hi,
can anyone here help me in making the flow diagram representing orders going through factory? I've attempted it to model it myself using system dynamics but I'm not sure if what I'm doing is right! Can anyone here help? Or knows someone who can?
ive attached my problem below.
new to system dynamics - require help with production model
new to system dynamics - require help with production model
- Attachments
-
- system dynamics production model.doc
- (28 KiB) Downloaded 418 times
Merry Christmas
I'm feeling in a festive mood, so have a look at the attached model, though I'm sure others will probably better ways of doing this. I've create the model in Vensim DSS, but it should open in any Vensim version.
But please take time to look through the equations and logic structure to understand how I have tried to map the model to the question posed and not just take them for granted. I may also have misinterpreted the problem outlined, so eqns may need revising. The main production planning driver in this simple representation is the "number of weeks of average orders to hold". Setting the "Part B Switch" to 0 runs part A, set to 1 runs part B of your question.
I learned a lot about System Dynamics from looking at how other people create models , some good, some bad !Although it's no substitute for trying to do it yourself, everyone needs a helping hand at the start. SD books by Prof Sterman and Prof Coyle would be useful for you to read.
When your building models, always remember to enter and check dimensions. Mass balance checks (views 2 & 3) are also a useful to check for missing flows. The "Cost" elements are on view 4.
But please take time to look through the equations and logic structure to understand how I have tried to map the model to the question posed and not just take them for granted. I may also have misinterpreted the problem outlined, so eqns may need revising. The main production planning driver in this simple representation is the "number of weeks of average orders to hold". Setting the "Part B Switch" to 0 runs part A, set to 1 runs part B of your question.
I learned a lot about System Dynamics from looking at how other people create models , some good, some bad !Although it's no substitute for trying to do it yourself, everyone needs a helping hand at the start. SD books by Prof Sterman and Prof Coyle would be useful for you to read.
When your building models, always remember to enter and check dimensions. Mass balance checks (views 2 & 3) are also a useful to check for missing flows. The "Cost" elements are on view 4.
- Attachments
-
- production model.mdl
- (21.81 KiB) Downloaded 426 times
production model
Hi and Merry Christmas too to every one on the forum
My mood is not bad either and I feel that this an opportunity to verify to what point, two people starting from the same definition of the problem can build different models although not pretending to have the experience of a true SD profeesional like Wilson.
I will then not read Wilson's model not to be influenced by it and I will follow more or less strictly Coyle's method of analysing and building a model from a written description of the problem.
I just have to ask some questions.
Is your problem a real one or is it an academic one?
Or are you working or are you a student?
It would of course be much better if it was a real problem, the curse of SD being that there seems to exist mainly academic or research problems and hardly any real one's.
About the definition of your problem.
I think that the cost of holding finished or raw material in stock is the average cost and not a marginal one because there is not mention of any fixed costs in your problem.
Is it then the average cost or a marginal one?
I think that both costs are per week as it is not mentionned.
Another question is about the purpose of the model.
You mentionned building an appropriate production plan.
But what is an appropriate production plan?
One where there are no lost sales, or one where the average marginal profit is maximum or something else?
Best regards.
JJ
[Edited on 6-12-2008 by LAUJJL]
My mood is not bad either and I feel that this an opportunity to verify to what point, two people starting from the same definition of the problem can build different models although not pretending to have the experience of a true SD profeesional like Wilson.
I will then not read Wilson's model not to be influenced by it and I will follow more or less strictly Coyle's method of analysing and building a model from a written description of the problem.
I just have to ask some questions.
Is your problem a real one or is it an academic one?
Or are you working or are you a student?
It would of course be much better if it was a real problem, the curse of SD being that there seems to exist mainly academic or research problems and hardly any real one's.
About the definition of your problem.
I think that the cost of holding finished or raw material in stock is the average cost and not a marginal one because there is not mention of any fixed costs in your problem.
Is it then the average cost or a marginal one?
I think that both costs are per week as it is not mentionned.
Another question is about the purpose of the model.
You mentionned building an appropriate production plan.
But what is an appropriate production plan?
One where there are no lost sales, or one where the average marginal profit is maximum or something else?
Best regards.
JJ
[Edited on 6-12-2008 by LAUJJL]
hi,
thanks for the responses so far.
It's actually a mixture of academic and real - im in a graduate placement and spending 6 months in factory management. My boss asked me to look at the current system and suggested I check out systems dynamics to improve efficiency. Whether my findings will actually be used or not is a different matter! I googled it and came across this software and forum. So i thought i'd write a brief and see what models people would produce to understand how to approach it. I'll certainly look at those suggested authors and get the books.
The cost of holding raw materials is a marginal cost, and they are all weekly costs. The objective of the plan for me would be to build one which minimises the cost as much as possible, and not to fulfil as many orders as possible just to meet the 2 week delivery target. It would be nice to know how many orders are lost though, running the simulation. But jah the target is to minimise cost.
I've only been fiddling with the software for a few days now, so i dont have a clue at the moment, i guess i'm going to have to look through the readme of the software to figure out fully Wilson's model.
I'd really appreciate it if someone could just draw out the flow diagram showing the positive and negative loops, so I can then construct the model myself in the software and apply the equations myself, that's the only way I can learn it. And then i'd be able to understand Wilson's model a lot better also and make changes where appropriate. I just need that initial nudge in the right direction!
i would also like to use excel to put the equations in and run the simulation/graphs on that instead. is it possible to import the information into excel somehow?
[Edited on 12-6-2008 by derbun]
thanks for the responses so far.
It's actually a mixture of academic and real - im in a graduate placement and spending 6 months in factory management. My boss asked me to look at the current system and suggested I check out systems dynamics to improve efficiency. Whether my findings will actually be used or not is a different matter! I googled it and came across this software and forum. So i thought i'd write a brief and see what models people would produce to understand how to approach it. I'll certainly look at those suggested authors and get the books.
The cost of holding raw materials is a marginal cost, and they are all weekly costs. The objective of the plan for me would be to build one which minimises the cost as much as possible, and not to fulfil as many orders as possible just to meet the 2 week delivery target. It would be nice to know how many orders are lost though, running the simulation. But jah the target is to minimise cost.
I've only been fiddling with the software for a few days now, so i dont have a clue at the moment, i guess i'm going to have to look through the readme of the software to figure out fully Wilson's model.
I'd really appreciate it if someone could just draw out the flow diagram showing the positive and negative loops, so I can then construct the model myself in the software and apply the equations myself, that's the only way I can learn it. And then i'd be able to understand Wilson's model a lot better also and make changes where appropriate. I just need that initial nudge in the right direction!
i would also like to use excel to put the equations in and run the simulation/graphs on that instead. is it possible to import the information into excel somehow?
[Edited on 12-6-2008 by derbun]
Production model
Hi
Thanks for the informations.
But there is one important point to find out.
Is system dynamics the right methodology to use?
I do not think that making a model that does not generate any added value provides anything good.
You may think that the method has no interest while the reason might be that the problem does not require the method.
One must fist try to identify what kind of problems the production has. Is it a problem of not enough profitability?
If it so, is there any particular reason for it?
SD is particular usdeful where there are strong variations in some places in the business that generate high costs.
In your case it may be that the people running both businesses do not coordinate their decisions.
Most of the time SD helps lessen oscillations in systems that can happen for many reasons that can come from bad decisions within or outside the system or from an inadequate structure of the business.
If there is nothing like inventories or backlogs oscillating or if the oscillation is not very annoying, then one must eventually consider other methods.
SD is a method that helps things being smooth.
Example: SD should help at least in theory reduce global or particular economic fluctuations.
So what is the real problem?
You can certainly build the model in Excel, because it is not complicated and higly aggregated, and if you are used to Excel you may better understand how it works.
One more idea.
Sd is not an easy method to use and needs to be seriously studied. You cannot just learn it in a week or in a month.
The software is not easy to use too and needs to be studied.
To do that the best thing is to study the user guide.
To my opinion Sterman's book is not a good introduction book and after you have studied it you will still not know how to build a model.
Geoff Coyle's last book is much more practical and easier to study. It has the drawback of using a software which is probably not much used any more, but it is relatively easy to translate the examples and the exercises in Vensim.
I did it.
Best regards.
JJ.
[Edited on 9-12-2008 by LAUJJL]
Thanks for the informations.
But there is one important point to find out.
Is system dynamics the right methodology to use?
I do not think that making a model that does not generate any added value provides anything good.
You may think that the method has no interest while the reason might be that the problem does not require the method.
One must fist try to identify what kind of problems the production has. Is it a problem of not enough profitability?
If it so, is there any particular reason for it?
SD is particular usdeful where there are strong variations in some places in the business that generate high costs.
In your case it may be that the people running both businesses do not coordinate their decisions.
Most of the time SD helps lessen oscillations in systems that can happen for many reasons that can come from bad decisions within or outside the system or from an inadequate structure of the business.
If there is nothing like inventories or backlogs oscillating or if the oscillation is not very annoying, then one must eventually consider other methods.
SD is a method that helps things being smooth.
Example: SD should help at least in theory reduce global or particular economic fluctuations.
So what is the real problem?
You can certainly build the model in Excel, because it is not complicated and higly aggregated, and if you are used to Excel you may better understand how it works.
One more idea.
Sd is not an easy method to use and needs to be seriously studied. You cannot just learn it in a week or in a month.
The software is not easy to use too and needs to be studied.
To do that the best thing is to study the user guide.
To my opinion Sterman's book is not a good introduction book and after you have studied it you will still not know how to build a model.
Geoff Coyle's last book is much more practical and easier to study. It has the drawback of using a software which is probably not much used any more, but it is relatively easy to translate the examples and the exercises in Vensim.
I did it.
Best regards.
JJ.
[Edited on 9-12-2008 by LAUJJL]
production model
Hi Wilson
Your effort and good mood does not seem to have interested many people!
Here is joined 7 different models that try to optimize your production model.
I did not try to make one myself following Coyle's method because I do not think it will interest anybody.
In fact there are 13 models, I have joined 6 of the models with no subscripted parallel simulations.
To optimize I run 100 parallel simulations and I optimize the average pay off, otherwise the results vary when you change the seed and the parameters adjust to the particular random data.
I do not know, but I do not understand why the cost of lost units is counted two times. I took away the cost of lost units in the weekly costs as it is already counted in the fulfilled orders. Anyhow it does not change the problem bery much.
Your solution apparently behaves poorly especially with parallel simulations. It still behaves no very good with one simulation. I have not really understood why.
The best I have found is the production_model_6 but it necessitates to increase the time to transfer orders from 1 week to 2, and it gererates too a much more varying production rate. So that may not be the best solution.
Anyhow having no further information from Derbun and his problem, it is difficult to go further.
The fondamental problem is that orders vary very quickly and it is difficult to transmit this variation with a 2 weeks ordering time and a 5 weeks total production time. It is this difference of delay that makes things difficult to handle.
Joined the models with vpd and voc optimizations files.
Regards.
JJ
Your effort and good mood does not seem to have interested many people!
Here is joined 7 different models that try to optimize your production model.
I did not try to make one myself following Coyle's method because I do not think it will interest anybody.
In fact there are 13 models, I have joined 6 of the models with no subscripted parallel simulations.
To optimize I run 100 parallel simulations and I optimize the average pay off, otherwise the results vary when you change the seed and the parameters adjust to the particular random data.
I do not know, but I do not understand why the cost of lost units is counted two times. I took away the cost of lost units in the weekly costs as it is already counted in the fulfilled orders. Anyhow it does not change the problem bery much.
Your solution apparently behaves poorly especially with parallel simulations. It still behaves no very good with one simulation. I have not really understood why.
The best I have found is the production_model_6 but it necessitates to increase the time to transfer orders from 1 week to 2, and it gererates too a much more varying production rate. So that may not be the best solution.
Anyhow having no further information from Derbun and his problem, it is difficult to go further.
The fondamental problem is that orders vary very quickly and it is difficult to transmit this variation with a 2 weeks ordering time and a 5 weeks total production time. It is this difference of delay that makes things difficult to handle.
Joined the models with vpd and voc optimizations files.
Regards.
JJ
- Attachments
-
- production.zip
- (72.4 KiB) Downloaded 389 times
JJ,
Sorry I've only just noticed your post. I will try and have a look at your models over the festive period.
I initially assumed the problem was 'just' an academic course question. Now I see that I was wrong, like you I think a lot more questions need to be asked before anything realistic could be developed, there are too many peices of crucial info missing or open to misinterpretation.
The model description appears very much as a "push" process, with only random fluctuations in production rate. In these days of "lean" manufacturing, it might be a good example to compare with an alternative "pull" process, kanban card systems and all that goes with it.
Anyway, have a very Merry Christmas
Andy (aka Wilson)
Sorry I've only just noticed your post. I will try and have a look at your models over the festive period.
I initially assumed the problem was 'just' an academic course question. Now I see that I was wrong, like you I think a lot more questions need to be asked before anything realistic could be developed, there are too many peices of crucial info missing or open to misinterpretation.
The model description appears very much as a "push" process, with only random fluctuations in production rate. In these days of "lean" manufacturing, it might be a good example to compare with an alternative "pull" process, kanban card systems and all that goes with it.
Anyway, have a very Merry Christmas
Andy (aka Wilson)