人間工学リスク評価の最適化：ファジィ不規則セルオートマトンを用いた筋骨格系障害のモデリングへの新しいアプローチ

Musculoskeletal disorders (MSDs) and structural musculoskeletal abnormalities cause substantial work-related pain, disability, and productivity loss in industrial workforces; traditional screening tools and single-variable analyses can miss complex, interdependent risk patterns that arise from combined biomechanical exposures, workstation mismatches, and worker characteristics. We aimed to develop and validate a data-driven framework - Fuzzy Irregular Cellular Automata (FICA) - to integrate multi-modal ergonomic measurements, detect high-risk abnormality clusters, quantify their associations with MSDs and demographic predictors, and prioritize targeted ergonomic interventions. Comprehensive data were collected from 415 supervisory/administrative workers (15 objectively measured abnormalities; self-reported MSDs via the Nordic questionnaire; workstation anthropometry; exposure-time logs). FICA represented the inputs as a fuzzy graph and applied Mamdani inference with a voting/colouring routine (FICAVCA) to identify clusters and rank interventions. Model validation used repeated 10-fold cross-validation and benchmarking against alternative algorithms. FICA identified interpretable high-risk clusters (e.g., lumbar lordosis + dropped shoulder, 22.2%) and demonstrated strong predictive performance (accuracy = 0.92; stability index = 0.89). Multivariate analyses confirmed BMI, age, and work experience as key predictors (R² = 0.43, p < 0.001); BMI ≥ 30 increased odds of lumbar lordosis (OR = 2.3) and genu varum (OR = 1.9). Model-based intervention simulations estimated prioritized workstation and exercise interventions could reduce cluster risk by up to ≈ 34% (scenario projections). FICA provides an interpretable, scalable method to translate multimodal ergonomic data into prioritized, resource-efficient interventions. Longitudinal trials are needed to validate projected intervention effects.