This paper investigates score-based and constraint-based causal discovery methods, emphasizing their application to manufacturing processes. The research focuses on how algorithms such as the PC (Peter and Clark) and FGES (Fast Greedy Equivalence Search) can uncover causal relationships within complex manufacturing systems, which are essential for process optimization, fault detection, and informed decision-making.

The study was done in three key phases. First, a thorough literature review established the theoretical foundation for understanding causal discovery methods. These methods were applied to real-world manufacturing data in the second phase, focusing on assembly lines and production processes. The main goal was to find key variables impacting production efficiency, product quality, and equipment failure rates. In the final phase, the performance of the causal discovery algorithms was evaluated through regression analyses, where R-squared values measured how well the identified causal relationships explained observed outcomes. The results revealed that the PC algorithm can identify direct and indirect relationships among variables. Also, the FGES algorithm proved more effective at uncovering complex interactions in high-dimensional datasets.

This study suggests that causal discovery techniques can be integrated into automated decision-support systems within manufacturing. Organizations can refine existing systems, optimize operations, and develop predictive models for proactive issue management by better understanding the causal structure of production processes.

Future research should expand the application of causal discovery methods across a broader range of manufacturing environments. Additionally, using machine learning to enhance causal identification and assessing how well these techniques perform in large industrial systems will be crucial for maximizing their impact on manufacturing performance and innovation.