Uncertainty Estimate of Charpy Data Using a 5-Factor 8-Run Design of Experiments

Abstract

Scatter in laboratory data with duplicates on Charpy impact tests is analyzed by identifying several sources of variability such as temperature, manganese sulfide, initial strain, mis-orientation, and notch radius in order to estimate the predictive 95% confidence intervals of the mean energy of absorption for each specific test temperature. Using a combination of real and virtual data on a high-strength pressure vessel grade steel (ASTM A517) over a range of temperatures from −40 °C (−40 °F) to 182 °C (360 °F), and the concept of a statistical design of experiments, we present an uncertainty estimation methodology using a public-domain statistical analysis software named DATAPLOT. A numerical example for estimating the mean, standard deviation, and predictive intervals of the Charpy energy at 48.9 °C (120 °F) is included. To illustrate the application potential of this methodology, we enhance it with formulas of error propagation to estimate the mean, standard deviation, and predictive intervals of the associated static crack initiation toughness, KIc. A discussion of the significance and limitations of the proposed methodology, and a concluding remark are given at the end of this paper.

Design of a PYTHON-Based Plug-In for Benchmarking Probabilistic Fracture Mechanics Computer Codes With Failure Event Data

“…In a local DB-3 reported by Interrante and Hicho [67] and benchmarked by Interrante, Fong, Filliben, and Heckert [68], a strong variation of the Charpy V-notch energy with test temperature for an ASTM A517 Grade H steel plate is shown in Fig. 14. That in turn caused a strong variation of the estimated static crack initiation toughness value as shown in Fig.…”

“…(local), engineers need to know how to interpret uncertainty e 3 of a global DB-3 in the context of a "usable" uncertainty of a local one, when such input is needed for predictive fatigue life models such as PC-PRAISE. To illustrate this point, we offer three examples from recent papers [33,68,69] and a new tool in the third of this 4-paper series by Fong and Marcal [52] as detailed below: 4.1. In a global DB-3 reported by Gerberich and Moody [70] and discussed by Fong, Ranson, Vachon, and Marcal [33], a strong variation of the crack growth law exponent m with test temperature for iron and various steels at R-ratios near zero is shown in Figure 13.…”

“…Similarly, a 10% change in X3 ( K threshold) varies its value from 8.19 at its center point to a low of 7.37 and a high of 9.10. From our past studies [68,69] and the literature (see, e.g. [71,72] and Figs.…”