Esmaeil Naderi

A Multiple Model-Based Approach for Fault Diagnosis of Jet Engines

In this brief, a novel real-time fault detection and isolation (FDI) scheme that is based on the concept of multiple model is proposed for aircraft jet engines. A modular and a hierarchical architecture is developed which enables the detection and isolation of both single faults as well as multiple concurrent faults in the jet engine. The nonlinear dynamics of a dual spool jet engine is linearized and a set of linear models corresponding to various operating modes of the jet engine (namely healthy and different faulty modes) at each operating point is obtained. Using the multiple model approach the probabilities corresponding to each operating point of the jet engine are generated and the current operating mode of the system is detected based on evaluating the maximum probability criteria. It is shown that the proposed methodology is also robust to the failure of pressure and temperature sensors and extensive levels of noise outliers in the sensor measurements. Simulation results are presented that demonstrate the effectiveness and capabilities of our proposed multiple model FDI algorithm for both structural faults and an actuator fault in the aircraft jet engine.

Fault Diagnosis of Jet Engines by Using a Multiple Model-Based Approach

In this paper, a novel real-time fault detection and isolation (FDI) scheme that is based on the concept of multiple model approach is proposed for jet engines. A modular and a hierarchical architecture is proposed which enables the detection and isolation of both single as well as permanent concurrent faults in the engine. The nonlinear dynamics of the jet engine is linearized in which compressors and turbines maps are used for performance calculations and a set of linear models corresponding to various operating modes of the engine (namely healthy and different fault modes) at each operating point is obtained. Using the multiple model approach the probabilities corresponding to each operating point of the engine are generated and the current operating mode of the system is detected based on evaluating the maximum probability criteria. It is shown that the proposed methodology is also robust to the failure of pressure and temperature sensors and extensive levels of noise outliers in the sensor measurements. Simulation results presented demonstrate the effectiveness of our proposed multiple model FDI algorithm for both structural faults and actuator fault in the jet engine.Copyright

Fault Diagnosis of Gas Turbine Engines by Using Dynamic Neural Networks

This paper presents a novel methodology for fault detection in gas turbine engines based on the concept of dynamic neural networks. The neural network structure belongs to the class of locally recurrent globally feed-forward networks. The architecture of the network is similar to the feed-forward multi-layer perceptron with the difference that the processing units include dynamic characteristics. The dynamics present in these networks make them a powerful tool useful for identification of nonlinear systems. The dynamic neural network architecture that is described in this paper is used for fault detection in a dual-spool turbo fan engine. A number of simulation studies are conducted to demonstrate and verify the advantages of our proposed neural network diagnosis methodology.Copyright

A data-driven approach to actuator and sensor fault detection, isolation and estimation in discrete-time linear systems

In this work, we propose explicit state-space based fault detection, isolation and estimation filters that are data-driven and are directly identified and constructed from only the system input-output (I/O) measurements and through estimating the system Markov parameters. The proposed methodology does not involve a reduction step and does not require identification of the system extended observability matrix or its left null space. The performance of our proposed filters is directly connected to and linearly dependent on the errors in the Markov parameters identification process. The estimation filters operate with a subset of the system I/O data that is selected by the designer. It is shown that the proposed filters provide asymptotically unbiased estimates by invoking low order filters as long as the selected subsystem has a stable inverse. We have derived the estimation error dynamics in terms of the Markov parameters identification errors and have shown that they can be directly synthesized from the healthy system I/O data. Consequently, the estimation errors can be effectively compensated for. Finally, we have provided several illustrative case study simulations that demonstrate and confirm the merits of our proposed schemes as compared to methodologies that are available in the literature.

Fault diagnosis of gas turbine engines by using dynamic neural networks

The goal of this paper is to present an innovative methodology for performing fault detection in gas turbine engines by utilizing dynamic neural networks. The proposed neural network architecture selected belongs to the class of locally recurrent globally feed-forward networks. The envisaged network is structurally similar to a feed-forward multi-layer perceptron with the difference that the employed processing units are not static and possess dynamic characteristics. The developed and constructed dynamic neural network architecture is then used to perform fault detection of anomalies in a dual-spool turbo fan engine. A number of simulation studies are conducted to demonstrate and verify the advantages and capabilities of our proposed neural network diagnosis methodology.

Inversion-based output tracking and unknown input reconstruction of square discrete-time linear systems

In this paper, we propose a framework for output tracking control of both minimum phase (MP) and non-minimum phase (NMP) systems {as well as systems with transmission zeros on the unit circle}. Towards this end, we first address the problem of unknown state and input reconstruction of non-minimum phase systems. An unknown input observer (UIO) is designed that accurately reconstructs the minimum phase states of the system. The reconstructed minimum phase states serve as inputs to an FIR filter for a delayed non-minimum phase state reconstruction. It is shown that a quantified upper bound of the reconstruction error exponentially decreases as the estimation delay is increased. Therefore, an almost perfect reconstruction can be achieved by selecting the delay to be sufficiently large. Our proposed inversion scheme is then applied to solve the output-tracking control problem. {We have also proposed a methodology to handle the output tracking problem of systems that have transmission zeros on the unit circle in addition to MP and NMP zeros.} Simulation case studies are also presented that demonstrate the merits and capabilities of our proposed methodologies.

Subspace-based identification of linear systems using arbitrary data segments

Subspace-based identification processes are commonly performed using sequential data that lead to restrictive conditions for real applications. In most industrial processes, the available data are composed of few segments of excitatory input sections as well as large intervals of steady data characteristics. System identification using the entire available data only increases the computational cost without providing any significant or positive effects on the identification outcome. In this paper, we propose a novel methodology for identification and construction of system matrices from arbitrary data segments such that the current subspace-based methods will remain valid for the purpose of system identification.