The Innovative Approach to Real-Time Detection of Fuel Types Based on Ultrasonic Sensor and Machine Learning

Abstract

This paper presents an approach to prevent the incorrect transfer of fuel to the wrong tank during refueling. An experimental setup was developed to perform ultrasonic and dielectric measurements on diesel, gasoline, ethanol, and water. The fuel types were determined using an ultrasonic sensor and time-of-flight values were measured at various temperatures. Additionally, the dielectric coefficients of these liquids were measured to determine the liquid types using a dielectric sensor. The results obtained from both the ultrasonic and dielectric methods are systematically compared, and the advantages and disadvantages of each approach are thoroughly discussed. It was observed that dielectric method does not always yield accurate results. The proposed system effectively prevents erroneous transfer of fuel to the tank during refueling. The developed system may be used in practice to distinguish fuel types. In addition, a new approach using machine learning techniques to determine fuel type is presented. Fuel types were classified using 33 machine learning algorithms such as support vector machines, artificial neural networks and K-Nearest neighbors. It seems that artificial neural network with first layer size 25 and quadratic discriminant classifiers have achieved remarkable results with a success rate of 94% in classification. The results highlight the important and effective role of ultrasonic sensors in accurately identifying fuel types, leading to more efficient and safer storage and transportation of fuel. It  is also concluded that machine learning techniques can be used effectively in identifying and classifying fuel types. The approach involving ultrasonic and artificial intelligence techniques was particularly innovative in distinguishing fuel types.