The Research Of Freeze-Thaw Rock Damage Characteristics Based On A Modified K-means Clustering Algorithm

A variety of engineering problems, and even disasters, always have happened in Cold regions among which, the freeze-thaw cycle effect is the main reason of engineering structure damage. Because the rock is a kind of natural damage material, and it has very complex nonlinear characteristics, it's very difficult to analyze and solve the problem of freeze-thaw cycle with the angle of the general research methods or determinism. So this paper embarks from the phenomenological perspective, based on the experimental study, using the method of combining CT scanning image analysis with statistical damage theory, systematically studied the mechanical properties of rock under different freeze-thaw cycles and the damage evolution rule.The fresh red sandstone rock was obtained from the engineering field and processed into standard test specimen, and carried out different freeze-thaw cycles experiment in view of the test specimen; using CT scan technology to scan the rock that under 0,5,10,20,40 times of freeze-thaw progressing; In view of the original CT scan images, first, the median filtering technology was adopted as the noise reduce pretreatment, then, a kind of improved k-means clustering algorithm that combining histogram with fuzzy recognition was applicated to classify and recognize the content of the pretreated scanning images.Based on the data that come out from CT scanning image's clustering analysis, from the macro damage mechanics perspective, the damage variable that based on effective load bearing area reduction concept was introduced, combining with the data distribution characteristics, verified that the new damage variable obey the Weibull distribution which use freeze-thaw cycle as the variable, meanwhile analysis by combining the distribution parameters with the rock actual damage, indicate that the rocks' injuring under the effect of freezing and thawing is a loss type failure event, and the damage distribution of rock samples having the features of large at top and end part and small at middle part.Analyzing by combining the results of freezing and thawing rock test with rock freezing and thawing damage characteristics, think that take the freeze-thaw rock CT scan images as the damage object of study, and using statistical analysis tools to establish damage evolution model is a feasible and reasonable method, and the methods has the characteristics of non-destructive and simple, also has reference significance to the related research.