Research On Ductility Of Unbonded Partially Prestressed Concrete Beam With High Strength Based On Genetic Algorithm And Neural Network

Along with the construction market fast development, the low strength concrete already can not suffice the market request in china. The high strength concrete has applied up to the C100, some has achieved the C130 in advanced countries.The ductility of high strength concrete is bad,so its earthquake resistance is also obviously weaker in the unbonded prestressed concrete (PPC) structure. Therefore, how to improve the ductility of unbonded partially PPC beam is worth further studying.At present, the artificial intelligence, especially artificial neural networks and genetic algorithms aspect, has been applied widely to various fields. But, it is applied hardly in the structural engineering. In this paper, according to respective character of neural network and genetic algorithms, the prediction models of the ductility of unbonded PPC beam with high strength are built. Meanwhile, kinds of ingredients of unbonded PPC beam with high strength are analyzed,and the curves of displacemen t ductility ratio are obtained. The result indicates that displacement ductility ratio of the beam gradually reduces with increasing unprestressed index, prestressed index, span depth ratio and depth of section. But the reducing speed is the difference.For the convenience of designer, the formula of displacement ductility of the beam, based on predictable result of the model, is built. In order to prevent"premature convergence", a new way to change fitness value is suggested. The paper provides one new way to study ductility of the beam. But, the model need increase the sample to further enhance precision and scope the forecast result.