Assembly retrieval techniques are an important part of improving the reuse of assembly models and enhancing industrial productivity. To speed up the retrieval rate， the assembly similarity retrieval model with parts encoding technique and Hausdorff distance is proposed. Firstly， the assembly parts attributes are encoded to model the assemblies into a numerical set； then the distance between assemblies is calculated using the improved Hausdorff distance， and the similarity level is defined by the distance； finally， the validity and effectiveness of the method are verified by specific examples.