An RMS-based Approach for Leak Monitoring in Noisy Industrial Pipelines

A method for leak monitoring of noisy industrial pipelines is proposed in this work. This method is based on evaluating a single feature, the root mean square of the acquired acoustic signal of a pipeline and relies on defining updateable thresholds based on the previously acquired values. If at any point, these values increase above the active threshold, a check routine is performed to identify if that event was transient and short-term. The presence of a leak can be modeled through this procedure as a long-term increase in the acquired rms values. The algorithm behind the operation is associated with a set of parameters, that are adjustable in such a way that the system can adapt to the transient noise characteristics of the pipeline under inspection. These parameters can be fine-tuned based only on noise measurements without the need for a training phase with artificially generated leaks. The algorithm can begin operating after only a reference noise measurement during its initialization. The proposed algorithm's overall accuracy was evaluated with measurements performed on an operating pipeline in the facilities of an oil refinery.