A remanufacturing parallel disassembly sequence planning (RPDSP) method driven by multiple failures was proposed in order to overcome the influence of connection failure and component non-disassembly caused by failures on the disassembly sequence. A hierarchical multiple failures transfer chain model was constructed through analyzing the non-detachability and failure characteristics of the parts to describe the relationship among the failure type, failure degree, scrap level, recycling decision, and disassembly mode of parts. Furthermore, the model was formally described in mathematically language and mapped into a polychromatic graph model of multiple failures information transfer chain based on the polychromatic sets theory, which is convenient for computer programming and manipulating. On this basis, the disassembly mode of parts can be identified by polychromatic reasoning. The information of part failure was integrated into the disassembly information model, and then a destructible disassembly information model was established, and the disassemblability formula was derived. A multi-layer chromosome coding method including disassembled node layer and destruction constraint layer was proposed to represent the partial destructive information of the product. An initial population acquisition method considering the part recycling decision was designed to eliminate unreasonable sequences. The RPDSP was realized by chromosome evolution rules such as selection, crossover and mutation operator. Finally, a case study of washing machine was used to demonstrate the feasibility and effectiveness of the proposed method.