Task 1
Development of a robust and accurate numerical methodology for thermomechanical parameter identification using temperature and strain full-field measurements.
I. The thermo-mechanical material parameter identification with the a non-linear FEMU methodology, however, using temperature and strain full-field measurements
This methodology minimizes a cost function that expresses the discrepancy between the experimental and the numerical response. The numerical evaluation is performed through FEA [19,20] and the optimization strategy will use a least-square Levenberg-Marquardt algorithm and a differential evolution (DE) algorithm [7], avoiding initial solution dependence. The constitutive models will be implemented in the FEA commercial program Abaqus by means of a user routine. The project team has already a large number of constitutive models already written in Abaqus UMAT user routines. Nevertheless, very recent some constitutive modes developed for high strength steels must be developed.
II. The development of an original methodology with the virtual field method (VFM), based on the experimental strain and temperature fields
The extension of the VFM to additionally use the temperature and strain-rate fields is required for this task, being one of the major challenges of this task (and, consequently of the project). This project extend the VFM to non-linear thermomechanics. It is a very different approach from the FEMU because it uses the principle of virtual works to find the set of parameters and introduces the experimental data directly in this principle (that balance the external with the internal forces). While the FEMU uses external observations to find the parameters, the VFM uses internal balances. The results obtained, including methodology robustness, will be compared with the ones of the previous task.
III. The development of an automatic numerical methodology for material parameter identification, integrating both previous methodologies
The coupling of both the FEMU and VFM methodologies developed in the previous subtasks, as an integrated multi-objective optimization methodology, is the goal of this task. It is expected that this FEMU-VFM methodology gathers the advantages of both methodologies, solve or substantially decrease the problem of solution non-uniqueness and that it is less sensitive to experimental noise.
Both FEMU, VFM and coupled methodologies will be previously validated using virtual experimental data (obtained with known material parameters) for several heterogeneous experimental tests and constitutive models.