Bayesian parameter identification for empirical model of CLT connections. (10th September 2019)

• Record Type: Journal Article
• Authors: Cao, Jixing; Xiong, Haibei; Chen, Jiawei; Huynh, Alexandre
• Is Part Of: Construction & building materials. Volume 218(2019)
• Page Start: 254
• Page End: 269
• Publication Date: 2019-09-10
• Subjects: CLT connection -- Hysteresis curve -- Bayesian parameter identification -- Cramér–Rao lower bound -- EEEP curve; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2019.05.051
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: Failure modes of three types of CLT connections are presented. Bayesian method is employed to identify the Pinching4 model parameters. The Cramér–Rao lower bound is derived as the estimate uncertainty. The bounds of the estimated response are investigated. Abstract: Cross-laminated timber (CLT) is a popular construction material that is widely used in timber buildings. In this study, the mechanical properties of three types of CLT connections were investigated: (i) an angle bracket joint; (ii) a crossed self-tapping screw joint; and (iii) a simple butt joint with self-tapping screws. Three repetition tests for each type of CLT connection were conducted under two different cyclic-loading protocols. The experimental results showed that pull-out failure in the tension direction was dominant, whereas wood crushing and screw failure were typical failure modes in the shear direction. An empirical model was used to simulate the nonlinear behavior of the CLT connections under cyclic loading, where the pinching, cyclic stiffness, and strength degradation were considered. To estimate the model parameters associated with the uncertainty of these mechanical properties, the Bayesian approach was proposed, and the Cramér–Rao lower bound was derived as the theoretical lower bound. The parameter-identification results were evaluated according to the elastic stiffness, yielding point, and ductility via equivalent energy elastic–plastic analysis. Comparison of these mechanical properties between the test results and model estimates revealed good agreement. Additionally, the bounds of the estimated response were investigated according to the parameter variance. In comparison with the experimental data, the error bars covered two standard deviations, indicating the reliability of the estimated results. The values of the identified parameters associated with the standard deviation provide a good reference for further analysis when such types of CLT connections are incorporated into a finite-element model of a structural system. The proposed method of Bayesian parameter identification is also suitable for assessing other types of empirical models.