With the growing popularity of Permanent Magnet Synchronous Motors (PMSMs) in industrial and vehicular applications, a deeper analysis of their behavior under fault conditions has become an increasingly important field of study. In this context, the present work aims to present a methodology for the electromagnetic simulation of a PMSM operating under both healthy conditions and under Inter-Turn Short-Circuit (ITSC) fault, based on the solution of the potential equation corrected by a complex relative permeance function. All proposed analytical solutions are validated through comparison with finite element simulations. The main contributions of this work are a low computational cost solution of the magnetic field distributions when the machine operates under healthy conditions and under different ITSC fault levels, as well as a simple and alternative method to model the permeance effect caused by the stator slots.

This paper aims to compare the use of different signals, time-series decomposition methods, and classification algorithms for fault detection in rolling bearings. To reach this objective, time series extracted from vibration and electrical current data from real bearings were analyzed. Empirical Mode Decomposition (EMD) and Singular Spectrum Analysis (SSA) were used to decompose the time series, and features were extracted from the components and the original series using descriptive statistics. Then Support Vector Machines (SVM), Random Forest (RF), and a Neural Network (NN) were used to detect the faults. The results indicate that decomposing the time series can improve fault detection accuracy and that vibration signals yield better results. The highest accuracy (99.3%) was obtained in the combination of SSA decomposition with the NN classifier.

In recent years, Permanent Magnet Synchronous Motors (PMSM) have been gaining popularity in the industry, especially in applications related to electric mobility, due to the high torque for a wide speed range, high power density and high efficiency. In this context, the PSMS are subjected to very stringent requirements of reliability and safety which also demand accurate identification of faults for proper treatment and protection of electrical machine itself, surrouding area and interfaces and, finally, the safety of vehicle and occupants. One of the main types of techniques used for fault detection and diagnosis in electrical machines is based on mathematical and multi-physics models. These types of algorithms have the important advantage of not requiring as much training data as machine learning-based techniques and allow for a more direct relationship between diagnosis and the physical components of the machine. Thus, given the importance of studying failures in PMSMs and the use of mathematical and multi-physics models for this purpose, this work conducts a brief literature review on the use of parametric or non-parametric techniques of the motor to extract fault characteristics and perform detection and diagnosis.

Anomaly or fault detection on machines is relevant in many industries, like wind power generation. Bearing failures, a common cause of breakdowns, pose significant challenges, leading to increased downtime and maintenance costs. Recent machine learning techniques, such as convolutional neural networks (CNNs) and long short-term memory (LSTM), have been exploited to extract spatial and temporal relationships from data to detect failures. Many literature works use the Paderborn bearing vibration data in their studies, which mainly rely on either CNNs or LSTM models to accomplish the detection task. In this paper, we use a hybrid CNN-LSTM network architecture to achieve high accuracy for bearing fault detection using the Paderborn dataset. In addition, a two-sliding window technique is used to prepare the data to pass through both types of layers. Our study emphasizes the importance of selecting appropriate architectures and considering bearing characteristics and operation conditions for effective fault detection. Our results achieved up to 99.58% accuracy. However, when tested on different machine operating conditions, the model’s performance degraded, indicating the necessity of other methods to address this issue. In addition, we verified that an improper model tunning can degrade its performance, even if a complex hybrid method is considered, highlighting the importance of hyperparameters tunning in fault detection problems.

The development of induction motor models is a key theme for industrial development, due to the growing interest in vector control and in condition monitoring. In this context, subspace methods for systems identification are efficient tools to find minimum realizations of dynamic systems, using only experimental data. This article presents an application of the combined deterministic-stochastic method for determining states and matrices of a multivariable linear state space model, that has the three-phase voltages and currents of an induction motor as inputs and outputs, respectively. The identification algorithm was applied to data from healthy motors and motors with broken rotor bars. Then, the Markov Parameters from each model were used to train a k-nearest neighbors (k-NN) classifier to detect the fault occurrence.

In this paper, a new method to solve the time series realization problem is presented. This method is based on the statement of this problem as an optimization problem and on the development of an immuno inspired optimization algorithm dedicated to solve it. To evaluate the quality of the results obtained with the proposed method, they are compared to the results of two other known methods. The comparisons show that the proposed method presents a significantly better performance.

In this paper the Aoki method is used to identify a multivariable time series. During the identification procedure, the algebraic Riccati equation is solved using an immuno-inspired algorithm developed by the authors. The algorithm computational aspects are shown and the input parameters importance is highlighted. With the results obtained from algebraic Riccati equation resolution, a time series was generated using the model. This time series has agreed with the original one. The main objectives of this work are to focus on the new proposed approach computational aspects and to highlight the parameters choice importance.