TY - JOUR
T1 - Forward and inverse problems in two-phase fluid dynamics
AU - Zabarankin, Michael
AU - Grechuk, Bogdan
N1 - Publisher Copyright:
© 2017 Society for Industrial and Applied Mathematics.
PY - 2017
Y1 - 2017
N2 - Given an objective function, e.g., an error measure, which depends on an optimization variable and a parameter, a forward problem finds optimal value of the optimization variable as a function of the parameter, whereas an inverse problem finds optimal value of the parameter as a function of the optimization variable. In general, the two functions are not inverse one of the other. If the objective function is nonconstant on the solution set of either the forward or inverse problem, then it should necessarily be nonsmooth with respect to both the optimization variable and the parameter in order for the solutions of the forward and inverse problems to be inverse one of the other. In this case, a corresponding suficient condition is suggested. It involves only single-variable functions, which make its verification computationally eficient. Of particular interest is the problem of finding a "steady" spheroidal drop freely suspended in an ambient uid subjected to either an external linear ow or a uniform electric field-the ratios of the external ow stress to surface tension and of the electric stress to surface tension are defined by capillary numbers Ca and CaE, respectively. The spheroid's axes ratio and Ca (or CaE) are the optimization variable and parameter, respectively, whereas the objective function is a measure of the normal velocity on the interface chosen to be the sum of the absolute coeficients in the Legendre series expansion of a stream function in the prolate spheroidal coordinates. The solution of the inverse problem is found analytically. Remarkably, its inverse solves the forward problem for almost the same range of Ca (or CaE), where it is defined. This justifies determining "steady" spheroids in [M. Zabarankin, SIAM J. Appl. Math., 4 (2016), pp. 1606{1632] as those for which the first coeficient in the stream function series expansion vanishes. As a result, the computational process and further analysis are simplified considerably.
AB - Given an objective function, e.g., an error measure, which depends on an optimization variable and a parameter, a forward problem finds optimal value of the optimization variable as a function of the parameter, whereas an inverse problem finds optimal value of the parameter as a function of the optimization variable. In general, the two functions are not inverse one of the other. If the objective function is nonconstant on the solution set of either the forward or inverse problem, then it should necessarily be nonsmooth with respect to both the optimization variable and the parameter in order for the solutions of the forward and inverse problems to be inverse one of the other. In this case, a corresponding suficient condition is suggested. It involves only single-variable functions, which make its verification computationally eficient. Of particular interest is the problem of finding a "steady" spheroidal drop freely suspended in an ambient uid subjected to either an external linear ow or a uniform electric field-the ratios of the external ow stress to surface tension and of the electric stress to surface tension are defined by capillary numbers Ca and CaE, respectively. The spheroid's axes ratio and Ca (or CaE) are the optimization variable and parameter, respectively, whereas the objective function is a measure of the normal velocity on the interface chosen to be the sum of the absolute coeficients in the Legendre series expansion of a stream function in the prolate spheroidal coordinates. The solution of the inverse problem is found analytically. Remarkably, its inverse solves the forward problem for almost the same range of Ca (or CaE), where it is defined. This justifies determining "steady" spheroids in [M. Zabarankin, SIAM J. Appl. Math., 4 (2016), pp. 1606{1632] as those for which the first coeficient in the stream function series expansion vanishes. As a result, the computational process and further analysis are simplified considerably.
KW - Drop
KW - Electric field
KW - Forward problem
KW - Inverse problem
KW - Steady shape
KW - Stokes ow
UR - http://www.scopus.com/inward/record.url?scp=85039967643&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85039967643&partnerID=8YFLogxK
U2 - 10.1137/16M107726X
DO - 10.1137/16M107726X
M3 - Article
AN - SCOPUS:85039967643
SN - 0363-0129
VL - 55
SP - 3969
EP - 3989
JO - SIAM Journal on Control and Optimization
JF - SIAM Journal on Control and Optimization
IS - 6
ER -