Frequency domain tensor-based 1D-convolutional neural network and multilinear principal component analysis for machinery fault detection

This paper describes the fault classification and detection of SVC (Static VAR Compensator) integrated differential relay based double circuit transmission line applying Discrete Wavelet Transform (DWT) approach. The SVC is placed at mid-point of circuit-I of double transmission line. The other circuit (circuit-II) remains uncompensated. The differential relay based protection scheme is used for the analysis of fault detection and classification. The process starts by extracting current signals (A, B or C-phase) from both sending and receiving end of the transmission line synchronously and processes through DWT to obtain the spectral energy (SE) at both the end of the line. The Differential Spectral Energy (DSE) of each phase current signals are obtained by SE of current measured at sending end minus SE of current measured at receiving end. This DSE value is used to register the fault pattern.

Rotatingmachinery is the main component of mechanical equipment. Nevertheless, due to variation of operating condition results in important detection performance deterioration. Therefore, fault detection and diagnosis of rotating machines is very critical for the reliable operation. In this paper, a novel classification technique is employed for fault detection of rotating machines based on kernelled support tensor machine (KSTM) and multilinear principal component analysis (MPCA). The vibration signal is firstly formulated as a 3-way tensor using trial, condition and channel. In order to process the rotating machines faults and identify the information classes in tensor space, the KSTM is then introduced from sets of binary support tensor machine classifiers by the one-against-one parallel strategy. The MPCA is utilized for reduction dimensionality of the high-dimensional signature space and reservation the tensorial structure information. The performance of the developed technique in classification faults of rotating machinery has been thoroughly evaluated through collecting signals on bearing and gear test-rigs. Experimental results showed that the proposed method can achieve the highest classification results among the six classification techniques investigated in this study.

Fault diagnosis is important for the maintenance of rotating machinery. The detection of faults and fault patterns is a challenging part of machinery fault diagnosis. To tackle this problem, a model for deep statistical feature learning from vibration measurements of rotating machinery is presented in this paper. Vibration sensor signals collected from rotating mechanical systems are represented in the time, frequency, and time-frequency domains, each of which is then used to produce a statistical feature set. For learning statistical features, real-value Gaussian-Bernoulli restricted Boltzmann machines (GRBMs) are stacked to develop a Gaussian-Bernoulli deep Boltzmann machine (GDBM). The suggested approach is applied as a deep statistical feature learning tool for both gearbox and bearing systems. The fault classification performances in experiments using this approach are 95.17% for the gearbox, and 91.75% for the bearing system. The proposed approach is compared to such standard methods as a support vector machine, GRBM and a combination model. In experiments, the best fault classification rate was detected using the proposed model. The results show that deep learning with statistical feature extraction has an essential improvement potential for diagnosing rotating machinery faults.

Advanced wind turbine blade inspection with hyperspectral imaging and 3D convolutional neural networks for damage detection

This research presents a comprehensive approach for bearing fault diagnosis using artificial intelligence (AI), particularly through the application of artificial neural networks (ANNs). By integrating these networks into vibration analysis, the approach aims to meet the critical need for prompt fault detection. The methodology comprises three key steps: vibration signal acquisition, feature extraction, and fault classification. Experiments were conducted to acquire vibration signals for the test bearings on a machinery fault simulator. Six time-domain features were extracted using MATLAB, creating a comprehensive dataset for training the ANN models with three algorithms: Levenberg-Marquardt backpropagation (LMBP), scaled conjugate gradient backpropagation (SCGBP), and Bayesian regularization backpropagation (BRBP). The BRBP algorithm achieved the highest correct classification rate (97.2%), followed by LMBP (90%) and SCGBP (83.6%). To evaluate their efficacy in bearing fault classification, these three networks were simulated, revealing that BRBP could predict all four classes of bearings with zero errors.

Large margin-learning methodology from time-frequency maps and its physically interpretable weights for simultaneous machine health monitoring and fault diagnosis

The increasing demand of electricity has led to the growth in adoption of electrical microgrids in power system. Microgrids have many benefits for power systems, however their protection is one of the most challenging issues in power systems. This research focuses on the Artificial Neural Network (ANN), an artificial intelligence-based technique for fault detection, classification and localization in an AC microgrid. For normal and fault situations, a three phase AC grid is modelled in MATLAB Simulink. This research employs a backpropagation technique with a feed forward-based ANN. The performance is observed using Mean Square Error (MSE) and regression plot. The network is trained for all symmetrical and asymmetrical faults using faulty voltages and currents for different loads at different fault locations. The fault detection obtained a tolerable MSE of 0.0098 and regression as 0.98 having accuracy of 100%. The fault classification obtained a tolerable MSE of 0.043 and regression value as 0.99 with an accuracy of 75%. The same Simulink model is then realised in OP4510 RT-LAB Simulator to compare the real time results with MATLAB results as the speed of simulations in MATLAB depends on the central processing unit of PC.KeywordsANNAC microgridFault locationFault detectionFault classificationRt-lab

The availability of advanced sensors and computing enables early detection of faults and failures in large-scale solar photovoltaic (PV) systems. However, small variations in current and voltage caused by arcing or other faults/failures are often difficult to detect, potentially leading to a permanent reduction in power output and can potentially cause safety hazard. This work develops novel algorithm for detection and classification of DC arc and AC faults using Conditional Generative Adversarial Networks (CGANs). Three distinct CGAN architectures are developed: two for DC arc fault detection and classification, and another for AC faults. The CGAN networks generate output signals representing probability distributions based on voltage and current measurements at the inverter location of the PV system. The performance of the proposed CGAN architectures is evaluated using rolling precision scores on backcasted datasets and compared against a k-Nearest Neighbors (kNN) model, and also validated using real field data. Results show that the proposed architectures achieve high precision in detecting and classifying both types of faults in a 1 MW solar PV system modeled in a real-time simulator.

Three-dimensional convolutional neural network for leak detection and localization in smart water distribution systems

Induction machines are used in a wide range of industrial applications due to their simplicity, ruggedness, and low price. Despite their robustness, induction machines eventually fail due to a variety of mechanisms. Most faults exhibit specific frequency components in the motor current spectrum, which allows for fault detection. Many classical fault detection methods have been developed for grid-connected machines with relatively fixed operating points. In inverter-driven machines with a wide operating range, these methods cannot reliably detect and classify faults. Machine learning methods have been successfully used for various classification tasks. This study therefore combines classical fault detection approaches with various fault classification algorithms to reliably detect induction machine faults over a wide operating range.

Effective fault detection, classification, and localization are vital for smart grid self-healing and fault mitigation. Deep learning has the capability to autonomously extract fault characteristics and discern fault categories from the three-phase raw of voltage and current signals. With the rise of distributed generators, conventional relaying devices face challenges in managing dynamic fault currents. Various deep neural network algorithms have been proposed for fault detection, classification, and location. This study introduces innovative fault detection methods using Artificial Neural Networks (ANNs) and one-dimension Convolution Neural Networks (1D-CNNs). Leveraging sensor data such as voltage and current measurements, our approach outperforms contemporary methods in terms of accuracy and efficiency. Results in the IEEE 6-bus system showcase impressive accuracy rates: 99.99%, 99.98% for identifying faulty lines, 99.75%, 99.99% for fault classification, and 98.25%, 96.85% for fault location for ANN and 1D-CNN, respectively. Deep learning emerges as a promising tool for enhancing fault detection and classification within smart grids, offering significant performance improvements.

Switching LDS-based approach for process fault detection and classification

• Introduces a novel method using Hilbert-Huang Transform (HHT) with LSTM networks for fault detection in DC microgrids. • Targets marine and satellite DC microgrids, addressing challenges with no zero-crossing points in traditional methods. • Demonstrates superior performance, outperforming SVM and Decision Tree techniques in fault detection time and accuracy. • Validates the method using MATLAB/Simulink, effectively detecting faults with resistance up to 1.3 ohms. • Ensures compliance with IEEE 8570 standards, enhancing reliability and applicability in real-world microgrid systems. Fault detection in DC microgrids is a major challenge because traditional protection methods like wavelet and Fourier transforms rely on zero-crossing points, which the DC signal lacks. This paper introduces a new approach combining the Hilbert-Huang Transform (HHT) and a long short-term memory (LSTM) neural network to improve fault detection and classification in DC microgrids. The proposed method first uses Empirical Mode Decomposition (EMD) to break down the fault current signal into multiple intrinsic mode function (IMFs). These IMFs capture important fault-related information at different frequency levels. Next, the Hilbert Transform is applied to these IMFs to obtain amplitude and frequency features over time. These features are used to create a set of numerical descriptors that summarize the fault characteristics. These extracted features are then fed into a two-layer LSTM neural network that can learn patterns in time-series data. The LSTM model is trained to classify the input into three categories: no fault, Pole-Ground (P-G) Fault, or Pole-Pole (P-P) Fault. Simulation tests were carried out using a MATLAB Simulink model of a 20kWatt DC microgrid under different fault conditions, noise levels, and load transients. Results show that the proposed HHT-LSTM model detects faults faster and more accurately than conventional techniques like Support Vector Machines (SVM) and Decision Trees (DT). It also meets the fault detection time requirements set by IEEE8570 and aligns with IEC standards. The model remains robust even under noisy conditions, making it reliable for real-time fault protection in marine, satellite, and an effective solution for enhancing fault detection in modern DC microgrids.

Background: Fault detection is essential problem in machinery. The different type faults can occur in induction machine due to stator short circuits and rotor failures. Direct Torque Control (DTC) is a well known control design for induction motor drive in high power applications. Methods: Generally, the classification of faults is done by Principal Component Analysis (PCA) and Support Vector Machine (SVM) techniques. In this paper Discrete Wavelet Decomposition (DWD) technique is used to categorize the stator and rotor faults of DTC induction machine. Using the DTC Simulink model a variety of faults are generated and tested. Finding: By using wavelet decomposition, the information on the health and fault of a system can be extracted from signal over wide a range of frequencies. The extracted features are used to classify the healthy machine from faulty machine. Application: This method can be applicable in windmill, lathe and drilling industries to increase the efficiency of machines.

Application of advanced fault diagnosis and monitoring techniques allows more efficient, reliable and safe operation of many complex industrial systems. Recently, there has been a significant increase in application of various data-driven deep learning models for motor fault detection and diagnosis problems. Due to high computational complexity and large training dataset requirements of deep learning models, in this study, shallow and adaptive 1D convolutional neural networks (CNNs) are applied to real-time detection and classification of broken rotor bars in induction motors. As opposed to traditional fault diagnosis systems with separately designed feature extraction and classification blocks, the proposed system takes directly raw stator current signals as input and it can automatically learn optimal features with the proper training. The other advantages of the proposed approach are (1) its compact architecture configuration performing only 1D convolutions with a set of filters and subsampling, making it suitable for implementing with real-time circuit monitoring, (2) its requirement for a limited size of training dataset for efficient training of the classifier and (3) its cost-effective implementation. Effectiveness and feasibility of the proposed method is validated by applying it to real motor current data from an induction motor under full load.