The art of scaling VORTA with machine learning

A process that works well in the lab needs careful engineering to work well at industrial scale. Make a cavitation device larger and it can handle more flow, but its treatment performance does not grow in a straight line.

At VORTA, we use machine learning alongside engineering experience and testing to understand those changes. It helps us make better-informed choices as we scale our technology for the process it needs to serve.

Scale-up takes judgement

Inside a vortex cavitation device, rotating liquid creates tiny bubbles that collapse and produce physical and chemical effects. Those effects depend on how the liquid moves through the device. As the device grows, the flow changes too.

That makes scale-up an exercise in judgement. A larger device may deliver the capacity a customer needs while using energy differently from a smaller one. The right design has to balance flow, energy use and the treatment result.

There is an art to getting that balance right. Data helps us see where to look, and engineering experience helps us decide what to test.

How machine learning helps

Machine learning finds patterns in experimental data that can be difficult to untangle by hand. It allows us to explore how changes in device size and operating conditions may affect performance before choosing which options to test.

A recent open-access study in Ultrasonics Sonochemistry illustrates the value of this approach. Researchers combined experimental results from vortex cavitation devices of different sizes and trained models to predict their chemical activity. Their work showed that larger capacity does not automatically mean greater efficiency. Smaller devices could produce more chemical activity for the same energy input under the conditions studied. [1]

For an engineer, that is useful information. It helps focus the design work on the relationship between capacity and treatment performance, rather than assuming a successful small device will behave the same way when enlarged.

Turning predictions into a working process

We use these techniques to support our engineering decisions. Models help us compare options and plan trials. Testing shows how the selected approach performs on the actual material a customer needs to process.

That matters because each application has its own demands. A wastewater stream may need a particular level of contaminant removal. A biomass process may need better breakdown of the material. Both need a device that can handle the required flow at an acceptable energy cost.

Our approach brings machine learning, engineering judgement and practical testing together. The aim is to scale VORTA around the customer’s process, with performance checked against the outcome that matters to them.

Talk to VORTA Labs about a trial on your process stream.

[1] Pang, X., Sarvothaman, V.P., Kulkarni, S.R., Roberts, W.L. and Ranade, V.V. 2026. ‘Scale-Dependent Hydroxyl Radical Generation and Energy Efficiency in Vortex Diode Hydrodynamic Cavitation: Machine Learning Insights toward Industrial-Scale Applications in Water Treatment’. Ultrasonics Sonochemistry 131, 107932. Read the open-access paper.