基于音频特征的拖拉机发动机状况识别系统设计

Design of a Tractor Engine Condition Recognition System Based on Audio Features

拖拉机发动机是保证拖拉机正常运行的关键部件,目前主要采用振动信号开展发动机故障预测与状况识别。为此,提出了一种基于GRU的循环神经网络模型,通过对拖拉机发动机在不同作业条件下产生的音频信号进行分析,提取Mel作为主要特征,构建基于音频特征的拖拉机发动机状况识别系统。预测结果表明:系统能够准确地识别发动机的正常运行状态和不同类型的故障状况,对拖拉机发动机异常的识别率可以达到97.15%。研究结果可以提高拖拉机的运行安全性和可靠性,减少故障停机时间,提高农业生产效率。

The tractor engine is a key component to ensure the normal operation of the tractor.At present,vibration signals are mainly used for engine fault prediction and condition identification.This study proposes a GRU-based recurrent neural network model to build a tractor engine condition recognition system based on audio features by analysing the audio signals generated by the tractor engine under different operating conditions and extracting Mel as the main feature.The prediction results show that the system can accurately identify the normal operation status of the engine and different types of fault conditions,and the recognition rate of abnormal tractor engine can reach 97.15%.The results can be used to help improve the safety and reliability of tractor operation,reduce downtime and improve agricultural production efficiency.