Time Series Prediction Model of Concrete Corrosion in Sulfuric Based on SVM

In this paper, a long-team immersion of concrete in dilute sulfuric acid is carried out. On the basis of the experimental data, a time serious prediction model of concrete corrosion in sulfuric based on support vector machine (SVM) is developed. The design steps and learning algorithm are also given. Comparing with the test result, this model has good predictive function, with the suitable reconstructed space matrix and prediction step length. Using the latter 20 groups of test data which data spaces are reconstructed as the prediction set, most of the prediction relative error(Er) can be controlled within 15%, and only a small amount are within 30%. It implies that this model can be a new powerful way for study of concrete corrosion resistance.