| Abstract Scope |
Introduction:
Spot welding is integral to automotive construction. A typical automobile body-in-white design may consist of 10s to 100s of spot-welded metal stack-ups, including a range of sheet metal types, gauges, and material combinations. Modern techniques for characterizing vehicle performance (stiffness, NVH, crash) requires that each one of these stack-ups be characterized for mechanical performance. Historically, joint performance characteristics for each stack-up have been assessed experimentally. Such testing however is extremely time-consuming and cannot keep up with the rapid changes occurring in both the characteristics of the sheet materials as well as intended stack-ups in the design. As automotive manufacturers push for reduced time to market, development of the necessary stack-up specific performance data becomes more difficult. Recently, the application of design of experiment (DOE) methods have been used as a tool decrease the testing required to obtain critical data. However, traditional DOE approaches still address only the stack-ups included design, and cannot be extrapolated. To address this challenge, a new novel approach to extrapolate data from such designs of experiments to future stack-ups. This work has investigated a method of collecting data based on specific stack-ups, and mapping those results onto a new design space where material strength and gauge for the component materials can be assessed independently. This then creates tools that allow prediction of stack-up performance not included in the original experiment and entire reference databases to be created.
Experimental:
In this study a design of experiment was created including stack-ups with a diversity of material strengths and thicknesses. This range was selected to build the data-bases necessary for mapping onto the expanded material characteristics space. The created DOE included 2 testing configurations (tensile and peel), 2 weld sizes (4√t and 5√t) and 12 separate stack-ups. Cross tension and tensile shear testing were employed here consistent with existing automotive performance databases. The resulting design of experiment was a specialty design, consisting of 24 trials. Samples were then prepared and testing conducted in accordance with the DOE. Regression modeling and extrapolation of the results provided performance attribute data for all 12 stack-ups for the two test configurations and weld sizes. This included 48 individual performance attributes.
These results were then used as the input for a second regression curve fit. For this second curve fit, a general linear model was created that used as input variables component material strength and gauge, relative weld size, and the test configuration as its input variable. For this analysis steels ranged in thickness from .65mm to 1.6mm, and strengths from 270MPa to 1180Mpa. Analysis using this new general linear model created a map that allowed prediction of cross tension and tensile shear loads based on the strengths and thicknesses of the two attached steels along with the relative weld size.
Results and Discussion:
Initial curve fitting of these results provided a regression equation showed that the measured joint strengths correlated with the input variables with an R2 value of 89%. Significant factors include the specific stack-up, stack-up2 test configuration, weld size, and one two factor interaction, test configuration*weld size. This model was then used to create a new database for analysis, where each stack-up was decomposed into the component sheet thicknesses and strengths. This resulted in a 48-trial matrix. Regressing this new dataset against the expanded list of variables resulted in a correlation of 98% R2. The resulting model included independent terms for each of the main effects except one material strength. In addition, interactions between material strengths and gauges, as well as weld size and test configuration were noted. This then created the model that offered potential for predicting the performance of stack-ups with a wide range of material strengths and thicknesses.
Conclusions:
This technique of generating data for a defined set of stack-ups onto a space where the substrates are separated is a powerful first order tool for understanding the impact of newer materials and configurations on vehicle design. This analysis methodology further offers potential for continuous database expansion based on adding newer grades and a wider range of gauges. Further, this tool in combination with artificial intelligence strategies offers potential for quickly acquiring and analyzing data facilitating rapid manufacturing. |