Sensing, identification, fault detection and diagnosis, and performance recovery of systems

磨损、设计缺陷以及外部影响和干扰(如过载)会导致许多系统逐渐恶化，从而可能导致性能下降和/或灾难性故障。对系统进行满意的控制和监控需要有关系统及其各个组件和运行特性的准确信息。全面的信息往往不容易获得，特别是在系统运行时。此外，由于设备/机械制造商一般不会预见到需要确定和提供此类信息，因此他们没有提供未来提取此类数据的内置手段。在过去的几年里，申请人一直从事在线系统/参数识别、故障检测、隔离和恢复。包括神经网络、模糊逻辑、遗传算法和小波分析等技术在内的系统建模和仿真已被成功探索和应用。目前正在研究和未来感兴趣的具体平台(或故障案例)是“爆震”，即内燃机中的自燃现象，以及航天器中的执行器和传感器故障检测。在敲击的情况下，除了造成低效运行和污染外，还会产生超过设计规范的压力和力，这可能会导致机械损坏和故障。将小波变换作为处理非平稳信号的一种方法，成功地应用于发动机故障检测。我们建议开发补充分析工具，以提高准确率和故障检测率，并将当前的结果从单缸情况扩展到更实际的多缸应用。在航天器执行器/传感器发生故障的情况下，我们正在开发单部件和多部件故障检测技术，并设计系统重新配置和恢复方案。我们将审查新的软件/控制和硬件冗余，重点是航天器中执行器和传感器的最佳空间位置，以改善系统的全局第一级故障安全特性，以实现稳健恢复。

Wear, design flaws, and external influences and disturbances, such as overload, cause gradual deterioration in many systems, which can lead to inferior performance and/or catastrophic failure. Satisfactory control and monitoring of a system require accurate information about both the system and its individual components, and operational characteristics. Comprehensive information is often not readily available, particularly when the system is in operation. In addition, since equipment/machinery manufacturers do not generally foresee a need to determine and supply such information, they do not provide a built-in means to extract such data in the future. Over the past several years the applicant has been engaged in online system/parameter identification, fault detection, isolation and recovery. Systems modeling and simulation including such techniques as neural networks, fuzzy logic, genetic algorithms and wavelet analysis have been successfully explored and used.  Specific platforms (or fault cases) currently under study and of future interest are "knock", or the autoignition phenomenon in internal combustion engines, and actuator and sensor failure detection in spacecraft. In the case of knocks, in addition to causing inefficient operation and pollution, they also create pressures and forces above the design specifications, which could lead to mechanical damage and failure. We have successfully used wavelet transform, as a method of processing unsteady signals, in engine fault detection. We are proposing to develop complementary analysis tools to improve accuracy and the fault detection rate and extend the current results from single cylinder cases to more practical multi-cylinder applications. In the case of spacecraft actuator/senor failures, we are developing techniques for single and multiple component failures detection and devising schemes for system reconfiguration and recovery. We will examine new software/control and hardware redundancies with emphasis on optimal spatial placement of actuators and sensors in the spacecraft to improve the global first-level fail-safe characteristics of the system for robust recovery.