Table of Contents
Problem Statement
OLI Solution: Digital Twin Powered by First-Principles Thermodynamics and ML
Hybrid Model Workflow
Benefits to Users
Objective
Cooling towers are critical to industrial operations, but their management often suffers from limited visibility, unreliable measurements, and vendor-driven decisions. OLI’s Cooling Tower Digital Twin pilot solution integrates OLI Flowsheet: ESP, the OLI Process API, and machine learning (ML) to create a smart, self-optimizing digital twin of your cooling tower system. This unified platform empowers operators with actionable insights to reduce costs, improve reliability, and enable informed water chemistry decisions.
Problem Statement
Cooling towers are high-stakes infrastructure:
- Massive water usage (322 billion gallons/day in the U.S.)
- Significant operational costs (chemical, water, loss of tower efficiency)
- Serious reliability risks from scaling, corrosion, and poor heat transfer at the process heat exchangers
Operators often rely on:
- Infrequent lab testing and faulty sensors (e.g., pH drift)
- Limited tools to predict or prevent failures
-Minimal insight into the effect of real-time operating parameters on scaling
OLI Solution: Digital Twin Powered by First-Principles Thermodynamics and ML
1. OLI Flowsheet: ESP – The Chemistry Engine
At the heart of the solution is a robust, first-principles-based flowsheet model that:
- Simulates real operating conditions of cooling towers
- Calculates scaling tendencies based on full ionic speciation
- Forecasts the impact of blowdown, CoC changes, and heat loads
Figure 1. OLI Flowsheet: ESP model of a prototypical cooling tower system.
2. OLI Process API – Connecting Real-Time Data to the Cloud
The Process API makes Flowsheet’s powerful simulations available in the cloud, in real time:
- Pulls data from historians (flow rates, conductivity, weather, etc.)
- Triggers cloud simulations with up-to-date inputs
- Returns outputs such as optimized blowdown rate, scaling risk, and species concentration
For more information on the API, please see our Getting Started with OLI Process API article.
3. Machine Learning – Bridging Data Gaps
Data gaps are inevitable—pH meters drift, sensors fail, and some ions are rarely measured. This is where ML can serve as a soft sensor for pH and other unmeasured variables.
In OLI’s cooling tower solution:
- A Partial Least Squares (PLS) model built with the PLSRegression module from scikit-learn¹⁻³ predicts critical values like pH using reliable proxies (such as conductivity)
- Enables accurate scaling and precipitation risk assessment even when pH data is incomplete
- Supports hybrid modeling: using both measured and predicted values for higher model confidence
This ML model is trained on plant-specific historical data and several pieces of analytical data from the lab.
Hybrid Model Workflow
- The trained PLS model predicts pH.
- Process API runs the water analysis at the specified pH and ionic composition.
- This reconciled water stream enters the full system simulation in Process API, which runs the cooling tower model previously built in Flowsheet: ESP.
- An optimization calculation recommends the optimal blowdown rate for minimizing scale risk. For more details on this optimization process, please see our article on Realtime Cooling Tower Optimization with OLI Process API.
Figure 2. Cooling tower hybrid model solution workflow.
Benefits to Users
- Optimized Water and Chemical Use: Control CoC dynamically, based on real-time risk—not guesswork.
- Robust real-time predictions of scaling based on live operating conditions, even under the limiting factors of poor physical sensors, missing data, or infrequent water analysis.
- Seamless Integration: Designed to work with any historian or dashboard system.
- Insights into Tower Health: Quickly compare how your tower is performing against internal KPIs and design values to assess the efficiency of your system.
- Identify areas of improvement and optimize tower performance.
OLI’s Commercial-off-the-Shelf (COTS) Cooling Tower solution features an intuitive interface that highlights key variables such as CoC, scaling risks, and optimal blowdown.
Interested in learning more?
This solution is designed for engineers and operators seeking data-driven, cost-saving strategies in cooling tower operations. To request access or learn more about deployment options, please contact us at support@olisystems.com.
References
- Buitinck, L.; Louppe, G.; Blondel, M.; Pedregosa, F.; Mueller, A.; Grisel, O.; Niculae, V.; Prettenhofer, P.; Gramfort, A.; Grobler, J.; Layton, R.; Vanderplas, J.; Joly, A.; Holt, B.; Varoquaux, G. API Design for Machine Learning Software: Experiences from the Scikit-learn Project. In ECML PKDD Workshop: Languages for Data Mining and Machine Learning, 2013; pp 108–122.
- Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V.; Vanderplas, J.; Passos, A.; Cournapeau, D.; Brucher, M.; Perrot, M.; Duchesnay, É. Scikit-learn: Machine Learning in Python. J. Mach. Learn. Res. 2011, 12, 2825–2830.
-
Scikit-learn Developers. sklearn.cross_decomposition.PLSRegression — Partial Least Squares Regression. Scikit-learn Documentation. https://scikit-learn.org/stable/modules/generated/sklearn.cross_decomposition.PLSRegression.html (accessed May 12, 2025).