Learning Curve and Forgetting Factor

Presentation Transcript

Laboratory Exercise 1

IE 3269 – POM

R. R. Lindeke, Ph.D.

With proper design, the learning effect can be driven together to balance out the grow phase of the product lifecycle!
Learning is usually expressed as a % -- the lower the % the faster is the learning!
Learning represents the rate of improvement, if it is 20% each time the quantity is doubled, then the learning percent would be 80% (100-20=80).
While the learning curve emphasizes time, it can be easily extended to costs as well

The learning behavior of systems and individuals is consider to follow this model:
b is a decimal value (<1) that relates the improvement seen as repetitions increase

y = axb

y - time to assemble the xth unit

a - time to assemble the 1st unit

b - measure of "Learning Rate"

x - unit number

When building a Learning Curve, we should perform a linearization operation and find a “least squares” regression fit to our experimental data set
To linearize our data set, compute the Log (base 10 or base e) of observed times and unit number of repetition
The ‘intercept’ of the fitted line (log data plotted on a linear axis!) is the log(a) value for the linear fit (a as defined above)
The slope of the fitted line is the b value in the learning curve equation

This equation can be used as a spot check, real slope, and thus real learning rate, is computed by regression using all our observations!

Mathematically, we can see that if we are positively learning, the b value is a negative number (indicating a decimal value b<1)
Eventually, the time to assembly or perform any learnable task will approach an asymptotic value that is controlled by the time it takes the ideal worker to accomplish all the required steps – this is a number that can be computed using (micro) motion analysis and the various steps needed
Our learning curves will eventually fail to predict expected times for performing an activity (they will eventually underestimate actual times)

Like all human activities (consider your math or physics courses!!?!) if we stop doing something for awhile, we are ‘rusty’ when we start up again
Same ideas hold true for manufacturing!
This idea is called the forgetting factor (mathematically)

Forgetting Factor studies consider Production Rate (compared to production times as seen in learning curve)
The model of the forgetting factor computes the rate loss due to forgetting between batches
This value can be used (along with learning curve values) to project batch by batch production rates (and times)

This iterative computational method can be continued over any number of batches using a similar technique

This is one of the reasons for modern CAM to be profitable
Forgetting factors are less pronounced if the ways are remembered by design
Forgetting is less important if only similar are processed regardless of batch time and time between batches
Forgetting is affected greatly by the time away for a specific product