I need to create combined statistics for several stations. I did see this:
which helped a bit, but I still have more questions.
I attached an image showing a typical frame in my model, which contains different amount of stations. I want to have a final chart showing the utilization of each frame in the model. Think of one frame as, for example, the grinding area.
I also made and attached an example of an idea of mine (grinding frame have three machines, printing have two). Would this show accurate statistics? Can it be improved? Or should I divide each cell with the number of stations in the frame, so I get a proper average "per station" in the frame?
Highly appreciate any ideas regarding this!
I don't think I understand you completely.
First, some clarification. Each product follow a orderlist, so it will for example to to Grinding several times. Each time in grinding it have a different ProcTime. The stations in Grinding are parallell, so a product will only go to one of the stations and then continue to the next operation. All stations share a common buffer before the interface. So there can be different products in all of the stations in Grinding at the same time.
Unfortunately I don't understand what you mean by adding a new station for the statistics. I mean, if I understand you it should work if there is only one product in one of the stations, and I copy the ProcTime to that station and then get the statistics. But I don't understand how it would work if there are two stations used at the same time. Could you elaborate a bit more what you mean?
I think I understand what you mean now, and it is not exactly what I am after. You say I should test and evaluate each area separately to find out the "maximum production capacity" of each area. This is quite interesting, but for now I do not need that. I would also take a lot of effort because I have roughly 40 different areas in the model.
Consider the follwing, in the master production frame I have a chart. In the chart I can add all individual stations from the areas. For example the bars in the chart would be Grinding.Station, Grinding.Station1, Printing.Station etc.
This is basically what I am after, but the resulting Chart would have a lot of bars. To decrease the number of bars I want to group them per area instead. This way I can see how much the entire area is used with the current production plan.
In an area there is one buffer before the stations and one buffer after them. In the end I want to experiment with the production plan and buffersizes to see how the areas utilization is affected. Therefore I do want the utilization to show amount starved (as waiting) and blocking (as blocked) as well as all the other statistics (working, setup, stopped etc).
I don't know if it can exactly solve your problem, but I would suggest using the Statistics Wizard. (Insert a Chart object in frame and right click on it > Statistics Wizard)
You can include subframes (or not) in the statistics wizard and see a column for every station.
For grouping the bars according to area, how about creating different charts and then displaying them in a frame side by side(Chart > Display Tab > Display in Frame)
I guess this was my initial question, how should a mean value for all stations be calculated?
Now I have implemented the solution I mentioned in the beginning; in every cell I add all the stations attributes together and divide by the amount of stations in the area.
Example of the cell "Working" in the charts input channels:
Working (Grinding.Station.StatWorkingPortion+ Grinding.Station1.StatWorkingPortion)/2
Do you think this is a good way to calculate the statistic mean value for one area?
I think this might have been a good alternative, but I would prefer to have everything in one graph to get a good overview.
Just to explain, earlier in the model each area was modeled with a parallellproc. The parallellproc's YDim was changed depending on the amount of stations in each area. If I simply made a chart with all the parallelprocs I would get the graph I wanted.
But it was necessary to change the parallelproc to singleprocs instead. This obviously gave the graph a lot more bars, so I wanted to combine the stations in one area somehow. As I mentioned above I have implemented a solution for it now, and from my tests it seems to be working.