Smart IoT-Based Leak Forecasting and Detection for Energy-Efficient Liquid Cooling in AI Data Centers

TL;DR

This work tackles coolant-leak risk and energy waste in GPU-centric liquid-cooled data centers by proposing a smart IoT framework that fuses LSTM-based probabilistic time-to-leak forecasting with Random Forest-based instant leak detection, all orchestrated through MQTT, InfluxDB, and Streamlit. The approach delivers probabilistic forecasts with calibrated intervals and rapid detection, validated on ASHRAE 2021-aligned synthetic data, achieving up to ~87% accuracy for 2-hour ahead predictions within a $\pm$30-minute window and a 96.5% F1-score for detection. Key findings show humidity, pressure, and flow as strong immediate indicators, while enclosure temperature exhibits thermal inertia, guiding sensor deployment and alerting strategies. The integrated system demonstrates potential energy savings (~1,500 kWh/year for a typical 47-rack facility) via proactive maintenance, though empirical validation in real operational data centers is required before deployment.

Abstract

AI data centers which are GPU centric, have adopted liquid cooling to handle extreme heat loads, but coolant leaks result in substantial energy loss through unplanned shutdowns and extended repair periods. We present a proof-of-concept smart IoT monitoring system combining LSTM neural networks for probabilistic leak forecasting with Random Forest classifiers for instant detection. Testing on synthetic data aligned with ASHRAE 2021 standards, our approach achieves 96.5% detection accuracy and 87% forecasting accuracy at 90% probability within plus or minus 30-minute windows. Analysis demonstrates that humidity, pressure, and flow rate deliver strong predictive signals, while temperature exhibits minimal immediate response due to thermal inertia in server hardware. The system employs MQTT streaming, InfluxDB storage, and Streamlit dashboards, forecasting leaks 2-4 hours ahead while identifying sudden events within 1 minute. For a typical 47-rack facility, this approach could prevent roughly 1,500 kWh annual energy waste through proactive maintenance rather than reactive emergency procedures. While validation remains synthetic-only, results establish feasibility for future operational deployment in sustainable data center operations.

Figures (7)

• Figure 1: System architecture showing data flow from IoT sensors in rack enclosures through MQTT broker and InfluxDB storage to dual ML models (LSTM forecasting and Random Forest detection) with Streamlit dashboard for real-time monitoring and alerts.
• Figure 2: Correlation matrix showing pressure-humidity inverse correlation ($r=-0.50$), humidity-leak strong positive correlation ($r=0.70$), and temperature independence ($r \approx 0.01$).
• Figure 3: Pressure distribution showing clear separation between normal (leak=0, $\sim$2.0 bar) and leak conditions (leak=1, $\sim$1.7-1.9 bar).
• Figure 4: Humidity distribution showing dramatic separation: normal (leak=0, $\sim$30% RH) vs leak (leak=1, 35-40% RH spread).
• Figure 5: Flow rate separation: normal (leak=0, $\sim$1.5 L/min) vs leak (leak=1, $\sim$1.35-1.45 L/min).