In each of the following situations, determine which type of the error (physical model, discretization, iteration, or programming) is most likely responsible for poor performance. In each case, suggest a course of action (e.g. refining the grid, testing the code on a benchmark problem, comparing with experiment, etc.).

a) Steady-state flow and heat transfer in a small-scale (research prototype) glass melting facility is simulated. The flow is known to be laminar. The simulations are performed using a commercial general-purpose software based on finite volume discretization.

The verification procedure, in which we compute a laminar thermal convection flow in a box, shows good agreement with the known benchmark solution. The simulation results show large discrepancy with the experimental data. Refining the grid and decreasing the tolerance of the iteration solution does not help.

b) A code is developed for simulation of unsteady conduction heat transfer. As a verification, the problem of heat conduction in a two-dimensional rectangular plate with fixed temperature at the boundaries is solved. The solution is utterly incorrect and remains such as we refine the grid and decrease the time step.

c) Three-dimensional flow and heat transfer in a counterflow heat exchanger is solved using a commercial CFD software. A welltested model specially designed for analysis of heat exchangers is used. The simulations conducted on a grid consisting of 9301 finite volume cells produce incorrect results.

d) URANS solution for turbulent unsteady flows within a chemical reactor shows good agreement with experimental data. It also shows that the mean flow fields always converge to an asymptotic steady state. We decide to focus on them and solve the steady-state version of the RANS equations. The results are disappointing.