Abstract
ABSTRACT The exploration of offshore oil fields has reached water depths where manned maintenance systems cannot be used. Therefore, automated underwater production systems are required that can perform all necessary tasks. The ultrasonic technique is being developed for future tasks ofsensoring and nondestructive testing. It has the ability to support an automated underwater production process in several ways. One of them is surface layer recognition with pattern recognition methods to allow efficient adjustment of the water-jet cleaning system. The ultrasonic technique serves as a tool to recognize the surface of various organisms settling on the off-shore structure. A second important application is the automated nondestructive testing of welds. Fatigue cracks in the heat affected zone can seriously damage a construction when the y remain and grow undetected. These applications represent only two examples of the broad applicabilityy of pattern recognitionmethods. This paper describes the methods developed for recognition of surface profiles and for flaw detection in weldings. Applications of data sampling are included as wellas feature extraction and pattern recognition techniques, such as distance-classifiers, correlation classifiers and an artificial multi-layer perception neutral network. INTRODUCTION Fully automated inspection is becoming an important research field in quantitative nondestructive testing. The automated support of the operator should be done in every difficult inspection task where the reproducibility of the results, and especially the change of the test setup, will lead to different results. For automated inspection of offshore structures where the scanning task and the data evaluation are computer-controlled, pattern recognition provides the interpretation of the signals which are described by specific features, The traditional task of numerical classificationin ultrasonics is the recognition of signals from real or phantom weld defects. These are one-dimensional signals, but a two-dimensional application of the pattern recognition methods is almost identical to the one-dimensional. Only the important feature extraction by digital image processing is different. In any case the extracted features must give an appropriate description of the patterns. The better the features are, the better will be the recognition result. Examples of two-dimensional pattern recognition are theevaluation of B- or C-scans, e.g. the recognition of rail defects in B-scans or weld defect recognition in C-scans, which is similar to image recognition. The field of deep sea application of ultrasonic is growing, Oil resources found in water depths from 500-2000 m lead to exploration developments which are diverless and automatically controlled. The offshore area itself is predestined for ultrasonic applications because the coupling of the probe to the surface of the material to be tested is automatically realized by the sea water. The integration of the whole working process at greater depths, such as cleaning, cutting, welding, mounting and testing, is the aim of a research project supported by the German Research Foundation (DFG). The project is called "Automated Underwater Production". In Fig. 1 a scenario is presented which may occur during an automated testing task. A remotely operated vehicle approaches to an underwater structure and afterdocking the surface is investigated by ultrasound.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have