Research On Prediction Of Tsunami Climb Height Based On Neural Network

Tsunami is a kind of catastrophic sea wave,mainly formed by submarine earthquake,volcanic eruption and submarine landslide.It has the characteristics of long propagation distance,small energy loss and strong destructive power,etc.Once the tsunami occurs,it will cause huge losses to people and property in coastal areas.The first wave climb of tsunami wave is the main cause of nearshore damage.Therefore,it is very important to understand the climb of tsunami wave and its motion state for nearshore infrastructure construction and disaster prevention and mitigation.In the experimental study and numerical simulation of tsunami waves,the model is usually simplified to an isolated wave,which can be replaced by studying the properties of the isolated wave.Therefore,this paper simulates the climbing motion of isolated wave by numerical simulation method,analyzes the motion characteristics of isolated wave,obtains the motion parameters during climbing,and puts forward a prediction model of the climbing height of isolated wave based on BP neural network,It is expected to be a certain reference for tsunami wave climbing height measurement and engineering disaster prevention.The research work of this paper is as follows:(1)Theoretical study on solitary waves.The concept of isolated wave,propagation process,wave making method and climbing are described respectively.The Goring wave making method is analyzed with emphasis,and the theoretical wave equation when the solution of isolated wave is the first and third order is given.The slope climbing equation of isolated wave is deduced,and two empirical formulas for maximum climbing of broken and unbroken isolated wave are summarized.(2)Numerical simulation of solitary wave climb based on OpenFOAM.Three working condition models with different static water depths were established,and 16 groups of wave height and climb data were numerically simulated for the isolated wave climbing experiment.The results of numerical simulation are verified from two aspects: motion characteristics of isolated wave and verification of theoretical formula.The results show that the numerical simulation method reflects the actual motion law of isolated wave well,and the data obtained by simulation also has a certain accuracy.(3)Establishment of the prediction model for the climb of isolated wave.According to the motion parameters obtained from numerical simulation,a prediction model of isolated wave climb based on BP neural network was established.A total of 70 groups of sample data,fifty groups were randomly picked as the training patterns,twenty groups were pickeded as test patterns,and write the program in Python.The analysis of the results of training samples shows that the prediction model realizes the recursive convergence of the error,and the error finally converges to 0.003363.By drawing the comparison graph between the real value of training samples and the output value,it can be seen that the difference between the two is not large and the training effect is good.Using the trained model to forecast the output,the remaining 20 samples respectively from the error of the predicted values and the real value indicators and data contrast and graphics trend change were analyzed,and the analysis results show that the test sample for the predictive model showed better adaptability,latter is fitting,graphical evaluation index correlation coefficient of 0.9574,showing a high correlation,the estimates of the model with good generalization ability,has certain promotion value.