Simulation of complex manufacturing processes (like welding) needs to take phase transformations of elastoplastic materials into consideration, in order to account for the microstructural properties in the weld core and HAZ (Heat Affected Zone). Such simulations can predict peak temperatures, residual stresses and weld distortions after cooling. Welding defects like hot cracks can thus be predicted in the virtual design process, instead of correcting them in downstream manufacturing steps.
Multi-phase welding simulations can be conducted using Marc and Simufact, resulting in 3 dimensional computation of stresses, strains and phase fractions.
The heat influx can be modeled as a weld flux generated from a moving weld torch. This transient heat input, which causes the structural response, can be modeled as a 3 dimensional heat source with either (1) gaussian volume distribution or (2) gaussian surface distribution together with a constant key-hole volume source. The thermal boundary conditions can also be specified through weld filler elements, which are introduced at melting point temperatures. This approach of introducing filler material through 'activation-on-the-fly' as the weld torch passes across the work-piece, is typically useful for small deformation welds. Although the solidification process is not modeled completely, we do support latent heat of fusion on the thermal side, and addition of filler materials at annealed state (i.e., zero stress states) on the mechanical side.
However, for large deformation welds, addition of 'quiet' filler elements is often useful. These elements regain their full mechanical and thermal properties as the weld torch moves over the work-piece. It is to be noted, that the transient heat input can be modeled as a combination of moving weld flux and the activation approach of adding filler material at melting temperatures. Finite elements (usually hexahedrons) are used for spatial discretization and the nonlinear FEM algorithm of Marc solver is used as the core engine with its staggered thermo-mechanical coupling capabilities.
In each increment, a thermal analysis its conducted to compute the thermal gradients. Based on the temperature distribution, Simufact determines the time dependent material properties at integration points and phase fractions across the model. The subsequent mechanical analysis computes displacements, and the deformed mesh is passed to the next time step. The thermal loading for the next increment is determined along with the usage of finite difference method to treat the time derivative of the temperature distribution. This numerical discretization algorithm is useful to solve the governing differential equations by approximating them with difference equations in which the finite differences approximate the derivatives. Marc uses an unconditionally stable Backward Euler scheme for this purpose.
The material modeling procedure in Simufact is done through table-defined, temperature dependent mechanical and thermal property definitions like specific heat capacity, conductivity, density, Young's modulus, Poisson's ratio, enthalpy, etc. The flow stress can be defined as temperature-, strain- and strain-rate dependent. In addition, phase transformation data is also taken into account, like CCT and TTT diagrams, volume change per transformation strain, solidus and liquidus temperature, etc.
Such a welding simulation is first developed, and then validated by previously conducted experimental results. Then, the model is used for optimization of processing parameters using Genetic Algorithms (GA).
GA is an adaptive, heuristic search algorithm for solving both constrained and unconstrained optimization problems based on evolutionary ideas of natural selection and genetics. It represents an intelligent exploitation of a random search, and uses historical information to direct the search into regions of better performance within the search space; until it converges to an optimal solution.
GA can solve a variety of optimization problems, where the objective function is discontinuous, non-differentiable, stochastic, or highly nonlinear. In case of welding simulations, the induced distortion is considered to be the objective function and a minimum weld quality is set as manufacturing constraints. Processing parameters like weld torch speed, input current, arc voltage and welding direction are usually set as design variables.
The weld optimization program runs GA to produce a new population of design variables based on simulation results of previously evaluated models. It also checks for the stopping criteria and keeps record of simulation results, current model information and constraint violations in each iteration. In this way, the analysis loop repeats until the best solution does not change over a pre-specified number of iterations.
The combination of Simufact and Marc technologies thus offers the possibility to calculate welding stresses, distortions and evolution of material properties; and can also optimize the processing parameters based on the required weld quality.
For more information, please visit the product websites:
MSC Software – http://www.mscsoftware.com/en-in