This paper suggests an automated approach for fault detection, diagnosis and identification of roller bearings, which is based on optimized form of growing neural networks. In our recent work, we selected features according to their classification accuracy within supervised learning stage. Since each one of selected features has different effect on classification decision, a weighted feature selection is put forward in this paper to improve the network taxonomy. This is followed by a self adaptive growing neural network that optimizes its architecture by adding or updating hidden nodes to fulfill the training requirements. This pattern recognition procedure used to recognize between signals coming from normal bearings and those generated from different industrial bearing faults. The developed approach is compared with two different types of supervised neural networks. Results demonstrate that the developed diagnostic approach can reliably separate different bearing fault conditions at various rotational speeds.

 In this article a fault detection scheme for stator winding short circuit fault detection in the case of a three phase induction motor is being presented. The three phase motor is being modeled in the equivalent two phase motor ($q-d$) space, while the modeling of the faulty case is being also formulated. The motor is being identified by the utilization of Set Membership Identification (SMI) that has the merit of identifying both the parameters of the motor as also providing uncertainty safety bounds by calculating orthotopes which bounds the systems parameter vector. Based on the volume and the trend of these orthotopes, rules for identifying the existence of a fault are being presented. If the current values of the identified parameters do not lie inside the safety bounds in the healthy case, but lie in an area that is being defined by the model of the short circuit case, then a fault is being triggered. Detailed analysis of the proposed approach as also extended simulation results are being presented that prove the efficiency of the suggested scheme.

 The increased complexity of plants and the development of sophisticated control system have necessitated the parallel development of efficient online fault detection and isolation system. The detection and isolation of faults in industrial system has lately become of great significance. This paper proposes a new technique for online fault detection and diagnosis in dynamic system with multi inputs multi outputs. Numerous diagnosis schemes and architectures have been developed and applied to the benchmark DAMADICS. One of the key issues in designing a fault diagnosis system is the system modeling. Neural networks combined with other methods have been widely investigated for that purpose. The main contribution of this paper is to develop a new method for online fault detection and diagnosis schema with a bank of fault free and faulty reference models designed according to neural networks. Fault detection is obtained according to the comparison of measured signals with the behavior of fault free reference model. Then, calculation of Euclidean norms of the output error signals resulting from the faulty models leads to fault isolation. The effectiveness of this approach is illustrated with simulations on DAMADICS benchmark.

 This paper deals with the problem of the state estimation and the sensor faults detection for discrete time nonlinear systems described by Takagi-Sugeno (TS) fuzzy models with unmeasurable premise variables. Indeed, a TS observer is synthesized, in descriptor form, to estimate both the system states and the sensor faults simultaneously. The idea of the proposed approach is to introduce the sensor fault as an auxiliary variable in the state vector. Besides, the multiple model with unmeasurable premise variables is reduced to a perturbed model with measurable variables.

Convergence conditions are established with Lyapunov theory and the $pounds_2$ optimization in order to guarantee the convergence of the state estimation error. These conditions are expressed in terms of Linear Matrix Inequalities (LMIs). The gains matrices of the multi-observers are characterized using the solution existence of the LMI conditions. Finally, the model of an hydraulic system with three tanks is used to validate the proposed approach.

 A new framework, called multi-decision prognosis, has been recently developed, which consists in using several decentralized architectures working in parallel and whose prognoses are combined to obtain an effective prognosis. In this paper, we analyze and improve multi-decision prognosis in several ways. Firstly, we present a generic form of multi-decision prognosis which is independent on the architectures in parallel and on how their decisions a combined. Secondly, we solve in a rigorous way a problem of decomposing infinite languages which arises with multi-decision prognosis. Thirdly, we identify and solve a so-called hesitation problem that arises in multi-decision prognosis.

 The paper deals with the problem of estimating an unknown input distribution matrix for non-linear discretetime stochastic systems. In particular, it is shown how touse the unscented Kalman filter as an unknown input filter. Subsequently, an analysis of the impact of unknown input decoupling on the fault detection is performed and a suitable fault detection condition is developed. Based on the achieved results, a numerical optimisation-based approach is proposed that can be used to estimate the unknown input distribution matrix. The final part of the paper presents an illustrative example with an induction motor, which confirms the performance of the proposed approach.