In this paper, a probabilistic approach is presented to simultaneously evaluate all
strength and additional fitting parameters of a pre-selected composite first-ply failure
model using known uniaxial and multiaxial failure stress states. The method processes
all available mechanical test results in terms of failure stress states as one batch
of data. For parameter fitting it implements the Maximum Likelihood Method and controlled
numerical sampling techniques. The present work introduces the theoretical background
as well as investigates the effect of the selection of the underlying strength distribution
function type. The biggest benefit of this approach is the simultaneous handling of
all available test data as well as the generality in terms of assumed strength distribution
function and failure model type. In addition, the automatic evaluation of the uncertainty
of all strength parameters enables introduction of a safety factor quantification
related to uncertainty in material properties as one of the main contributors to scatter
sources. A local sensitivity-based stress-state determination process is also introduced
to design a set of input failure stresses that guarantee the identifiability of all
parameters thus, ensuring robustness of the fitted constants. The entire methodology
is demonstrated through the example of the Tsai-Wu model by processing experimental
data and by comparing the results against the literature.
