| With the building of the social market ecology system and increasing strengthening our open policy and the new condition of our country joining the WTO, the mineral resource of our country is no more limit in domestic, but we must face the world.Mineral processing mathematics model uses specially in mineral processing analysis and good mineral processing process design and mineral processing controlling etc. We have difficult to build a mineral processing model of fitting condition because many mineral processing are very complex and have many influent factors, so we must find a new building model method to realize data processing of mineral processing. Artificial neural network as a new scientist, it is a new data processing method for mineral corporation .BP neural network is a high nonlinear model, it has a superior performance for approaching the complex nonlinear system and it offers a method to choose for building model of complex process.In the mineral processing (specially in grinding-classify), there have many influent factors, and each factor has some connection, however we require linear independence between input variables and output variables for BP network in theory. So the author uses regression analysis of SPSS (Statistical Package for the Social Science) to change the definite connection into linear independence for input variables and output variables, at last we build model with BP network. If we want build a practical BP model, we must choose not only a best network construct but also a reasonable method of network stopping training, the article uses early stopping method based on the building mineral processing model of the combination regression analysis and BP network, a best precise is achieved for building model.What the author has done can improve the quality of building mineral processing model, and can enrich the theory methods & practice of complexnonlinear building model. |
PDF Full Text Request