Statistical joint evaluation of fracture results from distinct experimental programs: An application to annealed glass

In this work, the generalized local model (GLM), as a procedure to derive the Weibull primary failure cumulative distribution function (PFCDF) related to an adequate reference parameter is applied to predict failure of annealed glass specimens from two different experimental test types. The procedure allows a unique PFCDF, identified as a failure material characteristic, to be inferred irrespective from the test data being evaluated either for each single test type sample separately or for a unique sample pooled from both single samples as a whole, whereby the particular specimen features pertaining to the two samples concerning specimen shape and size are considered in the assessment. Nevertheless, higher reliability is ensured in the parameter estimation for the case of pooled evaluation as a result of the higher number of specimens being jointly evaluated. In this way, once a suitable generalized parameter is selected, failure of specimens or components may be reliably predicted independently of the specimen shape and size chosen for the tests. The quality of the generalized parameter selected can be also discerned using this methodology. The applicability of the approach proposed is demonstrated by a practical example comprising two different test types on glass plates.