基于MOPSO-RBFNN的小芯样试件抗压强度预测方法

Prediction Method of Compressive Strength for Small Core Specimens Based on MOPSO-RBFNN

针对实际工程中在钢筋密集或构件尺寸较小处使用钻芯法难以获取标准尺寸芯样的问题,本研究提出了一种基于径向基神经网络的小芯样试件抗压强度预测方法。首先,本文制作了82组直径50mm的细石混凝土芯样试件与同条件养护同龄期150mm立方体试件进行轴心抗压试验。其次,对两种试件抗压试验得到的强度数值进行比较与分析,采用MOPSO算法对径向基神经网络的超参数进行优化。最后,建立了基于MOPSO-RBFNN模型的钻芯法取样试件抗压强度预测模型,并与其他方法的预测结果进行对比。结果显示,MOPSO-RBFNN模型预测结果的MAE和R2分别为1.311和0.987,误差评价指标均优于其他方法,验证了本文所提出预测方法的有效性。研究结果为混凝土小芯样试件强度预测提供了技术支撑,并对提高实际工程检测的可靠性和准确性具有重要意义。

To solve the problem of difficulty in obtaining standard size core samples using drilling method in areas with dense steel bars or small component sizes in practical engineering,this study proposes a compressive strength prediction method of small core specimens based on radial basis function neural network.Firstly,this paper conducted axial compressive tests on 82 sets of 50mm diameter fine aggregate concrete core specimens and 150mm cube specimens cured under the same conditions and age.Secondly,the strength values obtained from the compressive tests of the two specimens were compared and analyzed,and the MOPSO algorithm was used to optimize the hyperparameters of the radial basis neural network.a compressive strength prediction model for core drilling sampling specimens based on the MOPSO-RBFNN model was established,and the prediction results were compared with those of other methods.The results showed that the MAE and R 2 of the MOPSO-RBFNN model were 1.311 and 0.987,respectively.This paper also verified the effectiveness of the proposed prediction method by comparing the error evaluation indicators of this method with other methods.The research results provide technical support for predicting the compressive strength of small core specimens,and are of great significance for improving the reliability and accuracy of engineering testing.