This work proposes an approach using a Kalman filter and an adaptive controller to control the product size in centerless grinding. Plant parameters are estimated in real time by a weighted recursive least squares technique. The Kalman filter receives these parameters, along with the current input and output signals of the plant, to estimate the main trend of the plant output signal. An adaptive controller uses this information and the estimated main trend signal to generate a control signal. However, this linear control signal cannot be directly used to control the product size in a grinding process because of plant nonlinearities. Thus, the linear control signal is converted to the appropriate compensation signal used by the machine tool.