Preliminary considerations for the computer analysis of fringe mapsgenerated when holographic interferometry is applied as an NDT tool for airframestructures,

摘要

Abstract: The temptation when presented with the requirement to interpret holographic interferograms of aluminum aircraft structures derived through a non-destructive testing technique is to examine the whole interferogram. Computers are renowned for their ability to process large amounts of data accurately and speedily, therefore there is a strong temptation to harness their particular powers. This is not only time-consuming and wasteful of computing resources, it is also unnecessary. However, before work can begin on interpreting an interferogram by computer the problem itself must be defined. In this particular example the interferogram is from a sample taken from one of the world's aging passenger airline fleet. The sample is from an aluminum stabilizer in which faults have been induced or have occurred during the service lifetime of the aircraft. All faults have been confirmed by destructive evaluation or by alternative techniques. Thus the problem domain is known. When a human expert examines an interferogram s/he concentrates on areas where faults are likely to occur namely, the areas immediately surrounding the stringers and frames as well as the stringers and frames themselves. The faults are typically caused through endless pressurization cycles or through corrosion. These faults have been induced to show themselves by a distinctive pattern of interference fringes across stringers and frames, where normally no fringes should be expected to occur. Therefore the human expert search concentrates on these areas, using the fringe count density or shape over the whole of the interferogram simply for comparison or control. The computer aims to emulate the human search. However, difficulties have been identified that could prove problematic for the computer that are elementary for the human brain. In our early work the sample interferograms for computer analysis have been selected because, to a human, they are uncomplicated and relatively noise-free sample in which faults are easily identified. This gives a good test case against which the computer can be compared, however the strength of the computer may finally by in interpreting holographic interferograms that are difficult for humans to interpret either though complexity or human consideration such as fatigue when hundreds either though complexity or human consideration such as fatigue when hundreds of such images are required where, for example, a large structure such as an aircraft is considered. !2