Finite Element Analysis of Unbraced Structural Wood I-Joists Under Construction Loads

The research summarized the experimental analysis and finite element modeling of the lateral and rotational response of unbraced wood composite I-joists to worker loads. All experimentation and modeling was conducted on simply supported I-joists varying from 11-7/8 inches to 14 inches in depth and 20 feet to 24 feet in length. I-joists were subjected to static and dynamic loads. The deflections of the top and bottom flanges, as well as the rotation, were measured or calculated at both one-half and one-quarter the span length. The overall goal of this project is to accurately model the lateral and rotational displacements caused by human load effects.

I-joists were first tested statically by subjecting each joist to a three point bending test, free from all lateral restraints. This test was necessary to prove that the performance of the joists was repeatable. Lateral and rotational stiffness of the joist were calculated at one-half and one-quarter of the span length. The static experimental tests results were statistically analyzed using an analysis of variance (ANOVA) test. The results from this analysis indicated no difference between repetitions of the same joist; however, the test did indicate that there was a significant difference between joists of the same manufacture and size. Dynamic testing was then conducted. Dynamic loads were induced by having test subjects traverse each I-joist. The resulting loads induced at the top and bottom flanges were recorded for use in the finite element model. The lateral deflections and induced loads were compared to the static weight of the test subject and analyzed with an ANOVA test. The results indicated an increase in both the induced load and resulting deflection with an increase in weight. The analysis also indicated an increase in load and deflection with a decrease in lateral and rotational joist stiffness.

The recorded load values from the dynamic test were used as inputs into a finite element model. The resulting lateral deflections of the midpoint and quarter point were generated. The rotation of the beam was calculated from the difference between the top and bottom flange. Experimental results and finite element model results were compared by calculating a running average of the error between the acquired data and the finite element model. The model was said to be valid until the average model error reached 10 percent of the maximum acquired test value. All six deflection readings were analyzed in this manner. The percent of beam at which the model no long represented the test data was determined for each data set. This point was averaged across all deflection readings of similar joists and across all data sets of the same joist type. The model predicted the 20 foot long 11-7/8 and 14 inch deep joists until 54.5 percent and 51.2 percent, respectively, of the beam completed by the test subject. However, the 24 foot long 11-7/8 inch deep joist was only accurate to 31.2 percent of the beam completed by the test subject.

Differences in peak values, and the time at which the peak values occurred were also analyzed using an ANOVA test. There was a significant difference between the peak values of the acquired test data and the deflections generated with the finite element model. However, there was no significance within the time that the peak values occurred between the model and experimental results.

A simplified pseudo dynamic analysis was conducted using a constant percentage of the test subject's static weight applied to the top and bottom flange. This approximation proved adequate for the lateral displacement and rotation of the 11-7/8 inch and 14 inch deep and 20 foot long I-joists. However, the model became un-conservative for the 11-7/8 inch deep and 24 foot I-joists.