Failure Detection System Research Articles

Détection distribuée par reconnaissance floue de chroniques

This article deals with the problem ofprocess failure detection for discrete event systems by a chronicle-based approach. Then, the behavior of the system is modeled through a chronicle which is a time model of the process evolutions. Each chronicle is composed of a set of events and of a set of timing constraints among these events. The observed behavior is checked on-line using this time signature. The detection principle consists in checking that each observed event is consistent with the chronicle in regards with the timing constraints. Attention is focused on distributed discrete event systems. Therefore, a chronicle is decomposed into a set of sub-chronicles distributed on several monitoring sites. The distributed context of the approach leads to consider the problem of communications delays between monitoring sites that induce some uncertainty in timing constraint checking.

Auxiliary signal design for active failure detection in uncertain linear systems with a priori information

A method for auxiliary signal design for active failure detection based on a multi-model formulation of normal and failed systems is developed which allows for a priori information about initial conditions and the possibility of having a known input in addition to the test signal. These results permit the consideration of additional types of failures, such as biases, that a previous approach could not handle. Both theoretical results and computational algorithms are developed.

Robust Integrated Flight Control Design Under Failures, Damage, and State-Dependent Disturbances

Ar obust integrated fault-tolerant flight control system is presented that accommodates different types of actuator failures and control effector damage, even while rejecting state-dependent disturbances. It is shown that a decentralized failure detection, identification, and reconfiguration system, combined judiciously with adaptive laws for damage estimates and variable structure adjustment laws for disturbance estimates, yields a stable system despite simultaneous presence of failures, damage and disturbances. The proposed system is well suited for the case of first-order actuator dynamics. The properties of the proposed algorithms are illustrated on a medium-fidelity nonlinear simulation of Boeing’s Tailless Advanced Fighter Aircraft.

Discussion on: “A Multi-Model Approach to Failure Detection in Uncertain Sampled-Data Systems”

The paper of Nikoukhah and Campbell addresses the problem of active failure detection for uncertain systems. The authors study the following framework. Suppose we have two uncertain system models M0 and M1. Both these models have the input vðtÞ and the output yðtÞ. Furthermore, let Aiðvð ÞÞ denote the set of all possible outputs yð Þ of the uncertain system model Ai with input vð Þ ði 1⁄4 0, 1Þ. The standard failure detection problem is to construct an input vð Þ such that

A Multi-Model Approach to Failure Detection in Uncertain Sampled-Data Systems

In this paper we propose a methodology for active failure detection for sampled-data systems in the presence of model uncertainty. Active failure detection consists of injecting a signal (called an auxiliary signal) into the system. Using a multi-model framework to represent normal and failed behaviors of the system,we develop algorithms for the design of optimal auxiliary signals and online detection tests. The tests give guaranteed detection. Any conservatism inherited from the multi-model framework increases the size of the test signal and is not reflected in missed failures or false alarms as with several other approaches.

An overview of the applications and solutions of a fundamental matrix equation pair

Equation TA− FT= LC ( F is stable) is necessary and sufficient for the output of a feedback compensator ( F, L, K Z , K y ) to converge to a state feedback (SF) signal K x (t) for a constant K, where ( A, B, C,0) is the open loop system and TB is the compensator gain to the open loop system input. Thus, equation TB=0 is (1) the defining condition for this feedback compensator to be an output feedback compensator. Equation TB=0 is also the necessary and sufficient condition to (2) fully realize the critical loop transfer function and robust properties of SF control if K is systematically designed. Furthermore, because B is compatible to the open loop system gain to its unknown inputs and its input failure signals, TB=0 is also necessary for (3) unknown input observers and (4) failure detection and isolation systems. Finally, this equation pair ( TA− FT= LC, TB=0) is the key condition of a really systematic and explicit design algorithm for (5) eigenstructure assignment by static output feedback control. This paper reviews the existing solutions of this equation pair, and points out that a general and exact solution is uniquely direct, simple, and decoupled. This paper also points out that these unique features also enable two decisive advantages: (1) the systematic compensator order adjustment and (2) a simple and approximate solution which is general to all systems ( A, B, C,0) and which can be simply added to the exact solution whether it exists or not.

Fuel and Fission Gas Behavior during Rise-to-Power Test of the High Temperature Engineering Test Reactor (HTTR)

The rise-to-power tests of the High Temperature Engineering Test Reactor (HTTR) have been carried out successfully by the Japan Atomic Energy Research Institute (JAERI). For the safe operation of HTTR, the continuous and reliable measurement of the coolant activity is required to evaluate the fuel performance during normal operating conditions. In order to measure the primary coolant radioactivity (PCR), the PCR instrumentation of the safety protection system, the fuel failure detection (FFD) system and the primary coolant sampling system have been installed in the primary circuit. The PCR was less than 103 MBq/m3, and measured Kr and Xe isotopes were less than 0.1 MBq/m3 during the rise-to-power tests. The measured fractional releases are constant at 2x10-9 up to 60% of the reactor power, and then increase to 7x10-9 at full power operation. The prediction shows good agreement with the measured value. These results showed that the release mechanism varied from recoil to diffusion of the generated fission gas from the contaminated uranium in the fuel compact matrix.

Robust Diagnosis of Switching Systems

This paper presents an approach for robust diagnosis of switching systems based on an extended version of the timed failure propagation graph model. The extended failure propagation graph model is a labeled graph used for the representation of failure conditions and their propagation modeled as causal relations with timing properties for a general class of systems with both timebased and event-driven dynamics such as hybrid and discrete event systems. We introduce the extended model and describe the structure and main components of the failure detection and isolation system based on the proposed model.

Failure detection, identification and reconfiguration system for a redundant actuator assembly

In the paper a design of a Failure Detection, Identification and Reconfiguration (FDIR) System for a redundant actuator assembly is presented. The scheme assures that the output of the redundant actuator assembly is close to its desired value despite failures of individual actuators in the assembly. The reconfiguration re-allocates the control signals to the healthy actuators locally to compensate for the failure. In the paper the modeling of a double-redundant actuator assembly characterized by first-order dynamics of individual actuators is presented first. This is followed by the design of adaptive FDIR algorithms, and corresponding proofs of stability. The proposed approach is illustrated through computer simulations.

Using a bayes classifier to optimize alarm generation to electric power generator stator overheating

This paper shows how a Bayes classifier can be implemented for a failure detection system where statistical failure data is not available for one of the classes. Results of field data obtained from a large electric power generator are shown. The classifier is further improved by the iterative re-evaluation of the prior probabilities, which results in the use of higher alarm threshold values when a good agreement between the monitored quantity and its estimated value is observed, while large disagreement values result in smaller thresholds. As expected, the proposed system is an improvement over a classical Bayesian implementation and a large improvement over a fixed, arbitrary value threshold classifier.

Adaptive State Filtering for Space Shuttle Main Engine Turbine Health Monitoring

Real-time estimation of system states or parameters that are dife cult or expensive to measure directly is often needed for adaptive control or health monitoring purposes. A practical algorithm is proposed for adaptive state e ltering in nonlineardynamic systemswhen the state equations areunknown or too complex to model analytically. The state equationsare constructively approximated by using recurrent neural networks. The proposed algorithm is based on the predictor-update approach of the Kalman e lter, but a least-mean-square e lter implementation with an adaptive e lter gain is used. Furthermore, unlike the Kalman e lter and its nonlinear extensions, the proposed algorithm makes minimal assumptions regarding the underlying nonlinear system dynamics and their noise statistics. The e lter is used to estimate the high-pressure turbine discharge temperature of the space shuttle main engine, during setpoint changes and turbopump failures. The e lter is developed by using simulated engine data, and its performance is tested on both simulated and actual recorded space shuttle main engine transients. When the complexity of the problem studied is considered, the resulting e lter accuracy is shown to be quite acceptable. Further use of the adaptive e lter gain developed is to enable real-time detection of certain system failures, such as the turbopump failures of the space shuttle main engine. This concept is demonstrated also by using both simulated and experimental failure data.

Detection of Component Failures for Smart Structure Control Systems

Uncertainties in the dynamics model of a smart structure are often of significance due to model errors caused by parameter identification errors and reduced-order modeling of the system. Design of a model-based Failure Detection and Isolation (FDI) system for smart structures, therefore, needs careful consideration regarding robustness with respect to such model uncertainties. In this paper, we proposes a new method of robust fault detection that is insensitive to the disturbances caused by unknown modeling errors while it is highly sensitive to the component failures. The capability of the robust detection algorithm is examined for the sensor failure of a flexible smart beam control system. It is shown by numerical simulations that the proposed method suppresses the disturbances due to model errors and markedly improves the detection performance.

Motion in Mimic Displays: Effects on the Detection and Diagnosis of Electrical Power System Failures

New power system displays have been developed to aid operators in the detection and diagnosis of faults. The enhancement of integrated one-line diagrams of power system data with motion and motion cues was examined in this experiment. Participants acknowledged and solved power system failures on a simulated power network across a number of trials of varying complexity. Participants performed these tasks using interactive displays indicating power flow with digits, stationary arrows, or moving arrows. For high complexity scenarios, results indicated a general advantage for the motion display in the diagnostic task (problem resolution) but a slight advantage of the digital display on the fault detection task (problem identification). Performance with the stationary arrow display was generally between the other two groups, being nearly as good as with the digital display in the detection task and nearly as good as with the moving arrow display in the diagnosis task. Further research is necessary to determine the conditions where motion is most beneficial.

Control and fault diagnosis in bottom fermentation systems using parity space approach.

The main objective of this note is to describe a real-time fault diagnosis and control for a cylindroconical fermenter to laboratory scale as an alternative to the classic systems. Development of a good controller for a fed-batch reactor is not enough without a fault diagnosis system. We are working to expand this idea. The failure detection system is based on the parity space approach joined with a fuzzy controller with more robustness and of superior performance in MIMO systems compared to conventional strategies. A 2-week period is reported.

DETECTION OF COMPONENT FAILURES FOR SMART STRUCTURE CONTROL SYSTEMS

Uncertainties in the dynamics model of a smart structure are often of significance due to model errors caused by parameter identification errors and reduced-order modeling of the system. Design of a model-based Failure Detection and Isolation (FDI) system for smart structures, therefore, needs careful consideration regarding robustness with respect to such model uncertainties. In this paper, we proposes a new method of robust fault detection that is insensitive to the disturbances caused by unknown modeling errors while it is highly sensitive to the component failures. The capability of the robust detection algorithm is examined for the sensor failure of a flexible smart beam control system. It is shown by numerical simulations that the proposed method suppresses the disturbances due to model errors and markedly improves the detection performance.

Neural networks-based scheme for system failure detection and diagnosis

In this paper, the neural networks of fault tolerant control scheme in flight control is presented. The scheme can be broken up into three steps: fault detection, diagnosis, and accommodation. Particularly emphasis is placed on fault detection and diagnosis employing neural networks in this paper. The testing of the scheme is illustrated through F-16 aircraft model for an actuator failure.

Reconfigurable NDI controller using inertial sensor failure detection & isolation

A modified derivation of nonlinear dynamic inversion provides the theoretical underpinnings for a reconfigurable control law for aircraft that have suffered combinations of actuator failures, missing effector surfaces, and aerodynamic changes. The approach makes use of acceleration feedback to extract information pertaining to any aerodynamic change and thus does not require a complete aerodynamic model of the aircraft. The control law does require feedback of effector positions to accommodate actuator dynamics. Both accelerometer and rate gyro failure detection and isolation (FDI) systems are implemented, allowing up to three independent failures for each FDI system as long as they are in different axes. Nonlinear simulation results show that the FDI systems improve the robustness to accelerometer/rate gyro uncertainties. An advanced tailless aircraft model is used to demonstrate the concepts. The simulation includes accelerometer and rate gyro noise and bias, failures due to accelerometers, rate gyros, and actuators, and modeled missing surfaces that cause airplane aerodynamic changes.

An intelligent system for failure detection and control in an autonomous underwater vehicle

Autonomous underwater vehicles (AUVs) have been used extensively in deep sea research. Failure detection and control is an important issue in maintaining the stability of an AUV. In most AUVs, the vehicle resurfaces in the event of minor failures such as in the depth sensor, the inclinometer, etc. The paper proposes an intelligent system for failure detection and control in AUVs where the vehicle could continue exploration in case of minor failures in the sensors and control surfaces. The intelligent system, based on the model proposed in Patel and Ranganathan (1996), integrates the adaptability of an artificial neural network (ANN) and the inferencing ability of a fuzzy rule based expert system on a single VLSI chip. The associative function of the ANN is used to recognize and detect the failures by observing the various changing parameters of the dynamic vehicle. The inferencing ability of an expert system suggests ways to control the failure and indicates the subsequent status of the vehicle. The entire system could be used as a low level diagnoser in an overall control system for AUVs.

Similarity Relations in Diagnosis Fuzzy Systems

This paper deals with the use of similarity relations and coefficients in fuzzy logic. Fundamental similarity relations and their use for three possible applications in diagnosis systems are described. They are shown on an example of a system for failure detection of rolling-mill transmissions.

Robust Failure Detection and Isolation for Input-Output Non-Linear Systems

The Unknown Input Observer approach is one of the well-established method for robust failure detection and isolation in linear systems. Although some extensions to non-linear systems have been reported, these extensions are limited to rather restrictive classes of systems. This paper presents an extension of the Unknown Input Observer approach valid for any non-linear system which can be expressed in a NARMAX form. It is shown that the disturbance and failure distribution matrices are composed of the partial derivatives of the model with respect to the corresponding variables. For non-linear systems, these matrices are time-varying, and therefore perfect failure-disturbance decoupling requires time-varying observers. The design procedure of such observers is detailed, and the method is illustrated on simple examples.