Integration of Innovative Paths for Permanent Magnet Motors in Electric Vehicles: Intelligent Control, Proactive Diagnosis, and Collaborative Design

1. Introduction

Today, in the face of the urgent need to decarbonize global transportation, the development of electric vehicles (EVs) has become one of the core strategies to address energy crises and environmental challenges. As the “heart” of an EV’s drive system, the performance of the traction motor directly determines the vehicle’s power, efficiency, and driving range. Among various technical routes, the Permanent Magnet Synchronous Motor (PMSM) has emerged as a preferred technology for drive systems due to its high power density, high torque density, high efficiency, and excellent speed regulation performance. However, its complex intrinsic electromagnetic characteristics, harsh external operating conditions, and the pressing demands for high reliability and a wide speed range make research in fields such as drive control, fault diagnosis, and multidisciplinary optimization design critically important, as these directly impact the vehicle’s performance, safety, and lifespan [1,2].

With the deepening of research, the performance limits of PMSMs are continuously being pushed through the introduction of advanced control strategies. Traditional vector control and direct torque control frameworks are being deeply integrated with intelligent algorithms. Model Predictive Control, with its superior dynamic response capability and flexibility in handling multiple constraints, shows great potential in torque and flux linkage control [3]. Furthermore, to achieve efficient and stable operation across the entire speed range, deep flux-weakening control algorithms, robust control considering parameter variations (such as the decay of permanent magnet flux linkage with temperature), and adaptive observer technologies have become research hotspots, aiming to enhance the system’s anti-interference capability and efficiency [4].

Parallel to high-performance control is the growing demand for condition monitoring and fault diagnosis. Potential failure modes in PMSMs, such as local/global demagnetization of permanent magnets, inter-turn short circuits in windings, bearing damage, and rotor eccentricity, pose major threats to their reliability. Current research focuses on the integration of model-based diagnostic methods (such as sliding mode observers and extended Kalman filters) and data-driven methods (such as wavelet packet analysis and deep learning) [5]. By analyzing and extracting features from multi-source signals like motor current, vibration, and noise in real-time, these methods enable precise diagnosis and localization of early-stage, subtle faults, paving the way for predictive maintenance and effectively preventing catastrophic failures.

In terms of the motor’s physical optimization design, the optimization objectives have shifted from solely maximizing electromagnetic performance to the collaborative optimization of multiple physical fields, including electromagnetics, thermal, mechanical, and materials. Using Finite Element Analysis as a core tool, researchers conduct detailed simulations and designs of the motor’s nonlinear characteristics, topology (e.g., rotor flux barrier shape, permanent magnet segmentation and skew), and material selection (e.g., low iron loss silicon steel sheets, high-temperature resistant permanent magnets) [6]. To balance conflicting design objectives—such as suppressing torque ripple and noise while maximizing output torque and minimizing permanent magnet usage—a variety of optimization algorithms, from response surface methodology to multi-objective genetic algorithms, are widely employed [7]. Multi-criteria decision-making methods are also applied to compare and select different motor technology paths to achieve a balance among multiple objectives [8].

Despite the significant progress made in PMSM research, challenges still remain. A trade-off exists between the real-time performance of control strategies and their computational complexity; the online deployment of fault diagnosis systems and the issue of high false alarm rates need urgent solutions; the high computational cost associated with multi-physics, multi-objective optimization limits its application in rapid product development [9]. In the future, the deep integration of drive control, fault diagnosis, and optimization design will be a key trend. A framework based on Digital Twins, interacting high-fidelity models with real-time data, holds the promise of achieving a closed loop from performance prediction in the design phase to health management and adaptive control during operation. This will ultimately propel automotive PMSM systems toward higher efficiency, greater reliability, and enhanced intelligence [10].

This Editorial refers to the Special Issue “Permanent Magnet Motors and Drive Control for Electric Vehicles”. The Special Issue aims to showcase the latest research achievements from scholars in the field of permanent magnet motors, particularly those related to PMSM and drive control applied in EVs.

A total of 12 papers (including 1 review paper) were finally accepted for publication and inclusion in this Special Issue. The contributions are listed below.

All the papers included in this Special Issue collectively reveal a core trend: the development of permanent magnet motor technology is shifting from isolated, point-specific advancements towards deep integration and convergence across three foundational pillars: Intelligent Control, Proactive Diagnostics, and Multiphysics-Aware Co-Design.

Section 2 categorizes these papers into three interrelated themes and provides a brief overview of each, thereby helping the reader better explore and utilize them.

2. Overview of Contribution

2.1. Towards Intelligent and Full-Speed-Range Sensorless Advanced Control

Research in motor control is evolving from single-algorithm improvements towards intelligent optimization of the entire “sensing-decision-execution” chain.

In high-performance sensorless control, contribution 1 innovated by proposing an Exponential Moving Average (EMA)-based filter structure to replace the traditional Band-Pass Filters (BPF) and Low-Pass Filters (LPF) used in high-frequency injection methods. Due to its computational simplicity and low phase delay, this scheme enables rapid extraction of rotor position error under demanding dynamic conditions like motor direction reversal, step speed changes, and load transients, significantly reducing convergence time. Addressing PMSM nonlinearity, parameter time variance, and external disturbances, contribution 2 developed a Prescribed Performance Model-Free Adaptive Fast Integral Terminal Sliding Mode Control (PP-MFA-FITSMC). This method establishes a data model via Compact Form Dynamic Linearization (CFDL) and utilizes a Discrete Small-Gain Extended State Observer for online estimation of lumped disturbances. Consequently, it achieves finite-time convergence of the tracking error, constraining it strictly within ±0.0028 rad, without requiring a precise physical model. Simultaneously, AI is deeply integrating with traditional control. Contribution 3 designed a tiny neural network named TinyFC, not to replace but to complement the traditional PI controller. This network learns nonlinear dynamic relationships within the control loops and generates compensatory signals. Crucially, the study applied intensive lightweight processing to the TinyFC model, including hyperparameter tuning, pruning, and 8-bit quantization, significantly reducing computational overhead while maintaining accuracy. Simulations successfully demonstrated the elimination of overshoot in the speed control unit, proving its feasibility for deployment on edge computing devices. Furthermore, for novel topology control, contribution 4 proposed a Bilateral Cooperation Control Strategy for a new Brushless Excitation (MCR-BE) system. This strategy employs phase-shift modulation on the transmitter-side inverter and impedance matching via a Buck converter on the receiver-side to achieve constant voltage output and maximum efficiency tracking, respectively. The adopted Gray Wolf Optimizer-Particle Swarm Optimization (GWO-PSO) hybrid algorithm ensures accurate real-time mutual inductance identification, providing a new approach for controlling motors with special topologies. Particularly noteworthy is the comprehensive review by contribution 5 on position sensorless technology, which systematically compares various methods based on models and saliency, providing a clear technical roadmap for researchers and laying a theoretical foundation for continued innovation in the field.

2.2. High-Precision Parameter Identification and Fault Diagnosis

Based on Intelligent Algorithms Enhancing system reliability and fault tolerance is another key focus of this Special Issue.

In parameter identification, contribution 6 addressed the issues of insufficient rank in traditional multi-parameter identification equations and the tendency to fall into local optima by introducing the Bacterial Foraging Optimization Algorithm (BFOA). Their core innovation was constructing a completely new set of four-equation fitness functions, instead of relying on traditional voltage equations. This enables effective compensation for distortions in four key parameters simultaneously: stator resistance, d-axis inductance, q-axis inductance, and permanent magnet flux linkage, significantly reducing d/q-axis current deviation and enhancing the robustness of the predictive control system. In the field of fault diagnosis, research shows a trend from “shallow” signal analysis towards “deep” feature mining. Contribution 7 pioneered the use of branch current, rather than three-phase current, as the diagnostic data source to enhance fault feature saliency. They constructed a dual-modal feature extraction module, utilizing Variational Mode Decomposition (VMD) and Fast Fourier Transform (FFT) for feature extraction and fusion in the time and frequency domains, respectively, achieving a nonlinear coupling of time-frequency features. These features were then fed into a meticulously designed Cascaded Convolutional Neural Network (Cascaded CNN) for diagnosis. This network incorporates dilated convolutional layers and multi-scale convolutional layers to enhance feature extraction capability, ultimately achieving a diagnostic accuracy of 98.6% with a very low misjudgment rate. Similarly, contribution 8 combined Wavelet Packet Transform (WPT) energy feature extraction with a Genetic Algorithm (GA)-optimized Back Propagation (BP) Neural Network to classify multiple common faults (e.g., turn-to-turn short circuit, inter-phase short circuit, demagnetization, inverter open-circuit, rotor eccentricity). The introduction of GA effectively optimized the initial weights and thresholds of the BP network, avoiding local optima, and finally achieved 100% accuracy in diagnosing normal operation, inverter open-circuit, and demagnetization faults. In system optimization design, contribution 9 established a complete automated design framework of “Parametric FEA—Surrogate Model—NSGA-II”. Using Latin Hypercube Sampling and Kriging surrogate modeling, they reduced electromagnetic performance evaluation time from days to seconds, making multi-objective optimization based on genetic algorithms feasible in engineering practice. The optimized motor design achieved an excellent balance between efficiency, torque density, and cost.

2.3. Motor Design and Hardware Co-Design for Performance and Reliability

Superior system performance is ultimately rooted in precise motor design and reliable hardware implementation.

Contribution 10, focusing on increasing motor power density, conducted in-depth analytical modeling of the Halbach Array PMSM. Their model accurately calculates key parameters like air-gap flux density, flux linkage, and back-EMF, showing strong agreement with Finite Element Method (FEM) results (error ~1.6%). Using this model, they systematically analyzed the influence of slot opening width, magnetization angle, and main magnetic pole width on electromagnetic performance. Its generality allows for rapid parameter adjustment in FEM software Maxwell2015 without remodeling, greatly facilitating early-stage optimization. Contribution 11, focusing on Noise, Vibration and Harshness (NVH) performance, proposed an electromagnetic noise prediction method based on the coupling of electromagnetic force and structural modes. By analyzing the spatial order and frequency characteristics of the electromagnetic force in a 6-pole 36-slot PMSM and comparing them with the modal frequency array of various orders, they successfully predicted the main sources of vibration and noise. The effectiveness of this analytical method was finally verified through multiphysics simulation. Finally, contribution 12, through their improved gate driver circuit, ensured the reliability of the entire control system at the hardware level. The study thoroughly analyzed the issues of switching ringing and signal spikes caused by parasitic elements in Insulated Gate Bipolar Transistor (IGBT)-driven inverters. By improving the gate driver circuit topology and adding a ringing suppression circuit, they successfully limited the signal spike in the motor current to less than 10%. This not only prevents potential switching losses and device stress but, more importantly, provides pure and reliable current feedback signals for sensorless control algorithms, forming the physical foundation for high-performance control.

3. Conclusions

The research presented in this Special Issue clearly indicates that PMSM technology is undergoing a profound paradigm shift. Future progress will no longer depend on linear advances in control, diagnostics, or design alone, but rather on the deep, systematic integration of these domains. A PMSM underpinned by Multiphysics Co-Design, precisely driven by Intelligent Adaptive Algorithms, and continuously safeguarded by Embedded Intelligent Diagnostic Systems, represents the ideal blueprint for the next generation of electric drives. We are moving towards an era of the “cognitive” e-drive, where the motor system is not merely an energy converter but an intelligent entity capable of self-perception, optimization, and decision-making. The work in this Special Issuespecial issue serves as a solid and encouraging milestone on the path to that future.