In preventive maintenance, the prediction of failure rate is of great significance for making accurate and reasonable maintenance plan, mastering the initiative of maintenance and giving full play to the use efficiency of equipment. The traditional failure rate methods and models rely on a large amount of statistics. However, statistics data are often unavailable, which creates difficulties in predicting failure rates. Aiming at the current situation of "little data, poor information", this paper summarizes existing research on failure rate prediction methods at home and abroad, and the GM(1, 1) model is used to predict the failure rate. This article analyzes the concrete examples of cranes. Firstly, the GM(1, 1) model and the discrete GM(1, 1) model are established based on the determination of the length of the model for the fault rate curve with large irregular fluctuation, and the models is compared and optimized. Secondly, the linear regression GM(1, 1) model is introduced for the curve with linear trend, and three models are compared and optimized. It has practical application value in the prediction of equipment failure rate.