A Novel Approach to Predict Earthquakes using Adaptive Neural Fuzzy Inference System and Conservation of Energy-Angular Momentum

Abstract

This paper presents an adaptive neural fuzzy inference system (ANFIS) approach to predict the location, occurrence time and the magnitude of earthquakes. The analysis conducted in this paper is based on the principle of conservation of energy and momentum of annual earthquakes which has been validated by analyzing data obtained from United Sates Geographical Survey (USGS). This principle shall not be violated due to the fact that the angular earth speed about its axis is fixed to keep the 24 hours daytime unchanged. Furthermore, it is assumed that the area under the moment curve of earthquakes in the north part of the earth balances the area under the moment curve in the south part of the earth due to the conservation principle. For automatically tuning Sugeno-type inference systems, a sample of training data is used to train the ANFIS system using 3 bell-shape membership functions with grid partition to generate the fuzzy inference system (FIS) along with 270 epochs. In training the earthquake ANFIS methodology, the location of the earthquake is used as an input, meanwhile the moment of the earth quake is assigned as the output. The resulted training error was stabilized after 250 epochs converging to an acceptable value of 0.84. To further enhance prediction of earthquakes, different data set is used to verify the validity of ANFIS output. The inputs to ANFIS are the latitude, longitude and date to predict the corresponding earthquake moments as an output. Surprisingly, the FIS system is found to be capable to predict most of the earthquakes moment-magnitude at the specified location with a 0.17424 converging error. Dynamic ANFIS earthquake predictor along with 3D meshed surfaces are found to be efficient as well. Finally, the ANFIS results are demonstrated to show the effectiveness of the approach.