Enhancing Human–Machine Collaboration to Reduce Production Waste in Customized Building Component Manufacturing

Frequent manufacturing changeovers in customized building component production create long time and material waste as operators struggle to identify proper process parameters for diverse product variants. Although human operators possess contextual knowledge, such as how material behavior impacts product qualities, the vast configuration space and uncertainties often exceed what purely manual trial and error can efficiently handle. Meanwhile, advanced artificial intelligence (AI)-based methods (e.g., Bayesian optimization and reinforcement learning) promise to automate parameter tuning but may overlook subtle domain insights crucial for achieving efficient changeovers. To address these challenges, this study proposes a human–AI collaboration strategy that combines the intuitive expertise of operators with data-driven algorithms. Humans guide the algorithms by providing informed priors and explicit domain rules, constraining the search space, and expediting process tuning. Conversely, AI explores beyond human intuition to refine and validate these parameters under different operating scenarios. Tested on the developed interactive simulator ProSimX, our findings highlight the potential of the human–machine collaboration strategy by merging operator knowledge with algorithmic power, offering a pathway to improved quality and efficiency in building component production.