Acoustic and vibration-based condition monitoring is well established. Existing solutions, however, record signals continuously and run the AI analysis offline or on central servers. This creates large data volumes, prevents immediate in-vehicle response and leads to misclassifications when terrain, speed or vehicle state change. NeuRoll (Neuromorphic Real-time Online Learning solution) builds on the Vehicle Data Analysis System (VeDAS), which has reached series production for military wheeled vehicles. VeDAS records vibration, orientation, speed, distance, temperature, humidity, structure-borne sound and CAN bus data. The training data comes from test tracks and Bundeswehr proving grounds. At NeuRoll’s core is an event-based processing chain that analyses relevant signal changes wit ...