Fault detection and isolation of actuator failures in jet engines using adaptive dynamic programming

https://doi.org/10.1016/j.amc.2021.126664Get rights and content

Highlights

  • A new kind of data-based fault detection method is proposed, which combines adaptive dynamic programming and multiple-modeling.

  • A stable working state is established by using ADP and history data. Once the system runs out of the stable state, it’s known for sure that fault occurs.

  • A whole stable working space of the system can be built with the method of multi-modeling and data of all working modes, as long as system states remain in the working space, system is working normally, exclude the disturbance of modes switching, and improve the accuracy of the fault detection algorithm.

  • The fault can be isolated, if the direction and magnitude of the deviate vector are solved, which can be achieved by approximation and iteration using the method of ADP.

Abstract

This paper presents a adaptive dynamic programming-based fault detection and isolation (FDI) scheme to detect and isolate faults in an aircraft jet engine. To this end, the weights in Actor-Critic neural networks are first tuned to learn the input-output map of the jet engine considering its multiple working modes. The convergences of the trainings in Critic-Actor neural networks are strictly proved without knowing the drift dynamics and the input dynamics in the presence of unknown nonlinearities and approximation errors. Using the residuals that are generated by measuring the difference of each network output and the measured engine output, various criteria are established for accomplishing the fault diagnosis task, that addresses the problem of fault detection and isolation of the system components. A number of simulation studies are carried out for combustion chamber of a single-spool jet engine to demonstrate and illustrate the advantages, capabilities, and performance of our proposed fault diagnosis scheme.

Introduction

The development of the jet engine as an aircraft power plant has been so rapid that it is difficult to appreciate that prior to the 1950s, very few people had heard of this method of aircraft propulsion. The jet engine although appearing so different from the piston engine-propeller combination, applies the same basic principles to effect propulsion, both propel their aircraft solely by thrusting a large weight of air backwards. Although today jet propulsion is popularly linked with the gas turbine engine, which will be mainly discussed in this paper, there are other types of jet propelled engines, such as the ram jet, the pulse jet, the rocket, the turbo/ram jet, and the turbo-rocket. The mechanical arrangement of the gas turbine engine is simple, for it consists of only two main rotating parts, a compressor and a turbine, and one or a number of combustion chambers [1].

Many researchers have been working in the field of fault diagnosis for jet engines [2], [3], and several methods and techniques have been raised. The problem of fault detection and isolation (FDI) can be solved mainly in two different ways, model-based [4], [5] and data-based respectively. Data-based fault detection method is studied and applied increasingly, for mass data is stored easily, as the application of computers and sensors in control systems [6], [7]. Generally speaking, fault-tolerant control allows systems to work smoothly in a certain range, but abrupt changes induced by actuator failures, which will be mainly discussed in this paper, could disable the fault-tolerant controller, such as loss in effectiveness (LIE), bias and stuck (lock-in-place).

In recent years, data-based fault detection and fault-tolerant control [8], [9] becomes one of the most popular research fields among all others. Meskin has proposed a novel real-time fault detection and isolation (FDI) scheme for aircraft jet engines based on the concept of multiple model [10], [11], which linearize the nonlinear dynamics to a set of linear models [12] and Abbasfard concerned the FDI problem of nonlinear systems by using a symbolic-based linear multiple model approach [13]. An intelligent optimal control scheme for unknown nonlinear discrete-time systems was proposed by Liu [14], [15], using an iterative adaptive dynamic programming (ADP) algorithm [16], [17] belongs to machine learning, for the corresponding Hamilton–Jacobi–Bellman(HJB) equation is too complex to solve directly. Yet there’s hardly no research of fault detection and isolation in nonlinear systems using the method of ADP.

In this paper, a new kind of data-based fault detection method is proposed, which combines adaptive dynamic programming and multiple-modeling. To start with, a stable working state is established by using ADP and history data. Once the system runs out of the stable state, it’s known for sure that fault occurs. Furthermore, a whole stable working space of the system can be built with the method of multi-modeling and data of all working modes, as long as system states remain in the working space, system is working normally, exclude the disturbance of modes switching, and improve the accuracy of the fault detection algorithm. Eventually, the fault can be isolated, if the direction and magnitude of the deviate vector are solved, which can be achieved by approximation and iteration using the method of ADP.

The rest of the paper is organized as follows. In Section 2, modeling of the jet engine and a brief theoretical background of the techniques used in this paper the will be discussed. In Section 3, the algorithm of fault detection and isolation of the aero-engine with ADP is proposed. A simulation example is given to prove the method this paper proposed in Section 4. Conclusions and future works are presented in Section 5.

Section snippets

Jet engine modeling and theoretical background

In this section, the modeling of jet engines and theory and techniques used in this paper will be discussed [18], [19]. To be specific, a typical turbofan engine was chosen in this paper, which represents a variety of advanced engines. The basic components of a turbofan engine are duct fan, intakes, compressor, combustion chamber, turbine and nozzle. The figure follows is the structure of a turbofan engine (Fig. 1).

To build the stable working modes of the system, an Integral Reinforcement

Fault detection and isolation algorithm based on integral reinforcement learning

In this section, an IRL-based fault detection and isolation algorithm for the jth mode of the system is proposed. Base on the critic and actor NNs built in last section, a further fault NN is proposed to approximate the faulty model of the jth environment of the system, and even isolate the fault by calculate the fault. Fig. 2 follows shows the process of the fault detection algorithm.

Simulation result

In this section, simulation results of different modes of the system are presented. The engine model is presented in chapter 2, and yr is chosen as yr0=[0,0,0]T represents the normal working state, and yr1=[0.08,0.06,0.04]T and yr2=[1.33,0.80,0.53]T represent two accelerant states respectively. In Critic-Actor NNs, chose ψj(x)=[x12,x1x2,,x1x5,x22,x2x3,,x52]T, and ϕj(x) is partial derivative of ψj(x). The initial condition of x0 and initial weights of Critic-Actor NNs are chosen randomly in [1

Conclusions

In this paper, a new kind of data-based fault detection method is discussed, which combines adaptive dynamic programming and multiple-modeling. The fault detection algorithm for the nonlinear system is designed for each operating modes, to approximate the system model and estimate the faults. The Critic-Actor NNs convergence was guaranteed without knowing the drift dynamics and the input dynamics in the presence of unknown nonlinearities and approximation errors. A simulation example of

References (25)

  • J. Blesa et al.

    Robust identification and fault diagnosis based on uncertain multiple input-multiple output linear parameter varying parity equations and zonotopes

    J. Process Control

    (2012)
  • R. Royce

    The Jet Engine

    (2015)
  • Z. Peng

    Aeroengine Fault Diagnostics Based on Kalman Filter

    Diss.

    (2008)
  • L. Yebo

    Research on gas fault fusion diagnosis of aero-engine component

    Acta Aeronaut. Astronaut. Sin.

    (2014)
  • T. Sedighi, A.J. Koshkouei, K.J. Burnham, Observer-based residual design for nonlinear...
  • S. Yin et al.

    Data-driven adaptive observer for fault diagnosis

    Math. Probl. Eng.

    (2012)
  • P.-L. Hsu et al.

    Diagnosis of multiple sensor and actuator failures in automotive engines

    Veh. Technol., IEEE Trans.

    (1995)
  • W. Yulei

    Research on Data-Driven Design Methods of Fault Diagnosis and Fault-Tolerant Control Systems

    Diss.

    (2013)
  • S. Jianan

    A Research of Automobile Engine Fault Diagnosis Based on Data-Driven

    Diss.

    (2013)
  • F. Yue

    Nonlinear Adaptive Decoupling Control Based on Multiple Models and Neurals Networks

    Diss.

    (2009)
  • G. Yuying

    Multiple Model Based Actuator Fault Diagnosis and Active Fault Tolerant Control for Flight Control Systems

    Diss.

    (2009)
  • N. Meskin et al.

    A multiple model-based approach for fault diagnosis of jet engines

    IEEE Trans. Control Syst. Technol.

    (2013)
  • Cited by (5)

    • Quantitative analysis of incipient fault detectability for time-varying stochastic systems based on weighted moving average approach

      2022, Applied Mathematics and Computation
      Citation Excerpt :

      The past several decades have witnessed a surge of research interest on the problem of fault detection (FD) owing to demands for safety and reliability of modern industrial systems [1], such as the power network system [2], the sensor network system [3,4], the multi-agent system [5,6], the three-tank system [7] and the aircraft jet engine [8].

    • Optimal Layout of Asphalt Mixing Plant Based on Dynamic Programming Algorithm

      2022, Proceedings - 2022 European Conference on Communication Systems, ECCS 2022

    This work was supported in part by the funds of National Science of Foundation China (Grant nos. 61873306, U1908213, 6162100, 61420106016), in part by the National Key Research and Development Program of China (2020YFE021100), in part by the Fundamental Research Funds for the Central Universities (Nos. N2004018), and the research fund of the State Key Laboratory of Synthetical Automation for Process Industries (Grant nos. SAPI2019-3, 2018ZCX19).

    View full text