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

Darwin Designer

Hello

I have designed pipes of a model with Dar. Designer. The results seems not to be optimized very well. For example if you have a look at pipe P-4, you will see 400 mm diameter is designed for it which is not an optimized diameter costly (it better to be 300 or 200 or less diameter pipe). Several pipes are designed like p-4 which can be designed better that what Darwin did. 

Who has an idea?

7674.New folder.rar

Parents
  • Any optimization gives you the best solution that minimizes the cost subject to the constraints. That doesn't necessarily mean that you should build that exact solution. Instead you should view the solutions from Darwin (or any other optimization) as suggestions of what a good solution might look like. You need to apply your engineering judgment to the solution. I've often said that the most powerful tool for optimizing a design is between the engineer's ears.

    In your particular case, there was some condition for which Darwin needed to use a 400 mm pipe to satisfy some constraint. You need to consider if the extra cost was worth the benefit or if you want to get by with a smaller pipe. I suggest going back and running some simulations for such a scenario.
  • Dear Tom
    What I've got of your opinion is software designing HAS TO be studied and modified by engineer and Dar. Designer can not offers the best solution ALWAYES. Am I right?
    About the particular model, both less cost AND satisfying the constraints can be got with smaller pipe that 400mm, which means smaller pipe is the best optimization NOT 400 diameter pipe. This behavior admits the first idea (Dar. Designer can not offers the best solution ALWAYES). Is my opinion true?
  • There are really several levels of answers to your question.

    1. Darwin Designer uses Genetic Algorithms (GA) to find a solution from the overall solution space. You may not be letting it run long enough and if you let it run longer, you may get better solutions.

    2. GA does not guarantee global optimality. That is why experts in GA encourage users to make multiple runs with just slightly different data. In encourage you to do some reading on GA.

    3. The solutions from any optimization, including GA, are only as good as the data you provide. Are you 100% certain that your cost functions, demand forecasts, scenarios to consider, component failures, etc. are perfectly accurate? At Bentley we often use the term "optioneering" instead of optimization because what things like Darwin do is enable you to screen many solutions/options very quickly. No optimization can give you the best solution always. If someone tells you they can, don't believe them. An optimal solution is the solution that minimizes the objective function subject to the constraints. You need to provide the data that goes into the object function and the constraints. How do you define the "best" solution?

    You as an engineer still need to make decisions, account for that uncertainty in an imperfect world.
  • Yes, Darwin will often get you a very good initial design (dependent on how well the Designer run was configured), after which engineering judgment will need to be used. There are some further tips on this in the Help topic called "Advanced Darwin Designer Tips" and at the bottom of this wiki article:

    Using Darwin Designer


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

Reply
  • Yes, Darwin will often get you a very good initial design (dependent on how well the Designer run was configured), after which engineering judgment will need to be used. There are some further tips on this in the Help topic called "Advanced Darwin Designer Tips" and at the bottom of this wiki article:

    Using Darwin Designer


    Regards,

    Jesse Dringoli
    Technical Support Manager, OpenFlows
    Bentley Communities Site Administrator
    Bentley Systems, Inc.

Children
No Data