Aiming at the shortcomings of existing agricultural machinery fault detection algorithms, such as low detection rate and poor classification accuracy of different faults, a detection scheme based on optimized ResNet model was designed. Firstly, high-precision sensors are arranged to collect the working state data set of agricultural machinery dynamically, and the dual denoising scheme of sample entropy and wavelet threshold is adopted to denoise the fault set to reduce the noise interference better. The ResNet network model with residual block as the core is constructed, and BN layer is added to improve the model"s data standardization processing ability and improve the model"s overfitting control ability. The optimized Sparrow search algorithm was used to determine the optimal parameter set of the model, which significantly improved the performance of the deep network, and SVM model was introduced to improve the feature classification ability of the model. The Dropout layer and support vector machine tool are introduced in the data output of the model to reduce the complexity of the model and simultaneously improve the classification accuracy of various faults. The experimental results show that the proposed fault detection algorithm model has strong noise reduction ability, the fault location accuracy in the training set and the test set are 99.2% and 99.1%, respectively, and the classification accuracy of different faults is better than the traditional fault detection algorithm. The ablation experiment results verify the effectiveness of optimizing each component of the ResNet network model.