FORECASTING THE QUALITY OF THE TECHNICAL OBJECT’S FUNCTIONING USING MACHINE LEARNING

Abstract

Background. The quality of functioning of complex technical systems is determined by many characteristics. Forecasting the values of these characteristics based on the results of monitoring the performance of the facility makes it possible to fulfill the ever-growing requirements for safety and reliability. The necessary accuracy of forecasting requires the construction of high-quality mathematical models. As a technical object, the burner device is considered: the quality of operation of such devices is evaluated according to one of the main characteristics – the temperature of the flame core. The purpose of the study is to develop a methodology for building a mathematical model that would provide a fairly accurate forecast of the characteristics of the functioning of a technical object. Materials and methods. To build models based on the results of observations of the object under study, both classical methods of regression analysis and machine learning methods are used. The paper compares two approaches: the use of linear regression analysis and the compositional method "Random Forest". Results and conclusions. The technology of mathematical modeling and forecasting of the characteristics of the quality of functioning of technical objects using two approaches has been developed: linear regression and the random forest method. In the example under consideration, the evaluation of the quality of the burner device from the temperature of the flame core, both approaches gave almost the same and quite acceptable result. Obviously, in other tasks, these results can vary significantly.

Keywords

• performance indicators
• regression analysis
• machine learning
• random forest
• statistica system