Machine learning based intelligent sensing using non contact ultrasonic sensor

Abstract

ABSTRACT newlineThis thesis presents the study and development of a non-contact intelligent ultrasonic newlinemeasurement for various environmental conditions. The idea of intelligent sensing is to newlinedeploy machine learning models on the resource-constrained micro-controller unit of a newlinesensing module to make it self-contained and carry out intelligent tasks. Non-contact newlineairborne ultrasonic sensors use ultrasonic sound waves to detect or sense the object without newlineany physical contact. The principle of ultrasonic sensor measurement is based on the newlinedetermination of the time of flight. In addition to time of flight based range measurements, newlinefeatures of ultrasonic echo signals can be used to recognize and distinguish the target newlineobjects and materials. Airborne ultrasonic sensor uses air as the transmission medium. The newlineultrasonic wave propagates in air medium at the speed of sound. The speed of sound depends newlineon the air medium, which depends on the medium temperature and humidity. Therefore, newlinethe accuracy of ultrasonic measurement system waves is very susceptible to variations in newlinetemperature, humidity, and other gases present in the environment. This research aims to newlinedevelop an ultrasonic-based intelligent sensing system to enhance measurement accuracy newlinein presence of environmental variations. In this work, we study both machine learning newlinealgorithms and ultrasonic wave characteristics to develop an intelligent framework for newlinedifferent types of applications. This work examines three research topics viz., a) accurate newlinelevel measurement, b) estimation of ambient temperature and humidity, and c) material newlineclassification using echo envelope signal. newlineThe liquid level measurement involves measuring the liquid level in a container or tank newlineiv newlineunder dynamic conditions. The variation of medium temperature and humidity between newlinethe sensor and liquid level influences the sensor measurement accuracy. In this work, newlinewe developed an adaptive intelligent ultrasonic measurement system using a modified newlineartificial neural network to accurately measure the water level