The correctness of model transformation is an import research field in model-driven architecture. Syntactic correctness and semantic consistency are hot topics in the field of model transformation. Syntactic correctness has many mature solutions. However the validation of semantic consistency has some problems. Therefore, how to validate semantic consistency of model transformation is a major problem in model-driven development. In this paper, we propose a validation approach for semantic consistency of model transformation, which is based on pattern. We analyze some patterns in models and make these patterns as transformation pattern. We define transformation rule with transformation pattern and analyze three parts of semantic transformation. We present two theorems to validate semantic consistency of model transformation. Finally, we give a case to illustrate the effectiveness of our approach.