Total: 1
In engineering, we often encounter an open-top, liquid-holding tank draining via gravity-driven flow through a small orifice in its side. Torricelli's law and a mass balance give a differential equation model governing the dynamics of the liquid level in such a draining tank. Herein, we leverage this forward model to tackle an inverse problem of reconstruction: infer the shape of an exogenous, heavy solid inside a draining tank from measurements of its liquid level over time. To quantify uncertainty, we employ Bayesian statistical inversion to obtain a posterior distribution over the solid's cross-sectional area as a function of height. (Because the solid displaces liquid, the rate of decrease of the liquid level provides information about the cross-sectional area of the solid at that height; as the liquid level drops, it "scans" the area of the solid as a function of height.) In our experimental setup, a tank drains of water through a small orifice in its side while a liquid level sensor collects time series data of the water level. First, we calibrate and test a forward model of the water level dynamics using data from two tank drainage experiments without an exogenous solid. Second, we conduct a drainage experiment with the tank containing an exogenous solid, then leverage the calibrated forward model and data to infer (i.e., obtain a posterior distribution for) the area of the solid inside the tank as a function of height. Our approach may be practically useful to infer the shape of an unknown solid, or the porosity of packed solid particles, inside of an opaque tank.