Abstract: It is difficult to use a general method to implement fault prediction of distributed sensor networks, a kind of fault prediction method based on knowledge discovery was put forward. Firstly, the method establishes mathematical description system of time information, in order to realize the knowledge discovery based on causal index; then, the method uses knowledge reasoning mechanism of causality relationship to realize fault prediction of distributed sensor networks system. A fault prediction experiment of gas drainage monitoring system was carried out, the experiment results prove that the method can accurately predict fault of distributed sensor networks system, and has advantages of simple algorithm, utility and high-efficiency.