The field of process control has changed radically over the last 20 years. Back in the day, it was not uncommon to view process control as simply the exercise of reading a time trend line to see if a defect rates were rising or falling based on outlier activity around the classic Shewhart control chart. While this level of analytics is a good starting point, it by no means completes the picture.
SPSS Text Mining with unstructured data helps complete that picture. If an operator of a machine or a production line performs routine maintenance on that entity, he or she will routinely collect information / comments in the form of inspection logs. In turn those inspection logs might have a common word theme that predates an imminent part failure or machine outage. Furthermore, groups of they key words can be used for characterizing and prioritizing defects by type (ie the old Pareto analysis idea that 80% of your defects are clustered within 20% of your defect types).
The enclosed Flowchart gives some indication of how this text mining might work for an air conditioner manufacturing firm with noticeable defects that are logged at the inspection phase:
This is an improvement on the old Quality Control paradigm of merely tracking trendline and get us closer to the concept of root cause analytics for further corrective action.
For more information on this concept contact
Tony Firmani
Director, Advanced Analytics
Tony.Firmani@Perficient.com
