This paper describes a controller design and also a sensor fault diagnosis based on a LTV model for a winding machine. It shows that, according to an experimental identification approach, a LTV model is ideally suited to improve the control of web tension. Moreover, based on this model, an innovative bank of interpolated LTI Kalman filters is synthesised to allow sensor fault detection and isolation (FDI). The advanced control law has been successfully applied to a winding machine. Also, software sensor fault has been generated to improve the accuracy of FDI algorithm in winding machine.