Discrimination Of Water Sources Based On Neural Network

Mine water inrush accidents is one of the most frequently occurring disasters in coal mines which seriously affect the safety of miners,resulting in the loss of mine equipment.The hydrogeological environment of many coalfields in our country is very complex and to preventing water inrush is quite difficult,Therefore,it is an urgent need to establish a reasonable and effective method to identify the source of water in water inrush.Only in a timely manner to identify the type of water,can we determine whether the water inrush accident happened while combine with the mine water inrush environment.This paper determined using neural network model to identify the types of water sources after research and analysis,and propose a new Mix improved Shuffled Frog Leaping Algorithm to improve the BP neural network,the main research work is as follows:1)Based on the analysis of hydrochemical characteristics to determine the input parameters of discriminant model,then the chemical composition of water samples were analyzed with Piper three line diagram.Last after deal with the zwitterion detection,data standardization and principal component analysis,the reliability and the reliability of the data are ensured.2)According to the analysis of several water source discrimination models and the comparison of the experimental results,The BP neural network model is decided to used to identify the source of water inrush,and the input parameters of the model are the main components after deal with data processing,the output of model is the types of water sources.3)Aiming at the shortcoming of BP neural network,the SFLA algorithm are proposed to training the weights and thresholds of BP neural network.After specific experiments,the improved SFLA BP neural network is proved to be superior in performance.In addition,by introducing opposition strategy Gauss mutation and chaos disturbance algorithm,the global search and local search ability of SFLA is optimized,a new Mix improved Shuffled Frog Leaping Algorithm(MISFLA)is proposed and the effect of MISFLA_BP neural network is tested.4)According to the deficiency of shallow layer neural network,this paper propose to use Belief degree neural network to identify water inrush type,then introduce the basic principle of Belief degree neural network and the superiority compared with the traditional neural network,the structure of restricted Boltzmann machine(RBM)is detail analyzed.last this paper put forward an Bayesian improved restricted Boltzmann machine(BRBM)to improve the generalization ability of the network and pile BRBM to form DBN,its effect is tested and verified with discriminant experiment.