Rapid communicationOn comparing the shape parameters of two Weibull distributions
Highlights
► Estimating Weibull modulus by using linear regression and unbiased probability estimators. ► Hypothesis testing on two estimated Weibull moduli. ► Percentile tables are generated for this hypothesis test.
Introduction
Weibull statistics are widely used to model the variability in the fracture properties of ceramics and metals, where the concept of weakest link applies. For the two-parameter Weibull distribution, the cumulative probability, P, that a part will fracture at a given stress, σ, or below can be predicted as [1];where σ0 is the scale parameter and m is the Weibull modulus, alternatively referred to as the shape parameter. The Weibull distribution has been successfully applied to materials in which fracture properties are determined overridingly by the defect distributions. For instance, Green and Campbell [2] showed that the tensile strength (ST) of A356 castings alloys follows a 2-parameter Weibull distribution and that the filling system design has a strong effect on the Weibull modulus. Since the results of Green and Campbell, the 2-parameter Weibull distribution has been used extensively in the casting literature to characterize fracture-related mechanical properties such as tensile strength (ST) [3], elongation [4] and fatigue life (Nf) [5], [6], [7]. According to Campbell [8], Weibull modulus, m, is often between 1 and 10 for pressure die castings, and between 10 and 30 for many gravity-filled castings. For good quality aerospace castings, m is expected to be between 50 and 100. Hence Weibull modulus has been used as one of the measures of the casting quality and reliability.
There are several methods available in the literature to estimate the Weibull parameters such as linear regression, maximum likelihood method and method of moments. Among these methods, the most popular method is linear regression mainly because of its simplicity. After taking the logarithm of Eq. (1) twice yields:Note that Eq. (2) has a linear form when the left-hand side of the equation is plotted versus ln(σ) with a slope of m and an intercept of −m ln(σ0). Hence, the estimates of σ0 and m, and , respectively, are determined from the slope and intercept in Eq. (2).
Processing engineers working on improving the quality of material need tools to measure the degree of effectiveness of their efforts. When the Weibull modulus is used as one of the measures of quality and reliability, the effect of process changes can then be evaluated by comparing the Weibull moduli of samples taken prior to and after the implementation of changes. Alternatively, Weibull moduli from two different processes and/or supplier may need to be compared. Therefore there is a need to develop a method to compare Weibull moduli from two separate samples. This study is motivated by this need.
Section snippets
Research methodology
Monte Carlo simulations were used to generate two samples from a Weibull distribution with σ0 = 1 and m = 3. The probability estimators (plotting position formulas) that yield unbiased estimates of the Weibull parameters reported by the authors [9] were used to assign each data point a probability. These estimators have the form,where i is the rank of the data point in the sample in ascending order, n represents the sample size, and a and b are numbers, such that 0 ≤ a ≤ 0.5 and 0 ≤ b ≤ 1.0. The
Results and discussion
Recently, the authors analyzed the distribution of normalized and and found that (i) the distribution of normalized is normal [9], and (ii) the distribution of normalized is neither normal, lognormal, 3-parameter Weibull nor 3-parameter log-Weibull [10]. Therefore it was expected that the distribution of would not follow these distribution. Analysis of several combinations of sample sizes confirmed this expectation. As a result, it is necessary to generate tables for the
Conclusions
The distributions of the ratio of Weibull moduli estimated from two samples in this paper are neither normal, lognormal, 3-parameter Weibull nor 3-parameter log-Weibull.
Therefore percentiles of the distribution of the ratio of two Weibull moduli have been developed.
The use of these percentiles for comparing two Weibull moduli has been demonstrated by using datasets from the literature.
References (11)
- et al.
Mater. Sci. Eng. A
(1993) - et al.
Mater. Sci. Eng. A
(2005) - et al.
J. Light Metals
(2001) J. Appl. Mech.
(1951)- et al.
Cited by (19)
Study on probability model and length effect of tensile strength of unidirectional CFRP plate
2023, Composite StructuresWeibull statistical analysis and experimental investigation of size effects on tensile behavior of dry unidirectional carbon fiber sheets
2020, Polymer TestingCitation Excerpt :This result proves that the Weibull modulus is not a constant that depends on the material type. This remark confirms the results of previous research [33–35]. In addition, the analysis of the results presented in Table 6 demonstrate that the characteristic strength value of DUCFS specimen decreases when increasing the length.
Improvement of mechanical properties of Y-TZP by thermal annealing with monoclinic zirconia nanoparticle coating
2019, Dental MaterialsCitation Excerpt :In order to statistically evaluate the improvement of reliability after thermal annealing, the Weibull modulus ratios, compared to the Group SB, were respectively calculated (mSB-TA/mSB = 2.593 and mSB-mZr-TA/mSB = 2.745). These ratios were respectively above 97.5% confidence level (i.e., mx/my > 2.069 at N = 15 [27]), indicating a statistically significant improvement in the reliability of Groups SB-TA and SB-mZr-TA, compared with Group SB. In this study, the Group AS was used as a control group to evaluate the mechanical strength increment [12] by sandblasting (Group SB), which was also used as another control group to evaluate the mechanical strength decrement and the reliability increment [16,17] after thermal annealing (Groups SB-TA and SB-mZr-TA).
A new estimation approach based on moments for estimating Weibull parameters in wind power applications
2018, Energy Conversion and ManagementCitation Excerpt :The Weibull distribution (WD) has been commonly-used to characterize different types of data such as wind speed data, reliability data, survival data and ceramics, glasses and solid catalysts data [1–3].