The paper investigates the use of ontologies built on meta-associative graphs to address machine-learning problems that are important for industry. It considers intelligent analysis of manufacturing data, predictive maintenance, quality control and automated decision-making. Particular emphasis is placed on how meta-associative graphs-based ontologies make it possible to create adaptive models and integrate heterogeneous data sources, thereby improving production efficiency. The article outlines a methodology for building such ontologies and coupling them with machine-learning algorithms. A case study from Dozator-Plus (Mogilev, Belarus)—a manufacturer of hydraulic steering metering pumps—illustrates the practical advantages of the proposed approach.