For low cost automation, a traditional manually operated milling machine with a lead-screw transmission system was retrofitted with an AC servo-motor. This old fashioned machining table has nonlinear time-varying behaviour caused by obvious backlash and irregular coulomb friction of the sliding surfaces. It is difficult to design an appropriate classical controller for this complicated dynamic system. Hence intelligent model-free self-organising fuzzy control and neural network control strategies equipped with learning ability are employed to control this machining table, to improve both the adaptability and the path tracking accuracy. These control approaches can be implemented without the trial and error process for selecting initial parameters and fuzzy rules. The experimental results show that these control strategies achieve satisfactory transient response and tracking accuracy under the influence of similar to 0.4mm of backlash on each axis and large stick-slip friction.