AI-augmented optimization of electric machine parameters via permutation importance and dynamic sampling: a motor study

Rapid Machine Learning–Driven Modeling for Large-Scale Validation and Optimization of Control Variables in Wireless Power Transfer Systems

Validating wireless power transfer (WPT) systems for electric vehicles (EVs) is a challenge due to efficiency variations caused by coil misalignments and height differences arising from various vehicle designs. Traditional simulation methods, such as finite element analysis (FEM), provide high accuracy but entail significant computational costs and calculation times, limiting the number of case studies and their optimization. This paper presents a methodology that integrates Machine Learning (ML) and Genetic Algorithms (GA) to overcome these limitations. An ML model rapidly and accurately predicts key electromagnetic parameters across a wide range of positions and frequencies. These predictions feed into a GA that optimizes control variables (voltages and frequency) with the objective of maximizing power transfer efficiency, while simultaneously ensuring component integrity at each operating point. Beyond drastically reducing simulation time and experimental effort, this methodology will enable knowledge extraction and its use for formulating design rules. These rules can lay the groundwork for developing simplified, real-time adaptive control strategies, facilitating the reduction of control variables and the narrowing of search ranges.

An integrated optimization method for multi-class parameters and multi-dimensional objectives: Application to an advanced transmission

Current transmission design methodologies often optimize single parameters or physical fields, while conventional heuristic algorithms suffer from premature convergence and lack interpretability. To address these limitations, this paper proposes an integrated optimization design method for planetary gear sets and wet clutches. The approach leverages multi-physics simulations to extract evaluation metrics, constructing a comprehensive evaluation index via radar charts and the Analytic Hierarchy Process (AHP). A neural network-based surrogate model is then employed to predict the global solution space and identify optimal designs efficiently. Applied to an advanced transmission system, this method evaluates feasible candidates through integrated system simulation, 3D modeling, and finite element analysis. Results demonstrate significant improvements, increasing the comprehensive evaluation indices of the planetary gear set and wet clutch by 32.9% and 23.8%, respectively. This study offers a practical reference for efficient multi-parameter, multi-objective optimization of complex transmission components.

Robust multi-objective optimization of helical gears for electric vehicle transmissions

An inverse design method for electric machines based on conditional variational autoencoder

Electric machines play a key role in electromechanical energy conversion across modern industrial systems. Conventional design methodologies typically involve many repetitive tasks that exhibit low computational efficiency. To address this challenge, this paper proposes an inverse design system based on conditional variational autoencoder (CVAE), which establishes a direct mapping relationship between predefined performance indicators and optimal structural parameters. The main contributions are threefold: First, a proportional modeling approach is conducted that ensures the inverse design outcomes are free from geometric interferences. Second, a surrogate model is integrated to enhance data quality, combined with a tailored loss function that simultaneously optimizes predictive accuracy and solution diversity. Third, an algorithm is implemented to satisfy nonlinear inequality constraints in practical design scenarios. The test results demonstrate that the proposed system can efficiently generate diverse machine designs that meet specified key performance indicators (KPIs), validating its effectiveness in accelerating electric machine design.

Multi-objective and cross-scale inverse design of temperature-control materials via physics-constrained machine learning

Temperature-control materials, notably composite phase-change materials (CPCMs), show great potential for the thermal management of next-generation power electronics. However, the synergistic optimization of multiple performance metrics still relies heavily on Edisonian trial-and-error experimentation. Herein, we present an artificial intelligence framework that incorporates a physics-constrained inverse-design system (PHICS) built upon a directed acyclic graph (DAG) architecture for CPCMs, integrating interface, phase, and carrier engineering. By encoding structural hierarchies and physical causality through DAG, PHICS framework couples forward predictive modeling with a diversity-enhanced NSGA-II optimizer to efficiently map Pareto-optimal design boundaries. Guided by these predictions, we successfully fabricate a high-performance CPCM composed of an oriented graphite fiber skeleton and an n-octacosane matrix with amorphous alumina (am-Al2O3) interfacial transition layers. Benefiting from the bifunctional role of the am-Al2O3 interlayer as both an interfacial phonon bridge and electron barrier, the resulting CPCMs achieve a superior balance of thermal conduction, thermal storage, and electrical insulation. These results demonstrate the accuracy of PHICS-guided multifunctional composite design, establishing a closed-loop platform that combines physics-constrained machine learning with experimental validation to solve key thermal–electrical trade-offs.

Robust Optimization of an Electromechanical Linear Actuator Under Experimental-Budget Constraints

The design optimization of electromechanical linear actuators (EMLAs) is challenged by coupled tribological, thermal, and dynamic effects. Such interactions can violate the additivity assumptions of compact orthogonal designs. Factor ranking may become unreliable when residual variance includes non-additive contributions rather than stochastic noise alone. This study presents an Extended Orthogonal Experimental Matrix (E-OEM) framework for robust factor screening under experimental-budget constraints. It preserves the reduced effort of an orthogonal design while using a latent allocation term to retain structured non-additive variation that would otherwise be treated as residual error. It combines compact experimentation, fitted probability distributions, Monte Carlo simulation, and weighted multi-criteria ranking under push force, robustness, and cost targets. The framework is demonstrated on 90 automotive EMLAs considering lubricant type, drive frequency, and operating temperature. E-OEM was benchmarked against the full-factorial experiment (27 combinations, 810 push-force measurements). Under the selected weighting strategy, both E-OEM and the benchmark identified the same nominal optimum (590 Hz, +25 °C, Addinol lubricant), with a measured push force of 77.32 ± 2.05 N (1.01% deviation from the prediction). The results show that E-OEM provides an efficient screening and decision-support approach for actuator design when full-factorial testing is constrained by time and cost.