Jianfei Dong

Data Driven Fault Detection and Isolation of a Wind Turbine Benchmark

Abstract This paper investigates data-driven fault detection and isolation (FDI) designs for a wind turbine benchmark problem. The benchmark is described by a SimuLink model, which contains nonlinear lookup tables and unknown wind disturbances. Based on classical FDI design methods, a linearization of the SimuLink model into a standard state-space form and describing the linearization errors as perturbations may be necessary. To avoid these difficult modeling procedures, this paper applies a data-driven design method, which produces FDI filters directly based on the simulated data from the benchmark SimuLink model. The fixed-value sensor faults therein are especially targeted. Moreover, we develop in this paper a new data-driven fault isolation scheme, via exploiting hardware redundancy in a plant. Based on this, a bank of robust data-driven detection filters are designed for the benchmark and implemented in parallel. The simulation results show the effectiveness of the applied data-driven scheme.

Closed-Loop Subspace Predictive Control for Fault Tolerant MPC Design

Abstract Subspace predictive control (SPC) is recently seen in the literature for joint system identification and control design. This combination enables automatically tuning the parameters in conventional model predictive control (MPC); and therefore provides a solution to the problem of fault tolerant MPC design. The existing SPCs either deal with open-loop data or depend on the information of the controller in a closed loop. In this paper we introduce a new closed-loop SPC method, which is independent of any controller information. Both the analytic solution to the unconstrained case and the quadratic programming problem for the constrained case are formulated. A recursive solution for updating the SPC control law is proposed. A fault tolerant MPC scheme is then developed based on the recursive algorithm, whose effectiveness is demonstrated on tolerating a fault in a steer-by-wire actuator.

Robust Fault Detection With Statistical Uncertainty in Identified Parameters

Detection of faults that appear as additive unknown input signals to an unknown LTI discrete-time MIMO system is considered. State of the art methods consist of the following steps. First, either the state space model or certain projection matrices are identified from data. Then, a residual generator is formed based on these identified matrices, and this residual generator is used for online fault detection. Existing techniques do not allow for compensating for the identification uncertainty in the fault detection. This contribution explores a recent data-driven approach to fault detection. We show first that the identified parametric matrices in this method depend linearly on the noise contained in the identification data, and then that the on-line computed residual also depends linearly on the noise. This allows an analytic design of a robust fault detection scheme, that takes both the noise in the online measurements as well as the identification uncertainty into account. We illustrate the benefits of the new method on a model of aircraft dynamics extensively studied in literature.

Identification of Fault Estimation Filter From I/O Data for Systems With Stable Inversion

Classical methods for estimating additive faults are based on state-space models, e.g., moving horizon estimation (MHE) and unknown input observers (UIOs). This paper contributes new direct design methods from closed-loop I/O data for systems with stable inversion, which do not require building a state-space model by first principles, nor require identifying it. Inspired by subspace identification, we use the input and output (I/O) relationship of a plant in a Vector ARX (VARX) form to parameterize least-squares (LS) problems for estimating faults. We prove that with the order of the VARX descriptions tending to infinity, the fault estimates are unbiased. Under lower relative degrees, we prove that our new methods are equivalent to system-inversion-based estimation for both LTI and LTV systems. We will show more general unbiased estimation conditions for higher relative degrees. These require that the underlying inverted system from faults to outputs is stable. Algorithms of identifying unbiased fault estimation filters from data will be developed in this paper based on single LS. Moreover, covariance of the fault estimates can also be extracted from data.

Subspace based Fault Detection and Identification for LPV Systems

Abstract This paper presents a new Fault detection and Identification approach Connected to Subspace Identification (FICSI) for a special class of nonlinear system casted to Linear but Parametrically Varying (LPV) form. The model based algorithm utilizes the specific nonlinear nature of the LPV system along the past horizon to construct the output predictor. Similar with its LTI counterpart in a companion paper ([Dong and Verhaegen (2008)]), the FICSI-LPV avoids projecting the residual vector onto the parity space of the extended observability matrix, and hence produces residuals more sensitive to faults than the parity space approach (PSA) for LPV systems does. Asymptotically unbiased condition and algorithm are also proposed for fault estimation. The difference of FICSI from the existing fault detection and estimation approaches based on PSA or moving horizon estimation (MHE) can also be attributed to the fact that FICSI does not require an LPV state space model, but a sequence of Markov parameters mapping the I/O measurements and the scheduling parameters to the residual, which can be estimated in closed-loop.

Model-free Fault Tolerant Control Approach for Linear Parameter Varying System

Abstract The paper investigates the reconfigurable features of the Subspace Predictive Control scheme for Linear Parameter Varying Systems (SPC LPV). This model independent predictive approach combines the predictor based subspace identification and (constrained) predictive control for affine LPV systems. By means of using recursive update of the identified closed loop Markov parameters, the SPC LPV can be used for Fault Tolerant Control. The proposed algorithm is implemented and applied in a rotation speed tracking for a nonlinear DC motor.

On the equivalence of closed-loop subspace predictive control with LQG

Subspace predictive control (SPC) is recently seen in the literature for joint system identification and control design. The existing SPCs parameterize H2 optimal control laws by the identified Markov parameters from data. It has been proved that the SPCs based on open-loop subspace identification are equivalent to the classical LQG design, when the data horizon goes to infinity. It is the purpose of this paper to establish this equivalence for the closed-loop SPC algorithm we have recently developed. When the data horizon is finite, we also present in this paper a state-feedback LQG control law based on the identified Markov parameters, where the states are estimated in an optimal sense from the past I/O samples of a plant.

Data driven fault detection with robustness to uncertain parameters identified in closed loop

This paper presents a new robustified data-driven fault detection approach, connected to closed-loop subspace identification. Although data-driven detection methods have recently been reported in the literature, attention has not yet been given to a robust solution coping with identification errors. The key idea of this paper is to analytically quantify the effect of the identification errors on the residual generator of a new data-driven detection approach, namely FICSI. The comparisons of the proposed robust FICSI detection scheme with both its nominal counterpart and the nominal data-driven PSA solutions have verified the effectiveness of accounting the identification errors in improving the performance of the data-driven detection scheme.

Cautious H2 optimal control using uncertain Markov parameters identified in closed loop

Abstract In this paper, we present a new cautious H 2 optimal design approach based on the noisy and biased Markov parameters identified from a finite number of input and output samples in a closed-loop plant. This approach not only links closed-loop subspace identification with optimal control; but also directly evaluates parametric uncertainties on the identified Markov parameters. Neither a state-space model nor its stochastic uncertainty has to be realized. The effects of the parametric uncertainties on the output predictor and a quadratic cost function are explicitly analyzed. An H 2 optimal control problem is formulated as a “ min max ” problem of the expectation of the cost function with respect to the stochastic noise in the identified parameters. Analytic solution to this problem is derived in a closed form, which avoids computing the empirical mean of the quadratic cost as required by randomized algorithms. An extension of the cautious design to the solution associated with an arbitrary probability is also proposed. The solutions hence lead to easily implementable control laws, robust to the uncertainties in the identified Markov parameters from a closed-loop plant.

Model free probabilistic design with uncertain Markov parameters identified in closed loop

In this paper, we present a new probabilistic design approach based on Markov parameters identified via subspace methods from a finite batch of input and output data. This approach not only links closed-loop subspace identification with optimal control; but also directly evaluates parametric uncertainties on the identified Markov parameters. Neither a state-space model nor its stochastic uncertainty has to be realized in this approach. The effects of the parametric uncertainties on the output predictor are analyzed explicitly. Analytic solution to the probabilistic design is derived in a closed form, which avoids computing the empirical mean of a cost function as required by randomized algorithms. The solution hence leads to an easily implementable cautious optimal design, robust to the uncertainties in the identified Markov parameters from a closed-loop plant.