便携式人工神经网络及其在砼疲劳寿命预测中的应用

Implication of Portable Artificial Neural Network and its Practice on Fatigue Life Estimation of Concrete

基于快速人工神经网络,探讨了利用线性函数对数据进行归一化的方法,结合数据归一化后的值域范围以及用于确定隐含层神经元数目的经验公式,依据网络均方差最小化原则,得到了人工神经网络的结构参数,然后将疲劳实验数据作为训练数据,建立了混凝土疲劳寿命预测的人工神经网络模型,并将其导出为一个独立的便携式模型文件.计算结果表明,该模型计算精确度高,可以方便地嵌入到各种工程软件中,能够解决混凝土疲劳寿命预测模型的准确性和实用性两大难题.

Three linear functions are used to normalize testing results based on Fast Artifi- cial Neural Network. Considering the range of normalized testing results and the number of neurons in bided layer determinated by empirical formula, key parameters of artificial neu- ral network are obtained according to the principle of mean square deviation minimization. Then, an artificial neural network model for estimation of concrete fatigue life is established when testing results are studied as training data and is exported into a portable file. The calculated results have a good agreement with the testing data, and this model can be em- bedded into engineering software conveniently. Therefore, it can provide higher accuracy and better practicability than general one.