Research On Dam Deformation Application Based On FOA-BP-AdaBoost Strong Prediction Model

With the rapid development of society and economy,In recent years,the construction of large-scale hydraulic facilities has entered a critical stage.In order to ensure its operational safety,deformation prediction of these facilities has become an urgent task.Of course,reliable and accurate monitoring data is the premise of accurate prediction,and the importance of appropriate and reasonable prediction methods is self-evident.Due to the fact that hydraulic structures such as dams are closely related to various influencing factors,including water level,aging,and water pressure,these factors generally have strong arbitrariness and randomness.Therefore,it is very important to build a reliable and effective dam deformation prediction model.Nowadays,a single model to predict deformation monitoring is common,which usually includes SVM model,Kalman filtering model,GM(1,1)model and so on.At present,many scholars have proposed various combinatorial models to solve the limitations of the single model dealing with different data.In this paper,by using BP neural network,genetic algorithm(GA),fruit fly algorithm(FOA),particle swarm algorithm(PSO),and Adaboost strong prediction model,and auxiliary dam deformation data,different combination models are used to analyze and predict the accuracy.The main research contents of this article are as follows:(1)The related theories such as GA algorithm,BP neural network and Adaboost strong prediction model are discussed.Based on the complex nonlinear relationship between dam deformation and multiple influencing factors,this paper establishes a robust Adaboost prediction model(GA-BP-Adaboost)based on GA-optimized BP neural network.Compared with the single BP model and GA-BP model,we can see that the model combines the global optimization of the GA algorithm and the local optimization of the BP neural network.At the same time,the Adaboost Strong Predictor combines the advantages of multiple predictive sequences by assigning different weights to the sequence of weak predictors.The Adaboost strong predictor has achieved the goal of “superiority and superiority” and has greatly improved the accuracy of prediction.(2)The basic contents of the fruit fly(FOA)algorithm and particle swarm optimization(PSO)algorithm were studied.The FOA-BP model was proposed basedon the FOA algorithm that appeared in recent years,The model uses FOA algorithm with less adjustment parameters to optimize the threshold and weight of BP neural network,Therefore,the probability of the BP algorithm falling into the local optimal solution is significantly reduced,which significantly improved the BP neural network global optimization level and generalization ability.After comparing with the PSO algorithm and the GA algorithm,it can be seen that the FOA-BP model has strong ability of rapid convergence and accurate mid-and long-term prediction capabilities.It is proved that the FOA-BP combined model has certain feasibility and validity in dam deformation prediction.(3)Combine the advantages of the above Adaboost strong predictor and fruit fly algorithm(FOA)to establish a combined model: Adaboost strong predictive model(FOA-BP-Adaboost)based on BP neural network optimized by Drosophila algorithm(FOA).Through comparative analysis of examples,this strong prediction model combines the advantages of the three.In addition to ensuring better BP neural network rapid convergence and mid-and long-term prediction capabilities,it also achieved the goal of “superiority and superiority”.It has higher accuracy compared with GA-BP-Adaboost strong prediction model,etc.