DETECTION OF MEASUREMENT ERRORS CAUSED BY FAULTS AND INTERVENTION IN NATURAL GAS METERS USING ACOUSTIC EMISSION AND MACHINE LEARNING.

Abstract

This invention relates to a method and system for analyzing acoustic and / or vibration signals in meters (e.g., gas, water, or steam meters) using advanced signal processing and machine learning techniques to determine mechanical condition, diagnose faults, estimate measurement error, and predict remaining useful life. Within the scope of this invention, the raw signal sequence received from the meter via at least one acoustic or vibration sensor is denominated, and time-domain and frequency-domain features are extracted from the cleaned signal to create a descriptive feature vector. This feature vector is fed into a pre-trained evaluation model to generate qualitative condition classification and / or quantitative regression predictions indicating the meter's mechanical condition or measurement accuracy, measurement error, remaining useful life, and maintenance priority index.