Rapid identification model of mine water inrush sources based on extreme learning machine

Abstract

In the process of disaster prevention of coal mine water inrush, it is necessary to quickly and accurately identify the types of water inrush sources. Based on the high sensitivity, rapid and accurate monitoring characteristics of laser induced fluorescence technology, the fluorescence spectra of water samples were collected on the experimental platform of water sample detection. After pre-processing spectra and extracting features, the multi-classification learning model is established by the extreme learning machine (ELM) algorithm. In this paper, it determines the sigmoid function as hidden layer activation function, and obtains the optimal number of hidden layer nodes by the method of cross-validation. ELM is compared with the conventional neural network classification model in different part, such as the average time and the average classification accuracy. The average classification accuracy of ELM combined with principal component analysis is about 98% and 93% in the training and testing set respectively. And the classification learning time is greatly improved. Therefore, the model is more suitable for rapid and accurate classification of water inrush sources.