Ultrasonic Leak Detection Using MEMS Sensors For Industrial Pneumatic Pipeline Monitoring

This study presents an ultrasonic leak detection system for industrial pneumatic pipelines utilizing MEMS-based sensors. The system incorporates a conical horn (electronic gun) design to enhance signal focusing and improve detection sensitivity. Controlled experiments were conducted using six leak diameters (1–6 mm) and six pressure levels (5–30 PSI). Fast Fourier Transform (FFT) analysis was employed for feature extraction, improving the system's robustness over conventional CWT-based methods. The CNN model achieved 90% accuracy for binary leak detection, while a reduced feature-based model maintained 88.9% accuracy with improved computational efficiency. Results indicate higher detection accuracy for larger leaks at elevated pressures, while small leaks at low pressures posed greater challenges. The integration of the conical horn significantly enhanced signal clarity, particularly in detecting minor leaks. The proposed system's effective balance of accuracy, sensitivity, and computational efficiency makes it suitable for real-time industrial monitoring applications.