Special Issue on Intelligent Diagnostic and Prognostic Methods for Electronic Systems and Mechanical Systems

Fault diagnoses and prognostics are important tools to improve system reliability. Quickly locating faults can reduce troubleshooting time, while predicting faults in advance can effectively reduce system downtime, which plays a pivotal role in the industry. Modern electronic and mechanical systems are becoming larger and more complex in internal structure, but fewer and fewer parameters are available for testing, which increases the difficulty of fault location and raises the cost of diagnosis. It is urgent to study new fault diagnosis methods and prediction methods.

This Special Issue will collect and present important research on intelligent diagnosis and prognostic methods, including complex electronic and mechanical systems such as analog circuits, lithium batteries, and gears. The research includes, but is not limited to, fault feature extraction, diagnostic reasoning methods, performance degradation, and life prediction.

This Special Issue contains 10 papers covering various areas of fault diagnoses and prognostics, including the generation of excitation signals for fault detection circuits, estimations of weak fault signature signals, fault diagnosis methods, and fault prognostic methods. Wang, Z. et al. [1] investigated an intelligent detection method for electrical connection faults in RF circuits, using three machine learning methods—support vector machine (SVM), logistic regression (LR), and gradient boosting decision tree (GBDT)—for a filter circuit. Chen, J. et al. [2] added uncertainties to the health assessment model of the flight control system and established a health assessment model of the flight control system under uncertainty conditions. The influence of outliers on the assessment results was reduced and the false alarm rate was decreased. Wang, C. et al. [3] proposed an Elman neural network based on GA (genetic algorithm) optimization for landslide displacement prediction and validated it on the displacement monitoring data of Jianshanying landslides in the rainy season in 2020. Long, B. et al. [4] applied the GRU model to hydrogen fuel cell life prediction and compared it with BPNN and LSTM. Liu, Z. et al. [5] used a square wave as the excitation of the circuit under test, then processed the raw data using improved empirical wavelet variations, and imputed it to a multi-input deep residual network to achieve the fault diagnosis of analog circuits. Wang, L. et al. [6] proposed a method to generate high-resolution and high-speed test signals without a suitable high-resolution DAC. This signal can be used as an excitation signal for fault diagnosis circuits. Qiu, G. et al. [7] designed a robust accuracy weighted random forest online fault diagnosis model to locate open-circuit faults in IGBTs of three-phase pulse-width-modulated converters. He, P. et al. [8] proposed a new method for shared state estimations of redundant switching power supplies, a low-cost method based on LSTM recurrent neural networks, which can estimate the output current of each power supply branch by simply detecting the voltage ripple at the switching frequency of the load side. Wang, Y. et al. [9] proposed a DFT-based accurate signal frequency estimation method to solve the problem of weak fault characteristic signals. Wang, H. et al. [10] proposed an MC penalty-function-based sparse decomposition shock vibration signal identification method from the sparse characteristics of the vibration shock signal for the diagnosis of bearing faults in rotor systems.

Although submissions to this Special Issue have ended, scholars will not stop their research on intelligent fault diagnosis and prognostic methods for electronic and mechanical systems, and intelligent fault diagnosis and prognostic methods will play an increasing role in the industry.