Fault Simulation Research Articles

Self-driven continual learning for class-added motor fault diagnosis based on unseen fault detector and propensity distillation

Continual learning is a high-potential technique that enables intelligent motor fault diagnosis models to extend new diagnosable fault classes without costly training from scratch. However, existing continual learning methods have the following limitations. (1) They manually detect new faults, which is labor-intensive, untimely, and more importantly, may lead to mistaken diagnosis results. (2) They adopt the traditional knowledge distillation to align the absolute responses of old and new models, which alleviates catastrophic forgetting but restricts flexible learning from incremental datasets. To overcome the above limitations, this paper proposes a novel self-driven continual learning framework for class-added motor fault diagnosis, which can spontaneously detect unseen faults and perform more flexible continual learning from incremental datasets. For the automatic detection of unseen faults, after collecting online samples, adversarial training with exemplars of each seen class is conducted to measure the class separability. The truth fault classes that are unseen for diagnosis models can be clearly distinguished from all seen classes, and correspondingly missed diagnosis or misdiagnosing can be avoided effectively and incremental samples with new fault types can be collected quickly. For the flexible continual learning strategy, a more flexible knowledge distillation is proposed to preserve the prediction propensity rather than the absolute response. This strategy not only keeps the recognition performance of old classes but also loosens unnecessary constraints and increases the diagnosis model plasticity to learn new knowledge from incremental datasets, thus improving the accuracy of motor fault diagnosis during continual learning. The effectiveness of the proposed method is verified by conducting fault simulation experiments of three-phase motors and its superiority is also demonstrated by comparing it with some state-of-the-art diagnosis methods.

Modeling and Fault Simulation of a New Double-Redundancy Electro-Hydraulic Servo Valve Based on AMESim

The feedback spring rod of the armature assembly was eliminated in the double-redundancy electro-hydraulic servo valve (DREHSV), which employed a redundant design in contrast to the typical double-nozzle flapper electro-hydraulic servo valve (DNFEHSV). The pilot stage was mainly composed of four torque motors, and the double-system spool was adopted in the power stage. Consequently, the difficulty of spool displacement control was increased. By artificially changing the structural parameters of the simulation model in accordance with the theoretical analysis through AMESim, this paper aimed to study the dynamics and static characteristics of the DREHSV. The advantage of redundant design was further demonstrated by disconnecting working coils and setting the different worn parts of the spool. On the test bench, the necessary experiments were performed. Through simulation, it was discovered that when the clogged degree of the nozzle is increased, the zero bias value increases, the pressure and flow gain remain unchanged, and the internal leakage decreases. The pressure gain changes very little, the flow gain close to the zero position grows, the zero leakage increases significantly, and the pilot stage leakage changes very little as a result of the wear of the spool throttling edge. The basic consistency between the simulation curves and the experimental findings serve to validate the accuracy of the AMESim model. The findings can serve as a theoretical guide for the design, debugging, and maintenance of the DREHSV. The simulation model is also capable of producing a large amount of sample data for DREHSV fault diagnosis using a neural network.

Multivariate intrinsic wave-characteristic decomposition and its application in gear fault diagnosis

In the early stages of gear faults, the background noise in the signal is pronounced, making it challenging to fully assess the health status of equipment based on a single-channel signal. Processing multi-channel signals proves beneficial for extracting fault information comprehensively. Adaptive multivariate signal decomposition methods, such as multivariate empirical mode decomposition (MEMD) and multivariate local characteristic-scale decomposition (MLCD), employ a fixed multivariate mean curve extraction method for signal decomposition. Consequently, these methods often exhibit suboptimal performance when decomposing different multi-channel signals. This study defines nine multivariate mean curve extraction methods and introduces the multivariate intrinsic wave-characteristic decomposition (MIWD) method based on the principles of mean curve optimization and an adaptive projection method. MIWD dynamically optimizes the multivariate mean curve during the decomposition process, resulting in superior performance in terms of decomposition accuracy, capability, and orthogonality compared to MEMD and MLCD. Furthermore, we apply MIWD to gear fault diagnosis, and simulation and experimental results affirm the superiority of MIWD.

Statistical Modeling of Soft Error Influence on Neural Networks

Soft errors in large VLSI circuits have a significant impact on computing-and memory-intensive neural network (NN) processing. Understanding the influence of soft errors on NNs is critical to protect against soft errors for reliable NN processing. Prior work mainly rely on fault simulation to analyze the influence of soft errors on NN processing. They are accurate but usually specific to limited configurations of errors and NN models due to the prohibitively slow simulation speed especially for large NN models and datasets. With the observation that the influence of soft errors propagates across a large number of neurons and accumulates as well, we propose to characterize the soft error induced data disturbance on each neuron with normal distribution model using the central limit theorem and develop a series of statistical models to analyze the behavior of NN models under soft errors in general. The statistical models reveal not only the correlation between soft errors and the accuracy of NN models, but also how NN parameters such as quantization and architecture affect the reliability of NNs. The proposed models are compared with fault simulations and verified comprehensively. In addition, we observe that the statistical models that characterize the soft error influence can also be utilized to predict fault simulation results in many cases and we explore the use of the proposed statistical models to accelerate fault simulations of NNs. Our experiments show that the proposed accelerated fault simulation provides almost two orders of magnitude speedup with negligible loss of simulation accuracy compared to the baseline fault simulations.

Fault Tolerant Brushless DC Motor Drive for Aerospace Applications

This article brings out a Fault-tolerant BLDC motor drive for aerospace applications using the redundancy concept. In a way, it brings out a fault-tolerant strategy that can be used to continue the regular operation of a BLDC motor drive even after the occurrence of faults. As BLDC motors are used in critical and dangerous control areas like military services and space vehicles, a fault-tolerant drive is essential to maintain drive operation and provide desirable output. This article compares fault simulation results in the software model of a BLDC motor drive to those of fault simulation results in hardware for three main types of faults. Fault simulation is carried out for three types of faults, viz. inverter device open circuit fault, motor winding open circuit fault, and rotor position sensor (hall sensor) open-circuited fault. Fault tolerance is ensured by introducing a redundant drive (drive-2), which operates the complete drive at the advent of any of the faults mentioned above in the main (healthy) drive-1. A fault-tolerant (redundant) hardware comprising dual stator BLDC motor and redundant controllers is realized and operationalized. Fault simulation is carried out in this hardware, and these results are validated with the results of fault simulation in the MATLAB SIMULINK model. Software and hardware results are comparable and form a basis for developing fault-tolerant electro-mechanical actuation systems for high-reliability, high-cost applications, mainly aerospace.

Study on cavitation fault features of centrifugal pumps based on principal component analysis

The centrifugal pump plays a key role in the ship water supply system, and cavitation is a common fault mode that leads to poor efficiency of centrifugal pumps. To select valid features as the criteria of cavitation fault diagnosis, we calculate RMS, peak factor, kurtosis factor, wave factor of raw data, and the data after empirical mode decomposition (EMD). Although the raw features can be obtained, they were still in a high-dimensional space and contain a lot of redundant information. So, Principal Component Analysis (PCA) method was used to decrease dimensions and extract sensitive features. As a case study, the data were obtained from a centrifugal pump fault simulation bench, and a variety of cavitation states were observed in the experiment. And a six-dimensional sensitive feature vector was determined through data analysis.

Shear‐Enhanced Electrical Conductivity of Synthetic Quartz‐Graphite Gouges: Implications for Electromagnetic Observations in Carbonaceous Shear Zones

AbstractGraphite is considered as a material that promotes fault weakening and electrical conductivity (σ) enhancement at fault zones. We studied how shear deformation may affect the evolution of friction and electrical conductivity of synthetic quartz (Qz)‐graphite (Gr) mixtures and, more importantly, whether the σ of the mixtures present visible changes at the beginning of the simulated fault slip. Long‐displacement friction experiments were performed on 1.2–2.3 mm‐thick gouge specimens of varied Gr volume fraction (XGr = 0–100 vol.%) under identical normal stress (2 or 5 MPa), slip rate (∼1.0 mm/s), and N2‐flushing conditions. The experimental results suggested that the σ of the specimens with ≥4.6 vol.% XGr abruptly increased under limited shear displacement. With continued shear, the steady‐state electrical conductivity (σss) increased by more than seven orders of magnitude when XGr > 3.4 vol.%, while the steady‐state frictional coefficient remained high (0.54–0.80) except for the specimens with XGr > 13.6 vol.%. The post‐mortem microstructures revealed that the high σss observed in the intermediate Gr content specimens (3.4–13.6 vol.%) is associated with an ad‐hoc fabric (graphite–cortex clasts) present in the principal slip zone. For high Gr content, excess Gr flakes fill the pores and help develop mechanically lubricated surfaces. We propose that low Gr content (i.e., as low as 3.4 vol.%) can cause high conductivity anomalies in natural shear zones. Overall, the findings suggest that the initiation of slips within carbonaceous shear zones can be detected by identifying unusual temporal signals using electromagnetic stations.

Experimental EMFI detection on a RISC-V core using the Trace Verifier solution

Physical attacks are powerful threats that can cause changes in the execution behavior of a program. Control-Flow Integrity (CFI) is used to check the program’s flow execution, ensuring that it remains unaltered by these attacks. The RISC-V Trace Encoder (TE) provides valuable information about the user program’s execution path, and is used as part of a CFI solution. An enhanced version of the TE specifications permits detecting intricate fault models such as the corruption of any discontinuity instruction, using an additional Trace Verifier (TV) hardware module. In this paper, we present a buffer overflow software attack simulation and experimental ElectroMagnetic Fault Injection (EMFI) attacks conducted on an Field Programmable Gate Array (FPGA) board that implements a RISC-V core linked to the enhanced TE and TV modules. Unlike existing CFI solutions, our proposed approach does not require modifications to the RISC-V compiler, user application code or the RISC-V core. The average overhead of our solution in terms of hardware area, memory and power consumption are equal to 13.6%, 3.5% and 9% respectively.

Optimal sensor placement for permanent magnet synchronous motor condition monitoring using a digital twin-assisted fault diagnosis approach

Efficient health monitoring for identifying and quantifying damages can substantially improve the performance and structural integrity of engineered systems. Specifically, new advances in sensing technologies have pushed the research of large sensor networks to monitor complex mechanical structures. Given the need for health state monitoring, designing an optimal sensor framework with accurate detectability of failure modes has great significance. However, there is often little to no experimental data available for newly proposed mechanical systems; so a digital-twin method would make fault detection feasible for this applications. In this paper, a data-driven reliability-based design optimization (RBDO) approach is employed for sensor placement and fault detection of a permanent magnet synchronous motor (PMSM), which is a relatively new system for high power engineering applications. This system suffers from inter-turn and inter-phase short-winding faults, which can cause catastrophic failure of the whole structure. For PMSMs, current sensing and magnetic field sensing can be utilized for the detection of faults, but actual sensor placement has not been considered in recent literature. In this study, the first step is to create an FEA model of the PMSM for the simulation of faults, which serves as the digital twin. Next, a data-driven approach is implemented for sensor placement and classification of faults. The proposed method utilizes distance clustering for identification of various failure modes, which is suitable for many applications due to its high accuracy and computational efficiency. In addition, a genetic algorithm is implemented to determine the minimum number and optimal placement of sensors. This framework simultaneously searches for the optimal placement of sensors while training the classifier for detectability of system health states. Ultimately, the proposed methodology shows convergence to a solution with high accuracy for detection of faults, and is demonstrated on the novel system of a PMSM with magnetic field sensors.

The application of fault diagnosis techniques and monitoring methods in building electrical systems – based on ELM algorithm

The reliability of modern building electrical systems are receiving increasing attention as they become more intelligent and complex. As the majority of building electrical systems use neutral point grounding, earth faults or short circuits can get worse over time and damage both the distribution system and the electrical equipment. To this end, the corresponding three phases and four categories, namely three-phase voltage, three-phase current after fault, three-phase voltage distortion rate, three-phase current distortion rate, a total of 12 dimensional fault feature vectors and 10 fault simulation types, were summarised and extracted in conjunction with the actual operating conditions of the system. Using traditional fault identification ideas and neural network algorithm as reference, a 12-dimensional fault feature vector is used as the model input to construct a building electrical fault diagnosis and detection model based on ELM algorithm. Results showed that the ELM-based model’s classification accuracy for this experimental sample was 97.56 %, its AUC was 0.92, and its RMSE was 0.3521. These figures were higher than the classification accuracy and performance of the BP algorithm and GA-BP algorithm fault diagnosis models, and they also demonstrate better robustness and generalizability. The model also has a 97.27 % correct rate in fault discrimination, while the computation time is only 0.201 s, and its fault identification and diagnosis speed is faster than other algorithmic models. At the same time, this research model has a good fault monitoring accuracy of up to 98.6 % for building electrical systems. The research can provide a more sensitive, accurate and rapid fault monitoring method for the current building electrical system. It also improves the reliability of the building electrical system in a complex environment and achieves better protection of the system. This has a certain significance for the development of the building electrical industry.

A Physics-informed, Transfer Learning Approach to Structural Health Monitoring

One of the main challenges for structural health monitoring (SHM) is a lack of failure data to make accurate health predictions. Obtaining desirable failure data is generally very expensive, given the required testing needed to measure all types of system failures, which may be unfeasible in many health monitoring applications. Machine learning has helped to improve health monitoring performance but is still limited by the availability, relevance, and quality of the training data. This data dependence impedes data-driven models from generalizing to unseen data, which is problematic for datasets lacking failure data. Physics-driven models, like finite-element models, are powerful tools for predicting structural responses when the governing physics are not clearly defined. These models can generate simulated fault data to address the data limitation without having to physically damage a structure, but are computationally expensive and susceptible to modeling errors that can prevent the data from being statistically comparable to experimental data.
 A new trend has been to develop physics-guided machine learning models (PGML), a hybridization of the two aforementioned models that have been shown to improve generalization of, and even outperform, pure data-driven models while using less training data. These PGML models can take many forms, but generally embed some form of physics into a data-driven model as physically relevant constraints. Our research plan is to utilize PGML to improve neural network capabilities to predict structural damage. The proposed PGML model will follow a neural network architecture found in related literature consisting of feature extraction, physics-informed, and label prediction layers. The physics-informed layer will consist of an aggregate of sub-networks trained from simplified structure models which have known governing equations and can be used to generate simulated training data. The full PGML model will use transfer learning to bridge the connections between the untrained layers to the physics-informed layer using experimental data from more complex structures. We will verify our model using publically available SHM datasets used in a variety of past literature experiments.

Comparison Study On Pinch-Hitting Vibration Signal Analysis for Automotive Bearing: I-KazTM and I-Kaz 3D

Rotating machines are now an essential part of the automotive industry. Meanwhile, a bearing is playing the most important component of rotating machinery. To sustain the system's smooth running, maintenance methods such as preventive maintenance, breakdown maintenance, and predictive maintenance are used. Under preventive maintenance, vibration analysis is used to diagnose machines bearing faults. The main objective is to recognize bearing defects in a mechanical device by acquiring signals from the bearing using data acquisition hardware. This analysis is conducted under various load torque conditions, speeds, and defect types. A modular hardware configuration consisting of an accelerometer acquires the vibration signal. The signals are analyzed by using I-kazTM and I-kaz 3D signal analysis and its main objective is to observe the degree of dispersion data from its mean point. This analysis resolves the issues associated with time domain analysis. This pinch-hitting analysis research was conducted in two stages. The first stage is an experimental process that uses 3 types of bearings, the healthy (BL), inner race fault (IRF), and defect at outer race (ORF) bearing on the Machine Fault Simulator and forces with a different type of speed (1000, 1500 and 2500 rpm) and load variation (0.0564, 0.564 and 1.1298 N-m). In the second stage, computing the coefficient value and plots of signal’s I-kazTM and I-kaz 3D based on the bearings type were done accordingly. As a result, the analysis for detecting inner race fault, the deviation percentage averages calculation obtained the I-kazTM coefficient shows a better result with 96.86% by comparing to the I-kaz 3D that achieves 94.20%. Similarly, for the outer race defect, I-kazTM lead with 65.40% compared to I-kaz 3D with only 54.82%.

Inverse transient analysis based calibration of surrogate pipeline model for fault simulation of axial piston pumps

Axial piston pumps are the key component in power hydraulic systems. Flow and pressure ripple signals, which can reflect the flow dynamics of the axial piston pump, are important information sources for condition monitoring. Therefore, the characteristics of flow and pressure ripple signals should be investigated for further condition monitoring and fault analysis. However, measuring high-frequency flow ripple directly can be difficult, and the pressure ripple is not only affected by pump dynamics but also affected by the transient pressure wave propagation. However, in practical applications, the hydraulic pipeline system is complex due to the space constraint of the equipment, and the influence of the pipeline system cannot be ignored when modeling pressure ripple. This paper proposes a novel two-stage method for fault simulation of the pumps that considers the influence of the pipeline system on the pressure ripple. Firstly, the surrogate pipeline model with free parameters is calibrated using ITA to capture the characteristics of the complex practical pipeline system. Secondly, the calibrated model is applied to the fault simulation of axial piston pumps. A detailed 3D CFD model of the pump is built up, and the flow ripple at the pump outlet from the numerical model is used as an approximation of the real flow ripple of the pump, which serves as one of the boundary conditions of the surrogate pipeline model. After calibrating the model, the CFD model of the pump is used to simulate internal failures in the pump. Two failure cases are studied: a faulty slipper and a faulty cylinder. The simulated signal of the proposed model matches the experimental signal well under the healthy and faulty conditions, which confirms the effectiveness of the proposed method.

Experimental Simulation and Electromechanical Characterization of Dynamic Air Gap Eccentricity Faults inPMSG

This paper presents a designed experimental simulation scheme for dynamic air gap eccentricity (DAGE) faults in permanent magnet synchronous generator (PMSG), along with the testing of their electromechanical characteristics. Unlike previous studies, this paper proposes and applies an experimental device setting scheme that enables accurate and convenient calibration of the DAGE fault degree, offering a novel solution for practical DAGE simulation. The experimental unit measures the electromechanical characteristics of the PMSG before and after the DAGE fault, taking into account the influence of load. The mechanical parameter considered is the stator vibration, while the electrical parameter is the circulating current parallel branches (CCPB) inside the stator winding. The characteristic frequencies of stator vibration and CCPB under the DAGE fault are analyzed based on the experimental results and verified through theoretical calculations and finite element analysis (FEA). The findings demonstrate the effectiveness of the proposed DAGE experimental device. Moreover, DAGE failure increases the strength of stator vibration and introduces new frequency components, namelyfrand 2f ± fr. Under normal operation, the PMSG exhibits no CCPB. However, DAGE faults cause CCPB with frequency components off ± fr. Moreover, the severity of the fault degree positively correlates with larger root‐mean‐square (RMS) values and characteristic frequency amplitudes of stator vibration and CCPB. Furthermore, the amplitude of stator vibration and CCPB decreases with increasing load. © 2023 Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.

Experimental Vibration Data in Fault Diagnosis: A Machine Learning Approach to Robust Classification of Rotor and Bearing Defects in Rotating Machines

This study builds upon previous research that utilised a vibration-based machine learning (VML) approach for diagnosing rotor-related faults in rotating machinery. The original method used artificial neural networks (ANN) to classify rotor-related faults based on optimised vibration parameters from the time and frequency domains. This study expands the application of this vibration-based machine learning approach to include the anti-friction bearing faults in addition to the rotor faults. The earlier suggested vibration-based parameters, both in time and frequency domains, are further revised to accommodate bearing-related defects. The study utilises the measured vibration data from a laboratory-scale rotating test rig with different experimentally simulated faults in the rotor and bearings. The proposed VML model is developed for both rotor and bearing defects at a rotor speed that is above the first critical speed. To gauge the robustness of the proposed VML model, it is further tested at two different rotating speeds, one below the first critical speed and the other above the second critical speed. The paper presents the methodology, the rig and measured vibration data, the optimised parameters, and the findings.

Analysis of Power System Transient Stability of the Nigerian 330-kV Electricity Grid

Faults on power system are inevitable and can occur at any time resulting into transient instability. Transient instability can cause loss of synchronism and possible damage to power equipment and consumer loads. Therefore, this study analyzed the transient stability (TS) of the Nigerian 330-kV, 34-bus power network. Power flow and swing equations describing the system steady and transient states were respectively analyzed using Newton-Rahpson and Runge-Kutta, fourth order numerical techniques to provide linear solution due to their non-linearity. Simulations were done and swing curves for different fault conditions obtained. The critical clearing time at which the simulated faults were cleared was determined. The results of load flow analyses revealed that buses 6, 10, 13, 14, and 17 with respective voltage magnitude of 0.937, 0.921, 0.938, 0.829 and 0.780 per unit violated voltage tolerance limit of 0.95 to 1.05 p.u. The system total active and reactive power losses were 57.62 MW and -70.62 MVar respectively. The obtained swing curve showed that for fault on bus 11 [line 9-11] generators 2, 3 and 7 lost synchronism after 1.0 s but other generators retained their stability. Similarly, for fault on bus 13 [line 10-14] generators 3 and 7 lost synchronism but other generators retained stability after 0.4 s. These results indicate that faults occurrence on generating stations of power system is unavoidable, but the clearance time must be short (few seconds) to avert total system collapse and loss of synchronism.

Electromagnetic Time Reversal Fault Location Method Based on Characteristic Frequency of Traveling Waves

In its application to locate ground faults, the electromagnetic time reversal (EMTR) method uses voltage from only a single point. However, when a high-impedance fault (HIF) occurs, the reflected signal from the fault point is weak, leading to an error between the simulated and actual fault location results. This paper proposes a new EMTR method based on the characteristic frequency of the traveling wave reflecting from the fault point. First, the transient signal spectrum under different fault resistances was analyzed, and the reasons for the inapplicability of the existing EMTR method under HIF conditions were discussed. Subsequently, a single-end signal expression with fault resistance was deduced, and the applicability of the EMTR method, which involves calculating only the energy of the fault current at the characteristic frequency, was verified. A reduced-scale experiment and a simulation in PSCAD were performed, and the EMTR method and proposed method were compared. The results showed that the proposed method can be applied to different complex lines, and it outperforms the conventional EMTR method in the event of HIFs.

Data structures for deductive simulation of HDL conditional operators

The subject of research in the article is qubit-vector models for combinational circuits’ description and procedures for faults deductive simulation based on these models. The object of research is the processes of diagnostic support creation for digital systems based on the usage of vector qubit data. The purpose of the work is to increase the speed and quality of diagnostic support creation for digital devices by creating optimal data structures and deductive fault simulation procedures based on structural-functional models of combinational circuits. The following tasks are solved in the article: analysis of concurrent and sequential conditional operators of hardware description languages and schematic structures into which they are synthesized; development of the procedure for the truth tables (Q-vectors) formation for schematic structures presented by HDL; development of an universal data structure for cubic and analytical deductive faults simulation; vector models improvement of qubit representation of structures and components of digital systems based on address coding of input signals to increase the manufacturability and speed of faults simulation; development of the procedure for obtaining Boolean derivatives by permuting the lines of the truth tables (Q-vectors) and using the XOR operation; development of a data structure for deductive fault simulation based on the cubic representation of digital circuit components. The following methods are used: deductive, cubic, deductive-parallel simulation of faults, faults simulation by deductive Q-vectors. The following results were obtained: the equivalence of concurrent and sequential conditional operators, as well as their schematic implementation in the form of multiplexers, was shown; method of obtaining truth tables of the synthesized circuit structure using TestBench (Xilinx ISE) was proposed; different technologies and data structures of deductive fault simulation for tabular, analytical and qubit methods of digital circuits description were considered; the software implementation of faults cubic deductive simulation is considered and the equivalence of the obtained results for the multiplexers circuits (MUX 2-in-1 and MUX 4-in-1) using the DCP software product was demonstrated. Conclusions: a new Q-method of interpretative faults simulation of digital circuit is proposed, which is characterized by the usage of compact Q-vectors instead of truth tables, which makes it possible to significantly increase the analysis speed due to the addressable formation of the functional primitives outputs and reduce the volume of data structures, which practically makes the method competitive with compilative simulation technologies.

Mechanical fault diagnosis of gas-insulated switchgear based on saliency feature of auditory brainstem response under noise background

The mechanical fault of gas-insulated switchgear (GIS) seriously threatens the security of the power grid. Recently, acoustic-based fault diagnosis methods, which have the advantage of non-contact measurement, have been applied to the GIS mechanical fault diagnosis, but vulnerable to the interference of the background noise. To improve the capacity of the acoustic-based GIS fault diagnosis under noise background, by simulating the sound feature extraction ability and anti-noise ability of human auditory system, a novel GIS mechanical fault diagnosis method based on saliency feature of auditory brainstem response (SFABR) is proposed. First, an auditory saliency model, which considers both the auditory periphery and the auditory nerve center was constructed by combining the deep auditory model and the saliency model. After processing GIS emitted acoustic signal, the auditory brainstem response (ABR) was obtained, and the saliency features of the ABR were extracted to obtain the SFABR. Then, the characteristic frequency distribution of the auditory saliency model was adjusted to make it more suitable for the spectral characteristics of the GIS sound signal. Finally, the SFABR was mapped to a two-dimensional CNN to train a model for GIS mechanical fault diagnosis. This method simulates the process of auditory response extraction and the selection effect of auditory attention on sound elements. The 110 kV three-phase GIS fault simulation experiment shows that for GIS mechanical faults, the diagnosis method based on SFABR can obtain 96.1% fault identification accuracy. In different noise environments, compared with the traditional acoustic-based fault diagnosis methods, this method has stronger anti-noise performance, and can more effectively realize the identification of GIS mechanical failure types. In future research, the method can be further extended to fault diagnosis of more types of power equipment.

Few-shot fault identification of complex equipment via metric-based features capture GAN combining prior knowledge-augmented strategy

In engineering, complex equipment during aerospace engine manufacturing fails to carry out large-scale fault simulations, which leads to limited fault signals collected, resulting in poor efforts at intelligent fault identification for equipment. In existing research, generative adversarial network (GAN) is considered to be an efficient solution to such issues. However, few-shot data and conventional generation mechanisms may lead to time-consuming training and subpar quality of the generated samples. To overcome these challenges, we propose metric-based features capture GAN combining prior knowledge-augmented strategy that applies a new generation mechanism. The network is made up of a feature sampler, a generator, a discriminator, and an assistant classifier. The feature sampler provides the potential features of the few-shot data to assist in the training of the generator, which can significantly reduce the learning difficulty of the network. Furthermore, to accelerate model convergence, a feature metric module is inserted into the generator to compel the model to achieve knowledge transfer. By conducting experimental analysis on three bearing failure datasets, the proposed scheme has been confirmed to be efficacious and superior. In addition, ablation experiments indicate that the proposed scheme is valid for reducing the time-consuming training of GAN.