Event-Driven Neuromorphic Vibration Processing for Ultra Low Power Micro Gas Leak Detection

Gas leakage in pipeline systems poses serious safety and environmental risks, particularly when micro-leaks remain undetected in early stages. Conventional gas detection methods rely on continuous sensor operation, resulting in high power consumption and limited suitability for remote or battery-powered deployment. This paper presents an event-driven neuromorphic vibration-based architecture for ultra-low power micro gas leak detection. Structural vibrations generated by pressurized gas escaping through small defects are continuously monitored using a piezoelectric sensor configured for event-based signal generation. Detected vibration events are processed using a Spiking Neural Network (SNN) implemented on an embedded microcontroller platform to discriminate between normal operational disturbances and leak-induced anomalies. Upon anomaly confirmation, an array of semiconductor gas sensors (MQ-2, MQ-4, and MQ-135) is selectively activated for concentration verification, reducing unnecessary heater power consumption. Experimental validation on a controlled pipeline setup demonstrates reliable micro-leak detection with reduced false alarms and significant power savings compared to continuous gas sensing approaches. The proposed hierarchical sensing framework enables early detection capability while maintaining energy efficiency, making it suitable for remote and embedded pipeline monitoring applications.