This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

WATERGEMS- Where I could see the number of generation? And the features in mfGA.

Hello,

How could I find out the total generated solutions in Darwin Designer? And I'm studying the New York Tunnel case using WaterGEMS. However, It's very difficult to find out the feasible solution which cost is around $38.64million. Any one have experienced on that?

Cheers

Michelle

  • Hello Michelle,

    You should be able to set Darwin Designer to save multiple solutions, though this is only available the the Optimized runs, not the Manual runs. When using Optimization, there is a Options tab where you can select the number of solutions you want to keep. 

    Once the run is completed, you can look at the different solutions. If you are using a manual run, only one solution will be saved.

    For more information on Darwin Designer, follow this link. There are also sample models that show some different setups for Darwin Designer. Those are typically located at C:\Program Files (x86)\Bentley\WaterGEMS\Samples\Designer. 

    If that doesn't help, please provide some additional information on the issue.

    Regards,

    Scott Kampa

    Bentley Technical Support

  • Hello Scott,

    when stop optimizing, the window does not show the total trials. Only when computing, the generations and trials were shown. However, the final number of generations and trials would not seen.   Any other notification would show the number of trials and generations?

    In mfGA, there are two method to stop operating, one is basing on the population size and Generation number. The other is the stopping criteria - Max.trials and non-improvement generations. Which is prior when optimizing?

    Cheers

    Michelle

  • Hello Scott,

    Another questions, in fmGA, how to calculate the generation number? and how to estimate the trials in term of  population size?  

    Cheers

    Michelle

  • Hello Scott,

    Completing computing, notification: specified number of generation and eras reached as below show. I don't know why? Please explain it.

     cheers

    Michelle

  • Quote:

    Hello Scott,

    when stop optimizing, the window does not show the total trials. Only when computing, the generations and trials were shown. However, the final number of generations and trials would not seen.   Any other notification would show the number of trials and generations?

    In mfGA, there are two method to stop operating, one is basing on the population size and Generation number. The other is the stopping criteria - Max.trials and non-improvement generations. Which is prior when optimizing?

    Cheers

    Michelle

    -----------

    Regarding if the number of trials or generations is reported, this is not reported at this time, though obviously you will know one or the other depending on the which stopping criteria is used. Was there a reason you needed the information?

    Regarding when the Darwin Designer stops, as you mentioned, there are two methods: Max Trials and Non-Improvement Generations.

    Max Trials is simply maximum number of design trials you want the GA to process before stopping. Non-Improvement Generations is the number of maximum number of non-improvement generations you want the genetic algorithm to process without calculating an improved fitness. If the genetic algorithm reaches this number without finding an improvement that is better than the defined Fitness Tolerance, the genetic algorithm will stop. Non-Improvement Generations works in conjunction with Fitness Tolerance.

    If the design run stops because the Non-Improvement Generations reached the maximum number, that means the program has found the most "optimized" solutions in relation to others that were generated and within the confines of the parameters used in the simulation.

    When the maximum number of trials is reached before the criteria for the non-improvement generations is attained. If this happens, it can be beneficial to increase the number of trials to see if solutions with a better fitness can be generated, though that will not be the case every time.

    Adjusting some of the GA Parameters (population size, era generation number, etc.) can impact the results, which can in turn impact when and how the calculations stop. In general, even if you find a suitable solution it is recommended to try different GA Parameters, even if it is just a different value for Random Seed. Genetic Algorithm is looking for the best results in relation to the results already generated. It is possible that making adjustments to the parameters can yield different results.

    Regards,

    Scott