Variability in Fatigue Crack Growth Rate Testing

TLDR

The study aimed to generate data for a recommended practice of fatigue crack‑growth‑rate testing by conducting an extensive interlaboratory program to assess variability and bias. Data were collected from 15 laboratories on 10 Ni‑8 Co‑1 Mo steel specimens with various geometries, and the results were statistically and graphically analyzed to quantify experimental and analytical variability and bias. The study found interlaboratory variability of about 3:1 and intralaboratory variability of about 2:1, with no significant geometry or data‑processing bias, indicating that the experimental procedure for obtaining raw crack‑length versus cycle data is the main source of variability.

Abstract

Abstract To provide the data necessary to develop a recommended practice for fatigue crack growth rate testing, an extensive interlaboratory (round robin) program was conducted and the variability and bias associated with the current state of the art of fatigue crack growth rate testing was determined. Fatigue crack growth rate data (expressed in terms of linear elastic fracture mechanics parameters) were generated for a 190-ksi (1310-MPa) yield strength 10 Ni-8Co-1Mo steel at 15 different laboratories with several test specimen geometries. The results were evaluated statistically and on the basis of a graphical comparison, and the variability and bias associated with both the experimental and analytical aspects of crack growth rate testing were determined. In general, the overall interlaboratory variability was found to be approximately 3 to 1 (on crack growth rate at a given stress intensity range). The intralaboratory variability was typically 2 to 1. No significant geometry or data processing bias was encountered. The results of this study show that the primary source of variability associated with fatigue crack growth rate testing is the experimental procedure used to obtain the raw test data (crack length versus elapsed cycles).