Flame Temperature Prediction Using Machine Learning Model

The intensity or amount of heat present in any material, substance, or object is known as temperature. The process of measuring temperature is a tiresome and complex task from any visible heat source. The process of measuring temperature is known as thermometry. It plays a vital role in various industrial and manufacturing processes. There are several devices or gadgets present which are used for measuring temperature like a thermistor, Resistance Temperature Detector (RTD), infrared thermometer, thermocouples, pyrometers, etc. Every temperature measuring instrument has its demerits. While measuring temperature in some devices, one must be very alert because it is a necessity to check that the temperature of the material or substance should be less than or equal to the instrument temperature. In some instruments, the high temperature reduces productivity, and the efficiency of the sensors present in it. Some devices face the drawback of difference in temperature because in such types of devices there is a threshold temperature. If the temperature exceeds the threshold temperature in such a case, the measured temperature will differ with the temperature of the system. Under such circumstances, it will deviate from the original heat transfer property. To overcome all these drawbacks a machine learning model is proposed to detect approx. temperature using the color-temperature correlation approach. In this proposed system, histogram backprojection is used for pre-processing of the input image to derive the color of the flame. To predict the temperature, Support Vector Machine (SVM) and Artificial Neural Network (ANN) have been used and compared. Simulation results show that Support Vector Machine outperforms Artificial Neural Network.