A specific clutch facing part number carried an unusually high customer specification for burst resistance. Meeting it consistently, rather than on average, meant improving the reliability of the characteristic itself — and that meant understanding what in the manufacturing process was driving it.
❗ The problem
Burst resistance is an outcome of many process variables acting together. Looking at them one at a time tends to hide the relationships, and treating the process data before analysis risks removing exactly the structure you are trying to find.
🔬 The approach
The decision was to use multivariable analysis tools — tree clustering diagram analysis and principal component analysis — applied to data taken directly from the process, without special treatment. Exploring the raw process data is what allows relationships to surface on their own rather than being assumed in advance.
Eight process variables were investigated.
📊 What came out of it
Two of the eight variables showed a correlation with burst resistance. The interesting detail was the direction: one variable correlated positively with the burst test results and the other negatively. That opposition was hidden inside the process and would have been very difficult to see in a conventional correlation study, since the two effects partially mask each other when variables are examined in isolation.
Knowing the sign of each relationship is what made the information usable for decisions. After changes were made to the process on that basis, burst resistance increased — and so did reliability, which was the actual customer requirement.
💡 Why it matters
This is multivariate analysis used as an engineering instrument rather than a reporting technique. The result was not a better description of the process but a specific, directional change to it, validated by the characteristic the customer was measuring.
📄 Full publication
Gregori, I. R. S.; Damatto, R.; Haertel, W. — Multivariate Exploratory Techniques Helping to Increase the Reliability of Facing Clutch Material Burst Resistance. SAE Technical Papers, April 2008.
Read the full publication on ResearchGate
📚 Related Articles
- Simulating Clutch Facing Burst Resistance with Finite Element Analysis - The same burst-resistance problem approached by simulation instead of process data — a natural complement to the statistical route.
- A Gold Standard for Torque: Separating Clutch Facings by Judder Behaviour - Multivariate statistics applied to clutch facing bench data again, this time to separate facings by judder behaviour.
- Sub-Scale Judder Testing: Correlating a Bench Machine with the Vehicle - Another clutch-facing characteristic moved onto a bench test, with the same concern for correlating lab results to real performance.
Comments (0)
Leave a Comment
Be the first to comment!