ahorsky

thank you , is it posible to change the values in the setup of the global optimization on response surface? for example i want to cange crossover_rate and mutation_rate

Yes, post-processing doesn't really help in my case because I'm trying to reduce the SBGO workload while also improving design space coverage. Having a ship LOA of 42.15766 or 42.04474 meters is not realistic in practice. I tried setting the design variable as an integer, but the optimizer seems to ignore it. Rounding the values afterward doesn’t really solve the issue either, since the surrogate model is still trained on the full-precision values.

I'm running a Surrogate-Based Global Optimization (SBGO) and I would like to ensure that certain design variables are only sampled at fixed step sizes, such as: LOA sampled every 0.5 meters L/B and B/T ratios sampled every 0.25 Deadrise and rake angles sampled every 1 degree What is the proper way to make sure that SBGO only samples design variables using fixed step sizes? Can this be enforced directly inside CAESES, through some feature or setting, or must it be handled entirely on the Dakota side? Any guidance or working examples would be highly appreciated! Thanks in advance!

hey when take a screenshot during my run using the design pre/post processing, it add to the model normal vectors to surface, how do i remove the them ? thanks

Hi Claus, Thank you for your response. I tried creating a COM interface with Maxsurf Motion similar to what’s available for Maxsurf Resistance. However, from what I understand, Maxsurf Motion does not currently support a COM interface. Could you confirm if that’s correct, or am I mistaken? At the moment, I’ve set up a software connection in CAESES with Python. This workflow exports the hull geometry as an IGES file, which I then use in Maxsurf Motion to run the seakeeping analysis (this step typically takes 5–10 minutes per design). Maxsurf Motion outputs the results to an Excel file, which Python uses to calculate the Seakeeping Index (SKI). Python then creates a CSV file containing the single SKI value, and CAESES reads this value back from the CSV. The issue I’m facing is that CAESES does not “wait” for Python to finish processing the Maxsurf results and calculating the SKI. As a result, it fills in the SKI objective with a NaN value. If I manually pause the run after each design is created, add the Maxsurf output to the design folder, and then resume, Python has enough time to compute the SKI and CAESES successfully reads the results. I also tried adding a time delay in Python to hold the calculation until the Maxsurf output is available, but CAESES seems to ignore this and continues running without waiting for the results. I hope my explanation so far is clear. I’m now looking for a way to: Delay CAESES from calling Python until I’ve added the Maxsurf output file. Or allow CAESES to generate x number of designs without calculating the SKI, and then calculate SKI for all of them in a batch process. Do you have any suggestions on how this could be implemented? As a workaround for LHS sampling, I can let the run complete and then add the Maxsurf outputs afterward. When I re-evaluate the designs in CAESES, it recognizes and incorporates the SKI correctly. However, I’m unsure if this approach would also work during SBGO, where iterative evaluations are required. I’ve attached my setup files for your reference. Best regards, simple_test.cdbc Single_SKI.py Single_SKI_timeout.py

Hi Simon, Thanks again for your earlier advice about using Design Pre/Postprocessing for geometry export — that worked well and I’ve now successfully exported the hull variants from CAESES and performed motion analysis in Maxsurf Motion. Here’s where I am now: ✅ I ran an LHS DoE in CAESES to generate an initial set of designs. ✅ I analyzed each design externally in Maxsurf to calculate motions and then computed a Seakeeping Index (SKI) in Python. ✅ I would now like to: Import these SKI results back into CAESES. Run an SBGO, where at each generation the optimization pauses: Exports the new hull geometry. Computes motions in Maxsurf externally. Calculates SKI in Python. Brings the SKI back into CAESES as an objective to minimize. I’m trying to set this up via the Software Connection feature, but I’m unsure of the recommended workflow for this external coupling: Can CAESES directly use a Python script to pull SKI values as a response during SBGO? Is there a way to make the optimization pause for external motion analysis and resume automatically? Should I use Postprocessing Commands in the design engine or a Custom Software Connector? I’d appreciate your guidance or any example workflows that show how to integrate external evaluations like this into the SBGO loop. Best regards, Avishai Horsky LHS, where every generation is stops i do the motion analysis and then it continues aftre the input of the SKI

I’m new to CAESES and currently working on a fully parametric ship hull model. I’ve set up the model and now I want to run a Latin Hypercube Sampling (LHS) to generate a set of design variants and then run a Surrogate-Based Global Optimization (SBGO) using resistance (as a feature), seakeeping (evaluated externally in Maxsurf Motion), and stability. For each design, I need to export the geometry as an IGES file so I can use it in Maxsurf Motion. As far as I understand, there is no direct software connection between CAESES and Maxsurf Motion. My question is: how do I export the designs automatically? Should I create an export feature, or write a script? Or is there another recommended approach for this? I’ve tried to run a simple LHS DoE using an export script, but it didn’t work. I don’t fully understand how the export process should be linked to the DoE run. Any guidance on the correct way to set this up would be greatly appreciated hull2.fsc