Identification of Mechanical Properties of Base Metal and Weld Metal for a T Joint Using FRF Data

In this paper, the mechanical properties namely Young's module and structural damping coefficients of a welded T joint's members and weld metal are identified by implementing model updating technique using FRF data. The model updating is performed by using artificial neural networks. The gap between joint members and also different material properties for HAZ, weld metal and base metal are considered to construct the finite element model of the weldment. The training sets are frequency response functions and the targets are Young's modulus and damping coefficient. Training sets for the network are obtained by modal analysis of the T joint's finite element model in free-free condition using different Young's modulus and damping coefficients (by varying theses parameters in a gradual manner). By applying the frequency response functions that are obtained from experimental modal analysis of the T joint in free-free condition to the trained network, Young's modulus and damping coefficient are identified. The identified properties for base metal show good agreement for the published data for the investigated material (here, ST-52 steel). Also, the identified values for the weld metal are greater than base metal that is in agreement with published results. These results elucidate the ability of the procedure in successfully identifying the mechanical properties of the weldments.