Search

This contribution provides an example of the possible advantages of adopting a Bayesian inversion approach to uncertainty quantification in nondestructive inspection methods. In such problem, the uncertainty associated to the random parameters is not always known and needs to be characterised from scattering signal measurements. The uncertainties may then correctly propagated in order to determine a reliable probability of detection curve. To this end, we establish a general Bayesian framework based on a non-parametric maximum likelihood function formulation and some priors from expert knowledge. However, the presented inverse problem is time-consuming and computationally intensive. To cope with this difficulty, we replace the real model by a surrogate one in order to speed-up the model evaluation and to make the problem to be computationally feasible for implementation. The least squares support vector regression is adopted as metamodelling technique due to its robustness to deal with nonlinear problems. We illustrate the usefulness of this methodology through the control of tube with enclosed defect using ultrasonic inspection method. [ABSTRACT FROM AUTHOR]

This study addresses two issues: i) the estimation of simulation-assisted POD, not only versus the size of the defect, but also versus the location of the defect; ii) the non-parametric estimation of simulation-assisted POD when classical Berens hypotheses are not verified. It illustrates on an EC basic case of study the advantages of considering both size and location as characteristic parameters for POD estimation when location has a strong influence. Gaussian kernel smoothing regression has been applied to a simulated data set. Due to the large number of simulations involved, a surrogate model has first been constructed from a signals database generated by CIVA. [ABSTRACT FROM AUTHOR]

POD curves estimations are based on statistical studies of empirical data which are obtained thru costly and time consuming experimental campaigns. Currently, cost reduction of POD trials is a major issue. A proposed solution is to replace some of the experimental data required to determine the POD with model based results. Following this idea, the concept of Model Assisted POD (MAPOD) has been introduced first in the US in 2004 through the constitution of the MAPOD working group. One approach to Model Assisted POD is based on a transfer function which uses empirical data and models to transfer POD measured for one specific application to another related application. The objective of this paper is to show how numerical simulations could help to determine such transfer functions. A practical implementation of the approach to a high frequency eddy current inspection for fatigue cracks is presented. Empirical data is available for the titanium alloy plates. A model based transfer function is used to assess a POD curve for the inspection of aluminum components. [ABSTRACT FROM AUTHOR]

The concept of Probability of Detection (POD) is generally used to quantitatively assess performances and reliability of NDT operations for in-service operations related to damage tolerant designs. Application of the POD approach as a metric for manufacturing NDT assessment would also be relevant but the very expensive cost of such campaigns generally prevents us from doing so. However the increase in NDT simulation capability and maturity opens the field for POD demonstrations for manufacturing NDT with the help of simulation. This paper presents the example of an automated phased array ultrasonic testing procedure of Electron Beam Welding on rotative parts, as part of the PICASSO European project. POD is calculated by using the uncertainty propagation approach in CIVA. The peculiarity of uncertainties in automated NDT compared to in-service manual operations is discussed and raises questions on appropriate statistics to be used for this kind of data. Alternative estimation techniques like Box-Cox transform or quantile regression are proposed and evaluated. [ABSTRACT FROM AUTHOR]

Structure reliability guaranty requires prior evaluation of non destructive testing methods. The concept of Probability of Detection (POD) is generally used to quantitatively assess performances and reliability of testing operations. Such probabilistic approaches take into account the uncertainties that appear during inspections and that are responsible for the output variability. POD curve determination is based on costly and time consuming experimental campaigns. Increasing demand of NDT configurations requiring POD evaluation makes cost reduction of POD campaigns a major issue. A new trend is to apply simulation in the context of probabilistic approaches in order to replace some of the experimental data required to determine the POD with simulation results. This paper presents results of simulation based POD curves of a high frequency eddy current inspection procedure obtained with the new POD module of CIVA. The methodology used for describing uncertainties on the input simulation parameters is described and comparisons with experimental results are presented and discussed. [ABSTRACT FROM AUTHOR]

Structural noise and attenuation can cause significant losses in detection performances in the ultrasonic non destructive testing of diffusive metals, such as some austenitic steels. Being able to predict these phenomena can help in designing better testing procedures. In this purpose, a computational method has been developed to take into account both structural noise and attenuation in the simulation of an ultrasonic testing. The method makes use of a single scattering model based on the Born approximation, accounting for anisotropic scattering and for mode conversions. It assumes that scattering is due to the variations in the orientations of anisotropic crystals. Attenuation effects are accounted for by convoluting a computed ultrasonic field with a filter issued from a scattering model. The backscattered noise is predicted by computing the response of a set of point-like scatterers randomly distributed in the inspected volume. The parameters of the statistical distribution of the scatterers are determined from the characteristics of the metallurgical structure, using the above scattering model. This ensures that it generates a fully developed speckle equivalent to the speckle caused by a polycrystalline medium. In this communication, the computational method is described and comparisons between its predictions and experimental results are shown. [ABSTRACT FROM AUTHOR]

Calculation of Probability of Detection (POD) curves based on simulation results are presented in the case of the Eddy Current Testing (ECT) configurations involving a flaw buried inside a riveted structure. In order to overcome some accuracy limitations and to optimize calculation speed, the semi-analytical model used in simulation has been improved in order to handle separate discretizations of objects. New possibilities consecutive to this development are presented, as well as two parametric models used for the computation of interactive POD curves with the CIVA software. [ABSTRACT FROM AUTHOR]