It is reported that Chengdu North Rolling Stock Depot of China Railway Chengdu Group Co., Ltd. (hereinafter referred to as "Chengdu North Rolling Stock Depot") has introduced the TFDS (Freight Car Fault Trackside Image Detection System) fault intelligent recognition system for railway freight car detection. After the system was launched and put into operation, the manual workload was reduced by nearly 90%, the fault recognition rate was increased by nearly 40%, and it only took 3-5 minutes to detect tens of thousands of components. In the past, the dynamic inspection work of TFDS relied entirely on manual execution. Whenever a railway freight car passes through the TFDS detection station, the electronic camera equipment quickly captures and uploads dynamic images of the car's bottom accessories and side parts to the TFDS server. Subsequently, professional TFDS dynamic inspectors analyzed each image and identified potential vehicle malfunctions. It is reported that TFDS dynamic inspectors at Chengdu North Vehicle Depot need to review over 1.5 million vehicle images daily. In order to effectively reduce the workload of operators and further improve the efficiency and quality of train technical inspections, Chengdu North Rolling Stock Depot has decided to introduce the TFDS intelligent fault recognition system. Regarding this, Jia Mingyong, a technician at the digital operation and maintenance workshop of Chengdu North Rolling Stock Depot, said, "Unlike the previous' needle in a haystack 'inspection method, now we only need to review the faults accurately identified by AI algorithms and pushed." The use of this system has greatly reduced the analysis volume of each inspector's train images from more than 600 to 100, reducing the workload by nearly 90%. At the same time, the average technical inspection time for each train has been reduced from 15 minutes to 8 minutes. It is reported that the system adopts the largest visual pre training model in the industry, the Pangu Big Model, which can automatically learn various types of truck fault image samples in the freight system, achieve automatic recognition of truck faults, automatically summarize component characteristics, and automatically find fault patterns. (New Society)