Study On Mineral Resource Classification And Identification System

Mineral resource is one of the essential precondition and basic power of nationaleconomic construction. A study of classification and identification in mineral resources isalso one of the most important aspects in Mineral Resource study and development, it has animportant meaning for the exploitation and using too. The existing classification andidentification in mineral resources are mostly realized by using chemical analysis method.However, this paper tries to realize it by using the method of Physical Property analysis.This paper takes8kinds of common metallic mineral resources as an object of study torealize the fast and accurate destination in mineral resources classification and identification.And a Research method which bases on the BP Neural Network of that system has beenpresented. This paper has used the MATLAB7.0Toolbox to create the classification andidentification System which based on the BP neural network, and try to realize the fast andaccurate destination finally. The main research contents:(1)The paper has collected and arranged the physical characteristic dates of mineralresources on the base of analysising the status of mineral resource classification andidentification at home and abroad.(2)The paper has analysised the physical characteristic dates of mineral resources,which contains color, steak, luster, fracture, density and hardness. These seven features arehighly important to mineral resources as experimental study samples. These analyses providethe foundation for classification and identification system study of mineral resources.(3)The paper is based on a BP neural network of principles and research steps inmineral resources classification and identification, and discuss some key technologies toapply these principles, including the selection and pretreatment of swatches, theselection of input-output variables, conforming the number of the nodes in hidden layers,the selection of the initial weight and value, the selection of activation function, trainingarithmetic as well as parameter.(4)The paper has divided the mineral resource feature dates into two parts, such astraining samples and testing samples. Firstly, the mineral resource training samples has beentrained by different training algorithms and implicit layers node numbers. Secondly, the welltrained network has been tested by the indiscipline samples. Finally, the ideal trainingalgorithms and implicit layers node numbers has been determined fleetly and accurately.The results show that, the mineral resources classification and identification system isfunctionally in the differentiation of8kinds of metals mineral. The identification rate is above95%and the test samples identification rate is more than80%. However, the authorfound through his studies that the inner interaction between similar characteristics and thecomplicated associated mineral features will lead to the identification rate reduction oftesting samples.