The Application And Realization Of Artificial Neural Networks And Genetic Algorithm In Testing Of Piles

The applications of neural networks,genetic algorithm and support vector machine in pile testing are studied in present paper. Following conclusions are drawn:(1) The programmers of BP neural networks and genetic algorithm were implemented on Microsoft Visual Studio.(2) A study on prediction of vertical ultimate bearing capacity of single pile based on genetic algorithm and artificial neural networks was shown in present paper, and the results of high strain dynamic testing were used. The maximum of impact,maximum of energy,maximum of dynamic displacement on top of pile,depth of pile in soil and diameter of pile are made as factors, and it is made as object the results of CAPWAP. The parameters of BP neural networks were obtained by improved genetic algorithm. The results of prediction was close to the results of static loading test, which shows that the method of prediction of vertical ultimate bearing capacity of single pile based on genetic algorithm and artificial neural networks is adaptive and has high precision.(3) The application of genetic algorithm and neural networks of low strain integrity testing of foundation piles was studied in present paper. First, the FFT is applied on the results of low strain integrity testing; secondly, the some points of frequency were chosen to be as factors, and the kind of pile integrity was be made as object. The neural networks had three layers, and the cells of each layer were 10,X,4, which the X was obtained by improved genetic algorithm. The rationality and precision of the method in present paper were shown by the results of prediction. Summing up the application of genetic algorithm and neural networks of low strain integrity testing of foundation pile, the method of intelligent classifying kind of pile integrity was raised in present paper. Finally, the software of intelligent classifying kind of pile integrity was implemented primary.(4) The prediction model of vertical ultimate bearing capacity of single pile is built by combining improved genetic algorithm and BP neural networks. The practical date are used to train the neural networks whose parameters are determined in global optimal by improved genetic algorithm .Thus, a prediction model with better generalizing capacity is built. The study shows that this method is effective for determining the parameters of neural networks and the results is satisfactory..(5) Many problems in piles foundation engineering are complex system problems, and they cannot be solved by traditional methods. In view of this, based on the intelligent methods, this paper presents a concept design of expert system of pile testing. Attempting to start from influential, combined with the construction monitoring date,the information of instrument etc. Thereby, based on this system, it can be used to realize the precision,visualization and intelligence of testing of piles.