Present work reports the development of a computational methodology that not only depends on the phenomenological laws but also able to identify uncertain input variables. The major efforts are put forward to develop conduction and transport phenomena based analyses, in weld pool and its surrounding area using finite element method. The conduction heat transfer based analysis is developed using a novel concept of adaptive volumetric heat source where the source term is defined in an adaptive manner without a-priori definition of its geometric shape and size. The numerical process model is integrated with a genetic algorithm based optimization scheme to identify the most suitable values of a set of uncertain input parameters through an sensitivity of the error in the predicted values of the weld dimensions. Moreover, the integrated process model is employed for the prediction of a multiple sets of process parameters for a target weld dimensions. The reliability that may be associated with the predicted values of the process parameters for a target weld attribute is also described.