ICEPAK FAQ: How to trouble shoot non convergence or slow convergence of a forced convection problem
Sometimes while solving forced convection problems the residuals flatten and never approach the convergence criteria. Following is a systematic approach to troublr shooting this. Natural convection problems are usually slow to converge because of the coupling of energy and momentum equations. There may be various causes for the slow/non convergence of a natural convection problem. Slowly converging natual convection problems can be addressed on a casebycase basis. Contact your support enginner for this. However, certain specific reasons have been identified for the slow/non covergence of forced convection problems. These causes and trouble shooting tips are described below: Check if one of the following is true: There are flow inlet but no outlets. View vectors on external object faces that allow flow (i.e. openings, fans, vents etc.,). When some of the vectors are in a direction that you don't expect it to be, it means that you have a recirculation at these openings. Based on your understanding of the flow, see if such a reciculation is realistic. You may expect flow recirculation on the downstream side of flow obstructions. If the recirculation is realistic, that means that some of the information needed to converge the solution lies outside the domain, and by truncating the domain the CFD solution has difficulty satisfying continuity. In such situations extending the domain (without changing your physical problem) is necessary. Please consult your support engineer on how to do this. Sometimes. even when the above mentioned recirculation is inside the domain, but not well refined with mesh, nonconvergence, or even divergence can happen. Such recirculations are common behind the hub of fans. Use mesh controls to refine the recirculation region. A number of times, simple ducted flow into heatsinks or perforated holes cause continuity to flatten. Such situations are often encountered while trying to characterize perforated vents or heatsinks. This could again be due to poor refinement of the downstream wake region. Somethimes if the flow is actually turbulent but you are modeling it laminar, continuity may have difficulty converging. If you are using under relaxation factors recommended for natural convection (0.7 for pressure 0.3 for momentum), that may slow down the convergence of continuity residuals. Usually 0.3 for pressure and 0.7 for momentum are recommended for forced convection. The natural convection settings may be used for some forced convection problems with sharp momentum changes (i.e., sharp flow turns). Other than the above recommendations there is rarely a need to fine tune under relaxation factors. Please contact support if you think your situation is special. Convergence can be unncecessarily slowed if you are including natural convection in a problem in which forced convection is the predominant heat transfer mechanism. Most fan driven flows fall under this category. TUrning off gravity vector will significantly improve convergence rate 

