Finite element formulation for solving a steady-state boundary
inverse heat conduction problem of homogeneous and two-layered
materials is presented. Firstly, finite element formulation is introduced
to solve boundary inverse heat conduction problem of
homogeneous material without function specification. Effects of
bias error and location of known temperature data are tested using
4 4, 8 8, and 16 16 meshes on square domain of 1 m1m.
The formulation is incorporated with modified cubic spline function
to stabilize predicted unknown boundary condition. The problem
is solved using cubic spline function with nine and five control
points. It can be concluded that using cubic spline function with
fewer control points is preferred because it reduces computing
time considerably but yields solutions at same level of accuracy.
The formulation is applied in predicting unknown boundary
condition of two-layered material. Physical condition at layer
interface which is characterized by thermal contact resistance
and heat source strength is handled by coincident nodes. Three
conditions at the interface of two-layered material are studied
with thermal conductivity ratio of 0.1, 1, and 10. The problem is
solved by using 32 32 mesh on square domain of 1 m 1 m with
upper layer thickness of 0:125 m. Temperatures from analytical
solution are specified on nodes at y ¼ 0:75 m and cubic spline
function using five control points is employed in solution process.
The original boundary condition can be superbly recovered with
less than 0.1% in RMS error for all thermal conductivity ratios
and all conditions at the interface when no bias temperatures are
specified. If no heat source exists at the interface, RMS error will
vary linearly with bias error of specified temperatures. RMS error
will be drastically affected by bias error of specified temperatures
if heat source exists at the interface. Therefore, the method is not
recommended for predicting unknown temperature on boundary
of two-layered material under this situation.
The concept in present study can be applied under two situations
in practice. In thermal experiments, temperature distribution
on a portion of material boundary can be predicted when temperatures
at some interior points are available. In thermal designs,
temperature distribution on a portion of boundary is determined
to attain desired temperatures at some interior points which are
constraints in the design.