Hierarchical control of complex manufacturing processes

The need for changing the control objective during the process has been reported in many systems in manufacturing, robotics, etc. However, not many works have been devoted to systematically investigating the proper strategies for these types of problems. In this dissertation, two approaches to such problems have been suggested for fast varying systems. The first approach, addresses problems where some of the objectives are statically related to the states of the systems. Hierarchical Optimal Control was proposed to simplify the nonlinearity caused by adding the statically related objectives into control problem. The proposed method was implemented for contour-position control of motion systems as well as force-position control of end milling processes. It was shown for a motion control system, when contour tracking is important, the controller can reduce the contour error even when the axial control signals are saturating. Also, for end milling processes it was shown that during machining sharp edges where, excessive cutting forces can cause tool breakage, by using the proposed controller, force can be bounded without sacrificing the position tracking performance. The second approach that was proposed (Hierarchical Model Predictive Control), addressed the problems where all the objectives are dynamically related. In this method neural network approximation methods were used to convert a nonlinear optimization problem into an explicit form which is feasible for real time implementation. This method was implemented for force-velocity control of ram based freeform extrusion fabrication of ceramics. Excellent extrusion results were achieved with the proposed method showing excellent performance for different changes in control objective during the process.