This paper aims at improving the braking ability and reliability of disc brakes. Based on some braking tests of the disc brake, an intelligent forecasting model for its tribological properties was established firstly by the artificial neural network (ANN) technology. Its input layer contains three braking cells: braking pressure, sliding velocity and surface temperature. And its output layer contains three tribological cells: friction coefficient and its stability coefficient, and wear rate. Secondly, an intelligent forecasting system was developed based on the model. It is mainly composed of three parts: sensing system, data collecting system, and data computing system. Finally, the disc brake used in mine hoists was tested on the system as an example. It is shown that the experimental data which contain nonlinear relationships between the braking conditions and tribological properties of disc brake are theoretical foundations of the tribological forecasting. The ANN is especially suitable for establishing the tribological forecasting model. And the optimized BP network is proved as a simple and effective computing model. The intelligent model and system established in this paper has quite favorable forecasting ability and practicability. By the contrast, it has higher precision for forecasting of the friction coefficient and its stability coefficient than the wear rate. The difference was attributed to man-made testing errors of data samples during tribological experiments.