Komi Midzodzi Pekpe

Identification of switching systems using change detection technique in the subspace framework

The paper describes an identification technique of switching system. The considered system is represented as a weighted sum of local models. To estimate the switching times, a change detection technique is applied. It provides the weights associated to the local models. The Markov parameters of these models are identified by a subspace method. This calculation can yield similar local models which are merged. The procedure of parameter identification an models merging is repeated until convergence. The performance of the approach is investigated on a simulation example.

Subspace Method for Sensor Fault Detection and Isolation-Application to Grinding Circuit Monitoring

Abstract Sensor fault detection and isolation method is proposed in this paper. The method is only based on the knowledge of the input and output data. Any parameter estimation nor system order determination are necessary. The proposed technique uses matrix projection in subspace framework. The sensitivity of the method to sensor faults is shown. The method is applied for sensor fault detection and isolation in a grinding circuit.

Identification of MIMO Takagi-Sugeno model of a bioreactor

The main contribution of this paper is to provide an identification method to a single MIMO (multiple-input multiple-output) Takagi-Sugeno (TS) model for nonlinear MIMO systems. The methods proposed in the literature identify several TS MISO models and this needs the outputs to be separable. But this condition is not necessary in the proposed method. The TS identification method proposed uses the general principle of TS identification method which consist to decomposed a nonlinear system into a set of less complex model. Therefore, the nonlinear systems are decomposed into a set of linear systems and the contribution of each local model is expressed by a weighting function.

Sensor fault diagnosis for bilinear systems using data-based residuals

The proposed data-based FDI method has the advantage to require only available data: control signals and measured outputs. The unique information that we have about the system is its structure but the parameters values are supposed to be unknown. This data-based residual method was described, by the same authors in previous publication, for linear systems. In this paper we apply this method to bilinear structure models. A particular focus is made on computational complexity reduction. It will be shown that a part of the system dynamics may be neglected with the consequence to simplify the on-line residual computation. This method is illustrated on a simulated Activated Sludge Process.

Subspace identification of switching model

Abstract Subspace identification of switching model is considered in this paper. Here the switching model is supposed to be a sum of weighted linear models. The method established uses recursive subspace identification to estimate the switching function and least squares method for local model Markov parameters estimation. To perform the computation of the weighting functions a two-steps algorithm (switching times determination and model merging) is given. Finally the local model parameter estimation is based on the estimation of the Markov parameters.

Vibration-based fault detection of sharp bearing faults in helicopters

Many signal processing tools have been developed by the mechanical and signal processing community to find the characteristic symptoms of sharp bearing faults (like localized spalling) from vibratory analysis. However the context of helicopter imposes a limited sampling frequency regarding the observed phenomena, many noisy vibrations and flight regimes. The performances of the classical methods are limited in such an environment mainly in identifying fault frequencies. Local bearing faults induce temporal periodic and impulsive patterns that produce redundant harmonics in the spectral domain. In this article four methods are proposed to take advantage of that redundancy. These methods provide an estimator of the fault frequency and an indicator of the quality of that estimation. These indicators are used to assess the severity of the fault. The four methods are then tested on synthesized and flight data in order to illustrate and discuss their efficiency.

Parameter-free method for switching time estimation and current mode recognition

This paper is concerned with switching systems which belong to a special class of Hybrid Dynamical Systems. The objective is to estimate switching times and to recognize the current mode. The general principle of the method is to use indicator signals, named residuals, which are generated by projecting the sensor measures into a subspace related to the input signals on a given time window. The advantage of this residual generation method is that the values of the model parameters are not needed, only the kind of model (linear, bilinear, …) is used. The set of residuals may be used, under discernability conditions, to detect mode switching. An extension is proposed to identify the new mode by using a database constituted by sets of collected data that characterize each mode. An illustrative example is taken in order to show the effectiveness of our method.

Vibration-Based Fault Detection of Accelerometers in Helicopters

Vibration-based monitoring is an approach for health analysis of helicopters. However, accelerometers and other sub-elements that convert and transmit vibrations to the recording system must not corrupt the signal. These elements are prone to defects because of external injuries during flights or maintenance. This paper will deal with a method to tackle problems of loosening and mechanical shocks. The objective is to perform a passive detection of accelerometer failures from the vibrations without knowledge of previous recordings. Experiments of mechanical failures have been carried out on a shaker to reproduce in flight vibrations, and it appears that the loosening and mechanical shocks introduce asymmetry and random peaks in the temporal vibrations. Loosening was successfully detected but mechanical shocks were much harder to detect as a result of strong dependences in the vibratory environment. Loosening data sets from flights confirm experimental observations and the proposed detection method allows for the detection of the fault with better performance than standard indicators.

A parameter-free method for sensor fault detection and isolation in bilinear systems

This paper is concerned with Fault Detection and Isolation (FDI) and more specifically it focuses on a parameter-free residual generation method. The residual signals are obtained by projecting the measured signals onto the kernel of an extended input matrix, which depends on the structure of the system model. The method was not easily applicable in real-world applications due to a high computational complexity. In that paper, fault indicators are constructed differently, using kernels properties, to avoid this complexity problem. A simulated electromechanical actuator example is taken to illustrate the applicability of the method.

ONLINE CLASSIFICATION OF SWITCHING MODELS BASED ON SUBSPACE FRAMEWORK

The paper deals with the modelling of switching systems and focuses on the characterization of the local functioning modes using online clustering approach. The considered system is represented as a weighted sum of local linear models where each model could have its own structure. That implies that the parameters and the order of the switching system could change when the system switches. The presented method consists in two steps. First, an online estimation method of the Markov parameters matrix of the local linear models is established. Secondly, the labelling of theses parameters is done using a dynamical decision space worked out with learning techniques, each local model being represented by a cluster. The paper ends with an example, in view to illustrate the method performances.