What type of study is this?

This is a Quantitative Study study.

October 8, 2025

Increasing the Wear Resistance and Reliability of Resource-Determining Parts of Transport and Agricultural Machinery by Engineering Methods Using Neural Networks

Key Points

High prediction accuracy was achieved with R² values between 0.98 and 0.99, demonstrating the efficacy of the proposed methodology.
The methodology leverages synthetic data generated from physical wear models to predict machine part degradation effectively.
Using neural networks, specifically multilayer perceptron and LSTM, offers an innovative solution to modern machinery reliability challenges.
The results support implementing predictive maintenance strategies in cyber-physical systems of agricultural and transport machinery.

Abstract

It is shown that in modern conditions of intensive development of Industry 4.0 and digitalization of machine-building processes, the problem of ensuring the reliability and durability of resource-determining parts of transport and agricultural machinery is of particular relevance. It is found that traditional approaches to predicting the technical condition of machines, based on calendar maintenance, do not meet modern requirements of economic efficiency and operational safety. A methodology is proposed, which is based on the hybrid application of artificial neural networks - a multilayer perceptron (MLP) for identifying the dominant mechanisms of wear of machine parts and their conjugation and a recurrent long short-term memory network (LSTM) for predicting the dynamics of part degradation based on time series of operational parameters. It was determined that the use of synthetic data generated on the basis of physical models of wear allows overcoming the limitations associated with the lack of real operational data of machine components, systems and assemblies. Validation of the developed algorithm on a representative data set (50,000 samples) demonstrated high prediction accuracy: coefficient of determination R² = 0.98...0.99, root mean square error RMSE = 8.12...15.67 μm, mean absolute percentage error MAPE = 2.5 3.9%. These results confirm the prospects of integrating the proposed approach into cyber-physical systems of modern transport and agricultural machinery for implementing the concept of predictive maintenance.

Mark Helpful

Bookmark

Relay