Improved ESN neural network model for mine water inrush identification
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
It is difficult to judge the type of water source with similar water chemistry characteristics by the traditional water prediction model. The similarity analysis of 26 typical water samples is carried out by using water chemical composition analysis and FCM (fuzzy C-means)cluster analysis method. Four samples with higher similarity are extracted as samples to be tested and loaded into the GA_ESN discriminant model based on damped least squares regularization method, and the GA_ESN discriminant model of the regularization method is compared with the improved GA_BP and the standard GA_ESN model. The results show that the improved GA_BP discriminant model has the worst effect and the prediction accuracy is only 50%. The back estimation rates and prediction accuracy of the standard GA_ESN model are 100%, however, the accuracy of the model requires a high degree of complexity of the model, and it's prone to have overfitting problem. And the improved GA_ESN discriminant model can make up for the shortage of the above model, which can not only simplify the model training process, but also improve the accuracy of water source identification. Therefore, the model can be used as a fast and effective method to identify the source of mine water inrush.
