Mechanistic modeling is based on the underlying physical principles governing the system, whereas empirical modeling relies on observed patterns in the data. The mechanistic approach provides a more fundamental understanding of the real-world processes involved. Our pursuit is to uncover the underlying laws of nature, often written in the language of mathematics.
EXAMPLE: MODELING OF INHERENT INDIVIDUAL BUBBLE SIZE DISTRIBUTION AT HIGH HEAT FLUX
During high heat flux boiling, it's very challenging to experimentally measure the discrete bubble behavior due to the bubble interaction and coalescence. The measured apparent individual (i.e., non-coalesced) bubble radius distribution of a site i differs from the inherent nucleating bubble size distribution due to neighboring effects (Fig. 1). An analytical model has been developed to reconstruct the size distribution of potentially non-interacting, inherent, discrete bubbles, leveraging experimentally measured quantities.
Fig. 1. Conceptual diagram for a nucleation site i (left) and inherent-
nucleate/apparent-individual bubble radius distributions (right).
Fig. 2. Schematic of the nonlinear solver.
The non-linear analytical equations were numerically solved by implementing the stochastic gradient descent (SGD) method. A new weight updating function was induced for the proposed non-linear system (Fig. 2). The overall mechanistic model (analytical eq. + numerical solver) was validated by stochastic modeling of synthetic boiling data (Fig. 3). In light of the current understanding, the mechanistic model would eliminate measurement bias arising from limited experimental observation. Additionally, the proposed model can be incorporated into the post-processing stage to calculate exact individual bubble quantities for other high heat flux mechanistic models, such as heat flux partitioning.
Fig. 3. Validation of the mechanistic model with Monte-Carlo stochastic simulation of synthetic boiling.
34141 대전광역시 유성구 대학로 291 (구성동 한국과학기술원 373-1) 기계공학동 www.kaist.ac.kr