Application of a multivariable simplex data analysis system for welding engineers

Abstract This paper describes the development and application of a user-friendly PC based software system (SNAP - Simplex based Numerical Analysis Package). The software is statistically based, employing the ‘Downhill Simplex Method’, and can implement both the Ordinary Least Squares (OLS) method as well as more recently developed Robust techniques. The use of Robust statistics means fewer assumptions concerning the distributional nature of scatter in data need be made and departures from the standard Gaussian distribution can be accommodated. The package also incorporates visual aids which allow the user to examine the structure of experimental data and thus fit the most appropriate mathematical expressions. Various plotting facilities and confidence interval routines are also provided. The benefit of using the SNAP system are described with reference to a re-evaluation of published experimental data. Specifically, the relationships between welding residual stress ( longitudinal and transverse ), weld groove area and heat input variables are examined and quantified for arc welds produced in a hot rolled structural steel. The advantages gained through use of the SNAP system are demonstrated by the development of a greatly improved quantitative description of the relationship between residual stress and other measured variables as compared to the original study.