动态系统故障检测的集员估计方法研究

The veracity, realtimeness, and robustness of fault detection (FD) methods are the key focus for the international research society, and they are also the foremost problems in the area of FD. Considering that one of the difficulties for dealing with the veracity and robustness of FD methods is that the statistical properties of the systems are unkown in many cases and that there are some unmodelled dynamics in many system models, also considering that set membership estimation (SME) methods can effectively deal with these problems, we have introduced some SME methods into the FD of dynamic systems, and thus made great progress in dealing with the the veracity and robustness of FD methods. And the real-time problem of FD methods is also solved because the proposed SME algorithms have already solved this problem..In this project, we have studied the methods of intergrated error bounds and order determination of models, the high-precise and realtime SME algorithms, the SME algorithms for systems with time-varying parameters or with outlier measurements, the methods of combining SME and FD, the FD methods for nonlinear systems, the methods of determining optimal adaptive detection thresholds, and some other related theories and methods. We have also developed a simulation software for FD of navigation systems. In all these aspects, the methods of intergrated error bound and order determination of models have effectively released the limits for SME algorithm -- the error bounds and the order of the model must be known. They have provided a efficient method for increasing the precision of SME algorithms, and also provided a way for balancing the conflict between the model simplicity and high precision. The high-precise and realtime SME algorithms and the SME algorithms or systems with time-varying parameters or with outlier measurements have resolved the conflicts among the requirements for robustness, realtimeness, and precision. The proposition of combining SME and FD methods has provided a new direction for one to deal with synthetically the precision, realtimeness, and robustness of FD methods. The FD methods for nonlinear systems, the methods of determining optimal adaptive detection thresholds, and the other related theories and methods have made it possible for many practical problems to be treated effectively. And the simulation software for FD of navigation systems has laid a base for the theorotical achievements from this project to be used in practical engineering systems.In conclusion, the achievements of this project are significant for the practical application of SME methods and the FD methods, especially those methods with precision, realtimeness, and robustness

集员估计与故障检测相结合获得准确、快速实时和鲁棒的故障检测算法的研究。随机分析与确定性分析相结合对集员估计算法鲁棒性与准确性关系的研究。神经网络快速实时实现集同估计算法的研究。集员估计故障检测方法在控制和导航系统中的应用技术研究。本项目研究工作的完成可推进鲁棒故障检测理论和应用技术研究的发展，且能完善某些集员估计理论。