Aimed to the shortages of algorithm and the complicated actuality of the warning system,the outliers detection of distance sum-based algorithms were used in the system of early warning of mine safety.The research analyzed the noisy data,the outliers based in the abnormal equipment and the abnormity data of the concentration of gas.It is showed that the model can find out the actual latent information by excavating outliers and the model could make sure the veracity of safety warning.