Ultrasonic Sensor-based Fuel Monitoring System using Clustering Technique

Fuel is a necessary element for the development of industry, transportation, and energy production. In general, fuel means liquid hydrocarbon that is used to run an engine of different transport i.e. bus, truck, ship, etc. The theft of fuel causes serious socio-economic problems and leakage of fuel causes evaporation of fuel in the atmosphere which affects the air. In addition, a reserve of fossil fuels is not unlimited and will run out after a couple of decades. The purpose of this research is to find a way to avoid leakage, and theft by observing the regular usage, and refill. In this research, the clustering technique is applied to the ultrasonic sensor data to find theft, leakage, and refill. At first, four ultrasonic sensors are installed on the top of the fuel tank and configured with the pins of the Arduino Uno microcontroller. Now the microcontroller is ready to collect sensor data. After driving the vehicle in different routes the sensor data are collected and preprocessed in order to find the average level of the fuel and the rate of usage of the fuel. A machine learningbased model is formulated by considering four different classes of data, regular use, leakage, theft, and refill. The processed data is applied to the Extended K-means clustering algorithm and the Agglomerative clustering algorithm. In both cases, the accuracy is 96.67 percent but the Average Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Index are respectively, 0.7189, 0.7044, and 0.430,0.447 and 2316.92, 2243.41.