AI-Driven Threshold Leak Remediation in Datacenter Liquid Cooling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Datacenter liquid cooling systems face inefficiencies with air cooling and are prone to damage from liquid leaks, which existing technologies fail to effectively detect and mitigate, especially for threshold leaks that do not immediately affect normal operation.

Innovation Solution

A remediation system incorporating a rack-mounted power distribution unit (PDU) with AI/ML-based sensors that detect threshold leaks by predicting parameter changes, triggering a power controller to adjust the computing component's power state and a flow controller to modify coolant flow, thereby preventing damage and maintaining normal operation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Temperature

If liquid cooling systems are used to draw heat away from server components, then cooling efficiency is improved, but the system becomes susceptible to leakage damage that can short and damage equipment

Engineering Contradiction:

Improvecooling efficiencyVSAvoidleakage damage

Core Design Contradiction:

TemperatureVSObject-affected harmful factors

Solution Approach 1:

The system performs preliminary detection of threshold leaks using sensors that monitor coolant parameters before actual leakage damage occurs. The learning subsystem analyzes parameter changes predictively, and the system initiates remediation actions (power state changes, flow adjustments) in advance to prevent equipment damage from full-scale leaks.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements continuous feedback monitoring through sensors that track coolant temperature, flow rate, and pressure. The learning subsystem processes this feedback data to detect threshold leak conditions, and the control system adjusts cooling parameters in response to maintain safe operation while preventing damage.

Inventive Principle:

Principle #23Feedback

2Reliability

If existing leak detection technologies are used, then some leakage can be detected, but threshold leaks that do not immediately affect normal operation cannot be effectively detected or mitigated

Engineering Contradiction:

Improveleak detection capabilityVSAvoidthreshold leak detection

Core Design Contradiction:

ReliabilityVSDifficulty of detecting and measuring

Solution Approach 1:

The system monitors multiple coolant parameters (temperature, flow rate, pressure) and detects threshold leaks by identifying subtle parameter changes that indicate incipient leakage conditions. The learning subsystem analyzes patterns in these parameter changes to predict threshold leaks before they develop into immediate failures, enabling detection of conditions that existing technologies miss.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system replaces traditional mechanical leak detection methods with AI/ML-based sensor systems that analyze parameter changes and predict threshold leaks. The learning subsystem uses computational algorithms to detect subtle indicators of threshold leaks that conventional mechanical or threshold-based detectors cannot identify.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If the computing component continues operating during a threshold leak, then productivity is maintained, but the risk of equipment damage increases

Engineering Contradiction:

Improvecomputational outputVSAvoidequipment damage risk

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The system takes preliminary remediation actions when threshold leaks are detected, adjusting power states and coolant flow before equipment damage occurs. These preventive measures allow continued operation while reducing risk, enabling the system to maintain productivity during incipient leak conditions without compromising safety.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements beforehand cushioning by modifying operating parameters (reducing power state, adjusting flow rates) in response to detected threshold leaks. These cushioning measures reduce the stress on the system and mitigate the risk of equipment damage while allowing continued operation, thus protecting equipment without completely shutting down productivity.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS11864359B2Intelligent threshold leak remediaton of datacenter cooling systems

Publication Date: 2024.01.02 NVIDIA CORP

AI summary

A remediation system for threshold leaks in a datacenter liquid cooling system is disclosed. The system includes a fluid controller and a power controller that are adapted to receive input from a learning subsystem that can determine that a threshold leak has occurred even though a computing component is functioning normally, so that a change in power state to reduce reliance on the coolant and so that a change of flow of the coolant may be effected.