12 Apr Data assimilation for modeling cavitation bubble dynamics
Cavitation is ubiquitous in nature and engineering systems and governs processes from corrosion in propeller blades to drugs getting through the endothelial cells. Since the collapse of the cavitation bubble is so violent that it generates an intense shock wave and induces high shear stress to the surrounding liquid, we need a high temporal resolution method to characterize this phenomenon. The low temporal resolution of experimental data suggests a model-based analysis of this problem. However, high-fidelity models are not always available and these models are highly sensitive to initial input parameters, which are not always experimentally measurable, making this method less reliable. Therefore, we propose a robust characterization tool based on a novel state observer-based data-assimilation technique to overcome these two challenges in the existing methods. In this new autonomous technique, we fuse time-resolved cavitation bubble diameter measurements with the governing model to yield enhanced spatiotemporal prediction of the cavitation bubble dynamics.
The original or modified Rayleigh Plesset equation (RPE) is often used to analyze cavitation bubble dynamics. The prediction accuracy of these equations is governed by the initial values of the physical parameters. However, even for higher fidelity models, deviations from experimental measurements are observed due to the models’ underlying assumptions. Here, we present a novel state-observer data assimilation technique designed to fuse time-resolved cavitation bubble diameter measurements with a governing model to yield enhanced spatiotemporal prediction of the cavitation bubble dynamics. This technique places an observer variable in the original or modified RPE and uses a proportional–integral–derivative (PID) control law on the difference between the predicted and measured cavitation bubble diameter. The data-assimilated modeling most accurately estimates the bubble diameter and far-field pressure as the deviation of bubble diameter and far-field pressure predictions from measurements decrease by up to 90% and 60%, respectively. Although the assimilated model is not a substitute for high fidelity models, this technique overcomes the inherent model assumptions and makes the model’s outputs more robust with respect to the physical parameters’ initial values.
No Comments