The modular building industry is increasingly challenged by small, highly customized orders, resulting in frequent changeover that can consume up to 30% of total production time. Reducing overall changeover time is crucial to address the high costs and scalability limitations in modular housing. Optimizing production sequences to minimize sequence-dependent changeover times is essential for reducing delays in customized building manufacturing. However, the lack of historical data and understanding of factors affecting changeover time for unobserved customization-driven changeover types hinders job-shop scheduling optimization and changeover time reduction. While data-driven methods are often employed to predict production-related times, they suffer from domain shift issues when models trained on standard products are applied to customized products. The absence of mechanisms to explain invariant causal relationships that drive changeover time changes during production reconfiguration further reduces the effectiveness of domain adaptation. This study addresses these challenges by introducing a causal-informed Generative Adversarial Network (GAN) framework that enhances prediction accuracy by incorporating causal relationships revealing the changeover time change mechanism. Integrating causal analysis, this framework identifies key factors that influence changeover times, enabling the GAN to generate synthetic data aligned with these relationships. The proposed method effectively adapts to domain shifts, thereby improving the accuracy of changeover time predictions, which in turn supports better job-shop scheduling and reduces time waste in customized manufacturing. In a case study involving the production of customized ductwork for modular buildings, the application of this causal-informed GAN framework demonstrated an improvement in prediction accuracy and generated optimized sequences that achieved 12%–25% reductions in total changeover time compared to historical production sequences. This indicates the practicality and effectiveness of the approach in highly customized manufacturing environments, providing support for optimizing production processes and lowering costs.

The performance of point cloud-based defect classification algorithms depends on the quality of the computed geometric features, which in turn is strongly affected by the selection of the neighborhood size of local 3D points. Many existing algorithms select a single scene-specific value as the neighborhood size parameter and apply it uniformly to all 3D points while computing features. By doing so, some defect features end up being smoothened out at locations where the selected neighborhood size is larger than the optimal choice for those points. Conversely, geometric features might not always capture the local 3D structure if they are calculated based on less than optimal number of neighbors. This paper investigates and assesses the relationship between neighborhood size selection and the performance of point cloud-based unsupervised spall classification algorithms in a quantitative manner. Among the presented neighborhood selection approaches, an entropy-based approach that incorporates a tailored optimal neighborhood size for every point in a point cloud, resulted in a significant improvement over the performance of current state-of-the-art approaches. The performed quantitative study also demonstrated the robustness of this approach to variables, such as subsampling percentage, maximum neighborhood size and different noise levels. The implemented research testbed is comprised of statistically significant number of spall defect datasets generated from five different bridges. The outcome of this research is expected to improve the reliability of point cloud-based condition assessment for concrete bridges.