AI-Driven Coupled Simulation of Unplugged Freezer Thermal-Biological Dynamics for Food Safety During Power Outages

Abstract

Power outages in resource-constrained environments create acute food-safety hazards by allowing frozen contents to warm through the bacterial danger zone. This paper presents a coupled thermal-biological simulation of an unplugged household freezer, conceived as an open-ended student project. The project extended into a research contribution, integrating a three-stage lumped-capacitance heat-transfer model with Ratkowsky microbial growth kinetics and an AI-assisted formative feedback framework. Analytical results validated by Python numerical simulation (RK45) confirm that freezer contents reach the 4 °C food-safety threshold in approximately 10.4 hours under standard conditions and in 7.65 hours under Libyan summer conditions (Troom = 35 °C). A Monte Carlo parametric study (N = 500, NumPy seed 42) quantifies spoilage probability under Libyan-specific ambient conditions, showing an 80% probability of unsafe food within a 10-hour outage in summer. An AI feedback layer, the Anthropic Claude API, whose operation is documented in full, including representative input-output exchanges, reduced student turnaround from 3.8 days to under 90 seconds and improved final report quality by 0.6 grade points. The framework is directly applicable to food safety, pharmaceutical cold chains, and vaccine storage in off-grid Libyan facilities.