ANN based surrogate model for key Physico-chemical effects of cavitation

Area of Science:

• Fluid Dynamics
• Chemical Engineering
• Computational Science

Background:

• Cavitation generates intense localized effects like hydroxyl radicals and high energy dissipation.
• Current single cavity models are computationally expensive and difficult to integrate with larger CFD models.
• This limits the simulation of cavitation in industrial applications.

Purpose of the Study:

• To develop artificial neural network (ANN) based surrogate models for cavitation.
• To accurately represent complex physico-chemical effects of cavity collapse.
• To enable integration with CFD models for device and reactor scale simulations.

Main Methods:

• A cavity dynamics model generated training data for acoustic and hydrodynamic cavitation.
• A shallow, three-hidden-layer dense ANN was developed and trained.
• ANN performance was evaluated using unseen data for interpolation and extrapolation.

Main Results:

• The ANN models accurately predicted jet velocity, hydroxyl radical generation, and energy dissipation.
• The models demonstrated strong performance on unseen data, both within and beyond the training range.
• Algebraic equations representing the ANN were derived for CFD integration.

Conclusions:

• ANNs provide a computationally efficient method to model cavitation's physico-chemical effects.
• This approach facilitates the development of high-fidelity CFD models for cavitation devices and reactors.
• The methodology enables better simulation of key applications like water treatment and waste valorization.