A heirarchical, knowledge-based controller has been developed and implemented on a CNC lathe. The system consists of an intelligent supervisor and two control subsystems. These elements are realized as concurrent tasks in a real time, multitasking system. For each tool pass, a parameter adaptive control subsystem maximizes tool feed within tool breakage and tool edge chipping constraints. A second subsystem adjusts the depth of cut on-line in reaction to the detection of the onset of chatter. Consequently, this event-driven subsystem aims to ensure a depth of cut that results in stable cutting. An intelligent supervisor coordinates these subsystems, combining a frame based knowledge base with a real-time inference engine. The intelligent supervisor, having precedence over the other system elements, directs the control objective of this hierarchical control system. Cutting tests are performed to verify the controller performance in real time. Experimental results are presented.