Integrating Risk into Skill-Based Manufacturing: A Modular Platform for MultiObjective Optimization

Modern manufacturing systems are increasingly characterized by reconfigurable machinery, volatile production demands, and exposure to operational disturbances such as delays, equipment degradation, and failures. While optimization- and AI-based production planning methods have advanced significantly, there is still a lack of integrated and reproducible platforms for modeling, comparing, and benchmarking these approaches under realistic, risk-informed manufacturing conditions. This paper presents a modular software platform for multi-criteria, riskinformed production planning in skill-based manufacturing systems. The platform models a factory as a composition of machines, skills, and products, represented through modular classes that capture executable capabilities and product transformations. A state-transition formulation maps the current factory state to feasible skill executions, enabling the generation of valid production plans that transform an initial state into a desired product configuration while respecting physical and logical constraints. Each skill execution is associated with execution time, energy consumption, and reliability attributes. This unified representation supports the joint optimization of competing objectives, including minimizing energy consumption and makespan, maximizing reliability, and reducing overall production risk. The platform enables systematic benchmarking of diverse planning and optimization strategies, ranging from classical methods (e.g., A*, mixed-integer linear programming, constraint programming, and genetic algorithms) to AI-based approaches (e.g., reinforcement learning, symbolic planning, and graph neural networks). A common modeling and execution interface ensures consistent problem definitions and reproducible performance comparisons across algorithms and scenarios. Key contributions include: (i) a machine-skill-product meta-model linking symbolic process descriptions with executable planning, (ii) a state-aware action generator that enforces physical and logical feasibility, and (iii) a benchmarking environment for analyzing trade-offs between competing objectives. The proposed platform provides a foundation for integrating risk and reliability into intelligent production planning, thereby advancing safer, more efficient, and more trustworthy manufacturing systems. The platform is open-source and available via a public Git repository.

Keywords: Risk-informed production planning, Skill-based manufacturing, AI-based scheduling, Multi-objective optimization, Reliability in manufacturing.