Safiuddin Md, Raman Sudharshan N, Abdus Salam Md, Jumaat Mohd Zamin
Angelo Del Zotto School of Construction Management, George Brown College, 146 Kendal Avenue, Toronto, ON M5T 2T9, Canada.
Department of Architecture, Universiti Kebangsaan Malaysia, UKM Bangi 43600, Selangor, Malaysia.
Materials (Basel). 2016 May 20;9(5):396. doi: 10.3390/ma9050396.
Modeling is a very useful method for the performance prediction of concrete. Most of the models available in literature are related to the compressive strength because it is a major mechanical property used in concrete design. Many attempts were taken to develop suitable mathematical models for the prediction of compressive strength of different concretes, but not for self-consolidating high-strength concrete (SCHSC) containing palm oil fuel ash (POFA). The present study has used artificial neural networks (ANN) to predict the compressive strength of SCHSC incorporating POFA. The ANN model has been developed and validated in this research using the mix proportioning and experimental strength data of 20 different SCHSC mixes. Seventy percent (70%) of the data were used to carry out the training of the ANN model. The remaining 30% of the data were used for testing the model. The training of the ANN model was stopped when the root mean square error (RMSE) and the percentage of good patterns was 0.001 and ≈100%, respectively. The predicted compressive strength values obtained from the trained ANN model were much closer to the experimental values of compressive strength. The coefficient of determination (²) for the relationship between the predicted and experimental compressive strengths was 0.9486, which shows the higher degree of accuracy of the network pattern. Furthermore, the predicted compressive strength was found very close to the experimental compressive strength during the testing process of the ANN model. The absolute and percentage relative errors in the testing process were significantly low with a mean value of 1.74 MPa and 3.13%, respectively, which indicated that the compressive strength of SCHSC including POFA can be efficiently predicted by the ANN.
建模是预测混凝土性能的一种非常有用的方法。文献中现有的大多数模型都与抗压强度有关，因为它是混凝土设计中使用的主要力学性能。人们进行了许多尝试来开发合适的数学模型以预测不同混凝土的抗压强度，但对于含有棕榈油燃料灰（POFA）的自密实高强混凝土（SCHSC）却没有。本研究使用人工神经网络（ANN）来预测掺入POFA的SCHSC的抗压强度。本研究利用20种不同SCHSC混合料的配合比和试验强度数据开发并验证了ANN模型。70%的数据用于对ANN模型进行训练。其余30%的数据用于测试该模型。当均方根误差（RMSE）和好模式百分比分别为0.001和≈100%时，停止对ANN模型的训练。从训练好的ANN模型获得的预测抗压强度值与抗压强度试验值非常接近。预测抗压强度与试验抗压强度之间关系的决定系数（²）为0.9486，这表明网络模式具有较高的准确度。此外，在ANN模型的测试过程中发现预测抗压强度与试验抗压强度非常接近。测试过程中的绝对误差和相对误差百分比均显著较低，平均值分别为1.74MPa和3.13%，这表明ANN可以有效地预测含POFA的SCHSC的抗压强度。
