Predictive maintenance with artificial intelligence in electromechanical equipment to minimize environmental risks

Abstract

The purpose of the research was to analyze the application of predictive maintenance based on Artificial Intelligence in electromechanical equipment, with the aim of reducing environmental risks generated by operational failures and deterioration that are not detected in a timely manner. The study aimed to demonstrate how the use of neural network models made it possible to anticipate breakdowns, optimize equipment performance, and reduce the environmental impacts caused by unexpected shutdowns or leaks of polluting materials. To achieve the objectives set, a mixed methodological approach was applied, combining experimental observation, historical data analysis, and computational modeling. Vibration, temperature, pressure, electrical current, and flow sensors were installed and integrated into real-time monitoring systems based on SCADA platforms and IoT networks. The collected data was processed using open-source tools, mainly TensorFlow and Python, which facilitated the construction of predictive models using Long Short-Term Memory (LSTM) and Multilayer Perceptron (MLP) neural networks. These architectures were selected for their ability to detect complex temporal patterns and predict incipient failures in electromechanical components. The results showed that predictive maintenance with Artificial Intelligence reduced critical failures by 35%, improved energy efficiency, and significantly decreased incidents with potential environmental impact. In addition, maintenance planning was optimized and operating costs were reduced. In conclusion, it was determined that the integration of LSTM and MLP neural networks, together with smart sensors and continuous monitoring systems, represented an effective and sustainable strategy for strengthening operational safety, increasing equipment life, and contributing to environmental protection in industrial settings.