Application of Machine Learning Methods to Solve the Problem of Classifying Materials Based on the Results of Measuring their Thermoelectric Power

The paper investigates the possibility of using machine learning methods in the analysis of temperature dependences of the thermoelectric power of structural alloys, their clustering and classification for the purpose of subsequent use of these methods in the construction of methods for monitoring the chemical composition of materials. Data on the thermoelectric power of industrial alloy grades used in welding technology are presented. The results of dividing the array of experimental curves into groups using the hierarchical clustering algorithm are presented. General patterns of change in thermoelectric power within each group of alloys are identified and a comparison with the chemical composition of the alloys is made. Training and test data sets for training are prepared and the possibility of classifying a material based on the results of measuring its thermoelectric power is investigated.