In industrial production, it is of great importance to have high availability in its production equipment. Well-functioning maintenance is a significant factor for a high level of availability. This can be achieved by minimizing the number of reactive maintenance stops and optimizing scheduled maintenance. New methods for predictive maintenance provide a good opportunity for this, but most technologies that are available today are designed for individual sub-systems and they are rarely designed for a complex, interconnected machine. In the process industry, raw materials are rocessed into a finished product in a continuous flow through several subsystems and if one subsystem stops, the entire process flow stops. For these processes, it is more important to optimize the maintenance efforts for subsystems so maintenance can take place synchronized. This paper describes a method of supervised control that includes maintenance aspects; health parameters indicating deterioration are included in a MIMO controller. The method is verified in a simulation of a rolling mill with three rollers. The results show that it is possible to optimize the whole complex process including several subprocesses by using a health parameter as a control parameter and broadening the controllability of the system by dividing the workload in a way that all the subsystems reach the desired degradation level for maintenance in a desired optimum time.