1. Author
Halim Alwi
Christopher Edwards
Chee Pin Tan
2. Content introduction
In safety-critical systems, there is an inherent requirement that, overall, a certain degree of possibly degraded performance must be maintained even in the event of a severe fault or failure within the system. The ability to handle failure and failure conditions was originally called "self-healing control." However, it is now more commonly referred to as "fault-tolerant control". The aerospace industry is often the driving force and focus of such research. As recent crashes in London and Madrid illustrate, failures, however statistically unlikely, still occur in civil aviation, and preventing significant loss of life depends almost entirely on the good judgment and good judgment of the pilot. Skill. Generally speaking, fault-tolerant control (FTC) schemes are classified as passive or active. Passive schemes operate independently of any fault information, essentially exploiting the robustness of the underlying control paradigm. This approach is usually less complex, but is conservative in order to cope with the "worst case" failure effects. Active fault-tolerant controllers typically react to the occurrence of faults by using information from a fault detection and isolation (FDI) scheme, and they invoke some form of reconfiguration. This represents a more flexible architecture. Early publications focused on so-called projection methods, where if a specific fault is detected and identified, the corresponding control law is selected from a set of pre-specified and pre-computed controllers and switched online. Subsequent methods tended to focus on online adaptation or online controller synthesis. In the event of a severe failure, such as a complete actuator/sensor failure, reconfiguration is often required. For example, if a sensor or actuator fails completely, no adaptation within this feedback loop can restore performance without modifying the selection (ie, reconfiguration) of the actuators and sensors coupled via the controller. Fault-tolerant control can be considered an intersection of multiple research fields and is essentially an open problem. As expected, many robust control paradigms have been used as the basis for fault-tolerant controllers. The possibility of exploiting the inherent robustness of sliding modes for fault tolerance has been previously explored for aerospace applications, and work in [128] argued that sliding mode control has the potential to become an alternative to reconfigurable control.
Observer-based methods are the most popular form of model-based fault detection filters. Typically (in linear observer schemes) the output estimate error, formed as the difference between the measured object output and the observer output, is scaled to form the residual error. During fault-free operation, this residual should be "zero", but in the event of a fault, the strain should be "large" and act as an alarm. The authors pioneered a series of work on the development of sliding mode observers for fault estimation. This is achieved by appropriate scaling and filtering of the so-called "equivalent output error injection", which represents the average value required for the nonlinear output error injection term to maintain sliding. This is a unique property of sliding mode observers, arising from the introduction of sliding motion that forces the observer's output to accurately track object measurements. Even in the presence of a faulty actuator, the sliding mode forces the observer's output to track the measurement perfectly and the state can still be accurately estimated. The fault reconstruction signal is not calculated from the residual error based on the estimated error of the calculated output (which will be zero during the sliding motion), but from the equivalent output error injection signal. Therefore, it is in principle possible to simultaneously achieve accurate state estimation and fault estimation from a single (sliding mode) observer. This is quite different from the situation with traditional linear observer designs for FDI, which require a trade-off between robustness with respect to state estimation and fault sensitivity using detection based on the residuals of the output errors. Robust state estimation while maintaining fault sensitivity is a unique property of sliding mode observers.
This book will cover the theoretical development and implementation of sliding mode schemes for fault-tolerant control. A key development in this book considers sliding mode control allocation schemes for fault-tolerant control based on integral action and model reference frames. Unlike many control distribution schemes in the literature, one of the main contributions described in this book is the use of actuator validity levels to redistribute control signals to the remaining healthy actuators when a fault/failure occurs. Strict stability analysis and design procedures were developed for this program from a theoretical perspective. Fixed control allocation structures are also rigorously analyzed in cases where actuator effectiveness level information is not available. The proposed scheme shows that faults and even some total actuator failures can be handled directly without reconfiguring the controller. Later chapters of this book present the results of a real-time hardware implementation of the controller on the 6-DOF SIMONA flight simulator at the University of Delft as part of the GARTEUR AG16 program.
Chapter 1 provides an overview of recent developments in fault detection and fault-tolerant control. It is intended to provide impetus for theoretical development in subsequent chapters.
Chapter 2 begins with a definition of the terms fault and failure and briefly discusses the different types of faults and failures that can occur in actuators and sensors – using specific aircraft as examples. This chapter introduces the concept of fault-tolerant control and provides an overview of different FTC and FDI research areas. The main concepts and strategies behind some FTC and FDI schemes in the literature are also discussed, as well as their advantages and disadvantages.
Chapter 3 briefly introduces the concept of sliding mode control and examines its properties. This chapter also highlights the benefits of sliding mode application in the FTC and FDI sectors. This concept is introduced using a simple pendulum as an example. Unit vector methods, slip surface design, and tracking requirements (integral action and model reference-based tracking) for multi-input systems are also discussed. Chapter 3 concludes with a discussion of the benefits and motivations of sliding mode control in the FTC and FDI fields.
Chapter 4 studies the application of sliding modes in observer design. A historical development is outlined, thereby describing a specific class of sliding mode observers that will be used throughout the book. It will be shown how unique properties associated with the so-called equivalent injection signal required to maintain slip can be exploited to reconstruct actuator and sensor faults, which are modeled as additional perturbations to the inputs and outputs of the device. A design method based on linear matrix inequalities is proposed. These methods exploit all available degrees of freedom associated with the choice of observer gain. This chapter describes a sliding mode observer that can reconstruct faults but is robust to perturbations/instabilities that may corrupt the observer due to mismatches between the model and the actual system for which it was designed. reconstruction quality. Initially, a design approach was developed for the case of actuator failure. The sliding mode observer scheme developed in this chapter is also compared with the more traditional linear unknown input observers popular in the literature.
Chapter 5 examines the assumptions that must be made for the observer scheme described in Chapter 4. (These are equivalent to a relative minimum phase limit on the transfer function matrix relating the unknown fault signal to the measured values.) This chapter explores ways to remove these limits at the expense of creating a cascaded observer structure. The building blocks of the cascade will be the observer formulas taken from Chapter 4, and explicit construction algorithms will be given to ensure that the overall scheme still works in the case where the relative degree between the fault and the measured value is greater than or equal to 2. Accurate estimation of actuator failure. The advantages offered by these schemes over traditional linear methods, especially UIO, will be demonstrated.
Chapter 6 will pay special attention to sensor failure. Different formulations will be considered in which the measured output signal is filtered to produce a "virtual system" in which sensor failure manifests itself as an "actuator failure". Therefore, the actuator fault reconstruction ideas in the previous chapters can be applied to virtual systems to reconstruct sensor faults. The results will also be extended to the case of unstable plants, which results in post-filtered non-minimum phase configurations.
Chapter 7 considers the real-time implementation of the sensor fault reconstruction scheme from Chapter 6 (for FDI and FTC) on a laboratory crane and a small DC motor platform. These rigs provide a cheap, safe, and practical demonstration of the ideas presented in Chapter 6. Data acquisition and (subsequent) controller implementation were implemented using MATLAB® and dSPACE®. The sensor fault estimates obtained from the online sliding mode FDI scheme have been used to correct the sensor's measurement output. "Virtual sensors" are used in the control algorithm to form an output tracking error signal, which is processed to generate a fault-tolerant control signal.
Chapter 8 proposes a new reconfigurable control sliding mode scheme. The controller is based on a state feedback scheme in which the nonlinear unit vector term is allowed to increase adaptively when the onset of a fault is detected. The scheme is applied to a benchmark aircraft problem. Compared to other fault-tolerant controllers previously implemented on this model, the controller proposed in this book is simple, but has been shown to work throughout the "up and down" flight envelope. Shows good suppression of a certain class of actuator failures. However, the proposed controller cannot directly cope with the complete failure of the actuator. In the second half of this chapter, fault-tolerant control is demonstrated using sensor fault reconstruction methods to correct fault measurements prior to control law calculation. Here, a formal closed-loop analysis is performed on the obtained solution. An example of the application of this approach to a benchmark aircraft problem is described.
Chapter 9 proposes an online sliding mode control allocation scheme for fault-tolerant control. When a fault or failure occurs, the control distribution scheme uses the actuator's effectiveness level to redistribute the control signal to the remaining actuators. This chapter analyzes the sliding mode control allocation scheme and determines the nonlinear gain required to maintain sliding. This allocation scheme shows that faults and even some total actuator failures can be handled directly without reconfiguring the controller.
Chapter 10 describes an adaptive model-referenced sliding mode fault-tolerant control scheme with online control assignment. As discussed in Chapter 9, when a fault or failure occurs, the control distribution scheme uses the actuator's effectiveness level to redistribute the control signal to the remaining actuators. At the same time, adaptive nonlinear gains and reference models provide online tuning for the controller. This chapter provides a rigorous stability analysis for the model reference scheme. The scheme has been tested on a linearization of the ADMIRE aircraft model to convey the ideas related to the proposed scheme and to show that various faults and even complete actuator failures can be handled.
Chapter 11 describes the implementation of the sliding mode allocation scheme in Chapter 9 on the 6-degree-of-freedom research flight simulator SIMONA at Delft University of Technology in the Netherlands. The controller in Chapter 9 is implemented in "C" and runs on the "Flight Control" computer associated with SIMONA. Real-time implementation issues are discussed, and a series of failure scenarios from the GARTEUR AG16 benchmark are tested and discussed.
Chapter 12 presents the scenario of ELAL flight 1862 (Bijlmermeer incident) – one of the case studies of GARTEUR AG16. The results in this chapter present the results of the "Flight Test" event and the GARTEUR AG16 final workshop held at TU Delft in November 2007. The results show that experienced test pilots successfully implemented the sliding mode controller on SIMONA in real time and flew and evaluated the controller.
Finally, Chapter 13 makes a summary and makes suggestions for future work.
3. Original book catalog
