再生混凝土收缩徐变预测新方法

New method for prediction about shrinkage and creep of recycled aggregate concrete

针对当前方法的局限性,提出基于高斯过程机器学习的再生混凝土收缩徐变预测方法。该方法采用少量试验数据作为学习样本,建立主要影响因素与收缩徐变的复杂非线性映射关系,对新的预测样本做出精准预测,获得对应的收缩徐变。实例研究结果表明:该预测方法是可行的,能够在小样本试验数据下对新样本做出高精度的预测,并且方法参数自适应获取、实现过程简单,为再生混凝土收缩徐变的合理确定提供一条新途径。

Aiming to the fact that it is still difficult to reasonably determine the shrinkage and creep of recycled aggregate concrete, the method based on Gaussian process（GP） machine learning is proposed for forecasting of shrinkage and creep of recycled aggregate concrete. Using a few test data as learning samples, the nonlinear mapping relationship between shrinkage creep and its main influencing factors is established by GP.Based on that, the new prediction samples can be accurately predicted and obtained corresponding shrinkage creep.The results of case study show that this method is feasible, that can high accurately predict new samples based on a few test data.The method also has merits of self-adaptive parameters determination and simple to implement.It provides a new way for reasonably determining shrinkage and creep of recycled aggregate concrete.