During the Juno Critical Design Review (CDR), a question was raised whether the power cycling of the TWTA (electronics box) on the Telecom Panel could adversely thermal cycle the other electronics boxes mounted around the TWTA. This action item was assigned to me to evaluate the expected thermal cycling affects could be expected during typical operations.
Using the Thermal Math Model (TMM) developed earlier, I set the model up for a transient case that would simulate a science pass expected for the Juno Spacecraft. The first step was to get a starting point of steady state temperatures. However, an unexpected problem happened when the steady state model run would not converge numerically for the first time. After careful review of all the parameters, my feeling was that it had something to do with the logic built in for the variable emissivity that simulated louver performance. Previously, all other configurations operated in a region where the louvers were fully open and at a steady boundary condition. However, with all of the electronics boxes powered off, the temperatures were expected to be in the region of a mid-point within the louvers operational regime.
To prove this point, two cases were run, one keeping the louvers open and one keeping them closed. All other boundary conditions in the TMM remained the same. As expected, both of these cases converged and produced results that were as expected. Once these steady state cases were run, they were used as the starting point (i.e., initial condition) for the transient case and a science pass was modeled. The fully-open and fully-closed louver cases bounded the problem of the partially-open louver case.
It is important to be able to model a partially-open louver so I discussed the situation with the software manufacturer and they recommended I try changing the default relaxation and stability criteria. I then ran the model with a coarse convergence criteria and this converged relatively quickly. Using these results as the new initial conditions, I incrementally tightened the convergence criteria and continued to re-run the model until I had achieved results from a fine convergence criterion. Essentially, I had manually forced the model to converge. The results obtained were then compared against the model runs where the louvers were fully open and fully closed as well as a few other parametric cases to ensure the new results were credible.