Wim's Water Wisdom - March 2025

In the midst of the AI hype, let’s not forget the laws of physics, chemistry and biology

[Wim’s Water Wisdom - AM-Team's monthly opinion piece - March 2025]

As a modelling company founded in 2017, we have made it part of our mission to show our industry that mechanistic models can describe systems very well and provide tremendous value in planning, design and operation. Mechanistic models are based on solid laws of physics, chemistry and biology, developed throughout centuries. In specific cases, these ‘white box’ models can be even more accurate than experimental data. The latter is something we avoid mentioning though, as it creates disbelief. Given the fact that these models have ‘a priori knowledge’ (i.e. they know stuff before they have seen new data), another important feature is the fact that these models can operate in data-scarce situations. I still remember a meeting last year in which data scientists couldn’t believe what I was saying.

Recently we witnessed the rise of AI. AI belongs to the other family: the data-driven models. In contrast to mechanistic models, these models are built based on big amounts of data and not based on human knowledge encapsulated in the natural laws. They are hence called ‘black box’. Their big strength is the fact that they can make useful sense out of enormous amounts of data without the need of understanding the mechanisms of the systems you are modelling. At AM-Team we are applying them in situations where no adequate mechanistic models are available and high quantity and quality data is available. And we are blending them with mechanistic models to create very powerful hybrid models. They can be ‘blended’ in multiple ways, but that is part of another discussion. A disadvantage of these black box models for curious operational staff or engineers is their structure that doesn’t provide explanation to humans (i.e. ‘data in, data out’).

In a very short timeframe (i.e. post covid), AI has shaped the discussions around digital water, to such an extent that digital water has almost become synonymous with AI or data science. The intensity is great and something we embrace, as it will accelerate the adoption of digital water. However, the focus on AI has been so big that our industry seems to be forgetting the white-box models, even though they have been applied for many decades for planning and plant design. We believe that a more balanced discussion would bring us faster to lasting value from digital because we would be applying the right tool for the right purpose and smartly combining these tools from the start.

A great thing we can now create is the blend between data science and water engineering. Let us embrace a mixture of tools, a ‘models-fit-for-purpose’ approach. And let the data-scientists work together with the process engineers. It will not only serve the business case. It will also lead to human enablement. Let’s continue surfing the wave of AI, and mix it with the power and transparency of the laws of physics, chemistry and biology.
‍

Wim's Water Wisdom is a monthly op-ed from Dr. Wim Audenaert, CEO & Co-founder of AM Team shedding light on today's and tomorrow's challenges, opportunities and trends in the water industry. This month's piece invites for a balanced approach to digitalisation in water and wastewater treatment.

AM-Team helped designing, optimising and digitalising over 150 treatment plants in 22 countries. The company is globally known for a unique approach on treatment plant modelling, combining vision, innovation and grounded practicality.

Visit https://am-team.com to consult educational content and sign up to AM-Team's newsletter and upcoming webinars.

‍