Use of Multiple Regression Analysis to Develop Predictive Models for Failure Times of Adhesive Bonds at Constant Stress. Part I. Titanium Adherends.
Abstract
Predictive equations were developed for failure times of adhesive-bonded titanium under a constant load. Two types of titanium, each with two different pretreatments, were studied. The independent variables in the equations were temperature, relative humidity, stress, and (initially) SURFACE PRETREATMENT. The predictive equations were developed using multiple regression analysis which was performed by computer. Many models with both theoretical and empirical bases were considered. Statistical methods were used to find the best model based on correlation coefficients and significance of variables. Correlation coefficients for the best models ranged from .79 to .94, indicating that the models represent the data with reasonable-to-good accuracy. A comparison of actual failure times with those predicted by the models gave at least an order-of-magnitude agreement. These equations should be quite useful in smoothing and interpolating existing data. (Author-PL)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1972
- Accession Number
- AD0754526
Entities
People
- David W. Levi
- Elise Mcabee
- Raymond F. Wegman
- William C. Jones Iii
Organizations
- Picatinny Arsenal