||The explosion in computing power and its application to complex multiphysics problems has led to the emergence of computer simulation as a new way of extending the inductive methods of science. Many fields, particularly combustion, have been greatly changed by the ability of simulation to explore in great detail the implications of theories. But problems have also arisen; a philosophical foundation for establishing belief in simulation predictions, particularly important for complex multiphysics systems where experimental data are sparse, is sorely lacking. Toward the end of establishing such a foundation, a comprehensive philosophical approach to model validation, called instrumentalism, is proposed. A framework for verification and validation/uncertainty quantification (V&V/UQ) of codes is presented in detail, and is applied to a novel entrained flow coal gasification model implemented in the massively parallel simulation tool Arches. The V&V/UQ process begins at the mathematical model. The novel coal gasification model, which utilizes the direct quadrature method of moments (DQMOM) for the solid phase and large eddy simulation (LES) for the gas phase and accounts for coupling between the gas and solid phases, is described in detail. A verification methodology is presented in the larger context of validation and uncertainty quantification, and applied to the Arches coal gasification model. A six-step validation framework is adopted from the literature and applied to the validation of the Arches gasification model. One important aspect of this framework is model reduction, creating surrogate models for complex and expensive multiphysics simulators. A procedure for constructing surrogate response surface models is applied to the Arches gasification model, with several statistical analysis techniques used to determine the goodness of fit of the coal gasification response surface. This response surface is then analyzed using two methods: the Data Collaboration methodology, an approach from the literature; and a Monte Carlo analysis of the response surface. These analyses elucidate regions of parameter space where the simulation tool makes valid predictions. The Monte Carlo analysis also yields probabilities of simulation validity, given input parameter values. These probabilities are used to construct a prediction interval, which can then be used to compute the probability of a consistent simulation prediction.