Image acquisition and processing in an attempt to automate the fluorescent penetrant inspection

摘要

Mechanical failure in aircraft workpieces is likely from happening, if open to surface discontinuities which comprise stress concentration regions are present. The fluorescent penetrant inspection (FPI) is a sensitive nondestructive evaluation (NDE) method capable to verify the presence of indications in surface of raw materials or of processed parts or parts submitted to service charges.ududIn general, when FPI inspection is being conducted, an inspector performs an evaluation of a treated surface based on his knowledge and experience in nondestructive testing, being the defect detection a qualitative decision according to his/her judgment. In this case, the inspector’s vision acuity, attitude and motivation can compromise the analysis in inspection. The automation in inspection may improve the repeatability of this NDE method, regarding that machine vision systems classify parts considering quantitative analysis of objective data. Algorithms for the feature extraction are run, providing features for a data analysis procedure which classifies the detected indications as relevant or irrelevant indications according to specification or code. Therefore, the machine vision systems, when applied to fluorescent penetrant inspection, improve the overall technique reliability, guaranteeing automatic storage, retrieval and feedback of data for controlling the manufacturing and maintenance of equipments.ududIn the literature, there are several examples of machine vision systems: For example, the laser scanning systems, as proposed in Tracy and Moore (2001), is a good approach for the inspection automation. In this system, a focused laser beam spot is translated over the specimen surface and a photodetector measures the amount of fluorescence in that illuminated area whose power is directly proportional to the quantity of penetrant trapped in surface cavities (Tracy and Moore (2001)). Other manner which was proposed by Armstrong (1986) is to illuminate a treated sample with a standard UV lamp in dark booth and acquire images formed by fluorescent bright indications using a camera.ududIn this work, an ultraviolet indication detection system was developed for the automation of the inspection stage in FPI. Only open to surface indications are capable to be detected through the system. It measures the area and maximum euclidean distance and classifies the shape of indications. They are feasible to be evaluated according to predefined quality standards. Metrics such as the probability of detection curves were traced, obtaining a capability of 21.6 microns of depth with 100% of reliability.