Detection of Pipeline Leaks using BiopiezoelectricEnergy Harvesting Sensors via RecurrenceQuantification and Neuromorphic Computing

This paper presents a study on water pipeline leak detection using organic glycineenergy harvesting sensors. The analysis involves examining open circuit voltageswith recurrence quantification analysis (RQA), complexity measures, and neuromorphiccomputing techniques. Time-dependent un-thresholded recurrence plotsreveal nonlinear characteristics in bio-piezoelectric sensed signatures, providinginsights into leak intensities, flow rates, and sensor locations over short and longtime scales. RQA measures successfully distinguish between leak states, withincreased flow rates resulting in notable changes. The results indicate that glycinesensors outperform PVDF sensors in accurate leak detection. The methodologyis adapted to a neuromorphic framework, estimating leak occurrence and sizeclassification. The spiking neural network-based model shows the superiority ofbio-piezoelectric sensors for structural health monitoring within a neuromorphicframework, compared to PVDF sensors.