|
|
|
10/11/2011 03:08 PM
|
|
|
Does anyone have any experience using computer simulated models such as ProModel or FlexSim in process improvement work? If so, what were the benefits and drawbacks?
|
|
|
|
|
10/12/2011 02:55 PM
|
|
|
Hi,
experience with Enterprise Dynamics and Witness as (dynamic) simulation tools (same type as ProModel and FlexSim).
In Industrial (flow optimization, factory design) as well as service environments (contact center, hospitals).
Main benefits:
- scenario and sensitivy testing in safe environment;
- insight in very complex, multi-variable problem settings, which cannot be analyzed with human brainpower only;
- when especially useful: queueing and servicing problems, multiskill environments, when effects of variability are high;
- no need to model "mathematically", you build a "virtual copy of real life environment" and you test it out;
- visualization
Main drawbacks:
- data need: often data intensive, and as always in modelling: garbage in, garbage out
- dependent on type of software: difficult to tailor, or when more easy to tailor: often high/specific skill in programming needed
- most often most difficult is validation of the current state
Hope this helps!
Gert
|
|
|
|
|
10/12/2011 02:55 PM
|
|
|
I won't say they are of little value, however I will say that a physical simulation, where people are actively engaged in moving things around, etc. engages them far better and flushes out a much richer level of experience and detail.
Only experts can set up computer simulations, and only experts can adjust them.
A manual simulation invites participation from many more people, and makes it easier for them to see how the people and process interact.
|
|
|
|
|
10/12/2011 02:55 PM
|
|
|
We occasionally used ProModel in the past to simulate lots of things - before our lean journey began. I have to say, I was quite mesmerized by what one of my simulation buddies could put together using that package. So we asked the same question of our Senseis. They always came back with a pretty similar answer in the form of a question; "Why would you want to simulate instead of just trying it?" Another point I might add is that generally speaking, the only one in a simulation who learns anything is the person building the simulation. When you actually try something, everybody learns something.
Needless to say, we don't use simulations much anymore.
Tom
|
|
|
|
|
10/12/2011 02:55 PM
|
|
|
In my opinion, computer simulated models and Lean are counter intuitive. In order to truly understand the process, you need to go to the Gemba and "get your hands dirty". Talk to the people doing the work. Unless the operators are using computer simulated models to do their work, this is the wrong path.
|
|
|
|
|
10/13/2011 03:49 PM
|
|
|
Originally posted by: GertLinthout
....- no need to model "mathematically", you build a "virtual copy of real life environment" and you test it out;..
Using a computer is modelling "mathematically." A "virtual copy" of anything is nothing more than a mathematical model.
You just don't see (and may not understand) the math. That doesn't mean it isn't there.
And it doesn't free you from the responsibility of understanding the risks and limitations of mathematical modeling.
pc2.
|
|
|
|
|
10/17/2011 09:27 AM
|
|
|
Hi Julie,
I have over 25 years of using computer simulation models for process improvement, with a great deal of success.
Firstly, there are many different and capable simulation tools: ProModel, FlexSim, AutoMod, Witness, Plant Simulation, AnyLogic, Simul8, Arena, Simio, SLX, ExtendSim, Quest and Enterprise Dynamics to name a few. They all have different strengths and weaknesses, process modeling capabilities and ease of use - and clearly, some are better than others. I'm not going to recommend any without knowing your circumstances.
Secondly, I strongly recommend that you have a simulation professional conduct, or at least supervise, your project. Simulation requires competence in a number of technical fields, aside from the obvious process modeling skills:
Interviewing skills, to tease out the often unspoken rules, priorities and routes that are integral to the process, from your process experts (designers, managers, supervisors, operators, maintenance staff, etc.).
Statistical analysis, to analyze and model process data, design experiments and interpret results.
Computer programming and testing, to create an accurate simulation model.
Presentation skills, to communicate the results of the simulation study effectively.
In an effort to win greater sales, many simulation software vendors encourage anyone to try their hand at simulation, and the results are what you might expect. Indeed, one of my biggest problems has been to overcome the skepticism of people who have been involved with a badly-conducted simulation exercises in the past.
Finally, you need to follow a good simulation methodology. For brevity, I will not outline mine here, but three points I will make:
You must identify the objectives of your simulation study. The objectives dictate the complexity and duration of your simulation study; it the model is not accurate enough to support your objectives, it will fail.
You must verify the simulation model: test that it does what you expect it to do, without bugs. Fix any bugs found and re-verify.
You must validate the simulation model: test that it accurately represents the process you are attempting to model. Fix any problems found and re-validate.
If you don't do at least the above, any results produced by your simulation model will be suspect and should not be relied upon.
There are many advantages:
You can choose to either simulate and then implememt, or just implement. If your process design is bad, and you discover that fact by simulating it, then your costs will be relatively slight. If you skip the simulation, and implement your bad design, your costs will be(usually) astronomically higher. If your process is expensive, or complex, or difficult to modify once built, or produces high value products, or any combination thereof, then simulation is a no-brainer.
Computer simulation models execute many times faster than real time, so you can cover far more ground with simulation models than you can by experimenting with your real-life process. Depending upon the complexity of the model, you can simulation years of production in a matter or minutes or hours.
Modeling a process requires that it be studied in detail, often highlighting its more obvious problems. Many of the benefits of the simulation come from this phase of a simulation study alone.
Simulation models typically produce a number of informative metrics that the real-life system does not, giving you far greater insight into the performance of the real-life system.
Every simulation study I have ever been involved with has highlighted at least one significant, but unanticipated, problem with the design of the planned process. Simulation models can test your designs thoroughly, under many different conditions and scenarios - far more thoroughly than any human process designer.
Once you have discovered a problem, you can model the alternate solutions, simulation them all, and identify the best objectively. There's no need to fall back upon guess-work or gut-feel.
You can use the many different simulation experimentation tools to tune your process' operating parameters for peak performance.
Most simulation tools generate intuitive animations - some in true-scale 3D - that allow you to use the simulation as a focus for your process team's creativity. Everyone can see the problem(s) and make suggestions or even make course corrections. You can then simulate these and identify their merits - politics never comes into it!
Of course, there's no such thing as a free lunch, so there are disadvantages too:
A professional simulation study will cost money - but you will typically get a very rapid payback.
Conducting a simulation model takes time - anywhere between a day to a few months - depending upon the complexity of the process you're modeling and your objectives. Unfortunately, this is often a critical path activity between completion of the pre-simulation design and its procurement and implementation. In my experience, project managers are loathe to make sufficient time for a full simulation study, but that time is an excellent investment. Think "Marry in haste, repent at leisure"...
If you do not have support for conducting a simulation project at an executive level, then it can become extremely difficult to execute the simulation study effectively; many team members will not regard a simulation study, or their involvement with it, as a high enough priority otherwise.
You must have good data to hand. If you don't have good data, then you can perform sensitivity analysis across a range of values to gain better insight as to whether the accuracy of that data is significant or not. However, that can take more time to accomplish (paralysis through analysis) and will not necessarily lessen the uncertainty in the results. That said, if you don't have good data, then you have bigger problems designing a process than simply creating an accurate simulation model.
Sorry for the lengthy response, but I hope this is useful to you.
Best Regards,
Mike
|
|
|
|
|
03/19/2012 04:46 PM
|
|
|
Our company has been using ProcessModel for there process improvement needs and I can tell you that we have been able to successfully implement processes that have been virtually simulated. There is always a benefit of using simulation, you can check and see what is workable and what is not. In one of our assembly lines we were able to increase production capacity by 38%. You can try a free evaluation copy of ProcessModel from there website http://www.processmodel.com/
|
|
|
|
|
03/28/2012 05:46 PM
|
|
|
I agree with most everything that's been said except the criticism of simulation being used as a substitute for Gemba and that you should just do your process improvements rather than simulate them.
I use a very simple Monte Carlo simulation engine (XLSim) that's an Excel add-on. The process I use is to facilitate a team in building a current state process map. The I depict that map in Excel. After that the team members gather cycle time and decision outcome data for the tasks and decisions in their swim lane using another Excel tool (TimeKeeper - which I developed) or by some other means. I work in government on mainly transactional processes, so this isn't as unrealistic as it might first appear. in addition, reducing lead time is almost always the goal of the process improvements I work on. I shoot for 30 transactions through the process in question.
I then gather up everyone's data and calculate the min, median and max values and build from that triangular distributions representing the data. I also calculate the probabilities at the decision points.
What's probably the most important step at this phase is I'm able to deduce from the team data where the process delays are occurring and how big they are. I am able to add these delays to the current state map just as if they are tasks, and each one will also have their own triangular distribution.
Using XLSim, I then run usually 10,000 simulated transactions through the model, and the output is a cumulative probability chart that shows what the process is capable of delivering and the probability of any given outcome. I validate the simulation model by comparing the simulated lead time curve to the actual lead time curve.
At this point the team is reassembled and identifies process improvements that are fast, cheap and easy to implement. The resulting improvements usually are tweaks to the current state map as opposed to wholesale process re-design. Using Excel and XLSim, I can then simulate the effect of the countermeasures individually or in any combination, and see how the change affects the cumulative probability curve. The curve can translate (shifts horizontally iindicating a change in lead time) and/or rotate (indicating a change in the process lead time variation). Each simulation run takes a few minutes to program and seconds to run on the computer. I usually do it real-time with the team. Provided they are comfortable that their process map really represents their reality, they tend to be very receptive to the results. In addition, if any of the suggested countermeasures requires a monetary investment, management can see if the projected shift in performance is worth it.
So I agree, by virtue of my experience, with the many benefits recounted by others on this thread. Done in this way the Gemba is already baked into the simulation by virtue of the team-built process map and their data. The biggest con is the data: it can be difficult to get quality data and a large enough sample to be confident it is representative. You have to be somewhat clever to design practical data plans. For these reasons, it is important that you're workiing high-leverage projects.
I don't use it all the time, but when it's warranted, simple simulation as I've described can be an important tool that gives team confidence to execute their action plans. Teams that have experienced it once don't want to go back to "gut feel." Their mantra becomes "Prove it!"
|
|
|
|
>
Community >
Forums
Send to a Friend
