The repair quality of train air-condition unit has certain effect on carriage’s thermal comfort, indoor air quality, and further affects passengers’ health and railway department’s economic benefit. The article discusses the shortcomings of present diagnosis methods to train air-condition comfort are discussed. The status and future of neural network fault diagnosis expert systems that are applied to train air-condition fault diagnosis are pointed out. Aiming at train air-condition’s faults and linking the author’s experiment device about train air-condition, the article introduces how to use back propagation (BP) neural network and fuzzy logic to establish fault diagnosis expert systems. The results of simulation show effectiveness of this system.