Development and application of sliding mode LPV fault reconstruction schemes for the ADDSAFE Benchmark

Alwi, Halim; Edwards, Christopher

doi:10.1016/j.conengprac.2014.05.003

Cited by 34 publications

(46 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, the application of linear parameter varying (LPV) concepts to system modelling, control and FDD have also received much attention (Balas, 2002;Bokor & Balas, 2004;Henry, 2008;Sato, 2010;Wei & Verhaegen, 2011a;Hecker & Pfifer, 2014;Alwi & Edwards, 2014;Henry et al, 2014;Varga & Ossmann, 2014;Vanek et al, 2014;Chen et al, 2016;Rodonto et al, 2015;Ossmann & Varga, 2015;Rotondo et al, 2015;Alwi et al, 2015). Nevertheless, the technical demands of model-based FDD, especially for the FDD problem based on using LPV, are still quite limited and restrictive in the aerospace industry (Zolghadri, 2012).…”

Section: Introductionmentioning

confidence: 99%

“…The ADDSAFE project benchmark was provided to several academic and industrial partners involved in this project to evaluate the efficiency of their FDD approaches on various fault scenarios (Alwi & Edwards, 2014;Henry et al, 2014;Varga & Ossmann, 2014;Vanek et al, 2014;Van Eykeren & Chu, 2014). The benchmark model is highly representative of a generic twin engine civil commercial aircraft including the nonlinear rigid-body aircraft model with a full set of control surfaces, actuator models, sensor models, flight control laws and pilot inputs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Application of model-based LPV actuator fault estimation for an industrial benchmark

Chen

Patton

Goupil

2016

Control Engineering Practice

View full text Add to dashboard Cite

To bridge the gap between model-based fault diagnosis theory and industrial practice, a linear parameter varying H − /H ∞ fault estimation approach is applied to a high fidelity nonlinear aircraft benchmark. The aim is to show how the fault estimation can provide robust early warning of actuator fault detection scenarios that can lead to abnormal aircraft flight configurations. The fault estimator state space solution is parameterized a priori using parameter-independent design freedom. Following this only constant free matrices are determined and the resulting affine linear parameter varying estimator has low computational load. The evaluation uses parametric simulation via an industry standard Monte Carlo campaign supported by a functional engineering simulator. The simulations are carried out in the presence of aerodynamic database uncertainties and measurement errors covering a wide range of the flight envelope.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Application of model-based LPV actuator fault estimation for an industrial benchmark

Chen

Patton

Goupil

2016

Control Engineering Practice

View full text Add to dashboard Cite

show abstract

“…Furthermore, in (17), it can be shownÃ 11 (ρ) = A 11 (ρ) + LA 21 (ρ). In the new coordinatesC = CT −1 L = [0 I], and consequently the output error can be expressed as…”

Section: Sliding Mode Observer For Uncertain Linear Parameter Vamentioning

confidence: 99%

“…Finally create a new state vector by augmenting d(t) from (3), andẽ 1 (t) and e y (t) from (17) and (18), such that  ḋ (t) e 1 (t) e y (t)…”

Section: Assumptionmentioning

confidence: 99%

“…Another advantage is many nonlinear systems can be naturally approximated by LPV systems [13], [14]. Recent results on LPV based fault detection schemes appears in [15], [16], [17] from work in the EU-funded ADDSAFE project. Recently several sliding mode observer structures have been proposed for LPV systems (see for example [18], [19], [20]).…”

Section: Introduction Fault Detection and Isolation (Fdi)mentioning

confidence: 99%

See 1 more Smart Citation

Fault detection in uncertain LPV systems with imperfect scheduling parameter using sliding mode observers

Chandra¹,

Alwi

Edwards

2017

European Journal of Control

Self Cite

View full text Add to dashboard Cite

This paper presents a sliding mode fault detection scheme for linear parameter varying (LPV) systems with uncertain or imperfectly measured scheduling parameters. In the majority of LPV systems, it is assumed that the scheduling parameters are exactly known. In reality due to noise or possibly faulty sensors, it is sometimes impossible to have accurate knowledge of the scheduling parameters and a design based on the assumption of perfect knowledge of the scheduling parameters cannot be guaranteed to work well in this situation. This paper proposes a sliding mode observer scheme to reconstruct actuator and sensor faults in a situation where the scheduling parameters are imperfectly known. The efficacy of the approach is demonstrated on simulation data taken from the nonlinear RECONFIGURE benchmark model . One approach to extending linear methods to make them work effectively across the whole plant operating range is to design (linear) observers at select operating points, and then schedule the gains with respect to certain parameters [4]. One of the main difficulties with gain scheduling is the selection of the operating points and how to choose the gains at intermediate points. Furthermore, formally, between the operating points the stability of the observer cannot be guaranteed.A more rigorous alternative approach is based on so-called linear parameter varying (LPV) system designs. In the LPV approach the gains are automatically scheduled with respect to the plant varying parameters. Another advantage is many nonlinear systems can be naturally approximated by LPV systems [13] [20]). However in all these papers the scheduling parameters are assumed to be perfectly known. However in reality this may not be the case: for example inaccurate sensors and/or faults can lead to imperfect knowledge of the parameters. The use of this corrupted information will affect the performance of the observer. The design of Luenberger-like observers in scenarios involving uncertain scheduling parameters has been investigated. In [21], using a Linear matrix inequality (LMI) framework an LPV observer was proposed. Later, a controller and a combined controllerobserver scheme for inexact continuous LPV polytopic systems was presented in [22], [23]. In [24], [25] H ∞ filters were employed to deal with uncertainty in the scheduling parameters; exploiting parameter dependent LMI methods. More recently, an H ∞ filter has been proposed to deal with both additive and multiplicative uncertainties in the LPV parameters in [26]. Recent applications of these ideas to aerospace systems have been explored in [27], [28]. A fault reconstruction scheme for LPV systems with perfect scheduling parameter knowledge, using H ∞ methods has been investigated in [29]. Very few papers have addressed specifically the fault reconstruction problem for LPV systems with uncertain scheduling parameters, notable exceptions are [30]. To the authors' knowledge,

show abstract

Simultaneous time‐varying actuator and sensor fault reconstruction based on PI observer for LPV systems

Zhu

2014

Adaptive Control & Signal

View full text Add to dashboard Cite

We consider problems of actuator and sensor fault reconstruction simultaneously for linear parameter varying systems expressed in polytopic forms. By extending the sensor fault as an auxiliary state, a polytopic unknown input proportional-integral observer in which the actuator fault signals are assumed to be time varying is developed to estimate the system states and the actuator and sensor fault at the same time. The existence conditions of the observer are derived in terms of linear matrix inequalities that can be readily handled via some efficient tools. An example is given to demonstrate the advantages of the proposed method in comparison to the existing results.R s ; f a 2 R q , and f s 2 R w are external disturbance, actuator fault, and sensor fault, respectively. A.Â.t // 2 R n n ; B.Â.t // 2 R n m , F .Â.t// 2 R n q , and D.Â.t// 2 R n s are continuous matrix functions that depend on time-varying parameter Â.t/ 2 R v , which can be measured online. Matrix M 2 R p w has full column rank, and matrix C 2 R p n has full row rank. Remark 1Both actuator faults and unknown inputs are unknown factors to the systems, and they usually have similar effects on the system behavior, but they are usually dealt with in a very different way in FDI designs. Specifically, for actuator faults that come from actuators, we usually wish that we can detect or reconstruct them when they occur, while the unknown inputs are the external disturbances or modeling uncertainties of the systems, which we only need to suppress them. In this paper, we reconstruct the actuator faults by designing an adaption law and attenuate the influences of external disturbances d by using H 1 technique, which can suppress them to a desirable level.Assumption 1 Â.t/ 2 R v is bounded and lies into a hypercube ‚:D 1; : : : ; vº and Â.t/ 2 ‚ D C o¹Â 1 .t/; : : : ; Â r .t /º where Â i and N Â i represent the maximum and minimum values of Â i .t /; r D 2 v , and Â 1 .t/; : : : ; Â r .t / are the vertices of the hypercube. Assumption 2The elements of the matrices in (1) are affine functions of time-varying parameter vector Â.t/. Definition 1Define $ D ¹$.t/º 2 L 2 OE 0 1 /, and its L 2 norm is Remark 2

show abstract

Development and application of sliding mode LPV fault reconstruction schemes for the ADDSAFE Benchmark

Cited by 34 publications

References 36 publications

Application of model-based LPV actuator fault estimation for an industrial benchmark

Application of model-based LPV actuator fault estimation for an industrial benchmark

Fault detection in uncertain LPV systems with imperfect scheduling parameter using sliding mode observers

Simultaneous time‐varying actuator and sensor fault reconstruction based on PI observer for LPV systems

Contact Info

Product

Resources

About