A lightweight deep learning-based method for health diagnosis of internal combustion engines on an internet of vehicles platformQuanli Dou, Hanbin Luo, Zhenjing Zhang, Yedong Song, Shilong Chu, Zhiwei Mao
- Mechanical Engineering
- Aerospace Engineering
The health status diagnosis method for internal combustion engines based on deep learning mainly focuses on the research of vibration signals, but the hardware cost required for vibration signal monitoring is expensive, and the model is also complex, which is not suitable for vehicle internal combustion engines. The Internet of Vehicle (IoV) platform provides a lot of thermal parameter data that can reflect the status and performance of internal combustion engines. However, there is currently insufficient study on thermal parameters, and the in-depth fusion analysis of multiple parameter associations has not been realized. At the same time, considering the numerous thermal parameters of internal combustion engines and the complex relationship between them. This article proposes a deep learning lightweight diagnosis method based on thermal parameters to explore the important value of thermal parameters in the health diagnosis of internal combustion engines. Firstly, a parameter grouping model based on Mutual information is proposed to realize the automatic grouping of parameters, reduce the complexity of the model, and realize lightweight processing. Then, based on grouping, a deep learning health status diagnosis model based on denoising autoencoder-attention mechanism-bidirectional gated recurrent unit (DAE-AM-BiGRU) is proposed to achieve the purpose of data dimensionality reduction, noise reduction, and obtaining key features and timing relations. Finally, a simulation model of an internal combustion engine was constructed using GT-POWER simulation software to obtain simulation data of thermal parameters, verifying the effectiveness of the proposed method.