Punching And Shear Performance Of Recycled Lump Concrete Slabs

Recycling demolished concrete in form of recycled lump concrete with demolished concrete lumps(DCLs)is an effective approach.Recycled lump concrete slabs have been successfully utilized in actual practice,however the punching and shear behavior of such members are seldom reported.To further extend the applications of recycled lump concrete slabs,preliminary investigations on the punching and shear behavior of such members were carried out in this thesis.The main works in this thesis can be summarized as follows:1.Punching tests on seven recycled lump concrete slabs and one conventional concrete slab were conducted,and the influences of the replacement ratio of DCLs,the reinforcement ratio,the thickness and the span of slabs on their punching behaviors were studied.The effectiveness of current design codes in predicting punching capacities of recycled lump concrete slabs were investigated and corresponding correction coefficient is suggested.It can be found that:(1)When the strength of demolished concrete is lower than that of fresh concrete,the initial stiffness of the recycled lump concrete slab is comparable to conventional concrete slab,but the punching shear capacity of the former is slightly lower than the latter;(2)When the reinforcement ratio and span-to-thickness ratio keep unchanged,it is more effective to enhance the slab’s punching shear capacity by increasing the thickness of slab,rather than reducing the span of slab;(3)It is necessary to take the longitudinal steel bars into account while calculating the slab’s punching shear capacity.When the span-to-thickness ratio is relatively small,the prediction method for the punching shear capacity specified in current design codes needs to be further investigated;(4)To ensure that the predicted punching shear strength of the recycled lump concrete slab has comparable safety margin to conventional concrete slab,a reduction factor of 0.95 is suggested to the former.2.Shear tests on six recycle concrete thick slabs and three conventional concrete thick slabs were conducted,and the influences of the replacement ratio of DCLs and the span of slabs on their shear behaviors were studied.The effectiveness of current design codes and theoretical formula in predicting shear capacities of recycled lump concrete slabs were investigated.It can be found that:(1)The failure process and the failure mode of recycled lump concrete slabs are similar to the conventional concrete slab;(2)When the strength of demolished concrete is lower than that of fresh concrete,the initial stiffness of the recycled lump concrete slab is comparable to the conventional concrete slab,and the presence of DCLs does not exhibit a monotonic impact on shear capacity of slabs;(3)The combination strength of demolished and fresh concrete could be directly applied to predict the shear capacity of recycled lump concrete slab,while such approach shows higher prediction accuracy than Muttoni’s formula.3.Three machine learning algorithms,Back Propagation Artificial Neural Network(B-P ANN),Least Squares Support Vector Machine(LS-SVM)and M3 Model Tree were used to establish models to predict the shear capacities of reinforced concrete beams without transverse reinforcement,and the effectiveness of these three models was investigated.These models were also extended to predict the shear capacities of their recycled lump concrete counterparts.It can be found that:(1)Three machine learning algorithms exhibit comparable prediction accuracy to each other but higher accuracy than conventional statistic method.When the prediction accuracy is comparable,LS-SVM model has higher stability and calculating efficiency;(2)When predicting the shear capacities of reinforced concrete beams with a shear span less than two,all three models evidently show higher prediction accuracy than current designs codes,while LS-SVM model is relatively better than the others;(3)For those with a shear span greater or equal than two,generally current designs codes exhibit higher or comparable prediction errors compared with these three models,while LS-SVM models make relatively better prediction than the others.