Aiming to the demand for efficiency and flexibility under multi variety and variable batch production and the problem that existing manufacturing units were difficult to support high quality consistency and efficiency of key parts under fragmented orders, a C_max and minimizing the difference of virtual manufacturing cell composition before and after reconfiguration, an improved genetic algorithm based on inherited reconfiguration decoding strategy was proposed, and a similarity calculation method between orders under known cell configuration and the original manufacturing cell was designed. It ensures the maximization of the manufacturing cell and carries out reconstruction by inheriting the configuration of the original manufacturing cell. Finally, the validity and feasibility of the proposed model and algorithm are verified by the actual production data of a photoelectric observation product.