Kwinkqubela ebalulekileyo kwicandelo lokuxilongwa koomatshini, uphando olutsha lubonise ukusebenza kakuhle kokudibanisa i-modulation signal bispectrum (MSB) kunye ne-convolutional neural networks (CNN) yokuchonga iimpazamo ze-iigiya ze-bevel ezijikelezayoLe ndlela intsha ithembisa ukuchaneka okuphuculweyo, ukufunyanwa ngokukhawuleza, kunye nenkqubo yokuxilonga ekrelekrele ngakumbi kwiibhokisi zegiya ezisebenzayo kakhulu ezisetyenziswa kwiukusetyenziswa kweenqwelo-moya, iimoto, kunye nemizi-mveliso.

I-Spiraliigiya zebhevelzizinto ezibalulekileyo zothumelo ezifumaneka kumatshini we-torque ephezulu, iihelikopta, iinkqubo zokuqhuba zaselwandle, kunye nezixhobo zokunciphisa imithwalo enzima kwimizi-mveliso. Ngenxa yejometri yazo eyinkimbinkimbi kunye neemeko zokusebenza, ukubonwa kwangethuba kweengxaki zegiya ezifana nokuqhekeka, ukuguguleka, kunye nokuqhekeka kwamazinyo kusengumngeni wobugcisa. Iindlela zemveli zokucubungula isignali zihlala zinengxaki yokuphazamiseka kwengxolo kunye neempawu zempazamo ezingezizo ezithe ngqo.

Le ndlela intsha iveza isakhelo sokuxilongwa kwempazamo yesigaba esibini. Okokuqala, imiqondiso yokungcangcazela eveliswa yinkqubo yezixhobo zokusebenza ihlalutywa kusetyenziswa i-modulation signal bispectrum (MSB) indlela yohlalutyo lwe-spectral ephezulu ebamba ngokufanelekileyo iimpawu zesignali ezingezizo ezithe ngqo nezingezizo zeGaussian. I-MSB inceda ekutyhileni iimpawu zempazamo eziguqulweyo ezifihlakeleyo ezihlala zifihliwe kwi-standard frequency spectra.

Okulandelayo, idatha yesignali ecutshungulwayo iguqulwa ibe yimifanekiso yexesha kwaye ifakwe kwinethiwekhi ye-convolutional neural (CNN) imodeli yokufunda enzulu ekwaziyo ukukhupha ngokuzenzekelayo iimpawu zempazamo eziphezulu kunye nokwahlula iimeko zegiya. Le modeli ye-CNN iqeqeshwe ukwahlula phakathi kwegiya ezisempilweni, iimpazamo ezincinci, kunye nomonakalo omkhulu kwiimeko ezahlukeneyo zomthwalo kunye nesantya.

Iziphumo zovavanyo, ezenziwe kwisixhobo sovavanyo lwegiya ye-spiral bevel eyenzelwe wena, zibonisa ukuba indlela ye-MSB CNN ifikelela ngaphezulu kwe-97% yokuchaneka kohlulo, iphumelela ngaphezu kweendlela zemveli ezifana nohlalutyo olusekelwe kwi-FFT kunye nezinye iindlela zokufunda ezinzulu ezixhomekeke kwidatha yokungcangcazela eluhlaza. Ngaphezu koko, le modeli ye-hybrid ibonisa ukomelela okukhulu kwingxolo yangasemva, okwenza ukuba ifaneleke kwizicelo zoshishino zehlabathi lokwenyani.

Ukuhlanganiswa kwe-modulation signal bispectrum ne-CNN akuphucula kuphela ukusebenza kokuqaphela iimpazamo kodwa kunciphisa nokuxhomekeka kubunjineli beempawu ezenziwe ngesandla ngokwesiko, inkqubo ethatha ixesha elide kwaye ixhomekeke kubuchule. Le ndlela inokwandiswa kwaye ingasetyenziswa kwezinye izinto zoomatshini ezijikelezayo, ezifana neebheringi kunyeizixhobo zeplanethi.

Olu phando lubonisa inyathelo eliya phambili ekuphuhlisweni kweenkqubo zokuchonga iimpazamo ezikrelekrele kwi-Industry 4.0 kunye nentsimi ebanzi yokuvelisa izinto ezikrelekrele. Njengoko ukwenziwa kwezinto ngokuzenzekelayo kunye nokuthembeka koomatshini kuya kuba yinto ebalulekileyo,