among various ac motors, induction motor (im) occupies almost 90% of the industrial drives due to its simple, robust construction and generally satisfactory efficiency as compared to dc motor. however, the control of im is complex due to its nonlinear nature and the parameters change with operating conditions. since 1980s, field orientation principle (fop) has been used for high performance control of im. due to the well-known drawbacks of the fixed-gain proportional-integral (pi), proportional-integral-derivative (pid) and various adaptive controllers, over the last two decades researchers have been working to apply artificial intelligent controller (aic) for im drives due to its advantages as compared to the conventional pi, pid and adaptive controllers. the main advantages are that these controllers can handle any nonlinearity of arbitrary complexity, and their performances are robust. also fuzzy rules and neural network (nn) can be used to model a process for model reference or model predictive control. meanwhile, the designs of these controllers do not depend on accurate system mathematical model. neuro-fuzzy controller (nfc), as a kind of artificial intelligent controller (aic), has attracted much attention by researchers as it takes advantages from both fuzzy logic controller (ftc) and nn by combining the expert human knowledge and the learning ability of the nn. despite lots of research on aic application for motor drives, industries are still reluctant to use aic for real-life industrial drives. the main reason is that most of aic require complex calculation and hence suffer from high computational burden. therefore, attention needs to be paid to develop aic which is suitable for practical applications. in order to achieve that, in this thesis, first, a novel, low computational and simplified self-tuned nfc is developed for the speed control of im drive. for the proposed nfc only the speed error is used as the input, unlike conventional nfcs, which utilize both speed error and its derivative as inputs. obviously, this simplification lowers down computational burden and makes the nfc easier to be implemented in practical applications. next, a faulty im with broken rotor bars (imbrb) is considered and a nfc is developed to minimize the speed ripple of that motor. the speed error and rotor electrical angle are used as two inputs o f the nfc. a supervised self-tuning method is also developed for the developed nfcs. the system error, instead of controller error, has been utilized to tune the membership functions and weights because the desired controller output is not readily available. also the convergences/divergences of the weights are analyzed and investigated. simulation models for indirect field oriented control of im incorporating both of the developed nfcs are developed in matlab/simulink. im drives based on both of the developed nfcs are successfully implemented in real-time using dsp board ds1i04. for the first nfc, comparisons with conventional nfc and pi are done both in simulation and experiment at different operating conditions for a laboratory 1/3 hp im. also the effectiveness o f the second nfc is tested for a laboratory 0.5 hp imbrb both in simulation and experiment, compared to a well-tuned pi controller. it is found from the experimental results that the proposed nfc reduces the fundamental and second harmonic components of speed ripple which are significant components as compared to high frequency components.