Abstract: In view of the outstanding problem that the newly recruited maintenance personnel for medical laboratory instruments are not competent for the maintenance work because of their insufficient experience，in order to assist for their early competence in maintenance work，the experiences of fault detection of various instruments were collected through interviews with industrial experts，literature collation and other channels，and they were sorted out and transformed into a Turing Robot Custom Corpus. On this basis，we used WeChat subscription as the interactive front- end and Turing Robot as the fault diagnosis back- end server，Flask Web server framework was used to build the interactive bridge between WeChat subscription and Turing Robot；an intelligent fault diagnosis system for medical laboratory instruments was designed. The actual operation test showed that it ran steadily，and it can provide reliable maintenance suggestions for maintenance personnel.

Key words: traffic image, vehicle detection, dynamic probabilistic location model, convolution neural network（CNN）