Search

Nowadays, the data-driven diagnosis method, exploiting pattern recognition method to diagnose the fault patterns automatically, achieves much success for rotating machinery. Some popular classification algorithms such as artificial neural networks and support vector machine have been extensively studied and tested with many application cases, while the random forest, one of the present state-of-the-art classifiers based on ensemble learning strategy, is relatively unknown in this field. In this paper, the behavior of random forest for the intelligent diagnosis of rotating machinery is investigated with various features on two datasets. A framework for the comparison of different methods, that is, random forest, extreme learning machine, probabilistic neural network and support vector machine, is presented to find the most efficient one. Random forest has been proven to outperform the comparative classifiers in terms of recognition accuracy, stability and robustness to features, especially with a small training set. Additionally, compared with traditional methods, random forest is not easily influenced by environmental noise. Furthermore, the user-friendly parameters in random forest offer great convenience for practical engineering. These results suggest that random forest is a promising pattern recognition method for the intelligent diagnosis of rotating machinery. [ABSTRACT FROM AUTHOR]

Due to the advancements in data measurement and computer technology, automated data collection from multiple sensors has become common in recent years. However, very few papers have dealt with the cost-optimal early fault detection of gearboxes, condition based maintenance policy, and remaining useful life prediction when multiple sensors are used for data collection under varying load. The novel approach presented here is based on vector autoregressive vibration signal modeling and continuous time hidden Markov modeling using the optimal Bayesian control technique. System condition is modeled using a continuous time Markov chain with three states, namely, unobservable healthy state 0, unobservable warning state 1 and observable failure state 2. Model parameters are calculated using the expectation-maximization algorithm. The optimal control policy for the three-state model is represented by a Bayesian control chart for a multivariate observation process. The chart monitors the posterior probability that the system is in the warning state 1 and the system is stopped when this probability exceeds an optimal control limit. Prediction of mean residual life using a posterior probability is also developed in this paper. The validation of the proposed methodologies is carried out using actual gearbox vibration data obtained from multiple sensors. [ABSTRACT FROM AUTHOR]

The fault diagnosis and prognosis of low speed machines remains a difficult problem despite remarkable advances in the conditional monitoring domain. The Rolling-element bearing is a vital part of these machines and its failure is one of the main causes of machine breakdown. In order to have an efficient maintenance strategy, fault diagnosis of a bearing and time estimation of its remaining useful life is needed. However, conventional vibration analysis at very low speeds generally fails to detect vibrations issued from a faulty bearing due to its low energy, high and variable loading conditions and to the noisy environment generated by other mechanical components of low speed machines such as gearing systems. In this work, instantaneous angular speed (IAS)-based fault diagnosis is introduced in order to compensate for the shortcoming of conventional monitoring techniques since it is strictly synchronized to shaft rotation and much less dependent on the transfer path between the defect and the sensor contrary to vibration and acoustic emission analysis. At very low speeds and in the case of a seeded spall on the bearing’s race, the shaft IAS reveals the shaft dynamical behavior when the rolling element passes into the spall. It is proven that this behavior is different when entering the spall than when exiting. The determination of entrance and exit moments allows a precise fault size estimation which is a critical step for bearing prognosis. The proposed fault size estimation method is tested on different seeded spall widths at different low speeds. The results gave a satisfactory fault width estimation and show that IAS measurement is a promising tool for the health monitoring of very low speed machines. [ABSTRACT FROM AUTHOR]

This paper presents robust fault detection based on adaptive thresholds for a three axis satellite. For this purpose, first, the attitude control system (ACS) is described as a quasi linear parameter model that includes both bounded parametric modeling errors and measurement noises. Next, using the interval arithmetic tools, an interval linear parametric varying observer is designed to propagate the effect of satellite parametric uncertainties into the alarm limits. This idea enhances the robustness of fault detection system at the decision making stage. In other words, the adaptive thresholds are generated for evaluating the residuals. Obtained results show that the missing alarm rates are minimized by the developed method; also this approach detects small or incipient faults more effectively than the classical robust fault detection algorithms with constant thresholds. In the next part of paper, an isolation algorithm has been proposed using the fault tree approach. Also, an accommodation system has been designed based on reconfiguration of available actuators. Accordingly, after isolation of faulty reaction wheels using the developed fault tree library, the accommodation system turned them off and replaced the suitable magnetic tourqers instead of faulty reaction wheels. Therefore, despite occurrences of several failures in the ACS, attitude control error is kept limited. [ABSTRACT FROM AUTHOR]

This paper presents an online fault isolation methodology for identifying faulty signals in a multiinput- multi-output dynamical system. It is hypothesized that faults in a dynamical system can be suitably represented via non-linear functions. The isolation scheme, which is implemented online, relies on adaptive non-linear estimates of these non-linear fault functions based on the system input-output data. The non-linear fault estimation is achieved using radial basis function neural network (RBFNN) architecture while the fault isolation is accomplished using extrinsic curvature of the learned RBFNN model. The proposed approach is implemented on an F-5A Freedom Fighter aircraft's lateral-directional model and the results are presented to illustrate the concept. [ABSTRACT FROM AUTHOR]