A Novel ANN-Based Adaptive Ultrasonic Measurement System for Accurate Water Level Monitoring

Abstract

Low-cost ultrasonic sensors are widely used for non-contact distance measurement problems. Speed of ultrasonic waves is greatly affected by environmental conditions such as temperature and relative humidity among a few other parameters. Presence of acoustic and electronic noise also influences an ultrasonic sensor based distance measurement system. Existing standard techniques assume that the temperature and relative humidity levels remain constant throughout the measurement medium. In our proposed system, we measure water level in storage tanks of different depths, which exhibits a gradient of temperature and relative humidity across the measurement medium. Hence, the standard ultrasonic measurement system (UMS) is not able to estimate distance accurately. In this article, we propose an algorithm based on modified neural network architecture to increase the accuracy of UMS and also to extend the standard operating range of the ultrasonic sensor used. This article presents a novel approach to reduce measurement error using Levenberg–Marquardt backpropagation artificial neural network (LMBP-ANN) architecture. The proposed model is validated by comparing the actual water level at various depths under different environmental conditions with the output of the trained neural network. The measurement error in the proposed model is bounded by ±1 cm in the distance measurements ranging from 2 to 500 cm. This model is able to extend the maximum standard operating range of the ultrasonic sensor (HC-SR04 model) from 400 to 500 cm. The proposed model is evaluated using mean squared error (MSE) and $R$ -values to establish the effectiveness.