Contrast Of SOM Neural Networks And Cluster Analysis Using In Grain-size Analysis

SOM and hierarchical cluster are used in classifying the soil samples of incompact deposits which were sampled from Sunan Coalfield of Anhui Province. The results can be used to explain the sedimentary environment by grading analysis.The paper comprises the following 3 parts:(1)The characters of the moisture content, the puddlability, and the moisture-releasing of the aquifer in the coalfield are got by testing the hydrophysical properties of the deposits of the Forth Aquifer.(2)Using SOM and hierarchical cluster in classification and discussing the differences of function between the two methods in the view of mechanism and applied ways, it is concluded that SOM network could be applied conveniently in non-supervisor classification, and identification of incomplete samples without any prior knowledge and with a unique result. It is believed that SOM network is superior to hierarchical clustering methods in the classification of contributing factor of sediment which is nonlinear problem.(3)The correctness of the results of the two methods analysis can be testified by constructing the C-M figure of the clastic deposits.