Fault Detection for Discrete-Time Nonlinear Impulsive Switched Systems
This paper investigates the fault detection problem for discrete-time nonlinear impulsive switched systems. Attention is focused on designing the fault detection filters to guarantee the robust performance and the detection performance. Based on these performances, sufficient conditions for the existence of filters are given in the framework of linear matrix inequality; furthermore, the filter gains are characterized by a convex optimization problem. The presented technique is validated by an example. Simulation results indicate that the proposed method can effectively detect the faults.
Fault detection (FD) is an important topic in system engineering from the viewpoint of the higher demands for safety and reliability of control systems [1, 2]. The basic idea of the model-based FD is to design observers  or filters  and generate an residual signal. Since the value counted by the residual evaluation function is larger than the predefined threshold, an alarm is generated. To date, there are many methods to solve the FD problem. As one of the typical methods, the FD problem is converted into a robust filtering problem; then technique is presented [5, 6]. For another method, FD systems have been directly considered to be sensitive to the faults and simultaneously robust to the unknown disturbance, then the technique investigates this important issue [7, 8].
On the other hand, switched systems which belong to hybrid systems consist of a finite number of subsystems and a logical rule that orchestrates switching between these subsystems . The primary motivation for studying switched systems comes partly from the fact that switched systems and switched multicontroller systems have numerous applications in control of flight control , missile autopilot design , chemical systems , networked control systems , and many other fields. Until now, a number of recent results are focused on stability and stabilizability under arbitrary switching , restricted switching (like dwell time and average dwell time [15, 16]), multiple Lyapunov functions method, and piecewise quadratic Lyapunov functions. As one of the special switched systems, impulsive switched systems produce impulses when the system is switching among subsystems, and There is also a wide range of actual systems such as engineering, economics, and biology. This kind of impulses will cause instability and oscillations and lead to poor performance. Recently, a number of papers have focused on stability problem of the impulsive switched systems [17–21].
However, the problem of FD design in switched systems schemes is still in the early stage of development and a few results have been reported in the literatures [22–26]. To the best of the authors’ knowledge, the FD problem for impulsive switched systems, especially discrete-time nonlinear impulsive switched systems, has not been investigated yet. It is worth noting that the FD approaches for switched systems without impulses are not appropriate for switched systems with impulses due to the effect of the impulse in switching point. Therefore, a new FD technique is needed to solve the impulsive case. Moreover, even if the mathematics model for the actual system are established with neglecting the impulse in switching point, the inaccurate mathematics model may reduce the robustness of the actual system, and the results may increase the risk of the false alarm. Thus, it is necessary to directly investigate the fault detection problem for impulsive switched systems. As the significance in theory and practice, the FD problem for discrete-time nonlinear impulsive switched systems should be investigated, which motivates us to study this interesting issue.
In this paper, we consider a general class nonlinear impulsive switched system with nonlinear impulsive increments, and the fault detection problem for this class of systems is investigated. Firstly, a weighted performance for discrete-time nonlinear impulsive switched systems is presented; meanwhile, the performance of discrete-time nonlinear impulsive switched systems is derived to reflect the effect on the residual signal from the faults. Subsequently, sufficient conditions for the weighted performance and the performance are formulated by linear matrix inequalities (LMIs) with less conservatism. Finally, the filters gains are characterized in terms of the solution of a convex optimization problem.
The paper is organized as follows. Section 2 introduces the problem under consideration and presents the design objectives. Section 3 illustrates the FD filter design approach in detail. An example is given in Section 4 to demonstrate the proposed method. Conclusions of this paper are given in the last section.
Notation. The superscripts stand for the transposition and the inverse of a matrix, respectively. The matrices () and () denote positive-definiteness (positive semidefinite matrix) and negative-definiteness (negative semidefinite matrix). and represent the identity matrix and the zero matrix with appropriate dimensions, respectively. The Hermitian part of a square matrix is denoted by . The symbol within a matrix represents the symmetric entries.
2. Problem Formulation
2.1. System Model
Consider the following discrete-time nonlinear impulsive switched systems:where is the state and is the measured output. and are the disturbance input and the fault, respectively, which are energy bounded; then they are demanded to belong to . , is impulsive switching time points. Denote . The switching signal specifies that th subsystem is activated when , and . , which is globally Lipschitz continuous, and for all , with and ; that is, the solution is left continuous. is nonlinear functions, and for all . The th subsystem is denoted by the matrices , and with appropriate dimensions.
The following assumptions for nonlinear impulsive switched system (1) are introduced.
Assumption 1. There exist nonnegative scalars , such that
Assumption 2. Denote by the spectral radius for each subsystem and ; then .
Remark 3. Note that owing to the presence of system nonlinearity, the above assumptions essentially draw from the analysis of the stability for the system. Moreover, they are two basic conditions for this kind of systems (see  for further discussion). Therefore we consider the nonlinear impulsive switched systems, which satisfies Assumptions 1 and 2.
For the purpose of the fault detection, the following FD filters are designed:where is the state of the filter, is the residual signal, and . The matrices , and with appropriate dimensions are to be determined.
To present the purpose of this paper more precisely, the following definition is introduced.
2.2. Problem Formulation
The Frameworks of FD Filter Design. Given nonlinear impulsive switched system (1), the FD filters (3) are designed such that nonlinear impulsive switched system (4) is stable, and the fault effects on the residual signal are maximized, while the disturbance effects on the residual signals are minimized. Our design objective of the FD filters can be formulated as the following performances:
Remark 5. Condition (7) is used for the disturbance attenuation performance, which minimizes the disturbance effects on the residual output and ensures that the disturbance is not disastrous. Condition (8) is expressed to maximize the effects of the fault on the residual output . That is, the residual output is sensitive to the fault .
After designing the residual generator, how to evaluate the generated residual is considered. One of the widely adopted approaches is to select an appropriate threshold and an appropriate residual evaluation function. Similar to that proposed in , the residual evaluation function can be chosen as where denotes the evaluation time step.
Let be the threshold. Based on this, the occurrence of faults can be detected by comparing and according to the following logical relationship:
3. The Fault Detection Filter Design
Before beginning this section, the following lemmas are needed to present our main results.
Lemma 6 (see ). Let be a given scalar and a matrix such that , where is an identity matrix with appropriate dimension. Then for all and .
Lemma 7 (see ). Let be a given symmetric positive definite matrix and let be a given symmetric matrix. Then for all , while and denote, respectively, the largest and the smallest eigenvalues of the matrix inside the brackets.
In this section, sufficient conditions on the existence of the FD filters for nonlinear impulsive switched systems (4) would be given, and the desired filter gains can be obtained.
3.1. The Disturbance Attenuation Performance (8)
Considering discrete-time nonlinear impulsive switched system (4) with , we have
Firstly, the weighted performance for nonlinear impulsive switched system (11) is given.
Lemma 8. Let , and be constants satisfying , and , and Assumptions 1 and 2 hold. Furthermore, suppose that discrete-time nonlinear impulsive switched system (11) switches from th subsystem to th subsystem as switched time point . If there exist and Lyapunov functions candidate satisfyingwhere , then nonlinear impulsive switched system (11) is stable with the weight -gain for any switching signal satisfyingwhere , and the function represents rounding real number to the nearest integer greater than or equal to .
Proof. See the appendices.
Subsequently, inequality conditions for the disturbance attenuation performance (8) are constructed.
Theorem 9. Let , , , and be constants satisfying , and . If there exist matrix variables , and and symmetric positive-definite matricessatisfying the following inequalities:where , then switched system (11) is asymptotically stable for any switching signal satisfying (14) and guarantees the weighted performance .
Proof. See the appendices.
3.2. The Fault Sensitiveness Performance (8)
Considering discrete-time nonlinear impulsive switched system (4) with , we have
performance for discrete-time nonlinear impulsive switched system (19) is given.
Lemma 10. Let , and be constants satisfying , and , and Assumptions 1 and 2 hold. Furthermore, suppose that nonlinear impulsive switched system (19) switches from th subsystem to th subsystem as switched time point . If there exist and Lyapunov functions candidate satisfyingwhere , then discrete-time nonlinear impulsive switched system (19) is stable with the -gain for any switching signal satisfying (14).
Proof. See the appendices.
Based on Lemma 10, the following theorem is given to obtain sufficient conditions by linear matrix inequalities.
Theorem 11. Let , , , and be constants satisfying , , , and . If there exist matrix variables , , , and and symmetric positive-definite matricessatisfying the following inequalities:where , then discrete-time nonlinear impulsive switched system (19) is asymptotically stable for any switching signal satisfying (14) and guarantees performance:
Proof. See the appendices.
Moreover, if (27) is feasible, then the FD filter gains can be given by
In this section, we present a numerical example to illustrate the effectiveness of FD design approach. Consider the discrete-time nonlinear impulsive switched systems (1) with two subsystems and two parameters:Given , and and choosing , we solve the convex optimization problem (27) and get the optimal sensitivity performance gain . The gain matrices of fault detection filters and matrix in Lyapunov functions are obtained as
Then, according to (14), the dwell time for each subsystem is obtained:
To illustrate the simulation results of the FD objective, two cases which include the fault for subsystem 1 and subsystem 2, respectively, are considered. Furthermore, the switching signal is generated by satisfying (31) as shown in Figure 1. The threshold can be determined as .
Case 1. The fault for subsystem 1 is simulated as a pulse signal with amplitude 1 that occurred from 90 to 120 steps. The generated residual signal and evaluation of residual evaluation function are shown in Figure 2. The simulation results show that when the fault for subsystem 1 occurs, the subsystem 1 is not activated. Since subsystem 1 is activated at 93 steps the residual signal varies sharply, and at 99 steps, which means that the fault for subsystem 1 can be detected 6 steps after subsystem 1 is activated. Hence, the fault for subsystem 1 can be detected.
Case 2. The fault for subsystem 2 is simulated as a pulse signal with amplitude 1 that occurred from 170 to 220 steps. The generated residual signal and evaluation of residual evaluation function are shown in Figure 3. It can be seen that when the fault for subsystem 2 occurs at 170 steps, the residual signal is changed sharply and at 182 steps. Thus, the fault for subsystem 2 can be detected.
In this paper, the problem of FD filter design for discrete-time nonlinear impulsive switched systems has been investigated. Firstly, the weight performance and the performance are presented, and sufficient conditions to characterize given performances have been formulated as the form of LMI. Subsequently, the gains of FD filters are obtained by a multiobjective optimization problem. Finally, the effectiveness of the proposed method for discrete-time nonlinear impulsive switched systems is illustrated by the example.
A. Proof of Lemma 8
When and th subsystem is activated, along the trajectory of nonlinear, impulsive switched system (4) givesBy Lemma 6, it is clear thatTherefore, when assuming the zero input and using (12), Lemma 7, and Assumption 1, we have the following condition: , where . If (13) holds, then , which implies that Thus, we have
Since (14) holds, that is, there exists such that , then
Therefore, we conclude that converges to zero as ; then nonlinear impulsive switched system (11) with is stable.
For any nonzero and zero initial condition . Let , we can haveIf (13) holds, it is equivalent to For , it can have the following inequality as the similar way:Under the zero initial condition, it implies thatMultiplying both sides of (A.10) by , one can obtainMoreover, it follows from the switching sinal that . ThenIt further implies that Therefore, we conclude that discrete-time nonlinear impulsive switched system (11) has the weighted performance for any switching signal satisfying (14), which completes the proof.
B. Proof of Theorem 9
Premultiplying and postmultiplying to (B.2), it is transformed intoThen, by using the Schur complement formula, we can see that (B.3) is equivalent to (13). Then, according to Lemma 8, if the conditions (16), (17) hold, the switched system (11) is asymptotically stable with a weighted performance for any switching signal satisfying (14), which completes the proof.
C. Proof of Lemma 10
Following the same lines as that for Lemma 8, the switched system (19) satisfying switching signal (14) is stable. Then, the performance defined in (6) for discrete-time nonlinear impulsive switched system (19) is established.
For any nonzero , consider the following index: . It has the following:where . If (21) holds, it is equivalent to . By iteration operation on the above inequality for , we have
Under the zero initial condition, one has and ; thus
Thus, from , we obtain thatThen we have When taking from to , we can further obtain that
D. Proof of Theorem 11
Denote , andTo establish the convex condition, (21) can be rewritten as follows:On the other hand,Based on Projection Lemma, it follows from (D.2) and (D.3) thatwhere introduced by Projection Lemma is the matrix variable of appropriate dimensions. Partition as By Schur complement, (D.4) is equivalent toLet , and and define , , , , and ; then (D.5) becomes (17). Hence if the conditions (16) and (17) hold, nonlinear impulsive switched system (19) is stable and guarantees the performance, which completes the proof.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
This work is supported by the Funds of National Science of China (Grant no. 61403075) and Natural Science Foundation of Jilin Province (Grant no. 20140520060JH).
J. Chen and P. R. Patton, Robust Model-Based Fault Diagnosis for Dynamic Systems, Kluwer Academic Publishers, Boston, Mass, USA, 1999.
G. S. Deaecto, M. Souza, and J. C. Geromel, “Discrete-time switched linear systems state feedback design with application to networked control,” Institute of Electrical and Electronics Engineers. Transactions on Automatic Control, vol. 60, no. 3, pp. 877–881, 2015.View at: Publisher Site | Google Scholar | MathSciNet