Abstract:
The article considers the problem of timely assessment of a vehicle’s technical condition based on the analysis of
informative wear indicators, enabling the prevention of critical failures without the need to visit a service center. Traditional
approaches to technical diagnostics, which rely on mileage or scheduled maintenance intervals, are often insufficiently effective,
as they do not reflect the actual condition of vehicle components and assemblies. Therefore, an intelligent approach based on an
ensemble of artificial neural networks is proposed, allowing the determination of the wear degree of major vehicle systems by
analyzing the dynamics of their operational parameters.
The purpose of this research is to develop a model that enables automated classification of a vehicle’s technical condition
based on a set of indicators signaling potential faults. To achieve this, a representative training dataset was formed using
statistical data on typical wear symptoms (such as reduced acceleration dynamics, unstable engine starting, increased fuel
consumption, engine knocking, etc.), enabling the timely detection of early failure signs and determination of optimal moments
for maintenance. The developed model is based on the Kolmogorov–Arnold theorem and implemented as a pattern recognition
task using supervised learning methods.
Experimental results confirm the high accuracy and practical applicability of the model. The proposed neural network
architecture can be adapted to different classes of vehicles. Practical application of such an analyzer reduces maintenance
costs, enhances operational safety, and ensures prompt response to emerging technical issues. The developed solution can be
integrated into existing hardware and software systems for vehicle condition monitoring, providing convenience, accessibility, and
reliability of the diagnostic process. The results of the study promote the broader adoption of artificial intelligence technologies
in the field of vehicle technical diagnostics.
