AtomMind Predictive Analytics System for Chepetsky Mechanical Plant

The enterprise loses profit due to non-compliance with the quality parameters of finished products and inefficient technical maintenance and repairs of production equipment. On average, about 25% of shell tubes do not meet the specified parameters during technical inspection, and some batches exceed 10%. Each month, there are about 3000 orders with 700 parameters each—requiring real-time tracking of over 2 million technological parameters. Thus, data collection and analysis were only conducted for batches with significant deviations in defects (reactive analysis on a narrow sample). The project's goal was to use collected data about production processes (more than 600 parameters) to improve product quality by reducing the final non-compliance rate from 23% to less than 1%. The project to forecast defect levels and provide recommendations using predictive analytics was launched in 2022 for the main product line—zirconium cladding for nuclear reactor fuel elements (tvels). AtomMind was conceived as a recommendation system for production that would provide intelligent decision-making support. AtomMind is a predictive analytics system using artificial intelligence technology. Through continuous monitoring and timely diagnostics, AtomMind facilitates the transition from scheduled maintenance to predictive maintenance. The system predicts and prevents equipment failures by determining the likelihood of anomalies before they affect production. AtomMind is an effective tool for managing the quality of finished products by analyzing factors affecting product characteristics, forecasting potential deviations based on equipment data, and providing recommendations for adjusting technological parameters of production processes for users. AtomMind is a low-code platform with a unified interface and a set of integration tools that allow the system to be embedded into the enterprise's IT landscape, providing quick development of services and applications to increase production process efficiency.