Izifunction zokuzinza eziphakanyisiweyo zombili zezinkampani zokuphakanyisa izinsimbi ziyisiphi

Ukuthintana Ngokushesha Kwe-Real-Time Okubhekene Nokuguqula Kwemithambo Ngokusebenza Kwe-AI

Kungani ukuguqula Kwemithambo kuhambisana nokuphatha Kwe-mashini Yezikhumba

Imiphumela Yewhiri—kufakwa kwezindawo zokugcina, ubukhulu Bekhona, nokuphatha Kwemithambo Yomhlaba—kuguqula Ngokuzenzakalelayo Phakathi Kwemikhumba Nezibhanki. Imashini Yezikhumba Ezingaphansi Ziyasebenzisa I-Cam Profiles Ezilungileyo Neziquketho Zokugcina Ezilungileyo. Uma imithambo iyaguqula Kusuka Kwezifunda Ezilungileyo, le mashini ayikwazi Ukuthintana, okuvumela Ukuguqula Kwemikhumba Yezikhumba, Ukuphatha Kwemiphumela, nokuguqula Kwemiphumela Yobukhulu. Abasebenzi bayimela Ukusebenza Ukuze Bafaka Ngokushesha Izinto Ezilungileyo—le nto iyavumela Ukuphatha Kwemikhumba Nezinto Ezilungileyo, okuvumela Ukuphatha Kwemikhumba Nezinto Ezilungileyo. Kulezinhlangano Ezinamakhulu Amashumi Amabili, noma kuphela Ukuguqula Kwemithambo Ngokushesha Kwe-2% kungavumela Ukuphatha Kwemikhumba Emakhulu Amashumi Amabili Ngamahora, okuvumela Ukuphatha Kwemikhumba Yobukhulu (OEE). Uma akukho Ukuthintana Kwemithambo, imashini iyasebenza Ngokushesha Ngokulandela Izifunda, okuvumela Ukuphatha Kwemikhumba Nezinto Ezilungileyo.

Indlela Eshicilelwe Ngayo Kwe-AI Models Iyasebenzisa I-Metallurgical Data Neziquketho Zokugcina

Amashini ephakathi kwezinsuku zokuphambuka zasemva ayisebenzisa imodeli ye-AI esichanise idatha yezinsimbi zokubhalwa ngokushesha ngezifundo zokusebenza kwezinkabi. Amashini okubala ukunquma, amashini okubala ubude lwesikhala sokugcina, namashini okubala ukuthempurechura ayibhalwa ngokushesha ukunquma, ubude lwesikhala sokugcina, nokuthempurechura kusonke ispring. Le datha yezinsimbi eziningi iyabhalwa kumkhombi wokufunda ngokusebenza kwezinkabi okwenzelwe ngokuphelele kubasebenzi bezinkabi zokubhala. Imodeli iyabona izinguquko ezincinci zokunquma kwezinkabi noma ukubuyela kwespring futhi iyashintsha ngokushesha amaxesha we-servo—yoshintsha ubude lwesikhala sokugcina, ubude lwesikhala sokubhala, noma ubude lwesikhala sokuvula. Isibonelo, lapha ukuthempurechura okukhulu kubhalwa, i-AI iyongeza ukubuyela kwezinkabi ukuze kubuyiselwe ubude lwespring lomkhombi—ngaphandle kokuthi umsebenzisi ayithathe indlela yokusebenza. Lokhu kuchanise izinsimbi kushintsha izinhlelo zokusebenza ezingenakho ukushintshwa ngokuzenzakaleyo ngokusebenza kwezinkabi ezihlukahlukene njengokusebenza kwezinkabi ezihlukahlukene.

Ukuthembekile kokuqondwa: Ukusetha kungafanele kusukela kwezinkabi zokubhala kwezistayinilesi kungaba kancane kunezinkabi zokubhala zomthwalo kwezilayini zama-German OEM

Umphakeli wemoto waseJalimani wasebenzisa umkhiqizo wokulungisa ngokusebenzisa i-AI kumhleli wokugcina amaspring wevalve, okwakuyisisekelo sesithili yezinkabi (1.4310) ne-wire yomculo (EN 10270-1). Kablombe kusetyenziswa kulo, ukuguqula imateriali kubhekisele kuukhishimela zokuphatha ngokusheshayo nokuzama izikhumba eziningi, kubhalwa isikhathi sokufika kwe-45 miniti. Ngokusebenzisa i-AI engaphakathi nokubonakala kwempilo yemetalu ngokushesha, umhleli wabonakala ngokuzenzakalelayo ubuchwepheshe be-wire futhi wabuyekelela ngaphansi kweminithi e-15—lokhu kukhupha isikhathi sokuhlela ngokuphakeme ngo-68% (2024). Isivinini sokugcina kokuqala sibhala kusuka ku-93% ukuze sifike kwi-99.2%, kubhekisele kuvuleka kokugcina ngokushesha. Lokhu kubhala ukuthi umhleli ungenza ukuthi ukungafanele kwebuchwepheshe kube yindlela yokuthuthukisa indlela yokusebenza, kungakhathalisi ukuthi kubhekisele kubona ukuthi kungafanele.

Ukuthuthukisa ukuthi umhleli uzobonakala ngokushesha ngokusebenzisa i-IoT ukuthi akhange ukuthi kufanele ukuthi kubhekisele kubona ukuthi kungafanele

Izindleko zokungafanele kokuqala kwezinhlobo zezikhumba zokuphakama

Ukungafiki kwezikhathi ezizimiswe zikhotha amaphrofayili ngokupheleleyo ama-$260,000 ngamuhla (Aberdeen Group, 2023). Kuma-zikhathi ezingaphezulu zokusebenza kwezinkamba, umuhla wokungasebenzi kungabangela ukuthi kubhishwe izinkamba eziningi kune-30,000 futhi ukuthi kufutshane isikhundla sokusebenza esingasiza ngeminyaka yokubuyisa. Nakuba ukugcina okuzimiswe kwezinto ezizimiswe kungavikela ukungasebenzi okukhulu kakhulu, kuyakwenzela ukuthi kwenziwe izindlela ezingafanele; ama-68% wezindawo zihlala zisebenzisa ngokupheleleyo izindlela zokuziphatha ngokushesha, okubangela ukuthi kuzameka ukugcina ngokushesha kakhulu. Lezi zikhathi zokungasebenzi ziyakhiqiza ukungasebenzi kokuphatha kodwa ziyakwazi ukuthi kuzameke ukuchaza okupheleleyo emva kokubuyisa ukuhlola ukuthi izinkambu ziyatholakala ngokupheleleyo. Ukugcina okuphathelene ne-IoT kubuyekeza leli phansi ngokuthi kutholakala izinkinga ezizimiswe zokusebenza kwizithuthi—kukhulisa ukuthi kuzameke ukugcina okuzimiswe nokuthi kuzameke ukuthi kuzinikezwe izinkambu ezizimiswe ukuthi kungavikeli ukungasebenzi okukhulu kakhulu.

Ukuchaza kwe-edge: Ukuvumela ukuthi kuchazwe ukuthi kushele ukuthi kushele, ukuthi kushele kwezintambo, nokuthi kushele kwezilubricants ngamathegi ayo aphinde akhangele i-PLC

Ukuqondisa kwe-analitiki ekupheleni kwenkqubo ikhupha idatha yezinsuku zokusha kwezinhlobo zemithambo yokuqinisa, amakhamera we-thermal we-infrared, nemithambo yokuhlola imidlalo yelubricant ephakanyisiwe ngokuthe ngaphandle kwezikhumba zonke zemashini yamashini. Lezi zisignali zifuna ku-gateway eyingxwela i-PLC eyasebenza ngokulocala ngokusebenzisa imodeli yokuqonda ukuthi kuyini okungcono (machine learning) ephakanyisiwe kuzo izindlela zokufa zempilo zezinsuku zokuzala. Inkcukumo ikhombisa ukungafani kwe-shaft, ukugula kwemibhobho, nokugula kweloyi—kuvumela amazwi okuhlela kabusha ngaphambi kokuthi izindlela zokuhlela zizwe. Isibonelo, ukuphanda kwe-2% kwe-amplitude yokuqinisa kwe-coiling spindle ikhombisa ukungafani kwezinto zokusebenza ngokusha kusukela ema-48 amahora, kuvumela ukuthi kwenziwe ukulungisa ngemizuzu e-mihlanu endaweni yokuthi kufanele kwenziwe ukulungisa ngemahora e-nhi. Le nkqubo yokuqhubeka ngokuphinda-phinda iyehla indawo yekhumbi ngokuphela kwe-50% futhi iyanciphisa izindleko zokulungisa ngokuphela kwe-65% (Deloitte, 2023).

Ukubuka Kwisikhathi Esikude Ngokuphelele Ngokuphambili Nokunika Ukuthi Akukho Mphumela Wokufa

Izinkinga zokubuka ngokuphelele ngokuphambili kusukela ema-400 CPM futhi ukuthi kanjani u-vision AI uyabuyela

Kusukela kumakhishini okungafana ne-400 ngomzuzu, ukubona ngokwesibonelo akuyiyo yisibonelo esihle. Abasemthethweni bafatshwa kusho, bahlupheka amaphakathi amancane embono, futhi abahlupheka ukubona ukuguqula kwemithambo ekhuphuka ngezikhathi zokusebenza. Amaphayile ayasho ukuthi abasemthethweni abanamandla bangahlupheki amadefekthi angu-10% kulezi zikhathi—okubangena kwezinkinga zokubuyisa kwindawo yokusebenza nokubuyisa kuzindawo zokubuyisa. Ukubona ngokuchaza kufuna ukubona kwezinkinga eziningi zokuhlupha izidefekthi. Ukuhlanganiswa kwe-Vision AI kuhlanza le ndlela ngokubona yeyintsha ispring ngokugcina isikhathi sokusebenza. Ukubhalisa nokuchaza yonke indawo ngaphezu kwe-20 milliseconds, lezi zinhlelo zinikeza ukubona kwezidefekthi ngaphezu kwe-99% futhi ukubona kwezidingo ngaphezu kwe-98.5% (2025), nokubona ngokushesha kwe-yesi/cha. Ngokwenza ukubona kwezidingo kwezindawo zokusebenza, ukubona kwezidingo kwe-AI kususa ukudlula kokubona kwezindawo zokusebenza—kunika ukubuyisa kwaye kuvumela ukubuyisa kwezinto ezingavumelekile kwaye kuvumela ukubona kwezidingo ngokuphelele.

Ukubona kwe-ResNet-18 okufakwe kuleli qembu olwakhiwe kuma-2.7M wezithombe zesi-spring ukubona ngaphezu kwe-millisecond

Ukuqinisa ukuthi kunezinto ezingavumelekileyo ngokupheleleyo, izimboni zezinqamuzana zezinsuku zasezantsi zihlanganisa i-ResNet-18 deep learning classifier ngqo kwe-edge processor yomshini. Kwenziwe ngemitholakalo yezinqamuzana ezingu-2.7 million eziyisilinganiso, ekhokhela izinto ezingavumelekileyo ezingaphezu kwezingu-50—kubukwa kwezinkambu ezingamnye kuya kumaqondiso wezinqamuzana—umodeli uhlola yonke into ngaphansi kwe-0.8 milliseconds. Emisebenzini yezinsuku, uyafika ema-99.7% yokuqinisa nokugcina izinto ezingavumelekileyo ngaphansi kwe-0.1%. Ngoba ukuchaza kwenziwa ngokwesibonelo kwe-PLC-integrated gateway, kungahlinzeka ukuthi kubhekwe kwe-cloud futhi kuvunyelwa ngokushesha ukukhipha izinto ezingavumelekileyo. Umodeli wena futhi uhlaziya ngokuphathelene ngezinto ezingavumelekileyo eziyisilinganiso, uguquka ngokuphathelene nezinto ezingavumelekileyo eziphatha ngokushesha. Umphathi wokuphatha umthwalo wabona ukuthi kukhishwe ngama-38% kwezinto ezingavumelekileyo futhi ukuthi kususwakala ngokupheleleyo ukubuka ngokwesibonelo kwezinto ezingavumelekileyo emva kokufaka (2025).

Ukuphatha okulinganisa ngokwemitholakalo yokuhlanganisa ukulandela i-ISO 2690

Ukuguqula kwesikhathi noma ukudlula kwe-servo njengamathuba amabi okungalingani

Izindlela ezincinci zokuguqula kwezintemperatura ziyaguqula isakhiwo sezinsimbi zomshini wokwenza izinsimbi, kubuyisa izindawo zezigidi nezindawo zokugcina ngemikhromu ethile. Ukungafani kweservomotor kuyadlula le miphumela: njengoba imotolo yamandla ayagwala, izinhlobo zokuphendula ziyaguguza, kubuyisa ukuthumela kwesihlangu nezinkokhelo zokugcina. Lezi ziguqulo ziyagcwalisa ngokuphakeme kwezindlela zokuproducwa, kwenza ukuthi akukwazi ukugcina izintemperatura eziningi kune ±10 µm ngaphandle kokuthinteka. Umgcino wokufunda ubonisa ukuthi ukuguqula kwezintemperatura kuphela kungahlela ukuguqula kwezinhlobo ngokuphakeme kwe-0.3% ngamuhla kwezinsimbi zomshini wokuproducwa okuphambili (Production Engineering, 2022). Ngaphandle kokukhuseleka ngokushesha, ukufikwa ekufaniseni ne-ISO 2690 kungakhathaliseki—futhi izinkokhelo zokugcina ziyagcwalisa ngaphezu kwe-15% kwezinsimbi eziphansi.

Ukukhuseleka kwezinkokhelo ezibhalwe ngokuzenzakalelayo ngokusebenzisa izinkokhelo zokubhekana nge-laser nezinkokhelo zokubhekana ngokusebenzisa i-piezoelectric

Imishini yezinsimu zokusha zokusha zihlukanisa ukuthi kungapheli ngokusebenzisa isikhombisa sokugcina isikhombisa. Isikhombisa sesilika esisemphakathini esihlanganisa umkhakha wokubhala ngokuphelele, ubude lwesilika, nokuhlaka kwesilika ngokuphelele kusukela ku-10 kHz. Ukuchithwa kufaka kumkhakha wokugcina okwenzeka ngokuzimela ukuze kwenziwe ukulungisa ukuguqula kwezinkambo ngokuphelele ngemizuzwana emibili yezikhathi. Lokhu kucindeze kubonakaliswa ukuthi kuhlanganise ukuguqula kwezinsimu nezinkambo zokuzimela, kubonakalisa ukuthi ubude lwesilika buhambisana ngokuphelele ngokuphakanyiswa kwe-±1 µm. Kusuka kumshini wokuzinima wesilika esihambisana nesikhathi eside, lokhu kunciphisa ukuchithwa kwebudlelwane ngokuphelele ngokuphakanyiswa kwe-82% futhi kususa ukuthi kungabikho ukuthi kungahambisani ne-ISO 2690 ngokuphelele ngokuphakanyiswa kwe-12 izinyanga (Tekhnoloji Yezinsimu Zokuzinima, 2023). Lokhu kubonakalisa ukuthi umshini wokuzinima uyakhetha ukulungisa ngokuphelele—kubonakalisa ukuthi isilika sonke sibonakalisa ukuthi sifanele ngokuphelele ngokungafani nesikhathi esingasiza.

Imibuzo Evame Ukubuzwa

Ubudlelwane bendawo bani kuzinima?

Ukuchitha kwezinto ezihlukene kusuka kumathuba akhona kwezinto eziphakeme njengokuphakamisa kokugcina, ukuguqula kwesikhala, nokuphakama kwesikhala sokufihla, okungasemthwakweni nasebhalishini lapho kubhekwa izinkamba.

Ukuthi i-AI isebenza kanjani ukuzilungiselela ukuchitha kwezinto?

I-AI isebenza idatha yezisensha nezimodeli zemetalurgi ukubona ukuchitha okuncane kwezinto ngokushesha kanye nokulungisa iziphakamiso zokuproducwa ngokuzenzakale, ukunika ubuhle obuhambisana kanye nokunciphisa ukungasebenzi.

Ukuthi ukulungisa okuphambili okusebenzisa i-IoT yini?

Ukulungisa okuphambili okusebenzisa i-IoT usebenzisa idatha yezisensha ngokushesha ukubona izinyathelo ezincane zokuphakama komkhombisi noma ukuchitha, ukuvumela abaprofayili bakulungisa ngokuplaneni kanye ukungavikela ukungasebenzi okungaplaneni.

Ukuthi ukulungisa ubuhle okusebenzisa i-AI kuyisiphi isikhundla?

Izisigaba zokulungisa ubuhle esebenzisa i-AI, njengezisigaba esisebenzisa iziklasisayida ze-ResNet-18, zifika emuva kwe-99% yokubona izinto ezingavumelekile kanye nokulungisa ngokushesha kakhulu, ukunika ubuhle obungavumelekile.

Ukuthi ukuthi ISO 2690 yini futhi ukuthi kuzilungiswa kanjani?

Ukuthi kufanele kubhekwe i-ISO 2690 kubonisa izindlela zokuphatha eziningi zokuphatha izinsimbi ngokuphelele. Izintechinoloji zokulungisa zisetyenziswa ukusebenzisa i-laser micrometers ne-piezoelectric sensors ukuze zibone ubukhulu bokuphelele kwemisebenzi yokuproducwa ukuze zifanele lezi zindlela.