Predictive control of interconnected machines

A series interconnection of two subsystems of a machine is considered, for instance a wetclutch with separately identified models for current to pressure and pressure to position of an engagement. Additional conditional logic on such structure leads to hybrid dynamics. First method computes input which drives the entire system into one piecewise continuous region, followed by merger of input/output disturbances by graphical move sum approach, thus enabling series connection of transfer functions for MPC of this reduced system. For the MPC formulations, please refer [1]. The Mixed Loop Predictive control (MLPC) first generates inputs offline for the outer loop based on the overall set-point. The inputs in turn are used for tracking by the inner loop thus producing control inputs for the machine with online feedback. This second approach is powerful tool as in principle both the loops can be under different control schemes. The results from a wet clutch engagement are plotted in fig.1 where the MPC tracks an Iterative Learning Controller. Prismatic movements in mechanics like of shafts can be marginally stable, rediscretization of the subsystem model with bigger sampling time leads to desired loop shap3500 4000 4500 5000 5500 6000 0 5 10 15