Development of Methods and Algorithms for Diagnostics of Multifunctional Indication Devices Using Predictive Analytics
Abstract
Development of Methods and Algorithms for Diagnostics of Multifunctional Indication Devices Using Predictive Analytics
Incoming article date: 15.05.2025The article discusses the problem of increasing the reliability of multifunctional display devices in the context of digital transformation of production processes. An approach to predictive diagnostics based on the analysis of operational and thermal characteristics of UMI using machine learning methods is proposed. The classification of UMI according to physico-technological principles and architectural levels was carried out, which made it possible to structure diagnostic models. Mathematical methods for predicting failures are considered, including logistic regression, gradient boosting (CatBoost), and residual resource estimation models. Special attention is paid to the development of the Thermal Emission-Based UMI Profiling (TEB-UP) method based on the analysis of heat maps and machine vision algorithms (PCA, autoencoders, CNN). It is shown that temperature unevenness is a sensitive indicator of degradation, ahead of traditional failure rates. The TEB-UP method demonstrates the potential for integration into monitoring and predictive maintenance systems within the framework of Industry 4.0 and 5.0 concepts.
Keywords: multifunctional display devices, predictive diagnostics, thermal profiling, residual resource, machine learning