Sensitivity of truss plate model parameters to parameter determination methods

A factorial analysis was performed to assess the sen­ sitivity of Foschi’s nonlinear structural truss plate per­ formance parameters to parameter determination meth­ ods. The factors considered were treatment of asymmet­ ric failures of joints, the end of the data, and constraints placed on the asymptotic slope parameter M1. Results suggest that the parameter values are sensitive to the end of the data and constraints placed on the asymptotic slope parameter. Estimates made by averaging the two linear variable differential transducer (LVDT) channels for an asymmetric failure of a joint are found to be con­ servative. We recommend a standardized parameter de­ termination method made by averaging the two LVDT channels, including all values of the curve up to the max­ imum load, and setting M1 equal to zero. Further study with other data sets is needed to confirm these findings. Analytical methods have been developed to model the structural system performance of wood structures. These structural models are becoming more and more prevalent as research and design tools. However, im­ proper use of these models can produce misleading and erroneous results. It is important to continually moni­ tor the validity and assumptions in these models so that they continue to be tools that enhance research and design. Foschi has developed a model for the nonlinear per­ formance of truss-plated joints in these system mod­ els (3). In order to use Foschi’s model, one must fit pa­ rameter values to experimental truss plate load dis­ placement curves. The parameters are the constants in a mathematical equation representing the load displace­ ment curve. Several standard test procedures presently exist for testing truss-plated joints (1, 2). However, no standards exist for determining parameter values from experimen­ tally obtained load displacement curves. Least squares nonlinear regression is usually used, but this is only FOREST PRODUCTS JOURNAL Vol. 38. No. 5 part of the total parameter determination method. For example, how does one define the beginning and end of each curve? What restrictions can be placed on the pa­ rameter values? Different researchers currently use different param­ eter determination methods to obtain parameter values. Foschi constrained one of his parameter values equal to zero (3) . Triche (5) fitted his curves up to 0.015 inch of displacement, while McCarthy and Wolfe (4) fitted their curves up to the observed maximum load. If parameter values are sensitive to parameter determination meth­ ods, comparisons between parameter values will not be very meaningful unless standardized parameter deter­ mination methods are established. The comparisons are important for the development of a standardized data­ base of parameter values. The objective of this study was to determine the sen­ sitivity of Foschi’s parameter values to different param­ eter determination methods and to make recommenda­ tions aimed toward a standardized parameter determi­ nation method.