In　the　large　cable　structure,　the　cable　as　the　main　force　component　bears　a　large　load.　How　to　accurately　analyze　the　influence　of　cable　rupture　on　the　structure　and　give　reasonable　maintenance　measures　has　become　the　key　to　health　monitoring　of　large　cable　structures.　In　this　study,　a　cable　truss　structure　(CTS)　is　taken　as　the　research　object,　the　model　experiment　is　established,　and　the　numerical　analysis　method　of　component　failure　mechanics　is　proposed.　Firstly,　the　component　failure　and　mechanical　response　acquisition　mechanism　are　designed　according　to　the　CTS　experimental　model.　The　measured　parameters　are　used　as　indicators　to　evaluate　the　simulation　accuracy.　In　order　to　improve　the　accuracy　of　finite　element　simulation,　a　transient　dynamic　analysis　method　is　proposed.　Based　on　the　simulation　model　and　calculation　method,　the　mechanical　response　analysis　and　parameter　analysis　of　typical　failure　conditions　are　carried　out.　The　most　unfavorable　failure　mode　is　obtained,　and　the　best　maintenance　measures　for　component　failure　are　given.　Considering　the　time　correlation　of　component　failure,　a　mechanical　response　prediction　method　based　on　convolutional　neural　network　(CNN)-　Bi　long　short-term　memory　(BiLSTM)　optimized　by　improved　particle　swarm　optimization　(IPSO)　is　proposed.　Combined　with　finite　element　simulation　data　samples　and　prediction　methods,　the　mapping　relationship　between　component　failure　and　mechanical　response　is　established.　Finally,　the　continuous　dynamic　analysis　of　component　failure　is　realized,　and　the　most　unfavorable　failure　path　is　obtained.　The　research　results　show　that　the　accuracy　of　the　established　simulation　model　and　calculation　method　is　more　than　95%.　The　improved　deep　learning　prediction　model　significantly　improves　the　prediction　accuracy　and　efficiency,　and　the　analysis　error　and　time　cost　are　reduced　by　7.5%　and　23.7%,　respectively.