Elevation Error Prediction of Continuous Beam Cantilever Construction Phase Based on LS-SVM

According to the construction of long-link continuous beam bridge, I choose field measured data as research samples and establish an elevation error prediction model by dealing samples, selecting the kernel function, selecting parameters, training, and predicting. I compare the Least Squares-Support Vector Machines (LS-SVM) prediction value with the measured value, the SVM model predictions, the BP neural network model predictions and the dimensionality reduction model predictions, so that predict elevation errors during cantilever construction phases by established models. According to the results of the comparison, the elevation error prediction model is highly accurate and has more high efficiency, good stability, and strong learning ability. Under the verification of the elevation control results in the cantilever stage, LS-SVM elevation error prediction is used for controlling the elevation of the bridge and solves the problem–predictive control successfully which is caused for few beam blocks in the cantilever phase of a continuous girder bridge.