Description
Nowadays electrofusion (EF) is widely used on the gas distribution networks in order to join PE components. Nevertheless the field feedback points out some divergences –clearly apart from poor field practice- whose influence on the long term performance of the weld could be questioning. One of the well-known consequences of these divergences is the “sticking” (aka “cold weld” or “kissing bond”) that is the result of an uncompleted or even inexistent interdiffusion of the macromolecules of the materials to join.
Polyethylene is one of the emerging materials that are explosively spreading into many industrial pipes. Regardless of the short history, thousands of kilometers of polyethylene pipes are newly installed every year.
Pipe saddles assemblies
The figure 1 illustrates two typical images obtained with two of the pipe/saddle specimens prepared in the framework of the design of experiment and as displayed by the ND system used.
fig1 Typical images of a good weld (left) and of a poor weld (right) obtained with pipe/saddle assemblies
As can be shown the wires of the heating coil are very reflective and then can be individualized easily in terms of position and dimensions on both displays.
As was said above, the good weld is characterized by a wire plan and a back wall well defined. At the contrary with the poor weld as illustrated here the non welded interface reveals as a more or less continuous line under the wires plan and no more back wall reflection is visible.
Nevertheless these images must be considered as ideal ones because many of the other signals obtained for both good and bad welds are sometimes far less readable.
In this project, an image processing algorithm in MATLAB is implemented to detect good and bad joint in NDT.
Two set of images are prepared : Joint , Disjoint.
joint
disjoint
wavelet and Neural network is used in this project.
sample of region of interest segmentation :
Kirill –
This project helped me a lot.