FLOW3D model spontaneously explodes
 

My FLOW3D model spontaneously became unstable after happily chugging away for 2 days. Analysis of flow fields prior to the explosion showed nothing amiss. I was solving for the long-term mixing resulting from a quasi-steady-state jet injecting fluid into a stratified reservoir. The velocity field developed within the first 10% of the model, and the remaining simulation is to capture the gradual temperature mixing. Thus, I am perplexed why the model suddenly became unstable.

My question is: Are there any important tips/tricks I should use to avoid spontaneous instabilities? And if the solution becomes unstable and blows up, are there any important tips/tricks for diagnosing/fixing the model?

Thanks!

(For reference, here are my solver settings... Pressure: Implicit, GMRES. Viscous Stress: Explicit. Heat transferL Explicit. Advection: Explicit. Momentum: 1st Order. Turbulence: RNG)

One thing to watch for is the size of the maximum residual from the implicit pressure solver. The convergence criterion for this solver is [usually] inversely proportional to the time step size, so if the time step drops the convergence criterion can become large. If it remains large for too long it will yield a poorly converged solution.

Check if there is a sudden increase in the maximum residual near the time that the results become unreasonable. If so, try reducing the convergence criterion multiplier (epsadj) or changing to a constant convergence criterion (epsi).