Application of automatic machine learning algorithms in fault diagnostic of ac electric motors in mineral resource industry enterprises

Introduction. The electric motor has a wide range of applications in the mining industry. During its operation, important issues are diagnostics of the actual state and timely detection of faults. Artificial intelligence methods used for these purposes have a number of disadvantages, for example, significant amount of time is required for developing new and retraining existing algorithms. Frameworks for automatic creation of artificial intelligence models can solve these problems. Their application and, as a result, a simplified implementation process will be a driver for the implementation of artificial intelligence systems in the mining industry. Materials and methods. During the experiments, data were collected on an experimental unit consisting of an electric motor, a PLC, a frequency converter, current and voltage sensors. During the experiments, the parameters of instantaneous values of currents and voltages in phases were recorded and the “misalignment” defect was simulated by changing the platform height. At each height, the engine operated in two different modes. Thus, the classes used in model training were formed. The Fedot Industrial framework found a more accurate model for a given set of parameters based on the solution pipeline. Results. The FedotIndustrial framework showed itself well in determining the operating state of an electric motor. The prediction accuracy was 100%. Discussion. During the study, a simple experiment was conducted to prove the hypothesis, but its main advantage is the repeatability and ease of interpretation of the results. By scaling and applying a special approach when using the FedotIndustrial framework, it is possible to achieve increased efficiency of technological processes in the mining industry. Conclusion. The proposed approach demonstrated the effectiveness of using AutoML Frameworks for automatic recognition of an electric motor defect during its operation based on extended historical data. Resume. The research results can be useful in the further development of predictive automation systems and protection of electromechanical equipment in mining enterprises. Suggestions for practical applications and directions for future research. On the way to further application of machine learning models, the authors see a number of obstacles and limitations to their use in real production conditions. However, the gradual introduction of automatic Frameworks will help to solve some of them, following the specified approaches to their use.