Filter optimizer results

Question

question

Mohamed Eldakroury asked Mar 19, '24 Mohamed Eldakroury commented Mar 21, '24

Filter optimizer results

Hello,

I am simulating a demand - supply problem here. A source will create tokens and work cells will process them. I am building the model in process flow where I have 15 work cells. Each work cell has limited hours per week (I am using zones to control the number of tokens that can enter per week). I defined my output as a performance measure and my cell hours as parameters (15 different parameters)

I am running the optimizer to tell me what are the minimum hours for each cell to get the maximum total output per week. So, I ended up with 16 objective functions. One for the output and 15 for the work cells.

I found out that It will take too many replications to get close to a good solution. However, I end up with thousands of results and hundreds of those have the star icon that indicate that this is a best iteration. Is there a way I can filter the results to only show the one result that have the highest output with the minimum hours for each cell? And does it make sense the way I am running the optimization problem to start with?

Software Version:

FlexSim 23.2.0

proces flow optquest optimizer

1710860266298.png (379.3 KiB)

Answer 1 · 2024-03-20T08:21:59Z

Felix Möhlmann answered Mar 20, '24 Mohamed Eldakroury commented Mar 21, '24

In 24.0 a feature to filter the results in the "View Results" window was added, though I haven't worked with it yet, so I can't say whether that would help in this situation.

In previous versions you'd have to copy the results table and filter for "useful" entries in that (in FlexSim itself through a Calculated Table for example or a third party program like Excel).

Though as you say, the optimizer works better with fewer objectives. I would try to sum the work hours of each into a single performance measure that has to be minized. While doing so you can apply an exponent to each cell's value.

If x > 1, the optimizer will try to find solutions where the work hours of the cells are similar. If x < 1, it will do the opposite and prefer solutions where some cells work more than others.

Edit:

Since reaching the maximum weekly output is a requirement for the solution you should consider adding it as a constraint to the optimizer job. Replications that do not reach the output will be marked as "unfeasible" and the optimizer will apply a malus to the respective parameter combination when determining new replications.

1710922778084.png (8.4 KiB)

· 1

Mohamed Eldakroury commented · Mar 21 at 01:02 PM

Thank you. I'll test it out and let you know if I have follow-up questions.

0 ·

question