Physics-Informed Diagnostics for Food Processing Technology Assets in Kenya: Out-of-Distribution Fault Detection and Lifecycle Cost Trade-offs

Food processing enterprises in Kenya operate technology assets under conditions that diverge sharply from the nominal operating envelopes assumed by equipment manufacturers. This conceptual and methodological article develops a physics-informed diagnostic framework that addresses the dual problem of out-of-distribution fault detection and lifecycle cost optimisation for these assets. The analysis formalises a hybrid diagnostic architecture in which physics-based residual generation is coupled with statistical change-point detection, thereby enabling fault identification when operational data shift beyond the support of training distributions. A lifecycle cost model is then derived to quantify the trade-off between diagnostic sensitivity and maintenance expenditure, demonstrating that optimal detection thresholds depend on asset criticality, failure consequence and the cost structure of inspection. The framework is developed as a set of formal propositions linking detection delay, false alarm rate and expected lifecycle cost, with particular attention to the constraints of low-instrumentation environments common in Kenyan food processing.