Best Practice Guideline for Statistical Analyses of Fatigue Results

Description

This book provides a comprehensive guidance for the use of sound statistical methods and for the evaluation of fatigue data of welded components and structures obtained under constant amplitude loading and used to produce S-N curves. Recommendations for analyzing fatigue data are available, although they do not deal with all the statistical treatments that may be required to utilize fatigue test results, and none of them offers specific guidelines for analyzing fatigue data obtained from tests on welded specimens. For an easy use, working sheets are provided to assist in the proper statistical assessment of experimental fatigue data concerning practical problems giving the procedure and a numerical application as illustration.  Preface Contents 1 Best Practice on Statistical Analysis of Fatigue Data 1.1 Introduction 1.2 Assumptions 1.2.1 Form of S–N Curve 1.2.2 Tests for Linearity of Relationship Between log S and log N 1.2.3 Tests that N is Log-Normally Distributed 1.2.4 Tests of Homogeneity of Standard Deviation of log N with Respect to S 1.2.5 Tests of Statistical Independence of Test Results 1.3 Fitting a S–N Curve 1.4 Treatment of Results Where No Failure Has Occurred 1.4.1 Introduction 1.4.2 Results Associated with the Fatigue Limit 1.4.3 Fatigue Tests Stopped Before Failure 1.5 Establishing a Design (or Characteristic) Curve 1.5.1 Prediction Limits 1.5.2 Tolerance Limits 1.5.3 Results Where No Failure Has Occurred 1.6 Predicting Fatigue Life 1.6.1 Individual Weld 1.6.2 Structure Containing Many Welds 1.7 Justifying the Use of a Given Design Curve from a New Data Set 1.7.1 Problem 1.7.2 Approach 1.7.3 Assumptions 1.7.4 Method 1.7.5 Practical Applications 1.8 Testing Whether Two Data Sets Are Consistent 1.8.1 Problem 1.8.2 Approach 1.8.3 Tests Performed at the Same Stress Level 1.8.4 Tests Performed to Produce an S–N Curve 1.8.5 Composite Hypotheses 1.9 Testing Whether More than Two Data Sets Are Consistent 1.9.1 Problem 1.9.2 Approach 1.9.3 Tests Performed at the Same Stress Level 1.9.4 Tests Performed to Produce an S–N Curve 1.10 Sensitivity of Design Curve to Sample Size Appendix: Statistical Analysis of Fatigue Data Obtained from Specimens Containing Many Welds References 2 Working Sheets 2.1 Introduction 2.2 Sheet 1: Can Two Data Sets Be Merged? 2.3 Sheet 2: Are the Variances of Two Data Sets Statistically Equivalent? 2.4 Sheet 3: Are the Means of Two Data Sets Statistically Equivalent? 2.5 Sheet 4: Are the Data Normal Distributed Using Henry Graph? 2.6 Sheet 5: Are the Data Normal Distributed Using a Likelihood Test? 2.7 Sheet 6: Does a Standard S–N Curve Fit with a Data Set? 2.8 Sheet 7: Are Data Weibull Distributed Using a Weibull Probability Graph? 2.9 Sheet 8: How Many Results Are Necessary to Validate a Selected S–N Curve? 2.10 Sheet 9: How to Determine a Design Basquin S–N Curve Slope Fixed (Prediction Limits)? 2.11 Sheet 10: How to Determine a Design Basquin S–N Curve Slope Estimated (Prediction Limits)? 2.12 Sheet 11: How to Determine a Design Basquin S–N Curve Slope Fixed (Tolerance Limits)? 2.13 Sheet 12: How to Determine a Design Basquin S–N Curve Slope Estimated (Tolerance Limits)? 2.14 Sheet 13: How to Determine a Mean S–N Curve Bastenaire Equation? 2.15 Sheet 14: Are Two Experimental Design S–N Curves Statistically Equivalent? 2.16 Sheet 15: How to Determine the Degree of Improvement Produced by a Post-weld Treatment Process? Appendix 1: One-Sided Tolerance Limit Factors k Slope m Fixed Appendix 2: One-Sided Tolerance Limit Factors k Slope m Estimated References Glossary