SOFMANN And FCA Model Applied In Concealed Mineral Location Prediction

This thesis is one part of sub-task: HuiZe Pb-Zn mine concealed mineral prediction multi-information processing and mathematics model research and development of mineral prediction expert work platform. Post-doctor, Han Runsheng, CAS Geochemistry Research Institute is responsible for province and institute science and technology cooperation item: HuiZe Pb-Zn mine bed deep and periphery concealed mineral location prediction research (item number:2000YK-04); this item includes foernamed sub-task.Self-organizing feature mapping artificial neural network, fuzzy comprehensive appraisement model and fuzzy neural network model based on neural network and fuzzy mathematics combined with traditional statistic method for concealed mineral location prediction are constructed in the paper on basis of modern mineralization theory and mineral prediction theory and method, at the same time, according to the important breakthrough achieved by Doctor Han RunSheng and mineralization law approved and prospect data acquired by long-term prospect practice in HuiZe Pb-Sn mine. These models are assistant tools for concealed mineral location prediction in resource crisis mine deep and periphery .Main works in the paper:1) Considering complex relation between control- mine factors and mineralization, SOFMANN (Self-organizing Feature Mapping Artificial Neural Network), FCA(Fuzzy Comprehensive Appraisement) and FNN (Fuzzy Neural Network)used in concealed mineral location prediction are constructed. SOMFANN is characteristic of self-adapting, self-organizing, learning without teacher control mineral multi-multi-information; FCA can take full advantages of experts and fuzzy mathematics processingfuzzy phenomenon and fuzzy action in concealed mineral prediction; FNN make use of advantages of fuzzy logic easily expressing humankind knowledge and distributed information memory and learning ability.2) SOFMANN and FGA are first applied in YunNan province Hui-Ze QiLing Plant Pb-Zn mine bed deep concealed location prediction. Prediction result meet with engineering drilling validation, models are proved to be valid.3) Various prediction data are acquired by grid dividing method to prepare quickly necessary data for prediction model.4) Analyzing results are visualized in order to provide convenience for experts analyzing mineralization target.5) Programs for three models are developed and system procedure for concealed mineral location prediction is built.