Scribd
Upload
5 views6 pages

Statistical vs Practical Significance in Bearings

The document discusses the distinction between statistical and practical significance, emphasizing that statistical significance does not always imply practical impact. It highlights the importance of understanding Type I errors and the need to assess effect size alongside statistical results. Additionally, it outlines assumptions related to Weibull distribution and failure mechanisms, noting that violations can lead to inaccurate life predictions and biased estimates.

Uploaded by

Ghaseeta Yadav
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
5 views6 pages

Statistical vs Practical Significance in Bearings

The document discusses the distinction between statistical and practical significance, emphasizing that statistical significance does not always imply practical impact. It highlights the importance of understanding Type I errors and the need to assess effect size alongside statistical results. Additionally, it outlines assumptions related to Weibull distribution and failure mechanisms, noting that violations can lead to inaccurate life predictions and biased estimates.

Uploaded by

Ghaseeta Yadav
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Chapter 5

Statistical vs practical significance


• Practical benefit doubtful: Even a 2% longer L₁₀ may be statistically
significant—but hardly impactful.
• Uncertainty: risk of a Type I error—concluding a difference exists
when it does not—with probability α.

Situation Statistically Significant? Practically Significant?


Very large sample reveals 1% ❌ Not unless meaningful operations
✅ Yes
difference gain
❌ Possibly not significantly different ✅ High practical value, but not
Small sample shows 20% difference
(wide CI) confident statistically
How to Interpret Statistical
Significance for Bearings?
• Define what "no difference" means (null/alternate hypothesis).
• Compute p-value or analyze CI overlap to test it.
• Understand α means risk of false positive (Type I error) at that level.
Usually set at 5% or stricter for critical applications.
• Always pair a significant result with an assessment of effect size and
practical benefit.
• Report both:e.g. “L₁₀ difference of +10% with p = 0.03, meets
statistical significance at α = 0.05. But system-level maintenance
savings are estimated at only 0.5%, so practical impact is marginal.”
Assumption
• Weibull distribution
• failure mechanisms deviate—like surface wear, misalignment,
corrosion, or contamination—the Weibull may not fit well
• Mixed Failure Modes : single dominant failure mode
• subsurface fatigue, surface wear, corrosion
• same steel batch, mounting, clearance, lubrication, contamination,
temperature.
• tests are censored.
• Proper handling
Assumptions:
• few failures are observed, estimates of β and η are unstable
• uncertainty propagates into wide confidence intervals for L₁₀ life
• ISO 281 (and related methods like modified life Lnm/Lnmh) rely on
assumptions:
• Linear damage accumulation (i.e., fatigue is sum of small stress cycles).
• Standard lubrication regime, viscosity, contamination factors, and
fatigue limit values
• complex environments like wind turbines where wear or evolving
contamination dominate wear mechanisms
Summary

Assumption or Limitation Impact if Violated


Weibull distribution appropriate Misfit reduces accuracy of life prediction
Single failure mode Mixed failures distort Weibull parameter estimations
Non-representative results may not apply to real
Representative data sample
service
Proper handling of censoring Biased life estimates if failures are truncated
Adequate sample size Weak confidence when β and η are unreliable
Damage model and life factor validity ISO 281 predictions can understate key damage types

You might also like