A key topic regarding TiNi alloys concerns the possibility to attain junctions that preserve the shape memory properties of material. Experimental tests, previously performed on TiNi sheet friction stir processed, have highlighted the need to develop an appropriate image analysis method to quantify the various phases percentages present in the characteristics zones of Friction Stir Welding process. A proper Image Processing procedure has been performed in order to quantify the amount of the martensitic phase and to detect its morphology modification along to the processed region. Particularly each micrographic image, firstly, has been denoised using the 2D Wavelet transform technique and, successively, a texture segmentation procedure has allowed to evaluate the amount of the martensite and austenite phases and to classify the morphological changes of martensitic regions. In this research, two different wavelet de-noising techniques have been tested in order to select the better way to preserve the highest content of microstructural information in the image. The optical micrographic captures were subjected to a denoising procedure, using wavelet transform at five level of decomposition, that allowed to obtain on each image only two uniform regions for the austenite and martensite phases. The obtained results, in terms of preserving of accuracy of the microstructuralcharacteristics, suggested the selection of stationary Haar wavelet transform method as de-noising method. The de-noised images were converted into grey scale images and enhanced by means of adaptive histogram equalization. Finally the texture segmentation procedures, followed by image binarization, allowed to distinguish and quantify the amount of each phase. The proposed image enhancement algorithm for optical microstructural characterization of Friction Stir Processed TiNi shape memory alloys constitutes a quick and cheap method to obtain a microstructural characterization of material strictly connected to the thermo-mechanical parameters of process.
|Numero di pagine||6|
|Stato di pubblicazione||Published - 2017|
All Science Journal Classification (ASJC) codes