Research On Prediction Model Of Dam Horizontal Displacement Based On Neural Network

In order to ensure the safe operation of the dam and ensure the safety of life and property of the downstream people,safety monitoring should be carried out throughout the dam’s life cycle.Practice has proved that this work has played an extremely important role.As one of the important directions in dam safety monitoring research,the research on mathematical model of safety monitoring has been a hot issue in this field.The prediction model is established through a large number of measured data,which can be used to monitor and judge the safety of the dam.However,due to the complexity of some monitoring factors,the influence mechanism cannot be expressed by a definite functional relationship,or the series of monitoring data are short and the information is missing,so that the commonly used prediction models have low accuracy and poor reliability and are difficult to meet the daily management and research needs.Therefore,it is of great theoretical and practical significance for improving the theory of monitoring data analysis and treatment and improving the dam safety management level to study high-precision and high reliability prediction model using new theories and methods.In this paper,neural network theory is used to study the nonlinear mapping relationship between dam horizontal displacement and influencing factors,and then to analyze and predict the dam deformation law.The main research contents and conclusions are as follows:(1)A neural network prediction model based on genetic algorithm is studied.Due to the shortcomings of BP neural network,such as slow training speed and easy falling into local minimum,this paper introduces a genetic algorithm with global optimization ability to optimize BP neural network,constructing a dam deformation prediction model.The model first uses the selection,crossover and mutation operation of genetic algorithm to optimize the initial weights and thresholds of the neural network.Secondly,the training sets of multi-source monitoring data of the dam are studied by using the optimized initial weights and thresholds.At last,the accuracy of the model is improved by using the established model.(2)An improved adaptive genetic neural network model is proposed.Its idea is based on the basic principle of genetic algorithm.On the one hand,the operator of selection is improved.The traditional roulette method and optimal individual preservation strategy are discarded,while the optimal individuals with different proportions are saved in different genetic manipulation stages for direct inheritance to next generation strategy,increasing the diversity of initial population and preserving the best individuals.On the other hand,the traditional crossover and mutation probabilities and adaptive crossover and mutation probabilities are discarded.Instead,an adaptive crossover probability with good initial population evolution is used,which enables the algorithm to adapt to its own environment.The improved strategy based on the above genetic algorithm greatly enhances the global optimization ability of the basic genetic algorithm.(3)A neural network prediction model based on particle swarm optimization is studied.Because particle swarm optimization has the ability of global optimization,it optimizes the initial weight and threshold of BP neural network,which can effectively overcome the inherent flaws of BP algorithm.The wavelet neural network model based on wavelet analysis theory and BP neural network is discussed.Wavelet neural network uses the wavelet elements to replace the neurons and establishes the connection between wavelet transform and network coefficients through affine transformation.It can further improve the generalization ability and reliability of the network.(4)The models that has been studied are analyzed and verified by the measured data.The analysis results show that the improved adaptive genetic neural network model is better than the basic BP neural network,genetic neural network and other models.It not only improves the prediction accuracy and convergence speed,but also has better stability,which has good practical application value.