Teaching DOE

John
 
I’m teaching DOE to GB’s next week for my very first time.  If you get any idea’s please post them as I feel I’m going to need all the help I can get.
If nothing comes through, I have some idea’s of the way to go.  I will gladly pass on my experieces from next weeks course.
 
Chin up
Trent

JOHN,
 
Call me at (860) 654-4800.  I have an excellent approach using the Catapult that never fails!!!  I will be happy to share with you.
 
Regards,
 
Pete Teti

PeteI was speaking to John Matthews our Six Sigma Blackbelt, and he said he had contacted you about doing a DOE using the catapult. I am John’s master black belt, working here in Belfast for Bombardier, on this issue, and it would be great to get some advice from yourself. Our problem is basically getting a repeatable process, and there is too much noise, to allow us do a DOE which is really convincing to classes we teach. We reckon if we cannot get good repeatability for a fixed set-up, then any model we derive from our DOE runs will not be very good.If I can phone you tomorrow (Thursday), it would be really good. My telephone number is 011442890462108
My fax number is 011442890733647
My e mail is [email protected]
My address is Bombardier Aerospace, Six Sigma Centre, Airport Road, Belfast BT3 9DZ, N IrelandI’ll phone about 4.00 pm our time, which is probably 11.00 am your time. Look forward to talking to you.Regards
Michael McGrath

Sounds extremely interesting. John, can you share what you have done in DOE and also in basic statistics. We are in process of setting up some activities for such training and this would help us a lot. Thanks.

Noises are there to discover, not to avoid. Its perfectly normal to have a DOE that cannot repeat itself, which is mostly becoz of the unstableness of underlying noise, or to put it another way, the noises are not true “white noise”, there are factors to be dig out, the closer the noises to true white noise, the better repeatability you have, and vice versa. When you encounter these repeatability problems, best thing you can do is to try to dig some factors from the noises, otherwise you can just tell your pupiles the story about the nature of noises, after all, it’s the mechanism that counts, not the result.

John
Jenga blocks sound interesting.  Can you elaborate more?  Its now fri afternoon and the brain is slowing quite quickly.  What do use for the factors?  The levels of interest?  What is the effect?   How big are the groups for training?  Do the figures work in Minitab?
Sorry if my imagination is’nt working but its been a long week.
Cheers
Trent

Hello.  We use a paper helicopter as the basis for the DOE instruction.  There are three variables:  Wing Length, Body Length and Body Width.  A variation also can include using or not using a paper clip.  This works pretty well for Green Belt level understanding and doesn’t take nearly as long as the catapult experiments typically do.
I hope that this helps.
Jim Johnson.

The paper helicopter is very simple, fast and involves all the particpants in the understanding of DoE. It has controlled and uncontrolled variables and is available with standard office supplies – ie, paper, paperclips, etc. Once I used the helicopter after using the catapult I have never gone back.
email me if you need further info.
[email protected]
http://www.qprs.com
 

Here is a link to an online catapult that is a lot of fun.  I am making it accessible through the end of the month to get comment for improvement.  We use these type of simulations as part of our online Six Sigma courses, and would appreciate feedback from others involved in DOE instruction.
http://www.moresteam.com/trebuchet.cfm
Make sure to turn up your sound to get the full effect.
 
 

I have read everyone’s messages and agree that teaching DOE does have room for simplicity and knowledge transference. It is far from boring. I find that it is too equipment dependent. I really would like to find an exercise which can be carried into the field without a shipping trailer, which is understabdable and most importantly easily transferable to more practical on the job applications. Granted the instructor plays a major role in making the tie, but someone must have developed a better DOE training tool. Can someone tell us more about the paperclip, paper plane or other desktop exercises.

I’ve seen and used a number of DOE teaching props, like catapults, paper planes, and the like.  My favorite portable physical props are paper helicopters.  Typical factors are things like wing length, body length, body width, # of weights (paper clips).  You can also use different paper thickness.  Measure time of flight for the response.  This same tool can also be used for GR&R, process mapping, C&E matrix, FMEA, etc.  Another approach is to use a computer simulated process.  You can make your own in Excel, or use a commercial package.  The advantage of this is that you can tailor the process behavior to what you want, but it doesn’t have a much hands-on feel to it.

I think that the only thing you need is, How to plan a DOE? no matters what process you are working on, this must give you what to do and how to do it.

I use to let the class to decide most of the exercise. OK, sometime they miss the objective, but in real life is what happen and in the class is a excelent opportunity to see this and talk about it. Normally end it up with “planning is the most important part of DOEs”.
There are another big issue, about any kind of excercise to teach DOEs, the measurement system. I normally asked to the class to include before the DOE a Gage R&R to validate the measurement system. Well, if they do not have an acceptable system, they must improve it before running the DOE.

Shree,
I could not see any attachment. Could you please try again to attach you pdf? Thanks,

John,
To a degree, I agree with Mike Carnell.  The tool cannot be the problem.  The prop used to teach the tool is just a prop, so it will not be perfect/exact.  You, first and foremost, must be comfortable with the tool. When you are comfortable, use another prop that is more fitting to you.  There are an endless number of examples:  marbles, golf putting, helicopter, miniature bowling, etc.

After learning DOE, I found a good tool rdExpert to analyze experiments
Manee

would you please email me a copy of teaching the doe I am not allowed to download due to security reasons at my facility but I would be ever so grateful.
best regards,
strugggling hahaha

Our company offers an instruction guide that addresses experiments in DOE as well as other methods.  The guide is specifically written for instructors and each experiment is broken down into facilitator and student sections.  The facilitator sections have all the data and the questions to pose to the students.

Sorry, ISP went down.  To continue:
Throw out a 6″ paper plate to a random location and ask the class if they know how to set up the Statapult to hit this target.  Probably not.  How many trial shots would be needed? Probably 4 or 5.  Have a volunteer do this.  Now, what if the customer moves the target?  Would you have gained enough knowledge from the previous trail shots to now hit the new target? Probably not.  Pose this: If you could hit the target the first time, would that be of some value?  Do customers ever change their specifications?
Make a bet with the class.  “If you give me 4 practice shots, I’ll bet I can hit any spot on the floor with the very first shot.”  You will gain more than shooting skills, you will gain something Deming called “profound knowledge”.
Draw a test matrix on a flip chart:
here are some actual distances plotted:
pull bk angle    stop pin        shot 1     shot 2     avg
160                     5               89          89.5      89.3
160                     3             126.5     127.5     127.0
180                     5              118       117.5     117.8
180                     3             173.5     171.75   172.6
Notice that I changed just a bit, I took four shots but two repetitions of each run.  Why might this be an advantage?  The additional repetitions allow me to reduce the effects of random errror on my results.  I can also use this data to calculate the standard deviation of each run.  I can use these standard deviations to model the shot distance as a function of pull back and stop pin setting.  We can determine if some process settings have a big influence over the standard deviation.
Create a simple coded matrix of the variables
pull bk angle   stop pin
160 = (-)         5 = (-)
160 = (-)         3 = (+)
180 = (+)        5 = (-)
180 = (+)        3 = (+)
To aid in understanding, make the minus setting the one that results in a shorter shot.  Next show how to build a regression equation.  Begin with the calculation of the coefficients for the equation.  Here is a step by step procedure for calculating the coefficient of the A term:
              coded levels                                 distance
pull bk angle (A)    stop pin (B)    A*B          avg
       –                          –                  +            89.3
       –                          +                  –          127.0
       +                         –                   –          117.8
       +                         +                  +          172.6
—————————————————                                                       Y =      126.7
                 (117.8 + 172.6)
avg of (+) = —————-   =  145.1875   
                             2
                     (89.3 + 127)
avg of (-) = —————-    =   108.125
                             2
avg + = 145.1857
avg – =  108.1250
Delta=     37.0626
Delta/2 =  18.53125
This process works because we have a two level design and we have standardized the levels.  The students can see how this is done if the following diagram is provided:
                                   |  distance
                                   | /    145
                                  /|        |
                              /    |        |
                   108  /       |        |
—————–|——|—–|——
                        -1       0      +1
                       coded levels of A
                              RISE     (145-108)
           SLOPE  =  ——- = ———  = 18
                              RUN           2
              coded levels                                 distance
pull bk angle (A)    stop pin (B)    A*B          avg
       –                          –                  +            89.3
       –                          +                  –          127.0
       +                         –                   –          117.8
       +                         +                  +          172.6
—————————————————                                                       Y =      126.7
avg + = 145.1857    |   149.8     |  130.9375 |
avg –  =  108.1250   |   103.5     |  122.4       |
Delta  =     37.0626  |  46.3125  |  8.5625     |
Delta/2 =  18.53125 | 23.15625 |  4.28125   |
We can therefore construct the model in this fashion:
                          =  delta A       delta B     delta AB
Shot Distance = Y+——-A+———B+——–AB
                                   2               2               2
Shot Distance = (126.7)+(18.5)A+(23.1)B+(4.3)AB
We can simply add the extra terms one after another because the test matrix was orthogonal.
We have an equation.  The plate can be placed at any random location and the equation solved for the set-up parameters.  There is one equation and two unknowns, so many solutions exist, but we can pick one.
The center point is where the input parameters are set to their mid-points.  The mid-points have been coded a”0″.  Let’s substitute this value into the equation for A and B, which results in the predicted distance.
126 inches.
Place the plate at this distance.  Insure your reputation is on the line if the ball misses the plate.  Build tension in the classroom.  If you carefully follow your SOP’s and have a consistent pull/release mechanizism, you will amaze the class with your powers.  Before your load and shoot, remind the class of the lessons learned of the essential elements of DOE:
– An orthogonal array
– Coding the inputs
– Building the model
– Confirming the model’s validity
Site the plate, load and shoot. Even you will be amazed the first time you do this!
Good luck.

Juan, many sources exist on this site and others. e-mail me for some specifics. [email protected]
Mike

Does the Crystal Ball catapult simulation still exist?
thanks,
George Dyson
 

Thank you so much, your reference is very valuable.
Joaquin

Hi, Mike
   i don’t know if you are still in the original company, now i want you give a favor about the DOE kits. Our company ever belongs to ITT China TianJin and now was bought by another EuroAmerican company last year. We have a lot of Black Belt resource and also want to get more in the future. This year, we will continue to organize BB training in China and want to buy DOE training kits. Do you have some good advice on it? hope you can answer me as soon as possible.

Rachel, e-mail me at [email protected] for more information.  Regards,  Mike