In order to quickly on-line locate and diagnose the fault of CNC machine tools,the fault diagnosis technology of CNC machine tools based on circulating neural network was proposed.By extracting network nodes,a gate discrimination structure based on circulating neural network was established;fuzzy boundary theory was introduced to classify the fault feature space of machine tools;the statistics and discrimination of rule credibility rate were completed by organizing fault diagnosis samples,so on-line fault diagnosis of CNC machine tools was realized.Taking CAK6150 CNC machine tool as the research object,through data induction,it could be seen that the actual output value of fault diagnosis data was very close to the theoretical value under the support of circulating neural network,and the convergence speed was fast,by which the problem of mechanical equipment application fault in manufacturing enterprises could be solved better.