Indonesia Conference Directory

Improving production system quality, especially in the efficiency of coal combustion, is a must to optimize the electric energy production of a power plant. To maintain customer trust, towards an international standard distribution process, it needs innovation in combustion monitoring. Overheating conditions frequently occur and could break the camera due to limited information on combustion temperatures received by the user. From these problems, this study aims to design a classification system for monitoring the combustion process in the boiler or furnace. Combustion area captured by the Adafruit AMG-8833 IR camera and continued with the extraction and segmentation of thermography analysis and neural network (NN). This study utilizes the features of temperature conversion in each image segment in the form of HSV (Hue, Saturation, Value). Hue parameters (H) and value (V) parameters are used in the classification process for its large degree of red to green differences with a significant range at each temperature. Those parameters are the input of the Artificial Neural Network along with the average & overheating temperature as the classification target. The average error of this system is 0.08559% for the image classification with training data of 64x45 inputs, 16 neurons, and the best performance at 10th repetition.