Convolutional transformer approach for engine spark plug fault diagnosis using acoustic signal
Detecting and rectifying spark plug faults are pivotal in preventing engine-related issues that can have substantial operational and financial consequences. To improve the accuracy and robustness of spark plug fault diagnosis, this research introduces a novel Convolutional Transformer approach that leverages the strengths of Convolutional Neural Networks and Transformers, which effectively capture both local and extended temporal dependencies within spark plug acoustic signals. The results of this groundbreaking approach, as presented in accompanying tables and figures, demonstrate its superior performance, achieving an impressive 97.1% accuracy in a challenging 4-class classification scenario using solely acoustic signals. This achievement signifies a significant advancement in spark plug fault detection, potentially ushering in more reliable and precise diagnostic methods, ultimately contributing to the prevention of costly engine breakdowns and the extension of engine lifespan. Deep learning techniques such as Convolutional Transformers offer a promising way to improve the reliability and performance of internal combustion engines as the automotive industry continues to evolve, highlighting the importance of this research for future automotive developments.