Experimental and machine learning study on a novel high-performance hybrid steel-grout connector for cross-laminated timber panels under pre-yield cyclic loads

Abstract

This work presents the results from quasi-static cyclic tests on a novel hybrid steel-grout connector for cross-laminated timber (CLT) panels. These test series are part of an experimental, analytical, and numerical research program to develop reliable and resilient connections for hybrid CLT mass timber structural assemblies. Each designated connector arrangement’s cyclic loading step path has been anchored to the yield point; this latter was obtained as the average of the seven replicates of monotonic tests. From the cyclic test results, mechanical characteristics, namely, the secant stiffness and the residual slip, have been evaluated and discussed. Furthermore, machine learning (ML) models based on deep neural networks have been developed to predict the mechanical characteristics of connectors in the function of the mechanical and geometrical properties of each material used. The developed ML models proved to be able to predict the connector’s stiffness and residual slip and were used to infer the effects of experimental variables on these performance parameters. It was shown that the steel rod and grout diameters are the most influencing parameters regarding the secant stiffness of connectors. As for the residual slip, it was found that the grout diameter and the steel rod strength class are the most influencing parameters. Furthermore, it was observed that a grout-to-rod diameter ratio of about 3.6 enables maximization of the secant stiffness while minimizing the residual slip. Lastly, polynomial equations were developed and found to be able to predict the secant stiffness and residual slip of connectors with a coefficient of determination of 0.99 and 0.98, respectively.