The　development　of　industrial　production　and　product　updates　boost　the　demand　for　waste　disassembly.　Rule-based　or　heuristic　scheduling　methods　have　been　applied　to　the　job-shop　scheduling　problem　in　disassembly　factories.　However,　dynamic　factors　and　unbalanced　waste　type　distributions　might　influence　the　machine　utilization　rate.　We　proposed　a　reinforcement　learning-based　disassembly　job-shop　scheduling　method　to　optimize　the　dynamic　scheduling　process,　which　takes　into　account　the　knowledge-specific　characteristics　of　disassembly　factories.　We　embedded　the　dynamic　features　and　requirements　of　the　disassembly　process　into　the　design　of　the　environment.　We　incorporated　the　waste　waiting　time　and　machine　utilization　rate　into　the　reward　function　to　improve　job-shop　scheduling　collaboratively.　We　conducted　experiments　in　a　disassembly　factory　layout　compared　to　rule-based　scheduling　methods.　The　experiments　showed　our　method　had　superior　performance　in　the　disassembly　machine　utilization　rate.