The industrial processes that require the use of the web require a control system which allows for preserving the properties of the web unaltered, avoiding the risk of wrinkling, tearing, breakage and other defects. This control generally takes place by detecting the tension and the speed in certain points of the system since these variables determine the stress state on the web, which, if altered beyond certain ranges, can lead to the defects mentioned above. The problem of tension and web speed control is very demanding because the system’s dynamic is a function of many process variables that often vary over a wide range. In this study, an experimental system consisting of 12 rollers, four motorised, was analysed. This system was divided into four subsystems according to the logic of decentralised control. The tension of the initial and final subsystems and the speeds of the two central subsystems were monitored. This study proposes estimating continuous-time transfer functions using experimental time-domain data. A nonlinear least-squares search-based method minimises a weighted prediction error norm for directly identifying the mathematical model used to describe the web transport system. To test the performance of the proposed strategy, experimental data were collected in an open-loop configuration with constant voltage given to the four servo motors. The collected data were subsequently processed to define an extremely simple system model composed of a very limited number of parameters representing the system through transfer functions. The model was further validated by comparing the results obtained through simulations with the experimental data obtained with different inputs, and was also validated with some closed-loop tests.