Pattern recognition techniques phases

A well-known tool for the estimation of reactivity hazards of organic material is called CHETAH [5]. The method is based on pattern recognition techniques, based on experimental data, in order to infer the decomposition products that maximize the decomposition energy, and then performs thermochemical calculations based on the Benson group increments mentioned above. Thus, the calculations are valid for the gas phase, but this may be a drawback, since in fine chemistry most reactions are performed in the condensed phase. Corrections must be made, but in general they remain small and do not significantly affect the results. 

To find out the number of solutes (perhaps less than ten) that are necessary to characterize the liquide phases. An appropriate pattern recognition technique is feature selection. 

In recent years, we have used this comprehensive strategy to make the computerized assessment of a series of phase diagrams. By this strategy, we have found that the coordination of atomic parameter method and various pattern recognition techniques (including SVM) is rather effective to find out the outliers or suspicious phase diagrams [30 23 26 32],... 

Heberger et al. [55] used principal component analysis (PCA) to reduce the amount of test solutes when calculating Flory-Huggins parameters x 2i- Subsequently, PCA became a popular technique in data analysis for pattern recognition and dimension reduction, as it can reveal several underlying components, and may also help to explain the vast majority of variance among the data [56,57]. PCA is particularly useful for classifying stationary phases [58,59], polarity [56], and interaction parameters [57]. Detailed descriptions of PCA are available in standard chemometric books and reviews [58,59]. Notably, the method should facilitate the solution of problems connected with the solute dependence of the x 2i parameter. 

The formation of intermediate compounds is one of the most important factors affecting the geometry of phase diagrams. Atomic parameter-pattern recognition method is an effective technique dealing with the assessment or prediction problems about the formability of intermediate compounds. 

As one of the early applications of the EBSD technique to the study of steel products, Randle found that the technique could be used to differentiate hematite, magnetite and wustite based on their respective pattern symmetry, Kikuchi band widths and pattern intensities, although no details of the phase recognition and analysis were provided [151]. In more recent years, Higginson and co-workers further explored the capabihty of EBSD in the analyses of iron oxide scale layers growing on low and ultra-low carbon steels, leading to a continuous stream of publications of their research results [153-162]. The technique was also used in the studies of other research groups [99,109,163-166],... 

---