Kausalyze is proud to be working with Huntsman Polyurethanes to apply causal AI to one of the most persistent challenges in process manufacturing - understanding and managing heat exchanger fouling.
Fouling is a common problem across process manufacturing, and is characterised by the gradual build-up of unwanted deposits on heat transfer surfaces. As the deposits accumulate, the thermal efficiency of the asset falls, driving up energy use and forcing plants into unplanned cleaning and maintenance interventions. The cumulative cost is significant, with estimates published in Heat Transfer Engineering putting it at around 0.25% of GDP in industrialised economies and as much as 2.5% of total carbon dioxide emissions.
To some degree, this is an unavoidable side effect of operation. But fouling is commonly accelerated by changes in upstream conditions, which ripple through the process and show up in dozens of correlated variables at once.
Conventional predictive maintenance tools are designed to help engineers spot when an asset's performance begins to decline. But they typically model the asset in isolation, extrapolating from its own historical behaviour with no view of the wider process around it. They cannot link deterioration to its upstream drivers or separate the conditions accelerating fouling from the ones merely moving alongside it.
The result is forecasts that hold only while operating conditions remain stable, but turn unreliable at exactly the moment they are needed most - when something upstream changes. Engineers are left to interpret alerts through experience and intuition, and that expertise is becoming harder to retain across industry.
Kausalyze takes a different approach. Rather than predicting failures from correlations in historical data, its platform learns the causal structure of the process itself, combining plant data with engineering knowledge to map how operating conditions actually drive degradation. The result is not just a prediction but an explanation, a causal pathway that engineers can interrogate, challenge and act on.
That distinction is at the centre of the collaboration with Huntsman, where the causal models have been tailored to Huntsman's process conditions and reviewed jointly with their engineers.
Dr. Ir. Ramon Scheffer, Data Scientist, Process Improvement, Global Excellence Team, Huntsman Polyurethanes, explains:
“Kausalyze gives us deeper insight into asset behaviour than conventional predictive maintenance tools. Together with our engineering teams, it is tailored to our process conditions and helps pinpoint key drivers of heat exchanger fouling and degradation.
“It also shows potential to detect fouling or failure-related events earlier, enabling better maintenance planning, fewer unplanned stops, and improved operations while we continue to validate performance across operating regimes.
“A key advantage is explainable causal pathways: combining data-driven models with engineering expertise clarifies complex interactions and how operating parameters influence performance loss - supporting corrective action. We will keep working together to validate accuracy, robustness, and workflow integration before considering broader deployment.”
The collaboration has been a joint engineering effort from the start, with both organisations contributing specialist knowledge and reviewing the evidence together.
Ir. Jasper Rutten, Digital Manufacturing Director, Global Excellence Team, Huntsman Polyurethanes, said:
“Kausalyze's AI-driven process fault detection and causal analysis has shown promising early results in helping us identify patterns in our process data and improve our understanding of drivers of heat exchanger fouling and degradation.
“The collaboration has been productive and well received by our technical and operational teams. We are encouraged by the initial validation of the causal models and the usability of the visualisation tools in supporting joint engineering review.
“At this stage, our focus is on further validation and on assessing whether the approach can be used to predict as well as extend heat exchanger lifetime. We look forward to continuing the collaboration and evaluating how the solution can be integrated into our workflows before considering any broader scaling.”
For Kausalyze CEO, Dr Louis Allen, the collaboration demonstrates how causal AI can help engineering teams extract greater value from the data already available to them, and how a focused asset model can provide the foundation for a much broader view of plant performance.
“Process manufacturers do not simply need more alerts. Engineers need to understand what is driving a developing problem, how assets and process conditions interact and what they can do about it. The progress we are making with Huntsman comes from combining its deep process and engineering expertise with Kausalyze’s causal modelling, then validating the results together against real operational experience.
“We look forward to continuing that work and exploring how the model can be extended from this initial heat exchanger application to other assets and process areas. Over time, that creates the potential for a connected causal model of the whole plant, showing how operating conditions, equipment and process interactions combine to drive reliability, efficiency and performance.”
The current heat exchanger work is both a practical application and a building block. Each validated model adds to the understanding of asset behaviour and the upstream and downstream relationships that influence it.
Extending the approach across connected equipment and process areas can provide a plant-wide causal view, helping teams understand not only what is happening to an individual asset, but how changes in one part of the process influence outcomes elsewhere.
Both Huntsman and Kausalyze remain committed to continuing the collaboration, assessing the accuracy and robustness of the models across different operating regimes, while also evaluating how the technology can be integrated into established engineering workflows.