基于神经网络的混凝土结构使用寿命评估

混凝土结构耐久性是目前建筑结构与建筑材料研究领域的热点.本文用神经网络的研究方法,研究了处于硫酸盐侵蚀下混凝土结构使用寿命与环境侵蚀介质间的关系.建立了混凝土结构寿命评估神经网络模型.网络输出给出了在预期的破坏等级下结构使用寿命的近似估计.

ASSESSMENT OF EFFICIENT LIFETIME OF CONCRETE STRUCTURE BY MEANS OF NEURAL NETWORK

This paper is intended to discuss the relationship between the lifetime of concrete structure and the environmental conditions, such as the sulfate and other corrosive media by using the neural network principle. And for this purpose, the author has established a four-level BP model of neural network with 16 nodes in the input layer and 1 node in the output one. Given the expectant deterioration level and environment and material parameters, the network output presents an approximate estimate of lifetime of concrete structure. The paper has also put forward some samples, like whole weight matrix, threshold value matrix and some error vectors. And finally, the author also brought about an example of practical application. As is known, most soil contains some sulfate. However, if the concentration of sulfate is too high, the underground part of a concrete structure, for example, bridge, tunnel, culvert or a highrise basement, will sooner or later get damaged, which may in turn directly influence the lifetime of its adjacent concrete structures. Eroded by sulphate, all the concrete structure may eventually get deteriorated. Thus, we have proved that there exists a complex physical and chemical procedure including such factors, as the mineral composition of raw material (especially, cement), concrete physical performance itself and in relation to the concentration of sulfate and other hydronium content. In addition, the concrete is also found to be influenced by the environmental factors like evaporation, the cycles of freezing and dissolving, the alternating drying and wetting state, hydraulic erosion, etc. In general, due to its own complexity and cognitive limitation, the lifetime evaluation and prediction under sulphate erosion will still remain more or less illegible and uncertain. Hence, it is still necessary to resort to some intelligent methods for its analysis.