Quantify production planning efficiency through predictive modeling in manufacturing systems

This paper proposes a management system designed to evaluate and enhance the optimization degree within manufacturing operations for improved business planning. The proposed model computes predictive data about production forecasts (times, yields, quantity of items produced) to assist operators in filling in these metrics for newly introduced items. It then assesses the discrepancy between the predicted values and the actual measured production data. This assessment aims to provide metrics for evaluating the efficiency of business planning systems, providing a quantified understanding of discrepancies for more accurate profit estimates and strategic planning. The proposed approach exploits shallow and deep machine learning models and transformer-based approaches, and it is experimentally evaluated on a real-world manufacturing dataset. One planned outcome that these metrics will enable is the provision of a tool that supports manufacturing workers by completing data that they cannot define themselves and highlighting potential discrepancies between the manually entered data and the model data, at an early stage of the manufacturing process, thus avoiding errors rather than correcting them afterwards. This approach aims to increase collaboration between humans and machines, in line with the principles of Industry 5.0. • Introducing a predictive modeling system for efficient estimation in new item planning. • A framework evaluating variance between predicted and production data to provide a quantitative planning assessment. • We introduce AI-enhanced blocks in the pipeline, comparing shallow learning models and transformer-based methods. • A final validation on a real-world dataset demonstrates the effectiveness of diverse machine-learning techniques. • This approach emphasizes enhanced collaboration, aiding workers and highlighting discrepancies for error prevention.