<![CDATA[Manual temperature recording during the heat treatment process is prone to human error and does not support continuous real-time monitoring. This research develops and evaluates an IoT-based temperature monitoring system featuring offline buffering, WiFi disruption detection-recovery, and catch-up synchronization mechanisms. The system uses a three-layer architecture (hardware, backend, frontend) based on an ESP32 microcontroller with dual DS18B20 sensors, a PHP/MySQL backend, and an HTML5/JavaScript dashboard. Five key technical designs were validated through quantitative testing on a controlled local network simulating the plant's operational conditions. The measured real-time data latency averaged 0.7 and 0.6 seconds for the two sensors, well under the 2-second target. Offline buffering and catch-up synchronization recovered 99.5% of buffered data (199 of 200 points) in the initial trials, the single failure attributed to a since-remediated firmware defect. Post-remediation trials showed no loss of retained data under near-capacity and overflow conditions and idempotent handling of REST API failures, although data buffered before a restart was lost. Zombie WiFi detection functioned reliably under both critical and self-recovering disruption conditions. REST API uptime reached 99.32% and 99.49% across day- and night-shift simulations, and DataMerger ran without producing duplicate entries or data gaps. An alternating timeout characteristic was additionally observed in the single-threaded ESP32 firmware architecture. The system met the proposed targets under the tested conditions, demonstrating that a layered network resilience approach can improve the reliability and integrity of industrial temperature monitoring. However, validation under actual heat treatment chamber conditions remains a direction for future work.]]>
