Probability of Detection (PoD) Curves Based on Weibull Statistics

Probability of detection (PoD) curves are a popular metric for the reliability assessment of Nondestructive Testing (NDT) procedures. However, the classical Berens method for signal response PoD analysis strongly relies on the hypothesis of Gaussian residuals which can be violated in practical conditions. In particular, data from sparse field trials can be scattered and or skewed. Hence, this paper studies the feasibility of assuming a Weibull distribution, which is known for versatility in representing several fundamental statistical states, for regression residuals without modifying the overall Berens framework for PoD curve determination. The proposed ‘Weibull-Berens’ PoD statistics is first shown to compare well with the classical Berens method for an ideal case of Gaussian residuals. The advantages of the method are further demonstrated using a synthesised dataset, as well as a practical case of non-Gaussian residuals arising from reduced number of experimental trials.

[1]  Scott R. Eliason.  Maximum likelihood estimation: Logic and practice. , 1994 .

[2]  Christina Müller,et al.  Investigating human factors in manual ultrasonic testing: testing the human factors model , 2011 .

[3]  S. F. Burch,et al.  Human factors in POD modelling and use of trial data , 2009 .

[4]  P. J. Green,et al.  Probability and Statistical Inference , 1978 .

[5]  Solomon Kullback,et al.  Information Theory and Statistics , 1960 .

[6]  D. Cox,et al.  An Analysis of Transformations , 1964 .

[7]  Mato Pavlovic,et al.  Probability of detection as a function of multiple influencing parameters , 2012 .

[8]  Anil K. Bera,et al.  A test for normality of observations and regression residuals , 1987 .

[9]  M. E. J. Newman,et al.  Power laws, Pareto distributions and Zipf's law , 2005 .

[10]  Nicolas Bousquet,et al.  Nonparametric Estimation of the Probability of Detection of Flaws in an Industrial Component, from Destructive and Nondestructive Testing Data, Using Approximate Bayesian Computation. , 2015, Risk analysis : an official publication of the Society for Risk Analysis.

[11]  Bertrand Iooss,et al.  Model Assisted Probability of Detection Curves: New Statistical Tools and Progressive Methodology , 2016, Journal of Nondestructive Evaluation.

[12]  Satyanarayana Gaddada,et al.  Statistical scrutiny of Weibull parameters for wind energy potential appraisal in the area of northern Ethiopia , 2015 .

[13]  Weixi Chen,et al.  Demonstration of probability of detection taking consideration of both the length and the depth of a flaw explicitly , 2016 .

[14]  Daniel Kanzler,et al.  Probability of detection for surface-breaking holes with low-frequency eddy current testing – a non-linear multi-parametric approach , 2014 .

[15]  M. Lowe,et al.  Scattering of the fundamental shear horizontal guided wave by a part-thickness crack in an isotropic plate. , 2008, The Journal of the Acoustical Society of America.

[16]  J. V. Bradley.  Distribution-Free Statistical Tests , 1968 .

[17]  Jeffrey C. Lagarias,et al.  Convergence Properties of the Nelder-Mead Simplex Method in Low Dimensions , 1998, SIAM J. Optim..

[18]  G. W. Snedecor.  Statistical Methods , 1964 .

[19]  Krishnan Balasubramaniam,et al.  Bayesian synthesis for simulation‐based generation of probability of detection (PoD) curves , 2018, Ultrasonics.

[20]  F. Jenson,et al.  Simulation assisted pod of a phased array ultrasonic inspection in manufacturing , 2012 .