Abstract Plastic waste (PW) poses a significant threat as a hazardous material, while the production of cement raises environmental concerns. It is imperative to urgently address and reduce both PW and cement usage in concrete products. Recently, several experimental studies have been performed to incorporate PW into paver blocks (PBs) as a substitute for cement. However, the experimental testing is not enough to optimize the use of waste plastic in pavers due to resource and time limitations. This study proposes an innovative approach, integrating experimental testing with machine learning to optimize PW ratios in PBs efficiently. Initially, experimental investigations are performed to examine the compressive strength (CS) of plastic sand paver blocks (PSPBs). Varied mix proportions of plastic and sand with different sizes of sand are employed. Moreover, to enhance the CS and meet the minimum requirements of ASTM C902-15 for light traffic, basalt fibers, a sustainable industrial material, are also utilized in the manufacturing process of environmentally friendly PSPB. The highest CS of 17.26 MPa is achieved by using the finest-size sand particles with a plastic-to-sand ratio of 30:70. Additionally, the inclusion of 0.5% basalt fiber, measuring 4 mm in length, yields further enhancement in outcome by significantly improving CS by 25.4% (21.65 MPa). Following that, an extensive experimental record is established, and multi-expression programming (MEP) is used to forecast the CS of PSPB. The model’s projected results are confirmed by using various statistical procedures and external validation methods. Furthermore, comprehensive parametric and sensitivity studies are conducted to assess the effectiveness of the MEP-based proposed models. The sensitivity analysis demonstrates that the size of the sand particles and the fiber content are the primary factors contributing to more than 50% of the CS in PSPB. The parametric analysis confirmed the model’s accuracy by demonstrating a comparable pattern to the experimental results. Furthermore, the results indicate that the proposed MEP-based formulation exhibits high precision with an R 2 of 0.89 and possesses a strong ability to predict. The study also provides a graphical user interface to increase the significance of ML in the practical application of handling waste management. The main aim of this research is to enhance the reuse of PW to promote sustainability and economic benefits, particularly in producing green environments with integration of machine learning and experimental investigations.
Read full abstract