Multi-echelon structure

[Lubo]

Hi

I have been working as a management consultant, applying Systems Thinking and System Dynamics and developing management flights simulators in tools such as Powersim, Vensim or Ithink for some time. I am new to Ventity and like the direction its development is taking.

As a learning experience to understand Ventity’s capabilities better I am trying to build a network of warehouses, distribution centres and shops. I think
I have built a simple working structure of entities with attributes – a basic pipework, which needs to be further developed with policies, rules and decisions to capture the network dynamics. Is my application of “entities with attributes” as opposed to “hidden array equations” correct?

I’m trying to add additional dimension and have multiple items flowing through this structure (multiple network nodes) where I currently only have a single “item”. For example adding Item 1, Item 2 and Item 3 flowing through the pipework. Or to make it more complex we could try to add a hierarchical (tree like) structure. For example having a “Group A” which includes Item 1 and Item 2 and “Group B” containing Item 3 and Item 4 – and all this items flowing through the model.

I am struggling a little with the concepts of attributes, entities, reference attributes and keys and keep on getting all kinds of errors in my attempts when I try to link entities with 2 keys (Stock and item).

What would be the most elegant and efficient way to build this structure?

Attached is the pipework model mentioned in the post. Looking forward to a fruitful discussion and ideas.

Lubo

[tomfid]

Here's a model that may inform your thinking:

SupplyChainNetwork5.zip

(24.4 KiB) Downloaded 431 times

[tomfid]

This is a network version of the Beer Game model that installs with the Ventity examples (Supply Chain \8). The SupplyChainLevel entity in that model gets broken up into two things: a "Level" entity with an inventory and a "Relationship" entity that contains the order book and shipments in transit.

Adding an additional commodity to the model would be straightforward (I think). You'd just have to add an additional attribute for the product type to the Level and Relationship entities. You'd also have to supply additional initialization data.

[tomfid]

It turns out to be a little tricky to add the product attribute, so I went ahead and did it. See attached.

SupplyChainNetwork6.zip

(31.49 KiB) Downloaded 404 times

In this version, there are two tiers, retailers and suppliers, and two products, beans and rice. Each retailer can buy from either supplier. As is, retailer 1 uses both, but retailer 2 uses supplier 1 exclusively until time 60. The network could be completely different for beans and rice. This would avoid computing a lot of 0s if each supplier has a unique portfolio.

Currently the data defining these are almost the same (I cheated and duplicated the beans to get the rice, then scaled by 10%). But presumably in a real situation you'd have a spreadsheet or database to initialize everything uniquely.

[Lubo]

Thank you very much for your reply, the examples you have shared are perfect. So one can use initiation sheets to define different distribution networks for different products – is there a limit on a number of initiation sheets?

Your model is much more elegant than what I have created but could you please have a look.
If we assume for a moment that I want to have nodes in the model (warehouses, distribution centres, retailer inventories) as submodels or components of different levels of granularity or complexity (I would be able to choose which complexity to use based on the type of a problem being addressed). Do you think it would be better to have each of the nodes as separate entities as in the attached model? Would such a structure be simpler for initiation purposes when one could be using data from other systems.

In the attached model I was tinkering with Ventity and I think I got the previous model version to work. I think the problem I faced was breaching the maximum of key attributes for a single entity where having 4 keys in a compound key (NDC, RDC, Shop, SKU) is one too many (somehow the deleted attribute key of a relationship entity might still count as the fifth?). I broke the structure up and only included 2 nodes and a product attribute in a relationship entity. And this is working fine. This way I could also easily add additional layers in the distribution network if required.

The issue I am facing now is initiation of the model. It is working fine with the built in data but it is not working with external Excel files. Not even with the “working built in data” exported into an Excel file and then used as an input. What am I doing wrong?

Thank you
Lubo

[tomfid]

There's no hard limit to initialization data; you're limited by memory and computation time for the entities they define. The limit to those things depends on the size of the entities (# of equations) but generally 1000 is easy and 1000000 is going to be very slow.

One limitation of the current architecture is that you can't create collections of dissimilar entities. So, if you want warehouses and distribution centers to have distinct structures, but you also want to mix and match them in arbitrary supplier/demander relationships, it may take some thought to figure out how to do that.

When you have multiple attributes (NDC, RDC, Shop, SKU) you don't necessarily have to use them all for the key. You could have a single key (ThingID) and use the attributes for other purposes. This could make it a little hard to ensure relational integrity, but that should probably be done in the data anyway.

I'll take a look at the model.

[tomfid]

With Model 3.xlsx, I see a bunch of errors like:
02/16/2021 12:15:11: Run 'Run' at time [0] - Null reference ndc_stock in ndc2rdc[NDC1|RDC1|SKU1] using "NDC1|SKU1". This may cause other errors and warnings.

The cause seems to be that there are no NDC Stock entities (you can see this in the entity picker, or by looking at the count of NDC Stock[]). I'm not sure why - the data looks fine. Thinking ...

[tomfid]

Figured it out. The problem seems to be a bug in the Excel reader when there are spaces in the entity/tab names. As a workaround, getting rid of the space gets this running.

Level 3.zip

(161.12 KiB) Downloaded 419 times