Abacwaningi e-New Jersey Institute of Technology (NJIT) basebenzise ubuhlakani bokwenziwa ukuze baxazulule inkinga ebalulekile yesikhathi esizayo sokugcinwa kwamandla: ukuthola ezinye izindlela ezingabizi, ezisimeme zamabhethri e-lithium-ion.
Esifundweni esanyatheliswa kuyi- Cell Reports Physical Science, ithimba le-NJIT, eliholwa nguSolwazi uDibakar Datta, lisebenzise ngempumelelo amasu e-AI okukhiqiza ukuze lithole ngokushesha izinto ezintsha ezinezimbobo ezingaguqula ukuthuthukiswa kwamabhethri e-ion amaningi. Lawa mabhethri, asebenzisa izinto ezivamile njenge-magnesium, i-calcium, i-aluminium, ne-zinc, anikeza enye indlela ethembisayo nengabizi kakhulu kunamabhethri e-lithium-ion, abhekene nezinselele zokuhlinzekwa komhlaba wonke kanye nokusimama.
Ngokungafani namabhethri e-lithium-ion avamile, athwala ukushajwa okukodwa kuphela, amabhethri e-ion amaningi asebenzisa izakhi ama-ion athwala amashaji amabili noma amathathu. Lokhu kusho ukuthi amabhethri e-ion amaningi angakwazi ukugcina amandla amaningi kakhulu, awenze athandeke kakhulu ezixazululweni zesikhathi esizayo zokugcina amandla.
Kodwa-ke, ngenxa yobukhulu bawo obukhulu kanye nokushajwa kukagesi okuphezulu, ama-ion ahlukahlukene kunzima ukuwafaka ngempumelelo ezintweni zebhethri—isithiyo ucwaningo olusha olunamandla e-AI lweqembu le-NJIT olubhekana nalo ngokuqondile.
Ukuze kunqobe lezi zithiyo, ithimba le-NJIT lenze inoveli embaxambili ye-AI: iCrystal Diffusion Variational Autoencoder (CDVAE) kanye neModeli Yolimi Olukhulu (LLM) ecushwe kahle. Ndawonye, lawa mathuluzi e-AI ahlola ngokushesha izinkulungwane zezinhlaka ezintsha zekristalu, into eyayingenakwenzeka ngaphambili ngokuhlolwa okuvamile kwaselabhorethri.
Imodeli ye-CDVAE yaqeqeshwa kumasethi edatha abanzi ezinhlaka zekristalu ezaziwayo, okuyenza ikwazi ukuphakamisa izinto ezinoveli ngokuphelele ezinohlu olubanzi lwamathuba esakhiwo. Ngasikhathi sinye, i-LLM yaculwa ezintweni ezinokuzinza okuphezulu kakhulu kwe-thermodynamic, okubalulekile ekuhlanganiseni okusebenzayo.
"Amathuluzi ethu e-AI asheshise ngokumangazayo inqubo yokutholwa, embule izinhlaka ezinhlanu ezintsha ze-oxyde metal transition ezikhombisa isithembiso esimangalisayo," kusho uDatta. "Lezi zinto zifaka amashaneli amakhulu, avulekile alungele ukuthuthwa okusheshayo nokuphephile kwalawa ma-ion angama-multivalent-impumelelo ebalulekile yamabhethri esizukulwane esilandelayo."
Ithimba liqinisekise izakhiwo zalo ezikhiqizwe i-AI lisebenzisa ukulingisa kwemishini ye-quantum nokuhlola ukuzinza, liqinisekisa ukuthi izinto ezibonakalayo zingahlanganiswa ngempela ngokuhlolwa futhi zibambe amandla amakhulu okusebenza komhlaba wangempela.
credits
New Jersey Institute of Technology
