Abstract: Abstract: In order to solve the false alarm caused by photoelectric fire detection technology in confined spaces such as civil aircraft cargo hold, a millimeter wave resonator sensitive to the change of flue gas environment is designed, and a combustible combustion experimental platform which can realize multi parameter signal detection is established. The frequency offset is selected as the main characteristic parameter of fire smoke detection for multi parameter data processing and information fusion. The machine learning algorithm based on YOLOv5 network model is used to train and learn the infrared image and smoke data of fire detection, evaluate and analyze their classification performance, and experimentally analyze the recognition effect of the detection system on the types of combustibles and combustion types. The results show that the multi parameter fusion fire detection system is superior to the existing single sensing smoke detector in smoke false alarm rate, missed detection rate, detection response time, combustion species identification rate and smoldering fire identification rate. It can be seen that the method of aircraft cargo compartment fire detection based on multi parameter fusion detection system based on frequency offset detection is effective, which can improve the accuracy of fire detection and reduce the occurrence of false alarm and missing alarm.

Key words: Key words: millimeter wave resonator, fire detection, frequency offset, multi parameter fusion, machine learning