Help! Need a simple way to analyze some data.

[NucEngineer]

Hello forum! First time poster here and I'm in a pinch. If you want to jump ahead to the question, it’s at the end. The short of it is that I have to figure out the simplest way to analyze a large amount of data that will be compared with other data in order to determine if a system is operating within a desired range. I will give some background first, and then plug my question in at the end. Due to the nature of my work however, details will have to be limited.

The long of it:
I have been responsible for development and implementation of a new ventilation system. Think of it as a saw-dust collection system that might be used in a shop in the area around your saws, sanders, etc. to control the saw-dust as it’s generated.
This system was tested extensively in “mock-up” conditions, where the inward air velocity was measured at specific locations over and over again, and in different configurations. The system was tested at a “high flow rate” and a “low flow rate”. Provided everything was sat at the high flow rate and sat at the low flow rate, I could by default say that as long as this system was operating anywhere in-between these conditions, it will work.
Once the system is installed and operational at the job site, I have to go test it and compare my job-site data with the data obtained during mockup testing to determine that it’s operating within the desired range. For reasons that I don’t want to discuss, I cannot use flow rates or differential pressures as monitored by the fans pulling the air through the system. Air velocity at the system opening is the only means of “evaluation”.
For one “test” I have 16 locations around the perimeter and at each location, I have 3 vertical “test points” that I’ll call A, B, and C. While the A, B, and C test points are expected to provide different velocities with respect to each other, as a whole, an A test point is expected to provide similar velocities as all other A points, B similar to all other B points, and same for C. In addition to that, one test location as a whole is expected to behave similarly with respect to the other 15 test locations around the perimeter.

The last engineer royally screwed this up by taking the lowest of all values and the highest of all values seen during mockup testing, and used that as a blanket “go/no-go” range on the job site. It was a large spread, and all values taken at the job site were within his specified range. He failed to notice that all values were close to his max number, said it was good to go, and a catastrophic failure occurred. As an over-all average, the system was actually operating way outside of the designed and tested range.

Analyzing the data:
I need a simple method in determining that not only each data point at the job site falls within a range observed during testing, but that the distribution of those numbers are also within acceptable range comparable to mockup testing conditions.
This is the best method I have been able to come up with for job-site acceptance criteria. Remember that I have tested the system SAT at a “low flow” and a “high flow” condition. For the A test point, I pulled the lowest value observed in the “low flow” and the highest value observed during the “high flow”. For acceptance criteria at the job site, this will be the MIN and MAX value allowable for each A test point. The same process will apply for B and C respective to data obtained during testing for those points.

I then took the average of all A test points at the “low flow”, and the average of all A test points at the “high flow”. This will be the MIN and MAX range for the average of all A test points observed at the job site. Same will go for B and C.

So during job site evaluation, all A/B/C test points will be recorded at the 16 locations. Provided each of my recorded A test points fall within the individual MIN and MAX range, That is the first stage of acceptance criteria. Second stage will be to add all of the A job site test points and take the average. Provided the A test point average falls within the average MIN and MAX range, I know that for the A test point, it is in fact operating somewhere within the “low flow” and “high flow” range that was tested in mockup conditions. Obviously the same would follow suit for B, and C.
I may even take it a step further and compare the average for ALL job site test points relative to the average for ALL mockup test points to make sure that it is operating within the same range as a whole.


QUESTION:
Is this the most efficient way to analyze this data for acceptance criteria, in order to say “YES” it is operating within a desired range or “NO” it is not? Is there a more simple way to do it?

Thanks for all the help!

 

[MikeHalloran]

It happens that a friend of mine is now buying a sawdust collection system.
It differs from yours in having a mere four inlets,
and in the presence of 'blast gates' at each of those inlets, so that any or all can be closed in order to reduce shop noise and allow more of the available pressure difference to be applied to the inlet that's actually in use.

Absent some active balancing mechanism, I do not think it is possible to 'balance' a one-to-many piping system such that all tributary flows are equal at any one combined flow rate, never mind at two or more.

Suppose you manage to balance the system as you describe it once. What happens when the 'sawdust' starts building up on internal discontinuities?

I.e., I don't think your most basic assumptions are supportable.








Mike Halloran
Pembroke Pines, FL, USA

 

[NucEngineer]

Hi MikeHalloran, thanks for your response. I should have clarified that this system does not have multiple inlets. It is multiple plenums that connect to one large "inlet," in which everything is symmetrical. All test points are symmetrical to the system "inlet" geometry.

Also, assume the "sawdust" collecting on internal discontinuities does not play a factor in system performance or consistency.

I may have gone too far into the weeds about the system itself instead of my actual question related to comparing the data. Based on design and intended use, I have to assume that airflow will be fairly consistent and symmetrical. However, based on the reality of airflow characteristics and other small anomalies, I know that I will see some variance between test points.

Bottom line, I will need to ensure that each individual test point falls within an accepted range, as well as overall performance as a whole also falls within an accepted range.

Thanks

 

[GregLocock]

You'd be much better asking this in the statistics forum. If you like you can red flag it and ask the management to move it over there for you.

Given that you have had one catastrophic failure, and I assume the results were more serious than a bagful of sawdust blown out the shed, it pays to be cautious.

First of all you may know enough about your system to assume linearity, so that if you are operating somewhere between "high" and "low" then the pressures won't be higher than the desired maxima, but that is an assumption.

"Provided the A test point average falls within the average MIN and MAX range (of A), I know that for the A test point" is redundant as it will always be true, I think.

The next thing you need to do is inspect the data. I'd plot all my As as a histogram for the three load cases, test low, test high, operating, same for Bs, same for Cs. This may well give you your answer, or suggest some analysis directly.

If all As are supposed to be similar then they are probably normally distributed, which you can check from the histograms. This means you can start doing some traditional stats on it and start to look at the probability of any A exceeding some desired maximum value, for example.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[tbuelna]

I have seen big problems with custom designed dust collection systems for large CNC mills that were used to machine foam or composite materials. The vendor designed the dust collection system based on their "experience", but during operation the system did not come close to performing as required. The main issues seemed to be loss of performance as the filter elements accumulated even a modest amount of dust, and the inability of the intake nozzle to extract a large percentage of the fine dust being flung from the high speed cutting tool.

 

[IRstuff]

"The main issues seemed to be loss of performance as the filter elements accumulated even a modest amount of dust, and the inability of the intake nozzle to extract a large percentage of the fine dust being flung from the high speed cutting tool."

That sounds like the intro to the Dyson vacuum commercial

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529


Of course I can. I can do anything. I can do absolutely anything. I'm an expert!
There is a homework forum hosted by engineering.com:

http://www.engineering.com/AskForum/aff/32.aspx

 

[tbuelna]

The machine was a 10'x40'x9' 5-axis CNC mill. It was mostly used to cut foam blocks for composite tool plugs. The dust collection system had both centrifugal separators and fabric filters. But what we had to do after machining each foam plug was sweep up the piles of debris by hand. The only thing the dust collection system was good for was creating enough of a vacuum in the room the mill was located in to contain the foam dust.

 

[dbill74]

My 2 cents worth: Abandon and forgo your mockup/model. That the system failed would indicate to me that there was something wrong with the model or that it did not scale as intended.

Instead focus on why the system failed and look at what would need to change to prevent the failure from happening again.
You said the system was operating outside the designed and tested range, how so? What caused the system to start functioning outside the designed range?

Once you get the system operating, then go back to your model and see what differed from your mockup's assumptions so you don't make the same error(s) again.

Why are you averaging the minimum and maximum values on the mockup? Since your objective is to create a vacuum for collecting dust, then logic tells me you needs a minimum air velocity for the target particulate size, a max value would be for safety purposes. By averaging the minimum values you are setting yourself up for failure by allowing some inlets to have velocities less than you actually need thus creating insufficient vacuum and duct collection potential. Same can be said of your maximums.

A better strategy would be to establish the minimum velocity as the maximum minimum value. Then to establish a general maximum velocity, use the minimum maximum value. Either that or set the min/max for each inlet separately.

 

[Dougt115]

What are you calling a plenum? with a single inlet? It sounds like you installed a sawdust collection chamber with a single inlet to create the flow? Is this under the floors or along the walls or at the machine throw area?

My first take is that your model complete missed the effect of the sawdust on the flow pattern of the air and all data analysis will be a waste of time.

Particulants in air will drastically alter flow and should not be ignored.

 

[NucEngineer]

Really the question was more related to analyzing the data. The system is not actually used to collect sawdust and it's not in a woodshop. The particulates that may be encountered are extremely fine, and will be be filtered via multi stage hepa filters. Again, for the purposes of this analysis, the impact on the system as a result of the particulates is negligible.

I know that the mockup system performs as desired. The one and only goal is analyzing the system at the jobsite and ensuring that it is operating within the same parameters as my mockup system. If it is not operating within the desired range, system adjustments will be made (VFDs and dampers) to bring it into range. This range being established as the high and low flow testing that was conducted in the mockup.

Thanks

 

[dbill74]

If you just want to compare 2 sets of data points side by side, could have just said so and my first choice is MS Excel. With its conditional formatting you can easily and visually find data that is in/out of a specified range.

But then you have me and I would venture the rest of the community wondering why the elaborate explanation?
To be honest, not sure what you expect to get out of the mockup at this point or your proposed analysis. You discussed min and mix flow rates, what exactly are you measuring/needing to maintain? Volumetric flow rate, velocity, something else? What is the purpose of the system? Based on your latest post here, the example you gave in your OP was a poorly chosen example.
I am also curious how the mockup compares to the actual system? Is it a scale model or a portion at full scale?

PS) Have you ever watched Mythbusters on Discovery Chanel? They often find that things done in scale model don't always translate well to full size experiments.
Point being it is risky to assume that just because your mockup works as intended the real thing will work just like the mockup.