VansAirForceForums  
Home > VansAirForceForums

- POSTING RULES
- Donate yearly (please).
- Advertise in here!

- Today's Posts | Insert Pics


Go Back   VAF Forums > RV Firewall Forward Section > Traditional Aircraft Engines
Register FAQ Members List Calendar Today's Posts

Closed Thread
 
Thread Tools Search this Thread Display Modes
  #11  
Old 10-08-2014, 07:49 AM
rv6ejguy's Avatar
rv6ejguy rv6ejguy is offline
 
Join Date: Mar 2005
Location: Calgary, Canada
Posts: 5,766
Default

Quote:
Originally Posted by grubbat View Post
OFAT (scientific method) is 19th /20th century stuff long abandoned by top quality experts. Today it's a DOE (design of experiment) set up as a fractional factorial or if money and time allows, a full factorial experiment. Either way, you learn so much more besides which variables are significant, but also if there are interactions between and within variables. All this while in a shorter timeframe and lower cost. (Interactions is really good info, and pretty common. OFAT will make you go broke before you get enough information about interactions)

Either way, all this publicity should be good for all parties assuming we can fail to reject the null hypothesis. (Null would be EFII is better, . )

Cj
I am not sure what you are saying here at all. Let's say we make a camshaft change and an intake manifold change simultaneously. The cam change gains us 5% hp and the intake loses 5%. We assume neither did any good so we discard and move on to something else- lost opportunity to learn. By changing only one variable, we can determine the relative change and build from there. Time consuming, yes, but still the way I and most others do it on the flow bench or dyno.

Wind tunnel and flight testing still done this way too to my knowledge. Hardly obsolete.
__________________

Ross Farnham, Calgary, Alberta
Turbo Subaru EJ22, SDS EFI, Marcotte M-300, IVO, Shorai- RV6A C-GVZX flying from CYBW since 2003- 441.0 hrs. on the Hobbs,
RV10 95% built- Sold 2016
http://www.sdsefi.com/aircraft.html
http://sdsefi.com/cpi2.htm



Last edited by rv6ejguy : 10-08-2014 at 10:12 AM.
  #12  
Old 10-08-2014, 08:02 AM
M McGraw's Avatar
M McGraw M McGraw is offline
 
Join Date: Dec 2012
Location: Greenback, TN
Posts: 534
Default Comparison

Quote:
Originally Posted by Kyle Boatright View Post
What wasn't "straight" about the story in Kitplanes?

I find it hard to challenge the results of a test where the rules were clear and everyone involved accepted the conditions beforehand.
From the way I read this thread, it appears that due to a testing equipment limitation the EFII system was not able to be properly plumbed. The manufacturer is specific about the need to have the return fuel returned to the fuel tank with a minimum seperation of four inches from the pick up port. This happens in the fuel tank before the fuel enters the fuel pump. the purpose of this limitation is to remove air bubbles and allow unheated fuel into the system. Neither a fuel pump nor an injector can be expected to perform effectively without a clean cool fuel supply.
__________________
Marvin McGraw, 5TN4
RV-14. #140039 Complete
Flight hours: 500+
2020 Dues Paid
The Dues Paid note is a reminder for me
  #13  
Old 10-08-2014, 09:53 AM
rcpaisley's Avatar
rcpaisley rcpaisley is offline
 
Join Date: Jan 2008
Location: Upland, CA
Posts: 286
Default EFII testing

At least we are back onto a useful conversation about this!

What no one expected at the Barrett test was that the majority of the time would be spent trying to make the Bendix injection and the mags work. Every part of these systems was either replaced or overhauled (mags) during the effort. And believe me it was trying and tedious to all.

In the end, there wasn't time to properly hook up the EFII system or to tune it. This is not a disaster by any means - welcome to testing. Processes rarely work perfectly on the first run.

After the fact, my request was to complete the test properly and continue the original goal of getting the Barrett facility set up to run modern engine controls. This is still my goal. It was unfortunate that a magazine article was thrown together to cover the results of an incomplete test - this didn't help builders or the magazine.

The bright side is that the same test was completed a couple days later at Titan on their Superflow dyno and the results were as expected. A nice smooth running engine that makes a little more hp. Not much of a surprise to those who operate modern vehicles.

LOTS more dyno results will be floating around soon as more testing completes.

Many thanks to all of our customers who have helped us get this far. We couldn't be working with a better group of people.

Robert
__________________
EFII www.flyefii.com
Protek Performance
  #14  
Old 10-08-2014, 10:11 AM
JDanno JDanno is offline
 
Join Date: Apr 2008
Location: Tennessee
Posts: 440
Default

I really hated to see this article in Kitplanes. Just starting to read it I could see there were too many variables introduced to come to any scientifically valid conclusions. It didn't do the EFII products or the Bendix any favors as they were not able to set the test bed up to factory specs.
It sure would be nice to repeat the test when everything is properly setup.
That said I have the ignition part of the EFII. It starts a lot easier, especially when hot. Runs really smooth (of course the Barrett IO-540 always runs really smooth) and doesn't have to be rebuilt every 500 hours.
  #15  
Old 10-08-2014, 10:37 AM
grubbat's Avatar
grubbat grubbat is offline
 
Join Date: May 2012
Location: Ga
Posts: 662
Default OFAT and DOE

Ross,
I don't want to hijack the thread nor do I want to offend you or your OFAT testing method. I apologize for both in advance.

If you take a step back and look at the testing process, the example you gave illustrated two variables. A camshaft change and an intake manifold change. However, in real life, those are not the only two variables. They may be the only two variables that we wish to acknowledge, but they are certainly not the only two variables. If i were to ponder for the next 30 seconds on this, I would add the following variables: Air temp, operator, fuel temp, run time, cam temp, oil temp, gasket thickness, valve spring rate, valve spring thickness, cam duration, cam lift, cam hardness, intake air temp, exhaust temp, intake hose length, intake hose diameter, etc. And I am only scratching the surface. The fact is that there are literially tons of input variables that can affect the output. So, the question is, which input variable is significant. Also, which input variable has an interaction going on? Also, how significant is the interaction? A person who is used to doing OFAT would choke at this point because there is simply too many variables. So the OFAT experiment has to limit stuff and make assumptions to eliminate a lot of these variables. Eliminating these variables reduces the knowledge you gain and your degrees of freedom. Not good. There are things that can be done that can reduce the input variables down to the critical few but a DOE is very comfortable with multiple variables.
Research MOEN on his book on Design of Experiments. Its a good read, if you like engineering stuff and this sort of thing.

Let me close with this example. In the IC world, the quality of the chips were never getting to the level that they needed. Several DOE experiments were conducted over time and one of the critical input variables identified was, well, the weather. WELL, everybody knows you can't control the weather so the old-timers laughed at the idea that weather, since it was a critical factor, should somehow be controlled. hahaha, Who thinks they can control the weather....hahahah. Well, someone did. They build a large building with strict temp controls and started mfg IC's and the reliability level when off the charts. It is true that some variables are uncontrollable, and so robustness has to be built into the design so that this variable does not have any significant impact to our desire output. But, hey, we learn something and become smarter and hopefully make a more competitive product.

Knowledge of input variables, no matter how many there are, and the determination of which are significant, which should be controlled, and the interactions between each is stuff that the smart folks are doing today. They are learning more at a lower cost, and at a faster rate with DOE's.

I doubt that anyone will abandon OFAT experimentation and progress to DOE based on my feedback. However, I trust that folks who are making innovation and progress in our aviation world, will at least open an ear and see what others have already learned in the Quality world. I for one am thrilled at the innovation we have seen not only on VAF, but in the experimental sector.

Believe it or not, it is this sector that is now leading the industry.
cj


Quote:
Originally Posted by rv6ejguy View Post
I am not sure what you are saying here at all. Let's say we make a camshaft change and an intake manifold change simultaneously. The cam change gains us 5% hp and the intake loses 5%. We assume neither did any good so we discard and move on to something else. By changing one variable, we can determine the relative change and build from there. Time consuming, yes, but still the way I and most others do it on the flow bench or dyno.

Wind tunnel and flight testing still done this way too to my knowledge. Hardly obsolete.
__________________
Craig

RV-3 Sold
RV-4 Sold
RV-6a Sold
RV-9 IO-360 CS, Built and Flying
Aerostar 600A, Family Hotrod
  #16  
Old 10-08-2014, 11:20 AM
Gaylon's Avatar
Gaylon Gaylon is offline
 
Join Date: Aug 2006
Location: Friendswood, TX
Posts: 82
Thumbs up Barrett IO540-X with Dual EFII Ignition

I can attest to the smoothness of the engine with the Dual EFII ignition and how quickly the engine starts with this system. I too have a Barrett built 540 that was on their dyno and outfitted with EFII Ignition system. On that occasion a couple of issues got in the way of a successful full power run but it's my understanding it had to do with the loading of the dyno and the mapping of the timing curve, hence the timing should have been disabled. Allen and Robert discussed this and obviously got it figured out since John's engine was tested at full power.

I can't speak to the injection portion of the kit but as mentioned before with Dual ECUs you are already half way there.
__________________
Gaylon Koenning
RV-10 #40468
N815DG, Flying
G3X, GTN750, GTS800
Barrett IO-540, 9:1, Cold Air,
Dual SDS Electronic Fuel/Ignition (EM5, V29.1)
Polly Ranch Airpark(7XS0)
Friendswood, TX
Dues Paid 2019
http://www.TexasRV10.com
  #17  
Old 10-08-2014, 01:55 PM
Mike S's Avatar
Mike S Mike S is offline
Senior Curmudgeon
 
Join Date: Sep 2005
Location: Dayton Airpark, NV A34
Posts: 15,420
Default Gotta go with Ross on this one.............

Quote:
Originally Posted by grubbat View Post
Ross,-----snip----- the example you gave illustrated two variables. A camshaft change and an intake manifold change. However, in real life, those are not the only two variables. They may be the only two variables that we wish to acknowledge, but they are certainly not the only two variables. If i were to ponder for the next 30 seconds on this, I would add the following variables: Air temp, operator, fuel temp, run time, cam temp, oil temp, gasket thickness, valve spring rate, valve spring thickness, cam duration, cam lift, cam hardness, intake air temp, exhaust temp, intake hose length, intake hose diameter, etc.
Ross did not talk about changing all the possible variables you mention, only two variables to illustrate how making multiple changes at the same time can created useless results.

All the other variables you mention would not be a factor--(ie, valve spring rate/thickness--as the springs are not changed) in his example as they are not effected with a cam or manifold change, or are factors that can be controlled/corrected for in the actual testing--(air or fuel temp etc). Or, as in the cam lift/duration---well that is what has been changed in the test Ross is using as an example.

Anyhow, long winded way to say despite the number of variables, to see what the result of changing any one of them is, you should change only that thing, then test against baseline. Then change something else and test.

In the example given where one change helps, and the other hurts for a net zero change----how will you ever know what each one of them did???

And, yes cam and intake changes are quite dependent on each other and often will not give optimum results unless changed together as a "system".
__________________
Mike Starkey
VAF 909

Rv-10, N210LM.

Flying as of 12/4/2010

Phase 1 done, 2/4/2011

Sold after 240+ wonderful hours of flight.

"Flying the airplane is more important than radioing your plight to a person on the ground incapable of understanding or doing anything about it."
  #18  
Old 10-08-2014, 03:20 PM
rv6ejguy's Avatar
rv6ejguy rv6ejguy is offline
 
Join Date: Mar 2005
Location: Calgary, Canada
Posts: 5,766
Default

Quote:
Originally Posted by grubbat View Post
Ross,
I don't want to hijack the thread nor do I want to offend you or your OFAT testing method. I apologize for both in advance.

If you take a step back and look at the testing process, the example you gave illustrated two variables. A camshaft change and an intake manifold change. However, in real life, those are not the only two variables. They may be the only two variables that we wish to acknowledge, but they are certainly not the only two variables. If i were to ponder for the next 30 seconds on this, I would add the following variables: Air temp, operator, fuel temp, run time, cam temp, oil temp, gasket thickness, valve spring rate, valve spring thickness, cam duration, cam lift, cam hardness, intake air temp, exhaust temp, intake hose length, intake hose diameter, etc. And I am only scratching the surface. The fact is that there are literially tons of input variables that can affect the output. So, the question is, which input variable is significant. Also, which input variable has an interaction going on? Also, how significant is the interaction? A person who is used to doing OFAT would choke at this point because there is simply too many variables. So the OFAT experiment has to limit stuff and make assumptions to eliminate a lot of these variables. Eliminating these variables reduces the knowledge you gain and your degrees of freedom. Not good. There are things that can be done that can reduce the input variables down to the critical few but a DOE is very comfortable with multiple variables.
Research MOEN on his book on Design of Experiments. Its a good read, if you like engineering stuff and this sort of thing.

Let me close with this example. In the IC world, the quality of the chips were never getting to the level that they needed. Several DOE experiments were conducted over time and one of the critical input variables identified was, well, the weather. WELL, everybody knows you can't control the weather so the old-timers laughed at the idea that weather, since it was a critical factor, should somehow be controlled. hahaha, Who thinks they can control the weather....hahahah. Well, someone did. They build a large building with strict temp controls and started mfg IC's and the reliability level when off the charts. It is true that some variables are uncontrollable, and so robustness has to be built into the design so that this variable does not have any significant impact to our desire output. But, hey, we learn something and become smarter and hopefully make a more competitive product.

Knowledge of input variables, no matter how many there are, and the determination of which are significant, which should be controlled, and the interactions between each is stuff that the smart folks are doing today. They are learning more at a lower cost, and at a faster rate with DOE's.

I doubt that anyone will abandon OFAT experimentation and progress to DOE based on my feedback. However, I trust that folks who are making innovation and progress in our aviation world, will at least open an ear and see what others have already learned in the Quality world. I for one am thrilled at the innovation we have seen not only on VAF, but in the experimental sector.

Believe it or not, it is this sector that is now leading the industry.
cj
Mike pretty well sums things that many of the things you listed here would not change hp output in the first place. I'll step in and add a few things based on 30+ years of doing this stuff for a living. First, QC and process control are vastly different things than dyno work. Secondly, I've learned that despite 30 years in the field, the interrelated things in engine performance are generally too complex to simplify and predict in an engine development program and you will often be wrong with a prediction. This is why we do actual testing- because we are always going to learn something.

The dyno room purposely controls the test environment and applies correction factors for almost all known and significant factors which would affect the results such as barometric pressure, IAT and humidity. A good operator will also do several baseline pulls to get a reliable average and he'll do all subsequent pulls at similar CHTs and oil temps to reduce experimental error to a minimum. This is how the industry does it and it's the best way to get the best accuracy. I think you'd find another 10,000 dyno operators worldwide who wouldn't agree with you on your view here and I doubt if all those guys are wrong...

Accurate baseline
One change
Observe result
Compare
Next change
Repeat
__________________

Ross Farnham, Calgary, Alberta
Turbo Subaru EJ22, SDS EFI, Marcotte M-300, IVO, Shorai- RV6A C-GVZX flying from CYBW since 2003- 441.0 hrs. on the Hobbs,
RV10 95% built- Sold 2016
http://www.sdsefi.com/aircraft.html
http://sdsefi.com/cpi2.htm



Last edited by rv6ejguy : 10-08-2014 at 03:31 PM.
  #19  
Old 10-08-2014, 03:31 PM
grubbat's Avatar
grubbat grubbat is offline
 
Join Date: May 2012
Location: Ga
Posts: 662
Default True

Hi Mike,
I was using the other variables just as an example. The point I was trying to make is that we think we know what the variables are and sometimes we do. However, a OFAT experiment greatly limits your knowledge gained because we just do not have the time and money to investigate all the corners of the frame in question. In the past, designers / innovators just had to choose through experience or past knowledge what they thought might be the significant drivers. Sometimes they were right. Sometimes, after a lot of experiments, time, and money, they learned something. Changing one factor at a time now a-days is just simply too costly. OFAT is still taught in some schools because it is easily understood. However, the top tier schools, NASCAR folks, and top quality institutions are utilizing/teaching Design of Experiments (DOE).

I was teaching this stuff back in 2000. I had a couple of great Quality mentors from Allied Signal that showed me the light back then. At the time, we were only a handful of folks (in the world) who knew about AND actually practiced DOE's. It's really can be a useful tool.

In real life, interaction within and between variables is real and very common and sometimes nonlinear. Try figuring out a non-lineral interaction with an OFAT experimentation. Gives me heart burn just thinking about it.

As far as how do you know which factor or variable is significant, which one isn't, which one is interacting with another one, and what variable that you thought was significant when in reality isn't so much, well, that's why the DOE was born. Software such as Minitab makes it easier than slide rule.

cj

P.s. Ross, you are correct that most people do not use it. However, the smart ones do. Of course, when you are talking about running a multi-million dollar engine in a test cell, time is money. (GE did it, Allied did, and others also. To compete, you have to.)


Quote:
Originally Posted by Mike S View Post
Ross did not talk about changing all the possible variables you mention, only two variables to illustrate how making multiple changes at the same time can created useless results.

All the other variables you mention would not be a factor--(ie, valve spring rate/thickness--as the springs are not changed) in his example as they are not effected with a cam or manifold change, or are factors that can be controlled/corrected for in the actual testing--(air or fuel temp etc). Or, as in the cam lift/duration---well that is what has been changed in the test Ross is using as an example.

Anyhow, long winded way to say despite the number of variables, to see what the result of changing any one of them is, you should change only that thing, then test against baseline. Then change something else and test.

In the example given where one change helps, and the other hurts for a net zero change----how will you ever know what each one of them did???

And, yes cam and intake changes are quite dependent on each other and often will not give optimum results unless changed together as a "system".
__________________
Craig

RV-3 Sold
RV-4 Sold
RV-6a Sold
RV-9 IO-360 CS, Built and Flying
Aerostar 600A, Family Hotrod

Last edited by grubbat : 10-08-2014 at 03:40 PM. Reason: address Ross's comments
  #20  
Old 10-08-2014, 05:22 PM
tadsargent's Avatar
tadsargent tadsargent is offline
 
Join Date: Jan 2006
Posts: 307
Default

Did I miss the horse power number in this thread?
__________________
Tad Stripes Sargent
GRT HXr Glass
TITAN POWERED IO-370
Hartzell's advanced composite propeller
Closed Thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -6. The time now is 12:43 AM.


The VAFForums come to you courtesy Delta Romeo, LLC. By viewing and participating in them you agree to build your plane using standardized methods and practices and to fly it safely and in accordance with the laws governing the country you are located in.