Julio Banks proposes - Mathcad - ABC (AEDC-Banks-Chauvanet) Outlier Removal Criterion

[JCBanks]

Dear colleagues, I have used the attached "Outlier Rejection

procedure" for over 20 years to detect outlier data from acquired

thermo-structural tests.

I have tried, unsuccessfully, offering the procedure to AEDC (Arnold

 Engineering Development Center) because I use one of the two

Outlier Detection Methods but they have not shown interest.

Therefore, I am donating it to the PTC Community.

The method is a simple average of the of AEDC's and Chauvenet's

Outlier detection methods criteria resulting in the formula:

C_ABC(n) = (C(n)_AEDC + C(n)_Chauvenet(n)/2

Where "n = sample size"

.Ultimately, entered the numerical results into TableCurve 2D

resulting in the formula provided in the attached workbook.

I have successfully used to determine the outlier of:

• Bolt torque (n < 16) in a pipe-to-pipe flange assembly.
• Metal temperature around turbojet engine inlets, n < 24.
• Metal temperature around turbojet engine nozzles, n < 18.
  

It is my hope that the community members will utilize the

ABC (AEDC-Banks-Chauvenet) Outlier Rejection Criterion. 

to your particular applications.

Enjoy it ☺

---

References

The proposed ABC Outlier Detection Criterion is a result of the

interest inspired by the reference article and reading about  the

life Leonhard Euler converting me into a maven or mavin).

1) "The Unreasonable Effectiveness of Mathematics in the

     Natural Sciences"- is a 1960 article by the physicist.

     Eugene Wigner"

2) “A maven (also mavin) is a trusted expert in a particular field,

     who to seeks pass knowledge on to others. The word maven 

     comes from Hebrew, and it means one who understands, based 

     on an accumulation of knowledge”

[DJNewman]

You should post this here for much greater appreciation: https://community.ptc.com/t5/Mathcad/bd-p/PTCMathcad

This board is for new ideas for the Mathcad software itself.

[JCBanks]

I visited the site you recommended but I am not able to post

and only posted on this site.

Would you provide me with an alternative site which I might

be able to post this idea?

I should be noted that the ABC Outlier Rejection criterion I am

proposing may be made into a Mathcad software function.

Please advise, DJNewman

[DJNewman]

Well, a moderator moved it to where it should be.

[JCBanks]

DJNewman, thank you for informing me that the

ABC Outlier Rejection file has been moved to the

appropriate PTC site.

Best regards,

Julio

[Fred_Kohlhepp]

How does this compare/correlate with Peirce's criteria?

[JCBanks]

Dear Fred, thank you for sharing the data outlier Peirce criterion.

I would like to share a "File naming method" I utilize which makes it

easier to locate files in my PC.. Any file converted into PDF format

I include the name of the software as the first word to easily locate

the files by "Application name". An example I propose with this method

is epitomize by the renaming of your file in my PC as follows::

Mathcad - Peirce Outlier Rejection Criterion.mcdx

This file, when converted to Adobe appears as follows:

Mathcad - Peirce Outlier Rejection Criterion.pdf

You can see that during a search in our PCs, the first word

appearing in a file list is "The name of the software".

Regarding your recommendation to consider the Peirce Outlier

Rejection Criterion,  henceforth, "the Peirce method". it is my intend

to consider my ABC Outlier Rejection Criterion as part of "a class

of outlier rejection criterion".

Essentially, the Non-Peirce criteria I mentioned in this response

are represented by LS (Least Squares) fit. The AEDC (Arnold

Engineering Development Center) is a -- Outlier rejection

criterion

The two (2) outlier criterion I utilized are "simple functions, C(n)"

where n is the "sample size" which for temperature data acquisition

is "Number of Thermocouple sensors" and for torque in a flange

assembly is "Number of bolts".

Fred, he Peirce method, does not provide a better estimation of

the effective three-sigma,. 3s, constant, C3S(n), compared to the

simple functions, C(n).

One of the rules of numerical methods is "Programming effectiveness"

which means "Use the simplest expressions in creating algorithms

especially data intensive processing such as in video animation"

for instance, addition and subtraction requires the least CPU

computational effort than multiplication and division. Another 

example is the Horner's rule which in mathematics, the term

Horner's rule refers to a polynomial evaluation method named:

after William George Horner. simply stated, the rule transform a

nth-order polynomial, Pn(x, order) into a series of nested  binomials

of the form (ai + xi) where i stands for index subscript.

Fred, hank you for your contributions to my understanding of outlier

rejection method.

Julio

[JCBanks]

I need to replace the file named: due to misspelled name:

"Mathcad - ABC (AEDC-Banks-Chauvenet) Outlier Removal Criterion"

to correct the misspelling of the name Chauvenet instead of Chauvenet. the 

Please find attached the corrected file.

Thanks,

JCBanks