Open access peer-reviewed chapter

Fuzzy Fault Detection Filter Design for One Class of Takagi-Sugeno Systems

Written By

Dušan Krokavec, Anna Filasová, Jakub Kajan and Tibor Kočík

Submitted: 10 November 2017 Reviewed: 23 January 2018 Published: 28 February 2018

DOI: 10.5772/intechopen.74328

Chapter metrics overview

1,153 Chapter Downloads

View Full Metrics


The constrained unitary formalism to fuzzy fault detection filter synthesis for one class of nonlinear systems, representable by continuous-time Takagi-Sugeno fuzzy models, is presented in the chapter. In particular, a way to produce the special set of matrix parameters of the fuzzy filter is proposed to obtain the desired H ∞ norm properties of the filter transfer function matrix. The significance of the treatment in relation to the systems under influence of actuator faults is analyzed in this context, and relations to corresponding setting of singular values of filters are discussed.


  • multiple models
  • continuous-time Takagi-Sugeno fuzzy models
  • fuzzy fault detection filters
  • fuzzy state observers

1. Introduction

Since the work of Hou and Patton [1], there has been much interest in the design of fault residuals for linear systems that use H/H optimization principle in transfer function matrix of fault detection filter designed to scale up fault detection punctuality and high sensitivity to faults [2]. While retaining these features, a novel class of fault detection filters are proposed in [3, 4], preserving the unitary implementation of the fault detection filter transfer function matrix and receipting residual signal directional properties. However, the use of this methodology for Takagi-Sugeno (TS) fuzzy systems hits the boundaries of the working sectors and requires special adaptations.

Considering the properties of TS fuzzy models [5, 6], and some specifics in frequency characteristic evaluation of multiple model structures, the approach proposed in the chapter reformulates the H norm technique suitable in TS fuzzy fault detection filter design. The problem is solved via unitary modal technique when every linear TS fuzzy filter part is designed to have the same singular values of the transfer function matrix. Since working sector constraints may cause that the stable linear filter component cannot be obtained for a linear part in TS fuzzy model, to maintain H norm of the filter, the LQ modal control principle [7] is used for additional stabilization. Because additional stabilization aggravates directional properties of the applied linear part, in general, if additional stabilization is necessary, the residuals are only quasi-directional. It is immediately apparent that the formulated problem is related to forcing the singular values conditioned as state observer dynamics. The chosen model of the system is selected for this chapter to be sufficiently complex in illustration of all these specifics of synthesis.

Throughout the chapter, the following notations are used: xT and XT denote the transpose of the vector x and the matrix X, respectively; for a square matrix X0 means that X is a symmetric positive semi-definite matrix; the symbol In indicates the n th-order unit matrix; IR denotes the set of real numbers; and IRn and IRn×r refer to the set of all n-dimensional real vectors and n×r real matrices.


2. System description

The considered class of the Takagi-Sugeno dynamic systems with additive faults is described as the following:


where qtIRn, utIRr, and ytIRm stand for state, control input, and measurable output, respectively; ftIRp is an additive fault vector; AiIRn×n, BiIRn×r, FiIRn×p, CIRm×n, and m=p and the matrix products Vi=CFi and ViIRm×m are regular matrices for all i.

The variables θjt and j=1,2,,o, bound with the sector TS model, span the o-dimensional vector of premise variables:


and [8]


where hiθt,i=1,2,,s is the set of normalized membership function. It is supposed that the measurable premise variables, the nonlinear sectors, and the normalized membership functions are chosen in such a way that the pairs AiBi are controllable and the pairs AiC are observable for all i.


3. Basic preliminaries from linear systems

Let the state-space description of a linear continuous-time dynamic systems take the form with equivalent meanings and dimensions as they are described in Section 2. The nature of the characterization of expected solutions to the system [(5), (6)] is given by the following results.


Definition 1 [9, 10] If A has no imaginary eigenvalues, the H norm of the system transfer function matrix




while the k th singular value σh of the complex matrix G is the nonnegative square root of the k th largest eigenvalue εk of GG, G is the adjoint of G, and σ1 is the largest singular value. The singular values of the transfer function matrix Gs are evaluated on the imaginary axis, and it is assumed that the singular values are ordered such that σkσk+1,k=1,2,,n1.

To apply in design methodology, the following result from [4] is quoted.

If m=p and V=CF are regular matrices, then the system matrix factorization can be realized such that


and the transform matrix TIRn×n takes the form


where V1CIRm×n, FIRnm×n, and F are the left orthogonal complements to F.

The idea of the following condition was derived originally as an approximation in the frequency domain for the fault transfer function matrix reflecting Eqs. (5) and (6) from [12]. Here, it is demonstrated that it can be simply adapted for fault residual filter design.

A linear fault detection filter to the system [(5), (6)] is stable and unitary if for regular V=CF and a given positive scalar soIR the square transfer function matrix Grs of the fault detection filter satisfies the conditions


where JIRn×r is the residual filter gain matrix, σ1 is the maximal singular value of Grs, the polynomial Pos of order nm is stable, and Gr0IRm×m.

Proof. Considering the fault transfer function matrix of dimension m×m as


and then regrouping terms using Eqs. (9) and (10), it yields immediately the expressions


respectively, where Ao is given in Eq. (15).

Specifying the following matrix product Ao=TMV1CT1, where MIRn×m is a real matrix, it yields


and, with the block matrix structure of Eqs. (15) and (21), it can be defined as




where soIR is a prescribed positive real value. The plus sign is introduced for the purposes that come to light in the stability ensuing development of the observer system matrix.



and it is evident that ΔAo is stable if Ao22 is Hurwitz, denoting here that


Rewriting the set of Eq. (22) to admit a stable solution




then Eqs. (20) and (21) must satisfy the following conditions:


Therefore, the observer system matrix Ae takes the form




implies Eq. (16).

Regarding the transfer function matrix Ges of the state error estimate as follows


then with Eq. (29), it is




Substituting Eq. (34) into Eq. (32), it can obtain


Thus, defining the fault detection filter transfer function matrix as Grs=V1Ges, then


and Eq. (36) implies Eq. (14). This concludes the proof.

Evidently, writing the fault residual vector as




and rtIRm is the vector of residual signals, then based on the following observer structure


the autonomous observer error equation is


where qetIRn is the observer state, yetIRm is the estimated system output, and JIRn×m is the observer gain matrix; the fault detection filter (37), (39) is stable and unitary if for given positive scalar soIR and the Hurwitz matrix Ao22 the conditions (15) and (16) are satisfied.

Practically, with understanding Eq. (30), the observer sensor subsystem for the fault detection filter can be designed as follows:


and, consequently, it yields


Another option is to design the observer sensor subsystem so that V=Im.

With existence of the system parameter transformation, the above structures really mean that the subset of transformed state variables whose dynamics is explicitly affected by the additive fault ft and the second one, whose dynamics is not affected explicitly by the fault ft, exists.

It is important to note the fact that the eigenvalues of A and of Ao are the same whenever Ao is related to A as Ao=TAT1 for any invertible T [11]. But this does not mean that if eigenvalues of the matrix Ao are stable then eigenvalues of the matrix Ao22 are also stable. Thus, as well as for a stable system, it can lead to an unstable matrix Ao22, and any additional stabilization is required.

To apply the above results, it is necessary to be able to design fault residual filter if an unstable Ao22 results such that Ae be stable without loss of unitarity.

Lemma 2 [7, 12] To change signs of unstable eigenvalues of the system matrix A, the gain matrix KIRn×r of the state feedback additive stabilization


is a solution of the continuous-time algebraic Riccati equation (CARE)


where the matrix QIRn×n is null matrix and RIRr×r and R=RT>0 are positive definite symmetric matrices.

Then, K is given as


It is in that form that is able to be exploit for specific properties of the problem in TS fuzzy fault detection filter design.

In view of the above, these results hold for continuous-time linear systems, and, in principle, Theorem 1 gives a practical method to design unitary fault residual filters for the given linear system. Similar results are obtained for unitary TS fuzzy fault detection filter design in the following section.


4. TS fuzzy fault detection filters

Using the same set of membership functions, the fuzzy fault detection filter is built on the TS fuzzy observer


where qetIRn is the observer state vector, yetIRm is the estimated system output vector, and JiIRn×m and i=1,2,,s are the sets of the observer gain matrices. Additionally, the output vector of the residual TS fuzzy filter is defined as


where rt,ritIRm, ViIRm×m. Evidently, Vi=CFi has to be a regular matrix for all i.

Formally, the following result can be simply derived.

A TS fuzzy fault detection filter to the system [(1), (2)] is stable and unitary if for the set of regular matrices Vi=CFi and i=1,2,,s, and a given positive scalar soIR every square transfer function matrix Gris of the fault detection filter satisfies for all i the conditions




JiIRn×r is the residual filter gain matrix, σ1 is the maximal singular value of Gris, the polynomial Pois of order nm is stable, and Gri0IRm×m and FiIRnm×n are left orthogonal complements to the fault input matrix Fi.

Proof. Because every sub-model in Eq. (47) is described by linear equations, Eqs. (15) and (16) imply directly the conditions (56) and (57), and Eq. (58) is given by Eq. (11). This concludes the proof.

In practice, an additive fault typically enters through a matrix F that does not depend on the sectoral boundaries defining the TS model. In this case, the synthesis is substantially simplified because V is a constant matrix, and so it yields


Since, independently on i, the condition (52) is satisfied (σ1=σ2==σm), all sub-filter transfer function matrices have the same H norm, i.e.,


Moreover, considering that i=1shiθt=1, then


That is, the H norm of the transfer function matrix of such defined TS fuzzy fault detection filter is independent on the system working point. Of course, this cannot be said about the dynamics of the time response of the sub-filter components.

Moreover, Gri0 implies that all residual components of TS fuzzy fault detection filter have the same directional properties, which ensure unitary properties of the filter.

Sectoral boundaries may cause a matrix Ai to be such, when transformed using Ti that Ao22i will not be Hurwitz matrix. Because the transfer function matrix of the corresponding filter linear component in this case is unstable, maintaining the unitary property requires changes in the signs of the unstable eigenvalues of the associated Aei°=AiJiC.

Applying the duality principle and inserting the additive observer gain component KsiT obtained as a solution of the Riccati equation (45) for Aei°T, according to the scheme given in Lemma 2, the observer gain matrix is changed as


This additive stabilization results that the consequential characteristic polynomial, taking also the form


is stable since Pois is now stable.

The price for such an additional stabilization is that if j signs are changing in eigenvalues of Ao22i to obtain the stable Ao22i, also j eigenvalues so of Gri0 change their signs and the resulting matrix Gri0 will not be diagonal. According to Eq. (8), this does not result in a change in H norm, but such filter component will arrive at the unitary directional residual properties.


5. Illustrative example

The three-tank system is described by the set of Eqs. [13, 14] as


where the measured output variables ykt are water levels in tanks qktm,k=1,2,3 and the incoming flows are considered as the inputs variables uktm3/s,k=1,2,3; the bounds of the state and input variables are


λk,ηkIR are positive scalars and sign is the sign function.

The model parameters of the system are considered as:

gFkα1α3α2the gravitational acceleration9.80665m/s2,thesamesection of tanks0.25m2,the equivalent section of the pipe between the first and second tank6.5×104m2,the equivalent section of the pipe between the third and second tank6.5×104m2,the equivalent section of the outlet pipe from the second tank6.5×103m2,

Minimizing the number of premise variables and excluding switching modes in controller work, the premise variables are chosen as follows


Computed from the input variable bounds, the sector bounds of the premise variables imply the numbering:


which is used in the system state matrix construction


and prescribed, moreover, that the matrix C is given in such a way that the product CB is the identity matrix. This regularizes the residual design conditions if B and C are diagonal matrices.

The sector functions are trapezoidal, and the membership functions are constructed as product of three sector functions with the same ordering as Ai.

The set of real scalars, λk, ηk, and k=1,2,3, is interactively optimized under limitations that all couples AiB and AiC are controllable and observable for the given set of indices i, where


Consequently, the TS model matrix parameters are


Since the orthogonal complement to a square matrix does not exist, three fault detection filters can be considered for single actuator fault detection. To illustrate the design procedure, the TS fuzzy fault detection filter for the pair (C23, B23) is considered, i.e.,


with the derived parameters


Note that in this case all Ai with index higher than 4 lead to an unstable structure of Ao22i° and the resulting observer matrices Aei need to be additionally stabilized, applying the principle given in Lemma 2.

Applying Eq. (56), the following structure of Ao1 for the initial matrix A1 is computed:


and Ao221=0.0163 implies that the associated TS fuzzy fault detection filter linear component can be designed directly.

Choosing so=5, it is resulting from Eqs. (57) and (58) that


where the eigenvalue spectrum of Ae1 and the steady-state value of the TS fuzzy fault detection filter transfer function matrix Gr10 are


respectively. It is evident that all diagonal elements of Gr10 take the value so1=0.2. The same structure of Gr0 is obtained solving with Al for l=1,2,3,4.

Analogously, designing for the matrix A5, it can be seen that


Since Ao222=0.0034, evidently, the associated TS fuzzy fault detection filter linear component with the unitary transfer function matrix has to be stabilized additively.

Solving also for so=5, then


It is evident that matrix Fe5 is not Hurwitz and has to be additively stabilized.

Thus, defining the weighting matrices of appropriate dimensions as


and solving the dual LQ control problem to change the sign of unstable eigenvalue of Fe5 using the MATLAB function Ks5=careFe2TQRSI3, then


It can be easily verified that


while, evidently, Gr50 is not diagonal and the eigenvalues of Gr50 are ±0.2=±s01.

Note that the same structure of Grl0 is obtained solving with the system matrices Al and l=5,6,7,8 when additional stabilization is required. Evidently, elements of this set of TS fuzzy residual filter linear components are stable, non-unitary, and without directional residual properties. Nevertheless, these properties guarantee the same singular values of the linear transfer function matrix components; as follows the result of Definition 1, the TS fuzzy residual filter will have all the singular values the same. To document this, the singular value plot of the TS fuzzy fault detection filter, as well as of all its linear parts, is equal to that presented in Figure 1. With respect to the structure of the matrices B and C, the comparable results are obtainable for the matrix pairs (C12, B12) and (C13, B13).

Figure 1.

TS fuzzy fault detection filter singular value plot.

The rest of gain matrices of the stable TS fault detection filter is as follows:



Since the matrices Ai of the TS fuzzy system are not Hurwitz, the system in simulations is stabilized using the local-state feedback control laws, acting in the forced modes. Adapting the method presented in [14] to design the control law parameters, the local controller parameters are computed as




while woIRn is the vector of the desired steady-state system outputs.

If necessary for any more complex system, PDS controller principle can be applied to stabilize the plant (see, e.g., authors’ publications [15, 16] or other references [17, 18]).

To display simulations in the MATLAB and Simulink environment, the forced mode control is established with local controller parameter given as above for the system initial conditions qT0= and woT= Fault detection filter is constructed on the couple (C23, B23) and the set of matrices Ai and i=1,2,8.

As the results, Figure 2 shows the TS fuzzy system output responses, illustrating their asymptotic convergence to the steady states, and Figure 3 presents the TS fuzzy fault detection filter response, reflecting a steplike 90% gain loss of the second actuator at the time instant t=60s. These examples illustrate the power that can be invoked through the prescribed H norm properties.

Figure 2.

System output responses.

Figure 3.

Residual signal responses.

It can verify that TS fuzzy fault detection filters created for the couple pairs (C12, B12) and (C13, B13) have similar properties as that defined for the couple (C23, B23). The difference is, for example, that in the occurrence of a single fault of the second actuator the responses of TS fuzzy fault detection filter defined for the couple (C13, B13) naturally do not have directional properties, since the second column of K is not included in its construction.

As can be seen from the solution, the sector functions defined in this way cannot create a unitary TS fuzzy fault detection filter, but the obtained orthogonal properties of the residual signals are sufficient to detect and isolate actuator faults.


6. Concluding remarks

The problem of designing the TS fuzzy fault detection filters for highly nonlinear mechanical systems representable by the TS fuzzy model is considered, to achieve the desired filter H norm property in all working point belonging to the assigned work sectors. The proposed method exploits features offered in TS fuzzy system models to design TS fuzzy fault detection filters. The rules and formulation are developed to generate residual signals with quasi-directional properties and to make the TS filter transfer function matrix with prescribed H norm properties. By a convenient choose of the sector functions, this purpose is reached using a relative small number of membership functions. If unitary definition for TS fuzzy fault detection filters is satisfied, the design methodology provides new opportunities for fault detection and isolation rules in fault tolerant nonlinear control systems, their analysis, and optimization.



The work presented in this chapter was supported by VEGA, the Grant Agency of the Ministry of Education, and the Academy of Sciences of Slovak Republic, under Grant No. 1/0608/17. This support is very gratefully acknowledged.


  1. 1. Hou M, Patton RJ. An LMI approach to H/H fault detection observers. In: Proceedings of the UKACC International Conference on Control; 2–5 September 1996; Exeter, England, pp. 305-310. DOI: 10.1049/cp:19960570
  2. 2. Noura H, Theilliol D, Ponsart JC, Chamseddine A. Fault-Tolerant Control Systems. Design and Practical Applications. London: Springer-Verlag; 2009. 233 p
  3. 3. Zhao Z, Xie WF, Hong H, Zhang Y. Unitary system I. Constructing a unitary fault detection observer. IFAC Proceedings. 2011;44(1):7725-7730.
  4. 4. Krokavec D, Filasová A, Liščinský P. On fault detection filters design with unitary transfer function matrices. Journal of Physics. Conference Series. 2015;659:1-12. DOI: 10.1088/1742-6596/659/1/012036
  5. 5. Takagi T, Sugeno M. Fuzzy identification of systems and its applications to modeling and control. IEEE Transactions on Systems, Man and Cybernetics. 1985;15(1):116-132. DOI: 10.1109/TSMC.1985.6313399
  6. 6. Chadli M, Borne P. Multiple Models Approach in Automation. Takagi-Sugeno Fuzzy Systems. Hoboken: John Wiley & Sons; 2013. p. 256
  7. 7. Solheim OA. Design of optimal control systems with prescribed eigenvalues. International Journal of Control. 1972;15(1):143-160. 208932136
  8. 8. Tanaka T, Wang HO. Fuzzy Control Systems Design and Analysis. A Linear Matrix Inequality Approach. New York: John Wiley & Sons; 2001. 320 p
  9. 9. Boyd D, Balakrishnan V. A regularity result for the singular values of a transfer matrix and a quadratically convergent algorithm for computing its L-norm. Systems & Control Letters. 1990;15(1):1-7. DOI: 10.1109/CDC.1989.70267
  10. 10. Green M, Limebeer DJN. Linear Robust Control. Englewood Cliffs: Prentice Hall; 1995. 558 p
  11. 11. Fairman FW. Linear Control Theory. The State Space Approach. Chichester: John Wiley & Sons; 1998. 318 p
  12. 12. Krokavec D, Filasová A, Liščinský P. Unitary approximations in fault detection filter design. Mathematical Problems in Engineering. 2016;2016:1-14.
  13. 13. Nagy AM, Marx B, Mourot G, Schutz G, Ragot J. State estimation of the three-tank system using a multiple model. In: Proceedings of the Joint 48th IEEE Conference on Decision and Control and 28th Chinese Control Conference; 15–18 December 2009; Shanghai, China, pp. 7795‐7800. DOI: 10.1109/CDC.2009.5400889
  14. 14. Krokavec D, Filasová A, Serbák V. Virtual actuator based fault tolerant control design for Takagi-Sugeno fuzzy systems. In: Proceedings of the 14th IEEE International Symposium on Applied Machine Intelligence and Informatics SAMI 2016; 21–23 January 2016; Herl’any, Slovakia, pp. 63-68. DOI: 10.1109/SAMI.2016.7422983
  15. 15. Krokavec D, Filasová A. Stabilizing fuzzy output control for a class of nonlinear systems. Advances in Fuzzy Systems. 2013;2013:1-9.
  16. 16. Krokavec D, Filasová A. Stabilizing fuzzy control via output feedback. In: Ramakrishnan S, editor. Modern Fuzzy Control Systems and Its Applications. Rijeka: InTech; 2017. pp. 3-25. DOI: 10.5772/68129
  17. 17. Zhang K, Jiang B, Shi P. Observer-Based Fault Estimation and Accommodation for Dynamic Systems. Berlin: Springer-Verlag; 2013. 181 p
  18. 18. Lam HK. Polynomial Fuzzy Model-Based Control Systems. Stability Analysis and Control Synthesis Using Membership Function-Dependent Techniques. Cham: Springer-Verlag; 2016. 307 p

Written By

Dušan Krokavec, Anna Filasová, Jakub Kajan and Tibor Kočík

Submitted: 10 November 2017 Reviewed: 23 January 2018 Published: 28 February 2018