and SPE based on the gradient arithmetic of kernel function are applied to diagnose complex faults. To demonstrate the performance, the proposed method was applied to the Tennessee Eastman (TE) process. The simulation results showed that the proposed method could effectively identify various types of fault sources.3. A novel fault diagnosis method is proposed by incorporating independent component analysis (ICA) strategy with support vector machines (SVM), which can effectively settle the puzzles of the correlative industry faults which arise from the nonlinearity, multi-loop, complex operation and so on. The proposed monitoring method was applied to fault detection and diagnosis in the butadiene industry distillation column. The simulation results clearly show the power and advantages of ICA-SVM method. 4. An active fault tolerant control scheme based on the inversion model of adaptive RBF neural network is proposed for multi-variable nonlinear systems. The extend Kalman filter (EKF) algorithm is used to on-line update network variables for learning fault dynamics and time-varying parameters. The optimal output value of adaptive controller is worked out by the iterative algorithm of RBF inverse model to maintain the system performances after fault occurrence. The simulation results on the three-tank process with leakage faults validate the proposed method.5. The tolerant control strategy for process control system should not only guarantee the stability of fault systems, but also satisfy some performance indexs or constraints. The satifisfactory fault-tolerant control methods with multi-indices constraint are investigated. For uncertain discrete time-delay systems against actuator failures, the state-feedback based H∞satifisfactory fault-tolerant control method and guaranted cost control method are investigated. For the fuzzy nonlinear system described by T-S fuzzy models against actuator failures, the satifisfactory tolerant control method with the constraint on decay rate, control input and output is investigated, and the satifisfactory tolerant control method with the constraint on pole, state variance and H∞consistent index is also studied. For nonlinear fuzzy time-delay system against actuator failures, the H∞satifisfactory fault-tolerant controllers with time-delay memory and memory-less are separately designed.Finally, based on generalizing full text, some problems which need to be further researched are discussed.