Papers by Keyword: Fault Diagnosis

Abstract: Induction motors are critical components in various industrial applications. Any faults can seriously affect the production system. Therefore, early fault detection is essential to prevent such occurrences. This research aims to develop a fault diagnosis model for induction motors. Raw signal data were obtained experimentally in the laboratory using two identical three-phase induction motors. There are eight different conditions categorized into single-combined faults. 18 features were extracted from each signal, consisting of 12 time-domain features and 6 frequency-domain features. These features were selected using the minimum Redundancy maximum Relevance (mRmR) algorithm. The selected features were then used as input to build a model using the Discriminant Analysis. The results indicate that the Discriminant Analysis model achieved very high accuracy across all condition classes. The computation time of the developed model is exceptionally fast, even below one second. Quadratic Discriminant Analysis (QDA) proved to be more accurate than Linear Discriminant Analysis (LDA) in classifying faults data.

Abstract: Early fault diagnosis is a crucial element in maintaining the optimal operation of rotating machinery and avoiding sudden failure resulting in material and non-material losses. This research aims to select the salient features to diagnose the induction motor faults using an SVM model. The induction motor is simulated experiencing three fault scenarios: single fault, double faults, and multiple faults. These scenarios consist of stator fault, rotor fault, bearing fault, stator-bearing fault, stator-rotor fault, bearing-rotor fault, and stator-bearing-rotor fault. Vibration signals for each of these conditions are collected using an accelerometer sensor with a sampling frequency of 20 kHz. The study utilizes 12 statistical features, comprising 7-time time-domain features, namely mean, standard deviation, kurtosis, RMS, skewness, peak value, crest factor, and 5 frequency domain features, namely mean frequency, median frequency, spectral entropy, power spectral density, and spectral centroid. These features are selected using the ReliefF feature selection algorithm, and the selected features are then employed as classification parameters. The results indicate that the most relevant statistical features used for classification parameters are RMS, Standard Deviation, and Power Spectral Density. Meanwhile, the performance of the Support Vector Machine is excellent for three cases of the induction motor faults. The accuracies for single faults, double faults, and multiple faults are 99%, 100%, and 99% respectively.

Abstract: Automating failure detection of infrastructure is essential to avoid unexpected downtime, which could impact both operations and user safety. However, frequent replacement of laminated rubber bearing (LRB) as a vital seismic isolation component can be inefficient and inspecting them for faults is often labor-intensive. This is a significant challenge in maintaining structural integrity, especially in critical infrastructure where continuous monitoring is necessary. Recent innovation in deep learning (DL) provides a promising alternative to traditional inspection methods, offering more efficient and accurate assessments. Hence, this paper explores the practical application of DL for detecting internal failure in laminated rubber bearings installed in structures. Neural network models, including a convolutional neural network (CNN), long-short term memory (LSTM), and their combinations (hybrid CNN-LSTM), were employed. An experimental setup was developed to simulate a bridge structure supported by down-scale LRB samples at its base. Samples with internal debonding failure were manufactured by reducing the bonding adhesive, to replicate failure conditions due to shear loading where the LRBs were forced to slide in an extreme condition. The vibration platform was actuated under different levels of frequency. Both Healthy and Faulty LRB conditions data were collected for 10 minutes each, which is adequate for 10 sets of data divided into training, validation, and testing with a fixed 6:2:2 ratio, respectively. Results revealed that CNN outperformed the other two models in average classification accuracy at 5Hz and 10Hz with 97.65% and 91.45%, respectively. Plus, CNN recorded the shortest training period among all models compared, with only 128 seconds at 15Hz, compared to 695 seconds and 1599 seconds owned by LSTM and hybrid CNN-LSTM respectively. In conclusion, neural networks have shown the capability in identifying LRB internal failure. CNN has the advantage in terms of both classification accuracy and training period compared to LSTM and hybrid CNN-LSTM models.

Abstract: This research uses a multi-domain technique to give a thorough analysis of mechanical gears health evaluation that includes time, frequency, and time-frequency signal analysis. The research seeks to discover patterns indicative of healthy, partially damaged, or fully damaged gear states using a variety of graphical representations, including time and frequency plots, the Short-Time Fourier Transform (STFT), and scalograms which are visual representations of the wavelet transform of a signal. Advanced machine learning models are used to improve diagnostic accuracy when manual identification of these trends becomes difficult. The goal is to achieve a validation accuracy greater than 70% a threshold selected based on prior studies indicating that this level ensures reliable fault detection for industrial applications while balancing computational constraints. The reliability and effectiveness of gear monitoring systems can be increased by integrating contemporary signal processing and machine learning approaches, as demonstrated by this research, which also advances the identification of gear faults. Among the conclusions are the outcomes of tests done to identify gear problems in which authors were able to train a model with more than 72% accuracy and able to propose Artificial Intelligence model for classification of faults in gears.

Abstract: The good functioning of a discrete event system is related to how much the temporal constraints are respected. This paper gives a new approach, based on a statistical model and neural network, that allows the verification of temporal constraints in DES. We will perform an online temporal constraint checking which can detect in real time any abnormal functioning related to the violation of a temporal constraint. In the first phase, the construction of temporal constraints from statistical model is shown and after that neural networks are involved in dealing with the online temporal constraint checking.

Abstract: In this work, we are interested in the faults diagnosis and the faults prognosis in discrete event systems described by sequences of generated events. Through this work, we aim the maximization of the efficiency of diagnosis/prognosis operations by combining two concepts. The first one is the approach already developed in one of our works which consider the k-last generated events to perform the diagnosis/prognosis. The second concept is the reliability that takes into consideration the life cycle of each component of the discrete event systems to give the failure probability. This combination will be made using some notions of fuzzy logic.

Abstract: This paper addresses the problem of fault detection and isolation (FDI) in wind turbine benchmark model using data driven and multi-class support vector machine (SVM) approach. Since, the fault detection is fundamental for any active system, isolation is similarly vital, and identification is decisive for fault reconfiguration as well as maintenance addition to monitoring purposes. The need for man-made dynamic system to work automatically when sensor, actuator, or system faults occur was constantly developed in order to increase reliability and decrease unavailability and maintenance costs. The key step of our approach based on extraction of mean features from sensors measurements by applying the statistical methods such as moving standard deviation and the exponential weighted moving average (EWMA). The fault detection step is invoked later based on the multi-class SVM classifier that decides the presence or not of the fault. Another important contribution of this paper is the simulation of combined sensor and actuator faults simultaneously for the first time in wind turbine benchmark model. The FDI performances are illustrated through simulation study for seven different scenario tests. The results demonstrate clearly the effectiveness of statistical and SVM approach to detect and isolate single, multiple sensor and actuator faults and outperforms many results reported in the literature for solving this problem.

Abstract: This paper presents a remote equipment virtual maintenance training system architecture. The system function modules are designed. We use the open source software to develop the environment to build a complete system which based on ThinkPHP and MVC framework. Multisim circuit simulation software is nested for circuit board components positioning and fault diagnosis. The system can be used for fault diagnostics in electrical system and it can also be used to train maintenance staff.

Abstract: Liquid ring vacuum pump which uses rotating liquid as piston to abstract and compress gas is a kind of rotating positive displacement pump. Its running state implies some fault informantion. To capture performance levels of liquid ring vacuum pump, a comprehensive performance monitoring system includes data acquisition, data reprocessing, data storage, abnormal judgment and fault alarm, performance displaying is developed in this paper. Hence, the performance levels of liquid ring vacuum pump can be observed anytime by the monitoring system, the maintenance cost can be reduced, pumps can be operated at the highest performance level as far as possible, the reliability and the maintainability of liquid ring vacuum pump can be effectively improved by this monitoring system.

Abstract: Complex mass faults diagnosis of the pantograph type current collector was difficult. Based on the analysis of the structure, working principle and failure mode of the pantograph type current collector, fault tree of the pantograph was established. A lot of expert knowledge has been collected to support this diagnose process. Some serious problems such as ambiguity, uncertainty and inconsistency exist in the knowledge. Focused on the deficiencies, ontology modeling was proposed in this paper. StrOnto, FaultOnto and FTOnto were established to standardize the knowledge and to improve the efficiency of fault diagnosing. Finally, combined with the example of the pantograph type current collector of CRH2 EMU-train, the proposed algorithms proposed in this paper were proved reasonable and effective.