Existing　optimal　resource　allocation　for　system　safety　mainly　concentrates　on　series/parallel　systems　or　systems　that　can　be　converted　into　series/parallel　models.　However,　for　some　distributed　complex　electromechanical　systems,　it　is　very　difficult　or　even　impossible　to　refine　them　into　a　series/parallel　model;　in　addition,　the　safety　of　some　system　units　is　immeasurable　because　of　the　coupling　relationship　complexity　in　the　system　composition　structure.　In　this　paper,　a　novel　method　based　on　complex　networks　and　path　set-based　dynamic　programming　is　proposed　for　the　optimal　resource　allocation　for　maximal　safety　of　distributed　complex　electromechanical　systems　with　non-series-parallel　structures.　As　a　measurement　of　the　system　safety,　safety　importance　is　defined,　which　is　a　function　of　two　safety　feature　parameters　-　accident　loss　and　accident　probability.　A　practical　system　is　taken　as　an　example　to　illustrate　and　verify　the　feasibility　and　applicability　of　the　proposed　method.