Optimization Process

Radioss parses all three decks (optimization, Starter and Engine decks) to construct an optimization model. For each optimization iteration, an updated/optimized Starter file (<name>_rad_s#_i###_0000.rad) and Engine file (<name>_s#_i###_0001.rad) will be written out, where ### is the optimization iteration number.



Figure 1. Optimization in Radioss Process Overview
The updated Radioss Starter and Engine files will be solved by Radioss and the results will be communicated to OptiStruct to update the OptiStruct optimization model.
Note: For Solid Topology Optimization using Radioss Optimization, direct density control is used. After Linear static response topology optimization is complete, solid elements having values from 0 to 1 as topology design variables are updated in the finite element model. Each solid design element is directly assigned a filling percentage which has a value equal to the associated topology design variable. This utilizes the Radioss option, /INIBRI/FILL, to map the filling percentage with the topology design variable.

Implementation Sequence

The following steps illustrate the sequential process used to implement optimization in Radioss:
  1. Pass the optimization file <name>.radopt to OptiStruct using the syntax specified in Optimization Process. OptiStruct will automatically search for <name>_0000.rad and <name>_0001.rad in the same directory.
  2. OptiStruct parses the Radioss input files, constructs an equivalent OptiStruct model, and continues with the execution of this OptiStruct model.
  3. OptiStruct applies the optimized design variables and writes out the updated/optimized Radioss Starter/Engine files (<name>_rad_s#_i###_0000.rad and <name>_s#_i###_0001.rad). In the first outer loop, the design variables are not updated yet, and the Starter/Engine files will be identical to the initial user input.
  4. Radioss solves the updated Starter/Engine files generated in Step 3.
  5. OptiStruct reads the Radioss results obtained in Step 4 and starts the next optimization run (inner loop).
  6. If the results from Radioss vary beyond the threshold as compared to the previous outer loop, then the optimization process did not converge.
  7. The optimization process is then continued (using the ESLM method) on the updated results from the Radioss analysis and new optimized design variables are generated. Steps 3 to 6 are repeated to verify if the process converges after the subsequent iteration. If convergence is achieved, go to Step 8.
  8. The process ends and optimized results are output.


    Figure 2. Optimization Sequence in Radioss and OptiStruct