On Identification of Material Models When Nonrepeatability of Test Data is Present: Application to Textile Composites

Abstract

Engineering test data occasionally violate assumptions underlying standard material model identification. Consequently, one has to apply appropriate remedies with respect to each violation to enhance the reliability of identified material parameters. This paper generalizes the use of the signal-to-noise weighting scheme when heteroscedasticity of test data are suspected. Different mathematical and practical aspects of the approach are discussed. Additionally, the ensuing weighted identification process is simplified to an equivalent standard form by means of a space transformation. Finally, the approach is applied to the identification of a nonlinear material model for textile composites, on both qualitative and quantitative levels.