Optimizing Infrastructure Inspection Spans Under Assessor Bias: A Stochastic Simulation Algorithm for Mitigating Fat-Tail Social Risks

The rapid aging of social infrastructure constructed during Japan’s high-growth period presents a critical national challenge. Following a major tunnel collapse in 2012, Japan mandated a uniform “once-every-5-years” visual inspection policy. While establishing a safety baseline, this static span, largely an administrative compromise, fails to address the severe non-linear degradation dynamics exacerbated by extreme environmental shocks in disaster-prone regions. This study challenges the adequacy of this uniform interval by developing a probabilistic framework that integrates non-linear degradation dynamics with Multiple Dependent Competing Failure Processes (MDCFPs). To bridge the gap between engineering predictions and real-world operations, we incorporate assessor bias through Bayesian modeling of imperfect inspections. Furthermore, we transition from conventional agency-cost minimization to Social Life Cycle Cost (LCC) optimization, explicitly quantifying public health risks and socio-economic externalities. Large-scale Monte Carlo simulations (N=107) reveal a critical divergence: while conventional agency-cost optimization misleadingly suggests an optimal preventive intervention point of 5–6 years, our rigorous evaluation identifies the absolute physical safety limit at exactly 2.79 years. This demonstrates that the practical 5-year default already severely breaches physical thresholds, exposing society to catastrophic fat-tail risks. These findings substantiate the urgent shift toward condition-based dynamic scheduling and highlight the necessity of implementing policy safeguards to ensure social equity for vulnerable communities.