Lab #2 Loss Function
Objective:
To better understand how to calculate loss with an actual data set. How to interpret the results using the Taguchi Loss Function and comparing these to traditional methods.
Overview:
Devise a process to cut a series of one-inch length parts from a three-foot long wooden dowel. Use the formulas below in your calculations. The failure cost for each peg is 20 cents. That is, when a peg falls outside of the specification limits it must be discarded or reworked, which costs 20 cents. The specification limit for the peg is 1.000 +/- .005 inches.
To calculate Taguchi Loss Function:
Loss at a point (x) is L(x) = k
(x-m) 2
Loss of the sample set is L = k (s2 + (pm – m)2)
Where:
|
k = the loss coefficient (a scaling factor) | |
|
x = a measured value (i.e. 0.992²) | |
|
m= the target value (i.e. 1.000²) | |
|
s= the standard deviation of the sample s = [S (x-pm)2]/(n-1) | |
|
pm= process mean (average length) |
Equipment:
|
Calipers | |
|
Wooden dowel, 3 foot long | |
|
Safety glasses | |
|
A process to cut the parts |
Procedure:
Perform the following tasks in any shop available: using the band saw, the cut-off saw, any hand tool, a pocketknife, or plastic spoon. You design the process!
|
With the wooden dowel, measure a one-inch piece. | |
|
Cut the one inch piece and number it. | |
|
Repeat the operation 32 more times. | |
|
Be sure to use the same process for all the parts being measured and cut!! |
Results:
|
After all parts are cut, measure each part length to .001² accuracy. | |
|
Record below the measurements of 11 samples, each with a sample size (n) of 3. | |
|
Calculate the mean and standard deviation of the sample. | |
|
Calculate the loss for each manufactured peg (loss at a point) | |
|
Sketch traditional and Taguchi loss curves using all data point losses. | |
|
Calculate the overall loss of the sample set (traditional and Taguchi). |
Sample
Number
|
N size |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
10 |
11 |
|
1 |
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
|
|
|
|
|
|
|
|
|
|
Discussion:
|
How can loss function data be used to improve your process? | |
|
Why are there losses attributed to a part that is within specification limits? | |
|
Compare the costs of traditional and Taguchi loss functions? | |
|
If a part is in spec, where is the lost money coming from? | |
|
What do your calculations tell you in regards to your process and your parts? |