As liquid-cooled data centres become larger and more power-dense, coolant condition becomes an increasingly important part of reliable infrastructure management. Small changes in coolant chemistry can create serious operational risks when they are not detected early enough.
This project involved a major internet company’s core computing hub in East China. The facility operates 1,200 liquid-cooled servers, uses PG25 coolant throughout the cooling system and supports rack power densities of up to 25 kW. Its infrastructure handles critical e-commerce and cloud computing workloads, making cooling reliability particularly important to maintaining its 99.99% service-level agreement (SLA).
The existing monitoring arrangement had several weaknesses. Conventional pH sensors experienced substantial measurement drift in low-conductivity PG25 coolant, while weekly manual testing provided limited visibility between sampling intervals. Short sensor life, demanding calibration procedures and inadequate pressure resistance also increased operating costs.
To address these problems, the facility deployed 120 RK500-12 liquid cooling-specific pH sensors. The sensors were integrated into the existing environmental and power monitoring system through RS485, establishing continuous online pH monitoring across the liquid cooling infrastructure.
Following deployment, pH accuracy remained stable at ±0.1, the false alarm rate dropped to 0%, fault detection time fell from 48 hours to 10 seconds, and no equipment damage caused by pH imbalance was recorded during the first 12 months of operation.
Project Background
The data centre serves as a core computing hub for a leading internet company in East China. Its infrastructure includes 1,200 liquid-cooled servers supporting critical e-commerce and cloud computing workloads. Rack power density reaches up to 25 kW, placing significant demands on the cooling infrastructure.
The facility uses PG25 coolant throughout the liquid cooling system. In such an environment, reliable liquid coolant quality monitoring can help operations teams understand whether important coolant parameters remain within expected conditions.
Among these parameters, pH deserves particular attention. Changes in coolant pH may indicate conditions that could contribute to corrosion or other unwanted chemical interactions within the cooling loop. If the coolant becomes excessively acidic or alkaline, components such as cold plates, pipework, fittings and seals may face additional risks.
The data centre had already installed conventional pH sensors. However, those sensors were not well suited to the operating characteristics of the PG25 liquid cooling environment.
As the number of monitoring points increased, the weaknesses of the existing system became increasingly costly. The facility therefore required a data center coolant water quality sensor solution specifically suited to continuous liquid cooling monitoring.
Challenge
1. Serious Measurement Drift and Frequent False Alarms
The first major problem involved measurement reliability.
The general-purpose pH sensors previously installed at the facility experienced substantial data drift when operating in low-conductivity PG25 coolant.
Measurement errors reached as high as ±0.6 pH.
For a system relying on automated alarms, inaccurate measurements can create operational problems even when the coolant itself has not reached an abnormal condition. In this facility, the unreliable measurements resulted in a false alarm rate of 42%.
False alarms were more than a minor inconvenience. They contributed to an average of three unplanned shutdowns every month.
This demonstrated why sensor selection needs to consider the actual coolant rather than treating all water-quality applications as identical.
2. Weekly Manual Sampling Could Not Detect Rapid Changes
The second problem was monitoring frequency.
Manual coolant testing was carried out only once per week. This left long periods during which sudden changes could occur without being identified promptly.
The consequences became clear during one incident when coolant pH fell to 4.2, indicating strongly acidic conditions. The change was not detected in time.
According to the supplied project data, corrosion and leakage subsequently occurred in the cold plates of 32 servers. The incident resulted in a direct economic loss of RMB 1.86 million and caused four hours of business interruption.
For a facility supporting critical e-commerce and cloud workloads, this level of risk made weekly manual sampling insufficient as the primary method of pH monitoring.
The facility required continuous coolant water quality sensor data capable of identifying abnormal conditions much sooner.
3. Short Sensor Lifespan and High Maintenance Costs
Maintenance created another significant burden.
The traditional glass-electrode sensors were susceptible to hydrolysis and had an average service life of only five months. Frequent replacement resulted in annual costs of approximately RMB 380,000.
Calibration was also labour-intensive.
Trained personnel had to carry calibration reagents to the monitoring locations and perform the procedure on site. Each sensor required approximately 35 minutes for calibration.
With 120 monitoring points distributed across the system, maintenance consumed around 24 person-hours each month.
A replacement monitoring solution therefore needed to improve not only measurement reliability but also maintainability and sensor service life.
4. Insufficient Pressure Resistance and Coolant Leakage
The existing sensors were rated for a maximum pressure of only 0.5 MPa.
This was unsuitable for long-term operation under the pressure conditions of the liquid cooling system. During prolonged immersion, sealing failures and coolant leakage occurred frequently.
The resulting annual failure rate reached 18%.
For a liquid-cooled server environment, leakage around monitoring equipment can introduce an additional operational risk. Therefore, the new monitoring equipment required better pressure resistance and sealing for continuous coolant exposure.
Monitoring Requirement
Based on these challenges, the facility needed to replace intermittent and unreliable monitoring with a more robust online system.
The new monitoring arrangement needed to provide stable pH measurement in low-conductivity PG25 coolant while supporting continuous operation across 120 monitoring points.
Accuracy was particularly important because measurement drift in the previous system had created a high false alarm rate. The new sensors therefore needed to deliver more stable readings under the actual coolant conditions.
Continuous monitoring was another essential requirement.
Rather than waiting for weekly manual samples, operators needed 24/7 visibility into coolant pH so that sudden changes could be identified much sooner.
● Operate reliably in low-conductivity PG25 coolant
● Support 120 monitoring locations
● Compensate for changing coolant temperatures
● Withstand liquid cooling system pressure
● Provide reliable sealing during continuous coolant exposure
● Integrate with the existing environmental and power monitoring platform
● Reduce installation and maintenance requirements
● Support centralised management of monitoring data
These requirements formed the basis of the new liquid cooling water quality sensor deployment.
Solution
The facility deployed 120 RK500-12 liquid cooling-specific pH sensors across the data centre.
A combination of pipeline and side-wall installation was used to accommodate the different monitoring locations within the cooling infrastructure.
The sensors incorporated low-impedance sensitive glass membrane technology. This was selected to provide more reliable measurement in the low-conductivity PG25 coolant environment that had caused substantial drift with the previous general-purpose sensors.
Automatic temperature compensation was also enabled.
The liquid cooling system operated across a temperature range of 0°C to 55°C. Automatic temperature compensation allowed the pH monitoring system to account for temperature changes during operation and maintain more consistent measurement performance.
Pressure resistance and sealing were also important elements of the solution.
The deployed sensors were specified for pressure resistance below 1 MPa, equivalent to 10 bar, and used an end-face sealing design. This enabled them to remain continuously exposed to the coolant while addressing the sealing problems experienced with the previous equipment.
For installation, the sensors used a G3/4 standard threaded interface. This allowed them to be installed without modifying the existing pipework. Each sensor required approximately 10 minutes to install.
Deployment & Commissioning
The 120 sensors were deployed using both pipeline and side-wall installation methods according to the requirements of individual monitoring points.
The G3/4 threaded connection simplified physical deployment because major alterations to the existing liquid cooling pipes were not required.
Once installed, the sensors were connected to the facility’s environmental and power monitoring system using RS485 communication.
This integration was important because the customer required centralised management rather than isolated local measurements.
Following commissioning, pH information from the 120 monitoring points became available continuously through the existing monitoring environment. The facility could therefore move away from depending on weekly manual sampling for routine observation.
The system established 24/7 real-time online pH monitoring, giving operations personnel much greater visibility into coolant conditions.
Temperature compensation was also configured to support the system’s operating temperature range from 0°C to 55°C.
This combination of physical installation, RS485 integration and continuous monitoring created a more practical approach to coolant management across the 1,200-server facility.
Rated Products
RK500-12 Liquid Cooling-Specific pH Sensor
The RK500-12 liquid cooling-specific pH sensor was the main monitoring product deployed in this project.
It was selected specifically for pH measurement in the low-conductivity PG25 coolant environment. The low-impedance sensitive glass membrane technology helped address the measurement drift experienced with the facility’s previous general-purpose sensors.
Automatic temperature compensation supported operation across coolant temperatures from 0°C to 55°C.
The sensor also provided pressure resistance below 1 MPa (10 bar) and used an end-face sealing design for continuous operation while exposed to coolant.
A standard G3/4 threaded interface supported relatively straightforward installation without requiring modifications to the existing pipework.
RS485 communication enabled the sensors to transmit pH measurements to the facility’s central environmental and power monitoring system.
For liquid cooling systems where acidity and alkalinity are critical monitoring parameters, a specialised coolant pH sensor can therefore provide continuous data that is difficult to achieve through periodic manual sampling alone.
Supporting Coolant Water Quality Monitoring
pH is only one potential parameter in a broader liquid coolant quality monitoring strategy.
Depending on the coolant chemistry and cooling-system requirements, other water-quality parameters may also need to be observed.
For example, a conductivity sensor for coolant monitoring can provide information about changes in ionic content. Conductivity monitoring can complement pH measurement where the cooling-system design requires multiple indicators of coolant condition.
The exact combination of sensors should be determined by the coolant type, operating environment and monitoring objectives of each data centre.
Results
The new monitoring system produced substantial improvements across measurement reliability, fault detection, maintenance and system availability.
✔ More Stable pH Measurement
Following deployment, measurement accuracy remained stable at ±0.1 pH.
The false alarm rate, previously 42%, dropped to 0%.
This was particularly important because false alarms from the previous sensors had contributed to unnecessary operational disruption.
✔ Fault Detection Reduced From 48 Hours to 10 Seconds
Continuous online monitoring dramatically shortened the time required to identify abnormal pH conditions.
Fault detection time fell from 48 hours to only 10 seconds.
During the first 12 months of operation, no equipment damage caused by pH imbalance was recorded.
This represented a significant improvement over the previous arrangement, where weekly manual testing had failed to provide timely warning before a pH 4.2 incident damaged cold plates across 32 servers.
✔ Longer Sensor Life and Lower Maintenance Costs
The average sensor service life increased from five months to 12 months.
As a result, annual replacement costs fell by 71%.
The maintenance workload was also reduced by 75%, addressing the substantial amount of time previously required to calibrate and maintain 120 monitoring points.
✔ Leakage Failure Rate Reduced to Zero
The original monitoring equipment had an annual failure rate of 18%, partly because of insufficient pressure resistance and sealing problems.
After the new sensors were deployed, the sensor sealing failure and leakage rate dropped to 0%.
This removed one of the major reliability concerns associated with the previous equipment.
✔ Lower Losses From Unplanned Downtime
The financial impact was also substantial.
Annual losses associated with unplanned downtime decreased from RMB 2.1 million to RMB 150,000.
At the same time, the facility’s SLA compliance rate increased from 99.92% to 99.991%.
For a data centre supporting critical e-commerce and cloud workloads, even a small improvement in availability can be operationally significant.
✔ Lower Sensor Power Consumption
Each sensor consumed less than 0.2 W.
According to the project data, the monitoring system saved approximately 1,200 kWh of electricity annually compared with the original sensors.
This contributed to a 0.003 reduction in the facility’s Power Usage Effectiveness (PUE).
Although coolant monitoring represents only one small part of overall data centre energy consumption, efficient monitoring equipment can still contribute to wider infrastructure-efficiency goals when deployed at scale.
From Weekly Sampling to Continuous Coolant Monitoring
This East China data centre project demonstrates why liquid cooling monitoring equipment needs to be selected for the actual coolant and operating conditions rather than relying on general-purpose water-quality sensors.
The previous monitoring system suffered from measurement errors of up to ±0.6 pH, a 42% false alarm rate, short sensor life, high maintenance requirements and an 18% annual sensor failure rate. Weekly manual sampling also left the facility vulnerable to rapid changes between tests.
Deploying 120 specialised RK500-12 pH sensors changed the monitoring approach from intermittent testing to 24/7 online observation.
Measurement accuracy stabilised at ±0.1 pH, the false alarm and sealing failure rates fell to 0%, and abnormal conditions could be detected in seconds rather than hours or days. At the same time, longer sensor life and reduced maintenance helped lower the operational cost of managing the monitoring network.
For high-density data centres using PG25 and similar coolants, an appropriately designed data center coolant water quality sensor system can provide more frequent visibility into coolant conditions while supporting centralised monitoring, preventive maintenance and reliable liquid cooling operation.
