Discussion on Lean Manufacturing Simulation Webinar

[sugato_deb]

This thread is dedicated for discussion on the SME Simulation Tech Group seminar :

Topic: Simulation Modeling in Lean Manufacturing Programs

Abstract :

Simulation modeling is valuable to analyze and improve performance of materials, information, and financial flows in both discrete and process manufacturing enterprises. It incorporates variability in demand, supply chain and manufacturing processes to imitate the real world.

Jim Curry will demonstrate the value of simulation models in a lean program for:

evaluating things other tools cannot, such as impacts of product mix and setups,
analyzing impacts of variability in processes and demand,
replenishment and inventory strategy across multiple locations, and
analyzing impacts of process and capacity changes on performance.


Jim Curry is CEO of the OpStat Group, a company he formed in 1986. He has developed an analytical methodology and developed simulation models for lean manufacturing analysis for companies to optimize their global operations.

He has written material and presented for several professional organizations. He is also an Associate Professor at Fairfield University, teaching graduate engineering courses in supply chain and manufacturing process design.

[JimCurry]

If you have any follow up questions from today's session, I'll check this thread periodically. Jim Curry

[dbladecki]

Two questions that we did not get a chance to answer.


What are any limitations of this simulation tool?


How do you enter machine paramters and images?

[JimCurry]

Ref the first question. I haven't seen any limitations to the software.
- I personally have worked with models that have as much as 200,000 complex transactions (e.g., in a job shop)
- Very complex logic, resulting in a 300+mb model
- Thousands of products
- Linked to ~13 Excel workbooks averaging 5-6 worksheets each.

[JimCurry]

Ref the second question:
There are standard libraries of machine images and parameters for a variety of machines, e.g., automated storage and retrieval, carousels, conveyors, etc. You may also create your own machine blocks with some training.

I should also meniton that there is 3D animation coming available in the next month or so in the latest version of the software.

[JimCurry]

Simulation Packages I mentioned today:
Extend from Imagine That Inc. www.imaginethatinc.com (is the one I demoed)
Arena from Arena Software www.arenasimulation.com
ProModel from Promodel Corporation www.promodel.com

Here is a link to more extensive list http://www.idsia.ch/~andrea/simtools.html

[jimkw]

Hi Jim,

I was one of the attendees at the SME webinar. I'd like to thank you for
the effort and time you put into it. It was every well done and
informative.

I've been doing simulation of manufacturing processes for a little over ten
years using Simul8. If used properly, and with the right goals, I've found
it to be an extremely valuable tool. My best result so far has been an
increase in output on the order of 4000% with no addition of people or
equipment. That one actually got written up, and won me an ipod. It's at:

http://www.simul8.com/support/newsletter/Output_Repair_Dept.htm

By the way, with the exception of being a long time user and having won an ipod, I am in no way affiliated with Simul8 or their parent Visual Thinking.

Thanks again,

Jim

[dbladecki]

In a follow-up survey to the webinar, the following question was asked:

I would like to know whether the process time variance at one machine/workers can affect the variance of other(s)? I would also like to know whether we have to collect more data since the variance might change if we change other variance?

[dbladecki]

Another question asked in a follow-up survey:

On the initial phases of a simulation, how you handle process times before you have a chance to collect data yourself (or use extant data) to map out the simulation?

[dbladecki]

A third question from the webinar:

I would like to know which and how much (in range) of the variance of the elements in the process that we should change to reduce the variance of other machine so that we can produce output with minimal variance and maximum troughput.

[jimkw]

Typically I'll have an idea of what the process times are going to be when I'm going to start building a model. I will create a distribution for each one of the various process steps, many of which can be shared for multiple stations doing the same thing. I work in electronics, so we have two different kinds of processes. The first is mechanical activities, the second is test and/or calibration. Generally I will use two different distributions. For the mechanical side, I will use a bounded normal distribution and feed the parameters in off of a spread sheet. Later, after I have had a chance to collect data, I can just go in, change the spreadsheet, and have the model update itself. Test and calibration, based on my experience, will generally follow something like a Weibull distribution, often used for repair times. I'll do the same thing there that I did for the normal distribution, but I won't have the same concrete bounds that I do for the other.

I've never really thought about the variance in one process affecting the variance in another. I'd really have to give that some thought. What I've found, and remember, I'm talking about complicated electronic devices here, is that there are good units and there are bad units. When it comes to test and calibration time, a bad unit can be troublesome throughout the process. That's an interesting idea, use the prior history of a unit to set the variability of the follow on testing. I'd like to explore that a little bit.

[dbladecki]

If you are looking for information on this topic, check out Jim's technical paper available free to SME members. See http://www.sme.org/cgi-bin/get-item.pl?TP07PUB33&2&SME&.