AUTOMATED DEVICE FOR ANALYSING DROP SAMPLES OF MOTOR OILS USING NEURAL NETWORKS

Relevance Various types of oils are essential resources for the operation of many industries, machinery, and engines. In addition to their primary function as lubricants for technical equipment components, oils have long proven themselves as convenient and effective indicators that reflect processes occurring during the use of machinery. When it comes to motor oils used in vehicles, timely analysis of their chemical composition allows conclusions to be drawn about the need for replacement, as well as performing non-invasive diagnostics of the engine itself, predicting its wear rate, and forecasting further operational prospects. Each type of motor oil initially has a set of characteristics defined by regulatory documents. During engine operation, qualitative and quantitative changes occur in the composition of the oil. The accumulation of water, mechanical impurities, and changes in the amount of oil additives negatively affect its performance properties. It is recommended to replace the oil based on operating conditions according to its actual condition, i.e., when the established limit values of controlled quality indicators specified in relevant technical documentation are reached. Under sanctions pressure and limited access to motor oils from wellknown manufacturers, monitoring oil quality becomes even more critical. Domestic producers are forced to change the formulation of oils due to the unavailability of base oils and additives. There may also be an increase in counterfeits from less regulated sources of oil production. To gain consumer trust and conquer the machine oil market, domestic producers must possess appropriate devices and methods for ongoing technical control of oil quality using current means. Aim of research Development of an automated device for analyzing drop samples using artificial intelligence. Research methods Issues related to applying mathematical modeling and artificial intelligence for analyzing the residual resource of motor oils were examined. Criteria determining the quality of drop samples were identified. The structure of an automated device for analyzing drop samples was presented, and a neural network architecture with enhanced semantic convergence, recommended for data processing, was proposed. Results The article presents the results of developing an automated device for analyzing oil samples, which will reduce labor intensity, minimize subjective human involvement, and improve efficiency and accuracy in capturing dynamic images. The quality of produced images will benefit positively from the recommended use of special filter paper treatment. Applying the transformer architecture of neural networks will enhance the effectiveness of learning patterns in the dynamics of chemical reactions between oil and indicator paper reagents.