[chris15012]

I am using the multi-objective genetic algorithm (MOGA) within the Direct Optimization of ANSYS Workbench, applying this to a fluid physics problem computed in FLUENT. The application seems to be working very well, however I have some questions that either weren't explained in the ANSYS online guides or I must have just missed the answers to my questions or misinterpreted something. I have run several different optimizations and have a similar result. For some background, I have my initial samples set to 100, samples per iteration 50, maximum allowable Pareto percentage 70, convergence stability percentage 4 (usually I use 2), maximum number of iterations 20 and maximum number of candidates 3, thus, mostly the defaults. The confusion I have is that I don't see the correlation between my Raw Optimization Data and the History Chart, which in this case is history of minimizing a single parameter (disregard the multi- in the name, its really a single-objective problem). For my Raw Optimization Data in a minimization problem, the numbers always start very high, but those high numbers don't show up in the History Chart. There have been some failures with the meshes, but that doesn't seem to be a big deal. On another note, I decided to see if I could match an optimization result from the Raw Optimization Data with one on the History Chart, for instance looking at a result on the History of the last point, point 300 in this case, this point could be identified as a run number on the Raw Optimization Data, I think that number was in the 200 range, thus whatever number was there, it didn't correlate to the design point for that result, and there was only one of those numbers resulting on the History, so I know there weren't duplicate or similar numbers with other design points. I am just needing to have the results of Direct Optimization explained a bit. The History Chart also in my cases isn't showing the full marketability of the improvements made, because it cuts off many of the high numbers from the plot, which makes it appear the user is making less of an improvement from the changes in the geometry, and that the parameter being studied appears less sensitive, although even disregarding many of the big numbers, the algorithm has still done a sufficient job. I thought maybe the History Chart would start after the initial samples, but that doesn't seem to be true either.