A WEB-BASED DATA ANALYSIS METHODOLOGY FOR ESTIMATING RELIABILITY OF WELD FLAW DETECTION, LOCATION, AND SIZING

摘要

Recent advances in computer technology, internet communication networks, and finite element modeling and analysis capability have made it feasible for engineers to accelerate the feedback loop between the field inspectors of a structure or component for critical flaws by nondestructive evaluation (NDE) and the office engineers who do the damage assessment and recommendations for field action to prevent failure. For example, field inspection data of critical flaws can be transmitted to the office instantly via the internet, and the office engineer with a computer database of equipment geometry, material properties, past loading/deformation histories, and potential future loadings, can process the NDE data as input to a damage assessment model to simulate the equipment performance under a variety of loading conditions until its failure. Results of such simulations can be combined with engineering judgment to produce a specific recommendation for field action, which can also be transmitted to the field by the internet. In this paper, we describe a web-based NDE data analysis methodology to estimate the reliability of weld flaw detection, location, and sizing by using a public-domain statistical data analysis software named DATAPLOT and a ten-step sensitivity analysis of NDE data from a two-level fractional factorial orthogonal experimental design. A numerical example using the 1968 ultrasonic examination data of weld seam in PVRC test block 25 U, and the 1984 sectioning data of 25 U containing 15 implanted flaws, is presented and discussed.