Flexsim running performance vs Automode performance...

I received a question some days ago from my partner about above issue.
His question is about simulation running (time) performance followed conditions.

He had made the model of the shop floor of making semi-conductors amd he want to know the comparing flexsim features about it.

1. Resources (Fixed or mobile objects) are about 800 (approximate 1,000)
2. the quantities of a day (flow items) are 10,000.
3. the time unit like process time and setup time is seconds.

I want to know how much time does flexsim take to run the model in order to get the all resource's detail statistc report in that condition during one day work on?

In the case of Automode, it takes ten or twelve hours in order to get the one day working utilization data. Is it right?

I want to hear the right information and flexsim capability in same condition. It is an important issue to make the Aotomode users be changed to flexsim users.

I want to hear answers.

It really depends on how you setup and build the model. Unless you built the model in Flexsim and Automod and then compared them directly, there is no way to say one way or the other.

Syseo,

I've worked on scheduling and simulation for STMicroelectronics Corporate during 7 years and for AMD Dresden during 2 years.

I was in charge of modelling the front-end shopfloor of about 16 semiconductor fabs around the world (2 in Singapore).

It was in Autosched AP so no way to compare with Automod or Flexsim but I can imagine the real sense of your question because your customer should probably want to insert a simulation tool inside his producton control management. As we did with STM and AMD.

In AMD, we were using Automod to figure out how to handle the overhead vehicles of the fab30 and it takes such long time.

So to answer your question:

1) Bill is right. More details you have, more time you will have to wait the end of your simulation run.
2) Semiconductor issues are very detailed sensitive in the sense that you can figure out about 10 to 15 types of equipment processing (batch, wafers to wafers, cascading, robot cell bay, stockers, etc). Each processing type requires a set of specific data that you've to extract from the MES and to link to the model.
3) So you will spend months to develop such model as an external resource. And you have to know that a semiconductor product has a life cycle of about 3 to 6 months and a semiconductor technology about 1 to 2 years. At each technology, they will change all the equipments with new ones (either new processing types), etc.

So even if I am convinced that Flexsim can challenge Automod on the simulation run duration, you will spend a lot of time and money to get an accurate model in phase with the evolution of your customer shopfloor. Not sure you succeed on this task.

Only internal development can be considered in this case and they will need a good Flexsim developer to both design the data managment and the speed optimisation of the model.

Based on my own experience, Flexsim can win this challenge if you are able to design generic functions that can read data files and can design automatically the whole shopfloor. With a set of good interfaces, you can be able to design at least 50 to 60% of the model in one click. after that, you will have to finish the model manually.

Some universities have tried to do that but I have no feedbacks on their success.

In conclusion, the best way to use a simulation tool in a semiconductor fab is to design only the material handling system of the wafer transportation and to consider equipments as black-boxes delivering PODs at a defined cycle-time. If you go to much on processing details, the model will be too heavy even if in Flexsim.

Let me know if you need more information about this market.

I never worked with AutoMod, so it is difficult to tell about the differences to other simulation package. But I spoke with lots of AutoMod Users during their evaluation of different simulation packages. So I can only write here, what they told me and can not judge if it is right what they told me and if the current release might not be better or able to do things which were not possible in the past.
Anyway, I like to tell you about my experience during different evaluation processes with different AutoMod users. All of them worked several years with AutoMod.

What I figured out it that, even if you convince them that other simulation packages (like Flexsim) are really good simulation package, AutoMod users are the most difficult users of a simulation package to switch to another package. Even if the other Software is
- easier to use
- building models is much faster
- modify a model is much faster done
- other simulation packages can also handle big models
- other simulation packages can be fast
- spending less time on debugging in other simulation packages is much shorter
- etc.

I remember one time, when an AutoMod-User, who worked 15 years with AutoMod, send me a video of a transportation process (double cycle in a warehouse and I posted the model here at the community). I build a model for the presentation and it was exactly doing what he wanted. He asked how long it took me to build the model and I said it was in my eyes long, because it took me 4 hours, because I was thinking quite some time, which Flexsim modeling concept I like to use. He said, building this model in AutoMod would take him at least 7 times longer. Doing changes in the model like changing the entrance of items (in AutoMod called loads) would take him some hours which I was able to do in a minute in Flexsim (just change the position of the in-queue).
He said that if he travels to a customer he should be good prepared with different scenarios. Because if the customer wants to have a layout or a change in the layout and if it is only a small thing, he needs to go back home, prepare a scenario, which will take him at least a day. Then he needs to go back to the customer hoping that he now has all the scenarios. He sees that this is much faster done in Flexsim, for sure, if it is only a change in the layout, like adding processors, positions of queues, etc. That will decrease time and traveling cost.
I understood that this is because the concept of AutoMod is totally different compared with Flexsim. In AutoMod you model/program the flow of a load (in Flexsim an item) through the system. If you have any change, you need to reprogram the material flow of this item. If you add a processor and you have several items, you need to modify the material flow of all items. That’s why it takes so long to make changes.

Another AutoMod user told me, that if you can use the AutoMod libraries and building models with the standard, you can be very fast. But if you need something special (e.g. logic for the material flow), you need to program it. He was surprised what you can already do in Flexsim by just using the pick-lists in the trigger.

Another AutoMod user asked me during a presentation a lot of very specific and technical questions about how to do/model things in Flexsim. The most things I was already able to show him directly in my small model I build during the presentation. Other things I just explained how I would do them. At the end I asked if he could do all this specific things in AutoMod. Yes, most of them but with much more effort. This user was also surprised what can already be done in Flexsim with the delivered library.

Another AutoMod user did not believe that you can have in Flexsim a real big number of loads/items. We convinced him, that it is possible, if we use the same modeling concept as AutoMod is using (not having all the loads/items physically in the model, but using Arrays to handle information). At the end we got a project from him and build a Supply Chain Model with around 700.000 loads/items in the model where every load/item had several information (like said we used Arrays for the information and Messages to have events).
By the way, as far as I know, in AutoMod if a load/item enters a warehouse it is not displayed any longer and it does not have a location. In the material flow it is defined where it will get picked up, which is in a simple case always the same position or needs to program differently. So in Flexsim you would use a virtual rack (see models at the community) to use the same concept. But in Flexsim’s virtual rack the taskexcuters still pick the items from the "real" location.
So I guess, this answers your questions. Yes, it is possible to build big models and still have a good performance for the simulation runs. As Bill says, it really depends on how you setup and build models.

I have several more examples, but this post will already get long, so I will not speak about all of them.
Surprising for me was that none of these users was purchasing Flexsim or another simulation package. They all stick with AutoMod. I asked them why and they all really see the advantages of Flexsim but do not want to change.
Trying to figure out why they gave me different answers and to sum it up in my eyes these are the reasons I see and not what they told me, because their arguments to stick with AutoMod sounds more like an excuse not to buy Flexsim then real arguments.
Then they try to find specific things or let me say really specific technical things and even if you can show that it is possible in Flexsim to do it, they will not switch.

For me it looks like, that an AutoMod user, who needed years to be able to build nice models in AutoMod will not change, because he has the experience in his software (they are the AutoMod experts in the company), he knows what he can do and what he is not able to do (know all the gripes and goodies), he feels familiar with his software (why should I work with another software and spent time on learning it). So even if they see all the issues in their software and see advantages in other packages, they will not change.

What I am trying to say is that I spent a lot of time and effort in the past to convince AutoMod users and my experience is, that they are not willing to switch, even if another packages could be much, much better. Keep this in account if you have to do with an AutoMod user on how much time you like to spend on him.
That’s my advice: Only spent time and effort if he really is willing to switch and is not just doing an evaluation of other packages to convince others (in his company) that he is still using the "right" software.

If someone has more information about AutoMod or is able to correct what I wrote, please do so. I would really appreciate to learn more about AutoMod, not telling wrong things and understand much better, why AutoMod user will not change to a better package.

By the way, it does not mean, that we did not change any AutoMod user in Germany to Flexsim ...

I want to tell you first hand experience switching to Flexsim from Automod. We use all kinds of simulation/modeling tools at our company (www.stonge.com). Until 2 years ago, we were heavy Automod shop for all of our simulations till such time we came to know about Flexsim (although we experimented with the previous incarnation of this software). We were stuck in Automod for primarily one reason - Most of our models were based on CAD LAYOUTS. We had developed proprietary tools to quickly convert CAD layouts to be readily used in AutoMod with all its objects already geometrically accurately prepared for Automod world. It minimized the total effort required to get ready for simulating the logic.

When new projects came along for simulation, we decided to focus more on getting them done on Flexsim rather than the old Automod. So far in 2 years, the results have been mixed - primarily due to our learning curve of Flexsim - being Automod shop, old habits die really hard. But overall, I would say Flexsim projects have come up nicely for us whenever we have tried them. It definetly takes a completely different mind set while approaching simulation models using Flexsim vs. Automod. Coding in Flexsim is the single biggest area where Automod users find some difficulty adapting to. Message based thinking has to be there right from the get go to work with Flexsim. Automod unfortunately calls for LOAD based thinnking.

One great feature of Flexsim we have found when we compare against Automod is the out-of-the-box 3D visualization - which is very impressive when we take the model on the road to the client. We find that Flexsim quite easily allows us to integrate our custom 3D object files when in Automod world it would have taken some effort to position them accurately for visual effects.

Olivier Pellegrin , Tom David and Shankar Narayan.

Thank you very much for your kind explanations....

One of the biggest customers is SEC(Samsung Electronics Company) manufacturing 'Anycall' mibile phone and semiconductors and LCD TV, panel in Korea.
I have met many engineers working in some plants of it on call by them.
And also I have met many simulation engineers making model for SEC plants. They are all Automode users and SEC engineers are also all Automode users.

One of their common questions to me are the performance comparing to Automode.

I think also as same as you said. I know well about your meaning.
But according to your kind explanation, I got much deeper underestaning again.

I have other solution like APS(Advanced Planning and Sceduling) beside Flexsim and made one scheduling implementation for SEC using 'Preactor APS'.
I understand all your meaings.

Thank you very much again. if I need any yiur help, I will send it to you.

One of the prospects of Flexsim sales is SEC.
Yes. The engineers of them are same as you said.
They and Automode users of other said the same thing as you said but have not change it till now to other 3D simulation software like Flexsim.

But I hope and think they will change their mind and simulation software to Flexsim anytime.

Thanks a lot.