Multiple Model Based Actuator Fault Diagnosis And Active Fault Tolerant Control For Flight Control Systems

The need for increased flight safty and aircraft reliability leads to the design of reconfigurable fault-tolerant control systems. Fault diagnosis combined with fault-tolerant control for flight control system is a key enabling technology for increasing safty and reliability of control systems, and has become the focus of many investigations in recent years and received significant attentions. The primary benefit of its application in both military and civil applications is the ability to significantly enhance flight safety, besides this, it has the potential to recover the desired stability and performance characteristics, such that a crippled aircraft can complete its mission and land successfully. In this thesis, fault diagnosis and active tolerant control MM-based scheme is proposed to diagnose and accommodate multiple actuator faults.Firstly, the MM theory is introduced, and the two methods of generating multiple models, that is RRQR and multi-observer, are investigated. For the bank of controller, the existence of it is analysed and proved, and we analyse all faults in flight control system and their characteristics and provide their parameterized model.The case of second-order actuator dynamics with non-measurable rates is investigated, an adaptive observer is designed to generate residual error to diagnose fault and estimate some unknown parameter fault. According to second-order actuator dynamics, low order actuator dynamics is introduced to represent actuator fault.A reduced-order UIO based robust fault diagnosis is proposed to detect and estimate the fault of flight control systems with plant uncertainties and disturbance. The systems are decomposed into two different subsystems. One is decoupled from actuator fault and the other is affected by the fault, but its states can be measured directly. The developed approach can not only reduce the computations, but also improve the speed and accuracy of fault diagnosis. For the case of multiple faults, the scheme is developed based on a suitable combination of multiple model and UIO, the main idea of it is that MM are used to describe different fault scenarios, then a bank of UIO are designed to implement the disturbance de-coupling. By switching logic, the best model is selected to match the current system, the developed approach improve the system performance and can provide effective fault detection and magnitude estimation in the presence of fault and unknown disturbances. The design scheme of the proposed approach is straightforward and is easy to implement.For the case LOE, An active fault tolerant control strategy is developed for flight control system with uncertainty and input constraints. The design is based on a direct adaptive control algorithm andμ-modification is introduced in the model reference control architecture. Radial basis function network is used to approximate the model uncertainty, and plant-model matching conditions are derived, the developed method achieves both closed-loop signal boundedness and asymptotically track the output of a reference model. Moreover, the adaptive control laws for unknown parameters are designed and closed-looped stability is proved, and finally, some simulation results are given.Construct a bank of identification models to match the different fault scenarios, i.e., Lock-in-place, Float and Hard over. When fault occurs, devise a suitable strategy for switching among the models to search the closest model matching the current faulty system, and then apply the direct adaptive reconfiguration control method to accommodate fault. Under the MM framework, fault model is only based on the fault type, and it do not need the exact fault informations, so it do not have a high proior informance request on aircraft fualt model. Hence, under this framework, the proposed fautl diagnosis and tolerant control approach can detect, estimate and accommodate fault which may be unknwon, known or with those similar to known fault. Furthermore, using multiple models to cover different dynamics of the fauty system, it can not only allow a large class of fault conditions to be modeled, but also achieve fast and accurate fault accommodate and improve the performance of faulty system, and greatly reduce false alarm and missing detection rates.