Paper machines are complex applications of process industry which are typically controlled by model predictive controllers (MPC). The MPC predicts the output of the process by using process models that describe the interactions between the control variable and the desired manipulated variable. Typically, these models are identified by making step changes to the system, which are commonly known as bump tests. Sometimes these tests fail to give a response good enough for the identification due to disturbances or other interactions between the process variables.

The paper studies the method of using a Maximum Length Binary Sequence (MLBS) in the identification of low-order transfer function models such as FOTD (First-Order Time Delay) and SOTD (Second-Order Time Delay). MLBS is a deterministic periodic broadband excitation signal, which can be used to estimate the frequency response of the studied system. The estimated frequency response can then be used in a curve fitting problem, which gives the identified low-order model’s parameters as a result. The paper focuses on proper parameterization of the MLBS and the effects of selected disturbances to the identification results are studied.