Application Of Fuzzy Evaluation Method In Earthquake Prediction

The present situation,The earthquake prediction is one of the ten top problems in the world,it has a very important role in earthquake prevention and disaster reduction.Recent studies showed that the presence of a non-linear relationship between the factor induced earthquake.The Artificial Neural Networks(ANN)theory shows obvious advantages in solving some complex non-linear problems.Based on this understanding,the ANN is applied to the earthquake magnitude prediction based on the MATLAB platform.This paper analysed the limitations of traditional earthquake prediction methods the advantages and disadvantages of ANN were introduced.the conclusion prometed the necessity of combining genetic optimization algorithm with ANN.At the same time,the specific process of optimization is given.The study area was located at Sichuan Province,and the Longmen mountain fault zone in the north-south seismic belt passed through Sichuan Province,which causes Sichuan to become one of the most earthquake-prone areas in China.The ANN is extremely sensitive to the input factor.After analyzing some related literature,some seismological index indicators are selected as input,and the ground resistivity amplitude was used as input.Through analyzing the correlation between magnitude and input parameters,the correlation coefficient between ground resistivity and magnitude is 0.65.This verifies that ground resistivity can be used as a reliable precursor.The other related index presents weakly relative with the magnitude,which indicats that the combination of multi-aspect precursor index for earthquake prediction is a feasible future development.In this paper,self-organizing feature map neural network(SOM)was used to classify the selected magnitude,filling in the characteristics that the BackPropagation(BP)neural network is more similar in the sample,and the output is better.The clustering can increase the anti-noise performance and generalization ability of BP neural network and accelerate the convergence speed of neural network.This solves the defect of BP neural network for earthquake prediction from the input data.From the clustering results,it can be roughly divided into seven categories,each of which has some characteristics.For a sample whose clustering result is a seismic instance on the same fault,it is helpful to predict the time and location of the future.For samples with similar magnitudes,it will be helpful to predict the magnitude of the future.However,this study also shows that the current earthquake prediction methods have limitations,and any prediction method has a certain scope of application.According to historical data,the BP neural network has ability to forecast the future situation according.However,the BP neural network weight has a great influence on the output result,and obtaining a better weight is crucial to the output result.Genetic algorithm can perform global search and has strong robustness.Using genetic algorithm to optimize the weights of BP neural network,an optimal initial weight and threshold are obtained.This is to improve the output from the neural network structure.Finally,the output of the combined neural network(SOMGABP)is inverse normalized.At the same time,the commonly used neural network output results are compared to obtain error curves of output results of several neural networks.On the one hand,the output shows that it is possible to apply neural network theory to earthquake prediction.On the other hand,using the method proposed in this paper is applied to earthquake magnitude prediction.The accuracy rate of prediction obtained is 93.75%,the accuracy rate is higher than several other neural network study.It shows that this study can improve the accuracy of the magnitude prediction.In addition,although this article only predicts the magnitude of the three elements of the earthquake,this is also the limitation of this article.In fact,the time and location of the shock are implied in the precursor data.Since that is another complicated issue,this article did not discuss it.