In modern industrial robots the components in the transmission contain nonlinearities. These nonlinearities need to be to estimated either for better control or to use the parameters for diagnosis of the system. There is a lot of work done within system identification and mainly within the field of iterative parameter estimation.This thesis considers recursive grey-box identification for a nonlinear model of the transmission in an industrial robot. The nonlinearities that are identified are friction, spring stiffnes, hysteresis and backlash. These nonlinearities are a part of the models that are presented in this thesis.