الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 168 |  |  |
| | | from | 168 | | |
Abstract
Oil and gas transmission pipelines have a good safety record. This is due to a combination of good design, materials and operating practices. Therefore, only high quality pipe welds are okay in recent days, failure due to a pipe weld can cause a dangerous accident with possible loss of life and property. The inspection process is an important step in the quality assurance program for manufacturers and a variety of Non-Destructive Testing (NDT) methods were need for inspecting welding defects. It is essential that the manufacturer detect every fault, in order to avoid costly delays, ever increasing scrapping cost and customer dissatisfaction. As the cost of scrapping goes up, there is an increasing need for a cost effective advanced inspection system.
Radiographic inspection is a well-established testing and quality control method to detect weld defects. But the major challenging problem in the inspection process is the accurate (truthful) discovering of defects (or no defects) in a given radiography images, since the human factor still has a decisive influence on the evaluation of defects on the film.
In the way to reduce human factor (fault) and for the speedy improvement in productivity requires the improvement of new truthful and effective inspection systems. The research for the development of an automatic or semiautomatic system of analysis of radiographs of welded joints has grown considerably in the last years and especially in the last 10–20 years. As a result, automatic or semiautomatic systems depend upon integration among sciences and technologies, such as computer vision, programming languages facilities and neural network applications. The new system should emulate the skills, decision-making capacity and reduce both of cost and time for the process.
On the other hand, the progress in computational techniques, mainly the development of neural networks, has greatly stimulated the researches on the development of automatic systems for the inspection and the classification of defects. Neural networks consist of algorithms that learn how to mould some functions of the brain, although they are less complex than the human brain. The neural networks can process great amounts of data in a short period of time that typically could only be analyzed by one specialist. One of the most important characteristics of the artificial neural network is the training or learning by examples, exactly like the human brain.
This work introduced in this thesis depends upon computer vision system as a tool, image processing and neural network as a techniques and matlab as a programming language to have a prototype of automated visual system, which is capable of inspecting X-ray images of arc welding pipelines.
The vision system is used to capture an X-ray image for the object to be inspected and then the captured image is treated by the image processing and neural network techniques. The proposed system is an expert one and built (upon methodology) to do the duty by the two methods, the first is character vector method and the second is histogram method
Going over the previous and the recent researches in the domain of non-destructive testing especially automatic inspection system, little word about system components is introduced for pipelines and welding defects with presenting codes and standards (API-ASME-DIN-…etc.). Some image processing techniques were used to prepare input data (images), like resizing and gray level images then, dealing into artificial neural networks with its components.
The proposed system is an expert one and built (upon methodology) to do the duty by the two methods, the first is character vector method and the second is histogram method under three different images sizes. The hardware consists of a digital camera, X-ray images of welding samples from IIW-cards, backlighting box and suitable personal computer.
A special program was written using matlab 2011b, to open digital images, processing radiographic images, to build artificial neural network and to produce the executable program for inspection as a target of this work by using windows operating structure. Our system was developed to detect more than one type of defect.
Results were presented and discussed through tables and diagrams before and after improving the system upon radiographic images by two methods and for three groups of images sizes; and ended to say after improving the system, histogram method is better than character vector method under the work boundaries or limitations.
The name is declaring of being existence for that recommended name for this program is WDINProg (welding defect inspection using neural network program).This thesis consists of six chapters and it is organized as follows:
Chapter (1) Literature review
This chapter presents a literature review of recent and earlier researches with introduction about the importance for both of pipelines in oil and gas field and nondestructive test for inspecting welding defects. Subjects here were contemporary NDT methods, sounded NDT, normative concepts for NDT, and ended by the target from this research.
Chapter (2) Pipelines and welding defects
This chapter covers welding (process, codes and standards), pipe and tube, welding imperfections, detecting defects by X-ray (radiography), all that were aided by related researches
Chapter (3) Computer vision and neural networks
This chapter introduces the fundamentals of vision system and image processing as a technique followed by neural networks (phenomena and principals). It is built upon historical review and followed by details: image (capture, representation, storage), processing, and object detection, future of computer vision, neural network identification and finally neural network today.
Chapter (4) Experimental work and program
This chapter introduces the system used in the present study and how it can be used such as Work setup, technical specifications for hardware components, work plan by block diagrams, and the practical works were discussed for two main methods to create the data, with illustrating the WDINProg program.
Chapter (5) Results and discussions
This chapter presents the obtained results from the work and its improvements through output verification and illustration of the outputs by figures and tables. It presents work discussion which is serviced by tables and diagrams, and which offers a tool for inspecting X-ray images to detect defects.
Chapter (6) Conclusions and recommendations
This chapter introduces the conclusion of the presented work. It introduces points of utilities and is ended by the suggestions for some recommendation of the future work in the field of inspection of welded pipelines.