111 this paper. niaxinluni efficicnc!. operation of tu.0 t!.pcs of pcri~~ancnt magnct
s! ~ichro~ious machine drives. namcl!.. surfacc tqc permamit magiict s!~~~clironous
macliinc (SPMSM) and interior t! pc pcr~nancnt magnet s!mhronous machine
(IPMSM). are invcst~gated. Thc cf-'ficicnc!. of both drives is iilasimized b!
ni~nimizing cupper and iron losses. Loss minimization is implemented using flus
\wakening. A neural net\vork controller (NNC) is designed for each drix. to
~cI~~c\c loss niini~iiization at difi'crent speeds and load torque valucs. Data lbr
trainmg thc NNC arc obtained through off-line s~mulations of SPMSM and IPMSM
at di;-Yerent operating conditions. A ccurac!. and fast response of cach NNC is
po\.cd b!, appl!x~g sudden clianycs in spccd and ioad and tracking thc NhC output.
Thc motors' efficiency obtained by flus ~\cakciiing is coinpared \~ith thc cfficicnc!.
obtained whcn setting the d-axis current component to zero. \vliile varying tlie angle
of ad\!ancc "cp" of the PWM invcrtcr suppl! ing tlic PMSM drivc. Equal cfficiencics
21.c obtained at different valucs of cp. dcri\cd to be f~~nction uf sped and load
torque. A NN is also designed. and trarncd to \-ar! cp follo~\inz the dcr~~cdontrol
la\\. Thc nccurac!. and fast responsc ofthu NN co~itrolicr is also pro~.cd.