Abstract: Refrigerators are widely used in industrial enterprises. In order to timely find the operation abnormalities and faults caused by cylinder and crankshaft wear and equipment looseness during the long-term operation of the refrigerator, and ensure its long-term stable operation, a monitoring system for the refrigerator operation and fault based on wireless remote transmission of acoustic signal and voiceprint recognition is developed. The system uses GPRS mobile communication network to wirelessly transmit the refrigerator operation noise signal collected on site to the host computer, and then uses the voiceprint recognition method of Mel-frequency cepstral coefficient (MFCC) feature extraction and long short-term memory (LSTM) neural network to monitor and identify the operation status and faults of the refrigerator. The experimental results show that the combination of MFCC and LSTM neural network can effectively improve the recognition rate and diagnosis effect of the system, and has a good application prospect.