Shakedown analysis of PET blends with demolition waste as pavement base/subbase materials using experimental and neural network methods

Recycling and reusing construction and demolition (C&D) wastes for civil engineering construction activities has been identified as an energy-saving and sustainable solution. The purpose of this study is to evaluate the deformation behavior of two types of C&D waste materials, namely recycled concrete aggregate (RCA) and crushed brick (CB) when mixed with polyethylene terephthalate (PET) plastic waste. RCA and CB were mixed with 1%, 3%, 5%, and 7% of PET, and the permanent deformation behavior of the blends was evaluated using repeated load triaxial (RLT) test. The shakedown criterion was utilized for identifying the deformation behavior of blends. Most of the PET/RCA and PET/CB blends exhibited Range B (plastic creep) and Range C (incremental collapse) response, respectively, in the investigated stress levels. Shakedown analysis of the test results indicated that up to 3% PET could be mixed with RCA for base/subbase applications, while CB should be mixed with 1% PET, in the subbase layer. Artificial neural network (ANN) method was next used to simulate the permanent strain and shakedown behavior of the blends with consideration of the physical properties and stress states. The ANN model was found to be highly efficient for simulating the permanent strain graph and identifying the shakedown behavior of the blends. A sensitivity analysis was subsequently performed to investigate the impact of input variables on the permanent deformation behavior and the results indicated that number of cycles and confining stress were the most important factors.

Automated summary

Recycling and reusing construction and demolition wastes for civil engineering construction activities is an energy-saving and sustainable solution. This study evaluated the deformation behavior of recycled concrete aggregate and crushed brick mixed with PET plastic waste, finding that up to 3% PET could be mixed with RCA for base/subbase applications. The ANN model proved to be highly efficient for simulating the permanent strain graph and identifying the shakedown behavior of the blends, with number of cycles and confining stress identified as the most important factors.