Abstract
An innovative non-intrusive flow meter was designed for the water flow measurement at ambient temperature under steady-state conditions [1]. This paper aims to possibly expand the designed flow meter to cover a wider range of flow rates. In this regard, the most accurate machine learning model for predicting volumetric flow rates using the previously designed flow meter will is identified, and the achieved resolution, degree of uncertainty, cost considerations, and flow range capabilities of this novel flow meter will be benchmarked against an existing non-intrusive flow meter currently available in the market. The device features a band heater positioned outside of the pipe, complemented by two thermocouples that monitor the outer wall’s temperature. The procedure involves activating the band heater for 60 seconds, followed by deactivation and the recording of temperatures over the subsequent 120 seconds. Multiple tests are conducted for each mass flow rate, ranging from 8.5 GPM to 40 GPM. Arduino-based data collection is employed to record the temperature response for the system. Statistically, three temperature parameters are evaluated: maximum temperature, average temperature differences during heating, and average temperature differences during cooling. Regression learner methods are utilized to establish correlations between volumetric flow rates and temperature parameters.
