Study Of Processing And Visualizing The Data Of Sealing Coring Inspection Well

This　paper　was　completed　with　the　study　of　the　project　named　as　"Study　of　p rocessing　and　visualizing　the　data　of　sealing　coring　inspection　well"　supported　by　Daqing　Oil　Field　Cop.　The　main　works　focus　on　enclosing　the　method　of　processing　the　core　analysis　data　of　sealing　coring　inspection　well,　drawing　curves　and　statistics　graphics　of　computing　results　and　studying　a　self-adaptive　fuzzy　neural　network　model　applied　to　the　multiple　data　recognition　of　water　flooding　grade.　This　paper's　purpose　is　to　build　a　practical　and　well　human　computer　interactive　operating　environment,　to　compute　correctly　the　core　synthetic　data,　to　visualize　the　core　results　and　to　verify　the　recognition　effect　of　water　flooding　layers　by　utilizing　the　worked　results　of　the　core　data.　Firstly,　the　paper　summarized　the　essentialities　of　core　data　of　sealing　coring　well　in　Daqing　Oil　Field　and　the　status　in　quo　of　processing　core　analysis　data　recently.　I　depicted　in　detail　that　how　to　get　and　apply　the　core　data,　that　deployment　of　inspection　well's　location　and　it's　sealing　coring　technical　requirement.　In　addition,　I　presented　the　operation　process　of　every　disposal　method.　Then　the　paper　emphasized　a　self-adaptive　fuzzy　neural　network　model,　which　can　be　used　to　recognize　water-flooding　grade　of　the　multiple　data.　This　model　is　a　kind　of　forms　combining　fuzzy　logical　theory　and　neural　network　technology,　with　the　character　that　the　model　is　formulated　by　fuzzy　system,　but　the　rule's　extraction　and　membership　function　are　produced　through　self-study　of　neural　network.　Structure　of　the　model　includes　the　regulation　network　and　regulation　fitness　value　network.　Regulation　network　is　presented　by　single　layer　neural　network　and　its　output　reflects　the　results　of　input　specimen　X　calculated　according　to　the　rule　L.　The　structure　of　regulation　fitness　value　network　is　a　four-layer　BP　neural　network,　which　computes　the　fitness　value　of　each　rule　toward　input　specimen　X.　During　the　training　process,　the　system　adopts　mixed　algorithm,　that　is,　the　fitness　value　network　training　uses　unsupervised　Fuzzy　C-Mean　(FCM)　clustering　algorithm　to　get　membership　values　of　specimen　and　the　regulation　network　training　adopts　supervised　Least　Mean　Squared　(LMS)　algorithm　to　adjust　automatically　weight　values　of　every　specimen　input　space,　and　trains　the　input　specimens　to　improve　the　convergence　performance　of　the　system,　and　accordingly,　the　model　parameters　of　weight　value　vector　B　and　clustering　center　vector　Z　are　obtained　after　the　training.　Fuzzy　C-Mean　clustering　algorithm　is　sensitive　to　the　initialization,　that　is,　a　bad　initialization　may　cause　a　very　deviant　result.　This　algorithm　involves　initialization　of　two　principal　parameters.　They　control　the　clustering　number　of　the　specimen　from　many　aspects　and　are　also　chief　factors　affecting　the　velocity　of　FCM　clustering.　Firstly,　the　determination　of　c,　the　number　of　clustering,　is　a　problem　of　clustering's　validity　and　results　of　its　classification　are　related　to　both　specimen　set　itself　and　classification　algorithm.　In　1991,　Xie　and　Beni　defined　a　clustering　validity　function　called　Xie-Beni　index　by　using　object　function　J.　The　function　took　fuzzy　division　of　specimen　set　as　well　as　its　structure　into　consideration,　therefore,　better　results　can　be　obtained　and　the　number　of　clustering　can　be　determined.　Secondly,　optimization　selection　of　fuzzy　weighting　exponent λhas　not　been　resolved　theoretically　at　present　and　generally;　it　is　manually　selected　according　to　practical　requirements.　Generally,　the　value　ofλin　the　interval　of　[1.5,　2.5]　may　meet　the　requirement　of　clustering's　fuzzy　degree.　Ifλequals　2,　not　only　does　the　operation　velocity　rise,　but　the　optimum　clustering　results　can　be　obtained.　So,　we　can　see　that　this　model　combine　merits　of　BP　network　and　fuzzy　logic.　It　possesses　the　characteristics　of　good　adaptabilities,　strong　precisions　and　quick　astringencies.　Subsequently,　the　overall　design　of　this　system　includes　functional　framework　graph,　table　structures　of　core　database,　backup　and　recover　measurement　of　Access　database　and　the　key　sub-programs'　flow　chart　of　core　analysis　data.　Besides,　the　paper　introduced　the　development　of　visualization　in　scientific　computing,　the　visualization　technique　and　the　visualization　software　at　present.　The　visualization　part　of　this　system　consists　of　oil　layers'　watering　status　curves　of　sealing　coring　inspection　well　and　the　statistic　graphs.　The　parts　of　programming...