Towards Automated Quality Control in Industrial Systems: Developing Markov Decision Process Model for Optimized Decision-Making
Date Issued
2024-11-30
Author(s)
Mitkovska-Trendova, Katerina
Trendov, Simeon
Bogatinov, Dimitar
Abstract
In the context of rapidly evolving industrial environments, optimizing decision-making for quality control is
crucial. This paper develops a Markov Decision Process (MDP) model aimed at enhancing automated quality
control and reducing scrap in manufacturing systems, addressing challenges posed by complex and uncertain
decision scenarios. The study focuses on improving the sub-key element of quality-accuracy within a
Performance Measurement System (PMS) framework, specifically targeting scrap minimization and cost
reduction. The research employs a mathematical model that integrates vector random processes, each
representing critical factors such as machine condition, operator behaviour, tools, and materials. These factors
are modeled as individual one-dimensional MDPs, which are combined to create a multi-dimensional MDP
capable of monitoring and offering optimal policy for minimizing scrap rates and costs. The research
methodology leverages advanced data analytics, statistical modeling, and real-time monitoring to accurately
estimate transition probabilities and optimize policies. Different MDP models and methods are explored to
enhance adaptability and iterative learning, allowing for optimal policy refinement over time. The proposed
model is validated through its application to a real-world printing enterprise identified critical element,
demonstrating a reduction in scrap and costs. This improvement underscores the model’s effectiveness in
practical settings, offering structured, subsystem-specific interventions that enhance manufacturing quality
control. The results hold both theoretical and practical significance. Theoretically, the study contributes to the
body of knowledge on MDP modeling for industrial quality control, providing a scalable approach that
addresses complex interdependencies and decision-making under uncertainty. Practically, the model offers a
robust tool for optimizing manufacturing processes, supported by modern IT systems, integration of advanced
technologies, predictive maintenance, and data-driven decision-making. This integrated approach enables
manufacturers to proactively identify and mitigate quality issues, enhancing operational efficiency, reducing
waste, and driving continuous improvement in industrial systems.
crucial. This paper develops a Markov Decision Process (MDP) model aimed at enhancing automated quality
control and reducing scrap in manufacturing systems, addressing challenges posed by complex and uncertain
decision scenarios. The study focuses on improving the sub-key element of quality-accuracy within a
Performance Measurement System (PMS) framework, specifically targeting scrap minimization and cost
reduction. The research employs a mathematical model that integrates vector random processes, each
representing critical factors such as machine condition, operator behaviour, tools, and materials. These factors
are modeled as individual one-dimensional MDPs, which are combined to create a multi-dimensional MDP
capable of monitoring and offering optimal policy for minimizing scrap rates and costs. The research
methodology leverages advanced data analytics, statistical modeling, and real-time monitoring to accurately
estimate transition probabilities and optimize policies. Different MDP models and methods are explored to
enhance adaptability and iterative learning, allowing for optimal policy refinement over time. The proposed
model is validated through its application to a real-world printing enterprise identified critical element,
demonstrating a reduction in scrap and costs. This improvement underscores the model’s effectiveness in
practical settings, offering structured, subsystem-specific interventions that enhance manufacturing quality
control. The results hold both theoretical and practical significance. Theoretically, the study contributes to the
body of knowledge on MDP modeling for industrial quality control, providing a scalable approach that
addresses complex interdependencies and decision-making under uncertainty. Practically, the model offers a
robust tool for optimizing manufacturing processes, supported by modern IT systems, integration of advanced
technologies, predictive maintenance, and data-driven decision-making. This integrated approach enables
manufacturers to proactively identify and mitigate quality issues, enhancing operational efficiency, reducing
waste, and driving continuous improvement in industrial systems.
Subjects
File(s)![Thumbnail Image]()
Loading...
Name
3-2-ICAIIT_2024_12(2).pdf
Size
1.31 MB
Format
Adobe PDF
Checksum
(MD5):e2ef8e984e72f9721f17dfa88e58f716