To　avoid　the　deficiency　of　a　single　failure　criterion　and　deterministic　thresholds　of　limit　states　in　seismic　fragility　analysis　of　underground　structures,　a　method　of　fuzzy　seismic　fragility　analysis　of　underground　structures　considering　multiple　failure　criteria　is　proposed　by　combining　fuzzy-random　theory　and　copula　theory.　In　this　method,　probabilistic　seismic　demand　and　capacity　analyses　based　on　different　failure　criteria　are　first　performed.　Then,　the　fuzziness　of　the　limit　states　of　different　failure　criteria　is　characterized　by　using　membership　functions,　and　the　fuzzy　seismic　fragility　of　underground　structures　considering　a　single　failure　criterion　with　different　membership　functions　is　further　calculated.　On　this　basis,　the　copula　function　and　the　combination　of　membership　functions　are　optimized　to　characterize　the　correlation　between　multiple　failure　criteria　and　establish　fuzzy　seismic　fragility　curves　considering　multiple　failure　criteria.　Taking　the　Daikai　subway　station　as　the　background,　the　effects　of　different　failure　criteria　and　the　fuzziness　of　limit　states　on　the　seismic　fragility　are　discussed,　and　the　optimal　combination　of　membership　functions　that　characterizes　the　fuzziness　of　limit　states　under　the　failure　criteria　of　deformation　and　strength　is　also　presented.　The　results　show　that　the　failure　criteria　and　fuzziness　of　limit　states　have　significant　effects　on　the　seismic　fragility　of　underground　structures,　and　the　combination　of　trigonometric　membership　functions　is　the　optimal　combination　that　characterizes　the　fuzziness　of　multiple　failure　criteria.　Moreover,　the　fuzzy　seismic　fragility　of　underground　structures　considering　multiple　failure　criteria　is　more　sensitive　to　the　fuzziness　of　limit　states　under　frequent　earthquakes　and　basic　earthquakes,　while　the　multiple　failure　criteria　and　the　correlation　between　different　failure　criteria　have　more　significant　effects　on　the　fuzzy　seismic　fragility　considering　multiple　failure　criteria　under　rare　earthquakes　and　extremely　rare　earthquakes.　©　2024　Elsevier　Ltd