The critical cooling water system at Glass Futures St Helens, is a vital utility supporting pilot-scale glass manufacturing, research and innovation activities. The system must operate reliably to maintain thermal stability, protect assets and ensure safe operation of furnaces, forming equipment and ancillary processes.

Traditional monitoring approaches rely on fixed alarm thresholds and reactive maintenance, which can struggle to detect early-stage degradation such as fouling, pump inefficiency or valve performance drift.

Unplanned downtime or sub-optimal cooling performance presents risks to equipment integrity, experimental outcomes and operational efficiency.

Proposed Solution

• The Anomalyse anomaly detection platform has been deployed to provide continuous, model-based monitoring of the critical cooling water system.

• Using machine learning-driven behavioural models, Anomalyse learns what “normal” operation looks like across multiple operating modes and identifies any emerging anomalies before they trigger alarms or cause failures.

The solution is implemented in a non-intrusive manner by integrating with the existing Siemens SIMATIC PCS Neo control system via the Siemens SIMATIC Process Historian, ensuring no impact on real-time control or safety systems.

Benefits:

• Early Fault Detection: Identify developing issues before they result in alarms, trips, or equipment damage.

• Improved Reliability: Reduce unplanned downtime of critical cooling infrastructure.

• Optimised Maintenance: Shift from reactive to predictive maintenance strategies.

• Operational Insight: Gain a deeper understanding of cooling system performance and interactions.

• Low-Risk Deployment: Read-only historian integration ensures no impact on control or safety systems.

• Future Scalability: Architecture can be extended to other utilities or process systems across the Glass Futures facility.

Solution Architecture

Data Sources:

Process data is sourced from Siemens PCS 7 Neo, including key cooling water variables such as flow rates, supply and return temperatures, differential pressures, pump status & power consumption, and valve positions

These signals are already collected and stored in the Siemens SIMATIC Process Historian.

Data Integration:

• The Siemens data historian connects securely to the Anomalyse platform using the secure API.

• Historical data is used initially to train the models, while live data streams are ingested continuously for real-time monitoring.

• No direct connection to the control layer is required.

Analytics and Anomaly Detection:

The platform builds multivariate models that capture normal system behaviour across different loads, seasons, and operating regimes.

Rather than relying on static thresholds, the platform detects deviations in relationships between variables, enabling early identification of issues such as:

• Gradual pump degradation

• Blockages or fouling in cooling circuits

• Heat exchanger efficiency loss

• Control valve or sensor drift

• Abnormal operating patterns during transitions

Visualisation and Alerts

Detected anomalies are presented through the Anomalyse dashboard, which features:

• Anomaly scores indicating the magnitude of abnormality

• Score contributions from each model variable, explaining what exactly has changed

• User-calibrated alerts indicating events of interest

Alerts can be configured to notify operations and maintenance teams, supporting timely investigation and corrective action.

Want to learn more, or talk about how anomalyse can help your business improve productivity through Anomaly Detection? Contact sales@anomalyse.io today.