Enhancing Electric Vehicle Diagnostics Through Constant Speed Subrange Detection for Noise-Reduced Analysis

Abstract

Electric Vehicles (EVs) represent a paradigm shift in the automotive industry, offering a sustainable alternative to traditional engines. However, the efficient diagnosis of EV performance remains a formidable challenge, exacerbated by the intrinsic noise interference associated with speed variations. This paper introduces an innovative approach to mitigate these diagnostic challenges by harnessing constant speed subranges as a strategic diagnostic window. The methodology centers on the identification of the most frequent constant speed subranges, which serve as optimal periods for signal recording from EV electric motors. We propose a novel algorithm that identifies these subranges, initially trained, tested, and validated using the Worldwide Harmonized Light Vehicles Test Cycle (WLTC) dataset, the model is compared with four other models and show less variation ratio in the detected subranges. The crux of this approach lies in the recording and analysis of key motor signals - current, vibration, among others - within these identified subranges. The data derived from these periods are then utilized to train diagnostic algorithms, specifically tuned to the nuances of each subrange. The implementation of this strategy in real-world driving conditions involves the dynamic selection of appropriate diagnostic models aligned with the detected speed subranges. This targeted approach diminishes the impact of noise interference and paves the way for personalized data-driven diagnostics. The proposed method aims to reduces the complexity and enhances the efficacy of diagnostic algorithms, offering a tailored diagnostic solution that adapts to individual driving patterns.