Diagnostic method for performance degradation mechanisms in automotive emission systems under complex operating conditions

Artificial intelligence–based signal processing establishes an effective computational framework for diagnosing performance degradation in automotive emission systems under complex operating conditions. Emission signals during cold start, high load, and regeneration exhibit strong non-stationarity and feature drift, limiting the applicability of traditional physical models. This study aims to construct an intelligent diagnostic method integrating multimodal feature fusion with domain-adaptive learning to achieve accurate and interpretable degradation-mechanism identification. Synchronous multi-channel acquisition and dynamic time–frequency analysis extract temporal, spectral, and transient-event features, which are encoded through a hybrid Transformer–Convolution architecture. Adaptive weighting and feature-space rearrangement strengthen cross-condition stability during multimodal fusion. Experimental validation shows that the proposed model improves recognition accuracy by 8.4% compared with conventional approaches and effectively suppresses inter-domain distribution shifts. A lightweight inference design supports millisecond-level edge deployment. Overall, integrating intelligent algorithms with domain-adaptive strategies enables precise, robust, and interpretable degradation diagnosis, offering a transferable and deployable computational solution for emission monitoring and intelligent vehicle maintenance. Keywords: Artificial Intelligence; Signal Process.