The significant problem of mechantronics equipment built-in test (MEBIT) is its high false alarm rate. One of the important causes is its sensor channel's fault of BIT system. The method of neural network is adopted for the sensor channel fault diagnosis. The fault mechanism and types of sensor channels of MEBIT in a large ship power engine and the fault samples are obtained. After the training of neural network, the actual system's faults are diagnosed and identified. The experimental results show that the neural network can diagnose the faults and identify their types. The method is compact, effective and of practical value.