Using Q-YIELD™: A Case Study

In this case study, we assume that you are responsible for a semiconductor manufacturing process.

Recently you have been observing serious fluctuations in the number of device failures.

As a first step, you plot a graph of failures per wafer against time:

(Click image to enlarge)

There doesn't appear to be any pattern to this data. In particular, there aren't any obvious correlations with events such as equipment changes or preventative maintenance. Now you need to look harder.

From your production database you download all the process parameters of the batches in question. In this (simplified) example there are 61 process parameters and 367 process records.

You consider creating scatter plots of the number of failures against each of the 61 variables.
But perhaps the relationship is of more than one variable.