Stability classification of adjoining rock of underground engineering based on Hopfield network

The Hopfield neural network, with associative memory function, is used in the stability classification of adjoining rock of underground engineering. Five indexes, including rock quality designation ( R_QD), uniaxial compressive strength ( R_w ), integrality coefficient ( K _v ), strength coefficient of structural plane ( K _f ) and seepage measurement of groundwater ( ω ), are selected as the factors which affect the classification. The function provided by matlab toolbox is used to build a network and for stimulation. After memorizing the standard of classification, a Hopfield network is established for the stability classification of adjoining rock of underground engineering. Then the network is used in the classification of the measured samples of two projects, Manwan Hydropower Station and Guangzhou Pumped Storage Power Station, to detect the classification ability of the network. The research shows that the classification results based on the Hopfield network are reliable. The network has a fast convergence rate and high practicability.