Back Propagation Neural Network-Based Fault Diagnosis and Fault Tolerant Control of Distributed Drive Electric Vehicles Based on Sliding Mode Control-Based Direct Yaw Moment Control

Tianang Sun, Pak Kin Wong, Xiaozheng WangVehicles 2023 · DOI

TL;DR

This paper presents a novel fault-tolerant control system for distributed-drive electric vehicles (DDEVs) that combines a Back Propagation Neural Network (BPNN) for motor fault diagnosis with a Sliding Mode Control (SMC)-based Direct Yaw Moment Control (DYC) for fault tolerance. This integrated approach aims to maintain vehicle stability and safety, especially on slippery roads, by accurately detecting motor faults and dynamically adjusting torque distribution.

What Problem Does This Paper Address?

Distributed-drive electric vehicles (DDEVs), with their independent wheel-hub motors, offer advantages in control and efficiency. However, these motors operate under harsh conditions, making them prone to various faults (e.g., short circuits, open circuits). Such faults can significantly impact vehicle stability, particularly on low-friction surfaces like ice or wet roads, potentially leading to accidents. Existing fault diagnosis methods often struggle with the complex, time-varying parameters of Permanent Magnet Synchronous Motors (PMSMs) used in DDEVs. Furthermore, effectively compensating for these faults to ensure vehicle safety and control requires a robust fault-tolerant control strategy. This paper addresses the critical need for a reliable system that can diagnose motor faults in real-time and implement effective control actions to maintain vehicle stability.

Methods

The proposed system comprises two main components:

  • BPNN-Based Motor Fault Diagnosis: Instead of relying on complex physical models of PMSMs, which are difficult to parameterize accurately, the authors employ a model-free Back Propagation Neural Network (BPNN). This BPNN is trained offline using simulated motor data. It takes rotor speed and phase currents as input and, after processing through Fast Fourier Transform (FFT) to analyze frequency domain characteristics, outputs fault parameters. The system categorizes faults into three types: no fault, short-circuit fault, and open-circuit fault.
  • SMC-Based Direct Yaw Moment Control (DYC) for Fault Tolerance: A two-layer DYC controller is utilized. The upper layer, employing Sliding Mode Control (SMC), calculates the necessary yaw moment to maintain vehicle stability. SMC is chosen for its robustness to parameter variations, fast response, and insensitivity to disturbances. The lower layer then distributes the required torque among the healthy motors to achieve the desired yaw moment, compensating for the reduced or lost torque from the faulted motor. The system considers only single-motor faults for fault tolerance, as multiple simultaneous faults are deemed less probable and more dangerous to manage.

Datasets

  • Training and Simulation Data: The study utilizes simulated data generated from MATLAB/Simulink. Motor models for healthy and faulted PMSMs (turn-to-turn short circuits, open-phase faults) were constructed. Vehicle dynamics were modeled using a 2-DOF and a 7-DOF model.
  • Real-world Data (Implicit): While the paper mentions that motor parameters were collected from a real PMSM, the training and simulation primarily rely on these simulated models.

Key Insights and Ideas

  • Model-Free Diagnosis for PMSMs: The use of a BPNN for fault diagnosis bypasses the significant challenge of accurately modeling and identifying the time-varying parameters of PMSMs in DDEVs, offering a practical alternative.
  • Integrated Fault Diagnosis and Control: The paper highlights the importance of a tightly integrated system where accurate fault diagnosis directly informs the fault-tolerant control strategy, leading to more effective stability management.
  • SMC for Robust Yaw Control: The application of SMC in the DYC controller demonstrates its effectiveness in handling uncertainties and disturbances, crucial for maintaining vehicle stability on slippery surfaces.
  • Focus on Critical Faults: By focusing on common and impactful fault types (short-circuit, open-circuit) and considering the practical implications of multiple faults, the research provides a targeted and relevant solution.

Key Results

  • Fault Detector Accuracy: The BPNN-based fault detector achieved high accuracies in simulations: 97.5% for no fault, 98.75% for open-circuit faults, and 95.00% for short-circuit faults.
  • Improved Vehicle Stability: Simulations on slippery road conditions (straight-lane and single-lane change) showed that the proposed fault-tolerant controller significantly improved vehicle stability.
  • Trajectory Improvement: The average trajectory offset (ATO) was substantially reduced with the proposed system. For straight-line driving, improvements ranged from 73.272% to 82.201% compared to vehicles without the fault-tolerant system. For single-lane change conditions, improvements ranged from 37.152% to 44.663%.
  • Best Performance in Short-Circuit Faults: The system demonstrated its best performance in mitigating the effects of short-circuit faults.

Limitations and Future Work

  • Limited Fault Scenarios: The study primarily focuses on single-motor faults (short-circuit, open-circuit, demagnetization treated as shutdown). Future work should explore more complex fault scenarios, including multiple simultaneous faults.
  • Data Requirements for BPNN: BPNNs require substantial amounts of data for high accuracy, which can be time-consuming and computationally intensive. The paper acknowledges that computational speed decreases with large datasets, and accuracy can be affected by high-dimensional input variables.
  • Simplified Road Surface Model: The simulation used a simplified friction model for wet and slippery roads. More complex and realistic tire-road interaction models should be investigated.
  • Future Directions:
    • Testing under a wider range of driving conditions.
  • Collecting and comparing more data from real-world motor faults to improve the BPNN.
  • Exploring advanced machine learning approaches for fault detection.
  • Investigating other signal-processing methods for input data.
  • Considering more comprehensive road conditions and tire models.

Summary updated

Abstract

Distributed-drive vehicles utilize independent drive motors on the four-wheel hubs. The working conditions of the wheel-hub motors are so harsh that the motors are prone to failing under different driving conditions. This study addresses the impact of drive motor faults on vehicle performance, particularly on slippery roads where sudden faults can lead to accidents. A fault-tolerant control system integrating motor fault diagnosis and a direct yaw moment control (DYC) based fault-tolerant controller are proposed to ensure the stability of the vehicle during various motor faults. Due to the difficulty of identifying the parameters of the popular permanent magnet synchronous wheel hub motors (PMSMs), the system employs a model-free backpropagation neural network (BPNN)-based fault detector. Turn-to-turn short circuits, open-phase faults, and diamagnetic faults are considered in this research. The fault detector is trained offline and utilizes rotor speed and phase currents for online fault detection. The system assigns the torque outputs from both healthy and faulted motors based on fault categories using sliding mode control (SMC)-based DYC. Simulations with four-wheel electric vehicle models demonstrate the accuracy of the fault detector and the effectiveness of the fault-tolerant controller. The proposed system is prospective and has potential for the development of distributed electric vehicles.
BibTeX
@inproceedings{sun2023back,
  author = {Tianang Sun and Pak Kin Wong and Xiaozheng Wang},
  title = {{Back Propagation Neural Network-Based Fault Diagnosis and Fault Tolerant Control of Distributed Drive Electric Vehicles Based on Sliding Mode Control-Based Direct Yaw Moment Control}},
  booktitle = {Vehicles},
  year = {2023},
  doi = {10.3390/vehicles6010004},
}
Figure