Ngokwengxelo ye-Industrial AI kunye neNgxelo yeMarike ye-AI ka-2021-2026, izinga lokwamkelwa kwe-AI kwiiSetingi zemizi-mveliso linyuke ukusuka kwi-19 yeepesenti ukuya kwi-31 yeepesenti kwisithuba nje seminyaka emibini. Ukongeza kwiipesenti ze-31 zabaphenduli abaye bakhupha i-AI ngokupheleleyo okanye ngokuyinxenye kwimisebenzi yabo, enye ipesenti ye-39 ngoku ivavanya okanye ihlola iteknoloji.
I-AI ivela njengetekhnoloji ephambili kubavelisi kunye neenkampani zamandla kwihlabathi liphela, kwaye uhlalutyo lwe-IoT luqikelela ukuba imakethi yezisombululo ze-AI iya kubonisa inqanaba lokukhula lonyaka le-post-bhubhane elihlanganisiweyo (CAGR) le-35% ukuya kuthi ga kwi-102.17 yeebhiliyoni zeedola ngo-2026.
Ubudala bedijithali bubelethe i-Intanethi yeZinto. Ingabonwa ukuba ukuvela kobukrelekrele bokwenziwa kuye kwakhawulezisa isantya sophuhliso lwe-Intanethi yeZinto.
Makhe sijonge ezinye zezinto eziqhuba ukunyuka kwe-AI yeshishini kunye ne-AIoT.
Inqaku loku-1: Izixhobo ezininzi zesoftware ye-AIoT yamashishini
Ngo-2019, xa i-Iot analytics yaqala ukugubungela i-AI yemizi-mveliso, zazimbalwa iimveliso zesoftware ye-AI ezinikezelweyo ezivela kubathengisi betekhnoloji yokusebenza (OT). Ukususela ngoko, abaninzi abathengisi be-OT baye bangena kwimarike ye-AI ngokuphuhlisa kunye nokubonelela ngezixazululo zesoftware ye-AI ngendlela yeeplatifomu ze-AI zomgangatho wefektri.
Ngokwedatha, phantse abathengisi be-400 banikezela ngesoftware ye-AIoT. Inani labathengisi besoftware abajoyina imakethi ye-AI yeshishini linyuke ngokumangalisayo kule minyaka mibini idlulileyo. Ngexesha lokufunda, i-IoT Analytics ichonge ababoneleli be-634 betekhnoloji ye-AI kubavelisi / abathengi bamashishini. Kwezi nkampani, i-389 (61.4%) inikezela ngesoftware ye-AI.
Iqonga elitsha lesoftware ye-AI ligxile kwindawo yeshishini. Ngaphandle kwe-Uptake, i-Braincube, okanye i-C3 AI, inani elikhulayo lobuchwephesha bokusebenza (OT) abathengisi banikezela ngeeplatifti zesoftware ye-AI. Imizekelo ibandakanya uhlalutyo lwe-ABB lwe-Genix lwe-Industrial kunye ne-AI suite, i-Rockwell Automation's FactoryTalk Innovation suite, i-Schneider Electric's iqonga lokubonisana lokuvelisa, kwaye kutshanje, izongezo ezithile. Amanye ala maqonga ajolise kuluhlu olubanzi lweemeko zokusetyenziswa. Ngokomzekelo, i-platform ye-ABB ye-Genix ibonelela ngohlalutyo oluphambili, kubandakanywa izicelo ezakhiwe kwangaphambili kunye neenkonzo zokulawula ukusebenza komsebenzi, ukunyaniseka kwempahla, ukuzinza kunye nokusebenza kakuhle kwekhonkco lokubonelela.
Iinkampani ezinkulu zibeka izixhobo zabo zesoftware ye-ai kumgangatho wevenkile.
Ubukho bezixhobo zesoftware ye-ai kwakhona kuqhutywa zizixhobo ezitsha zokusetyenziswa kwesoftware ethile eyenziwe yi-AWS, iinkampani ezinkulu ezifana neMicrosoft kunye neGoogle. Ngokomzekelo, ngoDisemba ka-2020, i-AWS yakhupha i-Amazon SageMaker JumpStart, inqaku le-Amazon SageMaker elibonelela ngeseti yezisombululo ezakhelwe kwangaphambili kunye nezinokwenziwa ngokwezifiso zezona meko ziqhelekileyo zokusetyenziswa kwemizi-mveliso, njenge-PdM, umbono wekhompyutha, kunye nokuqhuba ngokuzimeleyo, Deploy with. ucofa nje okumbalwa.
Izisombululo zesoftware ezijongene nemeko ethile ziqhuba ukuphuculwa kokusebenziseka.
Iisuti zesoftware ezisetyenziswa ngokwemeko ethile, ezifana nezo zijolise kugcino oluqikelelweyo, ziya zixhaphaka. Uhlalutyo lwe-IoT luqaphele ukuba inani lababoneleli abasebenzisa isoftware yolawulo lwedatha yemveliso esekwe kwi-AI (PdM) linyuke laya kutsho kuma-73 ekuqaleni kuka-2021 ngenxa yokonyuka kwemithombo yedatha eyahlukahlukeneyo kunye nokusetyenziswa kweemodeli zoqeqesho lwangaphambili, kunye nokusasazeka ngokubanzi. ukwamkelwa kobuchwepheshe bophuculo lwedatha.
Inqaku lesi-2: Uphuhliso kunye nokugcinwa kwezisombululo ze-AI zenziwa lula
Ukufunda ngomatshini okuzenzekelayo (i-AutoML) iba yimveliso eqhelekileyo.
Ngenxa yobunzima bemisebenzi ehambelana nokufunda komatshini (ML), ukukhula okukhawulezayo kwezicelo zokufunda ngomatshini kuye kwadala isidingo seendlela zokufunda umatshini ngaphandle kwe-shelf ezinokusetyenziswa ngaphandle kobuchule. Inkalo yophando, i-automation eqhubekayo yokufunda koomatshini, ibizwa ngokuba yi-AutoML. Iinkampani ezahlukeneyo zisebenzisa obu buchwepheshe njengenxalenye ye-AI yazo yokubonelela ukunceda abathengi baphuhlise iimodeli zeML kunye nokuphumeza iimeko zokusetyenziswa kwemizi-mveliso ngokukhawuleza. NgoNovemba ka-2020, umzekelo, i-SKF yabhengeza imveliso esekwe kwi-automL edibanisa idatha yenkqubo yomatshini kunye nedatha ye-vibration kunye nobushushu ukunciphisa iindleko kunye nokwenza iimodeli zeshishini elitsha kubathengi.
Imisebenzi yokufunda koomatshini (ML Ops) yenza lula ulawulo lwemodeli kunye nokugcinwa.
Uqeqesho olutsha lwemisebenzi yokufunda koomatshini lujolise ekwenzeni lula ukugcinwa kweemodeli ze-AI kwiindawo zokuvelisa. Ukusebenza kwemodeli ye-AI ngokuqhelekileyo kuthotywa ngokuhamba kwexesha njengoko kuchatshazelwa zizinto ezininzi ngaphakathi kwesityalo (umzekelo, utshintsho kulwabiwo lwedatha kunye nemigangatho yomgangatho). Ngenxa yoko, ukugcinwa kwemodeli kunye nokusebenza komatshini kuye kwaba yimfuneko ukuhlangabezana neemfuno eziphezulu zomgangatho weendawo zoshishino (umzekelo, iimodeli ezinomsebenzi ongaphantsi kwe-99% Inokusilela ukuchonga ukuziphatha okubeka engozini ukhuseleko lwabasebenzi).
Kwiminyaka yakutshanje, abaninzi abaqalayo baye bajoyina indawo ye-ML Ops, kuquka i-DataRobot, i-Grid.AI, i-Pinecone / Zilliz, i-Seldon, kunye ne-Weights & Biases. Iinkampani ezisekiweyo zongeze imisebenzi yokufunda koomatshini kwi-software yazo ye-AI ekhoyo, kubandakanya iMicrosoft, eyazisa ukufunyanwa kwedatha kwi-Azure ML Studio. Eli nqaku litsha livumela abasebenzisi ukuba babone utshintsho ekusasazweni kwedatha yegalelo ethoba imodeli yokusebenza.
Inqaku lesi-3: Ubukrelekrele bokwenziwa busetyenziswa kwizicelo ezikhoyo nakwiimeko zokusetyenziswa
Ababoneleli besoftware bemveli bongeza amandla e-AI.
Ukongeza kwizixhobo zesoftware ze-AI ezinkulu ezikhoyo ezithe tyaba ezifana ne-MS Azure ML, i-AWS SageMaker, kunye ne-Google Cloud Vertex AI, iisuti zesoftware yemveli ezifana neComputerized Maintenance Management Systems (CAMMS), Iinkqubo zokwenziwa koMveliso (MES) okanye ukucwangciswa kwezibonelelo zeshishini (ERP) ngoku ingaphuculwa kakhulu ngokutofa ubunakho be-AI. Ngokomzekelo, umboneleli we-ERP we-Epicor Software wongeza amandla e-AI kwiimveliso zayo ezikhoyo nge-Epicor Virtual Assistant (EVA). Iiarhente ezikrelekrele ze-EVA zisetyenziselwa ukwenza ngokuzenzekelayo iinkqubo ze-ERP, njengokuhlenga-hlengisa imisebenzi yokuvelisa okanye ukwenza imibuzo elula (umzekelo, ukufumana iinkcukacha malunga nexabiso lemveliso okanye inani lamalungu akhoyo).
Iimeko zokusetyenziswa kwamashishini ziphuculwa ngokusebenzisa i-AIoT.
Iimeko ezininzi zokusetyenziswa kwemizi-mveliso ziyaphuculwa ngokongeza isakhono se-AI kwizixhobo ezikhoyo zehardware/software. Umzekelo ocacileyo ngumbono womatshini kwizicelo zokulawula umgangatho. Iisistim zombono zoomatshini bemveli ziqhuba imifanekiso ngeekhompyuter ezidibeneyo okanye ezidityanisiweyo ezinesoftware ekhethekileyo evavanya iiparamitha esele zimiselwe ngaphambili (umzekelo, umahluko omkhulu) ukumisela ukuba izinto zibonisa iziphene. Kwiimeko ezininzi (umzekelo, amacandelo e-elektroniki anemilo ye-wiring eyahlukeneyo), inani leengxelo zobuxoki liphezulu kakhulu.
Nangona kunjalo, ezi nkqubo zihlaziywa ngobukrelekrele bokwenziwa. Ngokomzekelo, umboneleli wombono womatshini woshishino uCognex ukhuphe isixhobo esitsha se-Deep Learning (I-Vision Pro Deep Learning 2.0) ngoJulayi 2021. Izixhobo ezitsha zidibanisa kunye neenkqubo zombono zendabuko, ezenza abasebenzisi bokugqibela badibanise ukufunda okunzulu kunye nezixhobo zombono wendabuko kwisicelo esifanayo ukuya. ukuhlangabezana neemfuno ezibango zonyango kunye neemekobume zombane ezifuna umlinganiselo ochanekileyo wemikrwelo, ungcoliseko kunye nezinye iziphene.
Inqaku lesi-4: I-hardware ye-AIoT yoShishino iyaphuculwa
Iitshiphusi ze-AI ziphucuka ngokukhawuleza.
Iichips ze-AI ezifakwe kwi-hardware zikhula ngokukhawuleza, kunye neendlela ezahlukeneyo ezikhoyo zokuxhasa uphuhliso kunye nokuthunyelwa kweemodeli ze-AI. Imizekelo ibandakanya iiyunithi zamva nje zokusetyenzwa kwemizobo ye-NVIDIA (i-Gpus), i-A30 kunye ne-A10, ezaziswa ngoMatshi ka-2021 kwaye zilungele iimeko zokusetyenziswa kwe-AI ezifana neenkqubo zokucebisa kunye neenkqubo zembono yekhompyutha. Omnye umzekelo yi-Google yesizukulwana sesine se-Tensors Processing Units (TPus), ezinamandla akhethekileyo-injongo eyodwa yeesekethe ezidibeneyo (i-ASics) ezinokufikelela kumaxesha e-1,000 ngaphezulu kokusebenza kakuhle kunye nesantya kuphuhliso lwemodeli kunye nokuthunyelwa kwemithwalo yemisebenzi ethile ye-AI (umzekelo, ukufumanisa into , ukuhlelwa kwemifanekiso, kunye nemigangatho yeengcebiso). Ukusebenzisa i-hardware ye-AI ezinikeleyo kunciphisa ixesha lokubala imodeli ukusuka kwiintsuku ukuya kwimizuzu, kwaye ibonakalise ukuba ngumtshintshi womdlalo kwiimeko ezininzi.
I-hardware ye-AI enamandla ifumaneka ngokukhawuleza ngemodeli yokuhlawula-ngokusebenzisa.
Amashishini e-Superscale ahlala ephucula iiseva zabo ukwenza izixhobo zekhompyutha zifumaneke kwilifu ukuze abasebenzisi bokugqibela bakwazi ukuphumeza izicelo ze-AI yezoshishino. NgoNovemba ka-2021, umzekelo, i-AWS ibhengeze ukukhutshwa ngokusemthethweni kweemeko zayo zamva nje ze-GPU, iAmazon EC2 G5, inikwe amandla yiNVIDIA A10G Tensor Core GPU, kwiinkqubo ezahlukeneyo zeML, kubandakanya umbono wekhompyuter kunye neenjini zokucebisa. Ngokomzekelo, umboneleli weenkqubo zokubona i-Nanotronics usebenzisa i-Amazon EC2 imizekelo yesisombululo sayo sokulawula umgangatho we-AI ukukhawulezisa iinzame zokucubungula kunye nokufezekisa amaxabiso achanekileyo okufumanisa ekwenziweni kwe-microchips kunye ne-nanotubes.
Isiphelo kunye neProspect
I-AI iphuma kwifektri, kwaye iya kuba yinto yonke kwizicelo ezitsha, ezifana ne-AI-based PdM, kunye nokuphucula isofthiwe ekhoyo kunye namatyala okusebenzisa. Amashishini amakhulu akhupha iimeko ezininzi zokusetyenziswa kwe-AI kunye nempumelelo yokunika ingxelo, kwaye uninzi lweeprojekthi zinembuyekezo ephezulu kutyalo-mali. Lilonke, ukunyuka kwelifu, iiplatifti ze-iot kunye ne-AI chips ezinamandla zibonelela ngeqonga lesizukulwana esitsha sesoftware kunye nokuphucula.
Ixesha lokuposa: Jan-12-2022