ISIHLULO ESIFUYWAYO se-DiesEL

Ngomkhwa wezixhobo zoomatshini ngokubhekiselele kubukhulu obukhulu, obukrelekrele kunye nokuchaneka okuphezulu, ukusetyenziswa kweendawo ezijikelezayo, ezifana nokuthwala i-roller, kuye kwaphuculwa kakhulu ukufezekisa ukuhanjiswa kwamandla, ukulungiswa kwendawo kunye nezinye iinjongo. Xa zonakaliswe okanye zisilela, ukhuseleko lokusebenza kwesixhobo somatshini kunye nenzuzo yemveliso iya kuchaphazeleka. Nangona kunjalo, ngenxa yendawo yofakelo ekhethekileyo yala macandelo ajikelezayo, kunzima kakhulu ukuphanda kunye nokugweba imeko yezempilo yezixhobo, kwaye iindlela zangaphambili ezixhomekeke ebantwini okanye amava azikwazi ukusebenza. Ke ngoko, ukuphuhlisa ukufumanisa okukrelekrele kunye nendlela yokuxilongwa ukuphumeza ukujongwa kwempilo kwezixhobo kuye kwaba sisihloko sophando esishushu.

Ngophuhliso olukhawulezayo lobukrelekrele bokwenziwa, ngakumbi nangakumbi iindlela zokufunda zoomatshini zenza ukuba izixhobo zoomatshini zokuxilonga ezikrelekrele zibe yinyani kwaye ziphumelele, njengokufunda okomeleza (RL) [1], [2], uthungelwano oluchasayo oluvelisayo (GAN) [3], i-autoencoder. (AE) [4] kunye nomatshini we-vector yenkxaso (SVM) [5], [6], [47]. Phakathi kwabo, i-SVM yi-algorithm yokuhlelwa okusekwe kwi-statistical learning, ekungekho lula ukuwela kwi-minima yendawo kwaye yahlula idatha yoqeqesho ngokusebenzisa i-hyperplane efanelekileyo ngelixa idatha yoqeqesho inokuthi ifakwe kwiimpawu eziphezulu ze-dimensional ngokusebenzisa iindlela zemephu ezingenanto, ezifana nemisebenzi ye-polynomial kunye imisebenzi yesiseko yeradial. Ukongeza, i-SVM inokubonelela ngesigqibo esichanekileyo se-hyperplane phantsi kweesampulu ezilinganiselwe, kwaye inamandla okwenza ngokubanzi. Ngokujonga ukusebenza kwayo okugqwesileyo, i-SVM isetyenziswe ngokubanzi kwiinkalo ezininzi. Wang et al. icetywe indlela ekrelekrele yokuxilonga impazamo esekwe kwindibaniselwano ye-generalized composite multi-scale weighted permutation entropy (GCMWPE) kunye ne-SVM [7], enokuthi ikhuphe iimpawu ezithwalayo ukusuka kwizikali ezininzi ukwakha ingqokelela yeempawu ezinomgangatho ophezulu. Bayati et al. icetywe indlela enempazamo yendawo ye-DC microgrid esekwe kwi-SVM [8]. Ngokusebenzisa ixabiso elilinganisiweyo lendawo ekupheleni komgca ngamnye, indawo echanekileyo ye-impedance fault ephezulu inokufumaneka, kwaye iziphumo zovavanyo zibonisa ukuba iskimu sinamandla kwingxolo kunye nezinye iziphazamiso. Ref. [9] icebise indlela ekrelekrele yokuxilonga impazamo yebhetri ye-lithium-ion esekwe kumatshini we-vector yenkxaso, esebenzisa isihluzo esicacileyo se-cosine ukuphelisa ingxolo.