Study of Damage Predicting Model on Subsurface Engineering Structure

As impersonality transforming sequence in subsurface engineering structure, damage predicting model was erected with the theory of least squares support vector machine. Estimating input-output relation in subsurface engineering structure damage problems according to learning samples, non-line implicit expression was constructed between structure damage problems and their factors, and then testing samples was predicted with the law, which was weighted by empirical risk minimization theory. Taking “crack” as a case, results show, LS-SVM model has effective small sample learning ability and higher matching and predicting accuracy, which exceeds predicting model of BP nerve network.