In order to solve the problem of insufficient data in equipment fault detection and difficulty to apply deep learning methods for improving performance, a method for improving equipment fault diagnosis accuracy based on optimized deep random forest is proposed. According to the characteristics of the dataset acquiring data, a re-sampled sample set decision tree model was constructed. The continuous value in the fault data was discretized by the C4.5 decision tree, and the extended convolution method was used to expand the sensing range of the sliding window, and finally training the model was verified by experiments. The experimental results show that the method of using deep random forest can be used as a reference for improving equipment fault diagnosis.