This project will develop a prototype wayside system for autonomous detection of train air leaks on moving trains. The project will be carried out in two stages. Work in Stage 1 will focus on sourcing, acquiring, and assembling the hardware necessary for the prototype design, as well as generating the initial data needed for work in Stage 2. Working with technology partners (Fluke and Sorama), hardware most suited to the project will be determined and acquired, including the type and number of acoustic sensors, visual spectrum cameras, etc. Using the hardware, an initial system will be assembled and installed at the SwRI Locomotive Technology Center rail yard for testing. Communication among various hardware components in the system will be established. Requirements for necessary data collection will be determined. Initial testing at the SwRI Locomotive Technology Center railyard will be conducted. Data will be generated on purposefully induced air leaks, varying leak rates and locations, using moving locomotives + railcar(s). This data will be used for hardware validation, fine tuning the sensor location and as model input for use in Stage 2 in which the machine learning AI will be paired with the existing hardware setup. A visual AI will designed and integrated in the developed prototype system. Initially, the AI system will be trained to identify air leaks using data generated in Stage 1. After the system is verified in bench testing that it is responding as expected to the previously generated data, final testing will begin. The system will again be installed at the SwRI Rail Yard, and data will be generated on purposefully induced air leaks, varying leak rates and locations, using moving locomotives + railcar(s). It is expected that the system will go through some iterations back and forth for further refinement. The refined system should flag air leaks and notify human users autonomously while minimizing or eliminating false positives. The final report will include all relevant data, methods, models, and conclusions.