Integrating Facility Layout Planning, Discrete-Event Simulation, Machine Learning, and Immersive Visualization: A Comparative Study of Two Manufacturing SMEs

Facility-layout decisions in small manufacturing firms are often evaluated with heterogeneous indicators, which complicates comparison and reproducibility. This study develops a traceable decision-support workflow and applies it to two manufacturing SME cases in Cotopaxi, Ecuador. Systematic Layout Planning and Guerchet were used as the common core, complemented by case-specific location and prioritization tools, three-dimensional discrete-event simulation in FlexSim, bounded surrogate-model screening, and Meta Quest 2-based immersive review. IPMATOAH’s model comparison increased output from 116 to 120 bales per working day (3.44%) while reducing flow-process-chart total time; VELPACK’s documented 112-to-360 bales/8 h comparison reflects the combined effects of layout redesign, machinery modernization, and capacity expansion and is therefore not attributed to layout alone. A 1200-scenario Latin Hypercube dataset was used only for exploratory screening: random forest reached 0.958 balanced accuracy with the full derived feature set, but a leakage-control analysis excluding direct target proxies reduced balanced accuracy to 0.656, confirming that field-level predictive interpretation is not warranted. Twelve domain professionals evaluated the two implemented immersive tasks (VRUC-01 and VRUC-02), yielding a SUS score of 86.25±9.74. The results support the workflow as a two-case methodological demonstration while identifying the need for replicated simulation experiments, longitudinal operational validation, and broader usability testing.