In this age of communication, organized thoughts are a rare commodity. This is a case in point.
A well recognized manufacturer had rash of warranty repairs that were occurring in a cyclical fashion. They knew exactly which part failed and what features of that part were not machined properly. However, they were not able to determine the root cause of the sub-specification machining.
Upper management had even larger problems caused by this component since a failure in the field typically resulted in the permanent loss of a customer.
The engineers easily identified a specific process and the associated central coolant system.
The engineers asked their new chemical manager for the concentration and other data relating to the metalworking fluid. The data from the previous chemical managers looked like this, but went on for pages covering years of sample collection and lab results.
Unfortunately, it was just a list of dates and numbers. No analysis with trending had ever taken place. When asked to do some analysis, they produced a graph of the data.
The interpretation was that the tramp oil is less than 1.5% (low), the concentration is usually 9% (high) and bacterial activity varied with no relationship to the tool costs. Tool cost data is kept in a separate database and not readily available.
When we added arithmetic weight to the coolant data, the direct relationship between tool costs (scrap rates), bacteria and total tramp oil became apparent.
The points in time when the slope of these trendlines reverse are the keys to determining the root cause of the warranty problem. The root cause of the problem in this case is the uptime of the tramp oil extraction equipment.
Compounding the difficulty in this data analysis were issues with the historical concentration test methodology, inconsistent sampling techniques, and accuracy of the data entry. Those are stories for another time.
Gary Rodak is the Owner of Machining Efficiencies, Inc. You can reach him at Gary.Rodak@MachiningEfficiencies.com.
Other articles from this issue: