This study investigates the deformation properties of recycled concrete aggregate (RCA) when blended with up to 70% of recycled glass (RG) for pavement base applications. A multi-stage repeated load triaxial (RLT) testing procedure was proposed and utilized for evaluating the permanent deformation behavior of RCA/RG blends. The resilient modulus (M_r) of the blends was examined by performing RLT test in different stress combinations using a proposed testing protocol. The shear strength response of the blends was also investigated. Shakedown theory was utilized to classify the permanent deformation behavior of the blends. Except for the RCA30/RG70 blend, all other blends exhibited either Range A or Range B response in the investigated stress levels. There was an increase in the permanent strain and a decrease in the M_r of blends as the RG content increased. The shear response of the blends exhibited a strain-hardening behavior in the post-peak zone when the RG content was more than 10%. Artificial neural network (ANN) models were developed for predicting the deformation properties of the blends and examining the effect of test variables on the deformation properties. The developed ANN models for prediction of permanent strain and M_r were converted to practical equations for pre-design purposes. Results of numerical modeling indicated that ANNs were robust for predicting the deformation properties as well as identifying the impact of input variables on the deformation properties of RCA/RG blends.

这项研究调查了再生混凝土骨料（RCA）与高达70％的再生玻璃（RG）掺合用于路面基础应用时的变形性能。提出了多阶段重复载荷三轴（RLT）测试程序，并用于评估RCA / RG共混物的永久变形行为。弹性模量（M _r通过使用建议的测试规程在不同的应力组合下进行RLT测试来检查共混物）。还研究了共混物的剪切强度响应。利用减振理论对共混物的永久变形行为进行分类。除了RCA30 / RG70共混物外，所有其他共混物在所研究的应力水平下均显示出范围A或范围B的响应。永久应变增加，M _r减小RG含量增加，共混物的数量增加。当RG含量大于10％时，共混物的剪切响应在峰后区域表现出应变硬化行为。开发了人工神经网络（ANN）模型来预测共混物的变形特性，并检验测试变量对变形特性的影响。已开发的用于预测永久应变和M _r的ANN模型被转换为实用的方程式，以进行预先设计。数值建模结果表明，人工神经网络在预测变形特性以及识别输入变量对RCA / RG共混物变形特性的影响方面具有鲁棒性。