TWI Industrial Member Report Summary 1022/2012
By C Schneider, R Sanderson, C Carpentier, C Nageswaran and L Zhao
Background
Determining the capability of non-destructive testing (NDT) procedures to detect the flaws they were designed for is an important part of implementing sound inspection programmes. During inspections there are a range of factors that influence the signals detected by the inspection system due to the presence of flaws. It is important to determine to what extent these factors influence the ability of the inspection to successfully detect flaws in the component being inspected.
The concept of the Probability of Detection (POD) has been used in industry to establish the capability of the inspection, subject to influential factors, to detect flaws. This is often expressed as a POD curve, which relates the likelihood of detection (along the y-axis) to a characteristic parameter of the flaw, usually its size. Hence, in general, as the size of the flaw increases (along the x-axis) the likelihood of the inspection system detecting a flaw of that size increases also; that is, those factors that adversely influence the capability of detecting a flaw will, in general, decrease as the size of the flaw increases.
Estimation of the POD by NDT typically relies on the manufacture of large numbers of realistic defect specimens, followed by practical trials of the inspection procedure. These are costly and time-consuming activities. POD curves could be generated more rapidly and more cost-effectively if theoretical simulation of PODs (called either model-based or model-assisted PODs) were to be made acceptably accurate, consistent and repeatable. The ultimate aim is to use software making use of Monte Carlo simulation concepts to numerically generate a distribution of flaw responses from postulated distributions of so-called 'essential parameters', which include locations of flaws in the components, their orientations etc.
This report presents the work done to investigate the use of two models to generate POD curves for a case that was used in previous unpublished work by TWI. In the previous work, TWI in collaboration with Rolls-Royce and British Energy Generation Ltd (now EDF Energy) devised a Case Study which was based on EDF Energy's Capability Statement for manual ultrasonic testing (UT). The Case Study was used in that previous project to identify limitations to an earlier (prototype) POD model and to recommend improvements to overcome those limitations; the Case Study has been modified for the same purpose of evaluating the limitations of the two models used in this project.
The first of the two models investigated in this project is based on the EDF Energy code termed PEDGE. TWI embedded the PEDGE software within some new code that runs a number of cases sequentially to generate the POD from individual modelling results output by PEDGE. The new code was named PODPEDGE. The underlying PEDGE code in PODPEDGE was released for use by the inspection group in British Energy in the early 1990s after thorough verification of the underlying theories and validation against experimental evidence. In essence, the code is able to evaluate the ultrasonic response from a specified smooth planar flaw in a pulse-echo ultrasonic inspection configuration.
The second model investigated in this project was developed by the Commissariat à l'Énergie Atomique (CEA) of France and is well known in the NDT industry as CIVA. It is designed as a simulation platform consisting of numerous software modules designed to model a broad variety of inspection tasks using ultrasonic, eddy current and radiographic techniques. The validity of the model has been investigated by TWI in previous research projects (Schneider et al, 2005; Nageswaran et al, 2008) and there has been a significant body of work by other organisations (notably CEA itself) to establish the capabilities of CIVA.
This report describes the case studies simulated by PODPEDGE and CIVA. The results from PODPEDGE and CIVA are presented separately; the performance of the two models is firstly considered separately and then the relative characteristics of each modelling package are discussed. This project aimed to establish the feasibility of generating realistic POD curves theoretically and makes recommendations for improvement to both the POD models and the underlying concept of model-assisted PODs.
Objectives
- Demonstrate the feasibility of estimating realistic POD curves by theoretical simulation of the ultrasonic inspection process.
- Compare the performance of different software tools for estimating POD curves.
- Provide guidance on how to use the software tools effectively.