Abstract
Dendrite structures have been studied from the standpoint of solidification theory. In this paper, they were treated from the standpoint of pattern recognition. Multivariate analysis was applied to the dendrite structures; length (L), width (M) and spacing (N) of the stalks and length (l), width (m) and spacing (n) of the arms were measured as variables, and then solidification rate (R), content of added element (C) and temperature gradient (G) at the solid-liquid interface were obtained.
Experiments were done on Ag-Pd alloys (investment casting) or Al-Cu alloys and Ag-Sn alloys (unidirectional solidification) and 10∼100 measurements were made on each specimen.
The analysis of the variance and of the correlation coefficient were done to determine these many variables as statistical problems. The results obtained are as follows: l is significant for C and has no significance to R in unidirectional solidification. M is not significant to C but significant to R in the investment casting. This means that R plays an important role in the pattern. The problems concerning the correlation matrix are as follows: M and m have a strong correlation, and the correlation between m and n suggests a relation to the distribution of solutes. N shows a strong independency. These correlationships are tabulated.
