An LSTM-based decision-making model for predictive manufacturing performance optimization

In today's industrial landscape, decision-making has become increasingly complex due to rapidly evolving conditions and the growing volume of data. Traditional Multi-Criteria Decision-Making (MCDM) methods have long been the go-to approach for industrial decision-making, offering structured frameworks for evaluating multiple factors. However, while these methods remain valuable, they are often static and reactive, struggling to process large datasets, anticipate future challenges, and adapt swiftly to changing industrial constraints. To remain competitive, industrials must transition toward dynamic and proactive decision-making strategies that enhance efficiency and responsiveness. This paper introduces a novel industrial decision-making framework that integrates Long Short-Term Memory (LSTM) networks to predict optimal decisions in real time and for future periods. By leveraging real-time industrial performance data and forecasting future performance trends, the proposed model enables continuous adaptation to changing industrial conditions. This approach offers a flexible, proactive, and responsive solution, transforming decision-making from a reactive process into one that anticipates and mitigates potential disruptions. The framework can be applied to various industrial decisions, including industrial process optimization and maintenance scheduling, ensuring critical operational aspects are effectively managed. Designed to be hierarchical and user-friendly, the model incorporates decision-maker evaluations to enhance decision quality. The framework was tested in the automotive sector to improve manufacturing lead time for ventilated brake discs. Initial results demonstrate its ability to anticipate performance trends and recommend optimal corrective actions, marking a significant advancement over conventional MCDM techniques. Additionally, the framework is intuitive to implement and easy for decision-makers to interpret, further facilitating its adoption in industrial settings. The findings underscore the framework's potential to help industries stay ahead of operational challenges. However, further validation is needed to assess its sensitivity to decision-maker preferences and refine the method for specific industrial applications. Future research should focus on expanding the model to diverse industries and integrating complementary tools to further enhance its predictive capabilities.