Аналіз та обґрунтування вибору моделі для моніторингу параметрів транспортного засобу та прогнозування технічного обслуговування

Today, vehicle equipment provides huge amounts of data for technical condition monitoring. The process of vehicle operation is constantly associated with the natural wear and tear of components, which in turn leads to a deterioration in technical and operational characteristics. For the reliability of the vehicle operation, high-quality component monitoring during maintenance and fault diagnosis is required. Monitoring and diagnostics mean the process of maintenance with the future detection of defects in vehicle systems. Analytics of the information obtained using Big Data technologies are becoming extremely important for analyzing and processing large amounts of data, especially for predicting the failure of mechanisms. Predictive maintenance works better than repair or preventive maintenance. Accordingly, data-driven forecasting is much more effective in terms of assessing the operational stability of units in real time and preventing possible failures in their operation. For better detection of possible defects, it is advisable to use neural networks, which can provide more data on the compliance of the technical condition of the vehicle with the indicators transmitted from the on-board computer. Using real data from multiple sensors and reports on malfunctions in the operation of vehicle components and assemblies, machine learning models can explore patterns of information and create predictive fault models based on real-time condition monitoring. Such information transmitted to the system from the vehicle is processed using neural networks and the output is a better analysis of the vehicle's condition for faults, which improves the quality of maintenance. The article discusses the peculiarities of diagnosing modern vehicles, provides possible options for obtaining data from vehicles and an example of developing neural networks with the setting of optimal machine learning parameters and for predictive diagnostics using available data