Physics-informed learning–based fault-tolerant control for robust wireless power transfer in electric vehicles

Saleem, Muhammad Haris; Amin, Arslan Ahmed; Alsuwian, Turki · 2026 · Crossref

DOI: 10.1038/s41598-026-57870-y

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the vulnerability of Wireless Power Transfer (WPT) systems in electric vehicles (EVs) to faults such as power converter failures and coil misalignment, which can severely degrade charging performance and safety. While WPT offers benefits like extended range and contactless charging, its complex architecture is prone to issues like inverter switch breakdowns and resonance mismatches. Existing control methods, including conventional observers and pure data-driven neural networks, often require precise mathematical models, large datasets, or lack physical constraints, leading to unrealistic responses or high computational overhead. To overcome these limitations, the authors propose a Physics-informed Neural Network (PINN)-based Active Fault-Tolerant Control (AFTC) framework that integrates physical laws into the learning process to ensure robust, real-time fault detection and compensation. The methodology employs a PINN observer with a six-layer Multi-layer Perceptron architecture (128 neurons per layer) and hyperbolic tangent activation functions. The network inputs include duty cycle, frequency, rectifier voltage, and time. Crucially, the PINN utilizes a composite loss function that combines data-driven mean squared error with a physics-based residual term derived from Kirchhoff’s Voltage Law (KVL) and Kirchhoff’s Current Law (KCL) equations for the series-series compensated WPT circuit. This ensures the model’s predictions adhere to electromagnetic principles. The system was trained using a dataset of 100,000 values generated in MATLAB/Simulink, covering normal operations, coil misalignment scenarios, and injected faults (open/short circuits, sensor noise). Lyapunov stability analysis was conducted to prove global asymptotic stability of the observer under disturbances. The results demonstrate that the proposed PINN-based AFTC achieves accurate current tracking and maintains residual signals within thresholds despite parameter variations and faults. The system exhibits a reduced settling time of 0.22 seconds and a steady-state error of 0.3 A, outperforming conventional control methods. The framework successfully distinguishes between multiple fault types, including inverter IGBT faults, sensor faults, and passive component issues, while enabling rapid resonant recovery through retuning logic. The integration of redundant switches allows for smooth vehicle operation during converter failures. The significance of this work lies in providing a resilient, data-driven control strategy that does not rely on extensive fault datasets or exact mathematical models, thereby enhancing the reliability of WPT-based EVs. By embedding physical constraints into the neural network, the approach prevents transient and unrealistic responses, ensuring stable power transfer even under severe faults like coil misalignment or switch failures. This contributes to the broader adoption of EV technology by mitigating range anxiety and safety concerns associated with wireless charging infrastructure.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• steering pattern