Gradient boosting method application to support process decisions in the electron-beam welding process

The purpose of the study is to develop a technological process mathematical model of creating permanent joints of dissimilar materials based on electron-beam welding using machine learning algorithms. Each of the connected elements is a responsible unit of the complex device, due to this fact, strict criteria are set for the quality of the welded joint. In essence, the set task is a regression task. There are many algorithms suitable for solving the regression problem. However, often the use of one algorithm does not provide sufficient accuracy of the result. One way to solve this problem is to develop a composition of algorithms to compensate for the problems of each of them. One of the most effective and potent compositional algorithms is the gradient boosting algorithm. This algorithm use will improve the quality of the regression model. The proposed model will allow the technologist to set the process parameters and to get an assessment of the final product quality, as well as by setting input and output values. The use of assessment methods and forecasting will reduce the time and labor costs of searching, developing and adjusting the process. A description of the gradient boosting algorithm is given, as well as an analysis of the applicability of this algorithm to the model and a conclusion regarding the areas of its applicability and the reliability of the forecasts obtained by its direct use. In addition, we consider the process of direct model training based on the data obtained as part of search experiments to improve the quality of final product. The results of the applicability analysis allow us to judge the admissibility of using the proposed method for processes that have similar statistical dependencies. The application of the proposed approach will make it possible to support the adoption of technological decisions by specialists in electron-beam welding during the development of the technological process and when new types of products are put into production.