Machining Planning: A Model of an Expert Level Planning Process

Abstract

Understanding and automating planning is a problem that has been of interest and of economic importance in manufacturing domains over the past 25 years. I recent years, there has been an increasing need for the automation of manufacturing planning for machining processes due to the reduction in the number of new trainees entering the profession. The specific motivations behind the work in this thesis are to provide tools for process planners and designers which will: (1) allow process plans for small batch manufacturing to be produced in a more timely and cost effective manner. This will make the production of both prototype parts and custom manufacturing faster and more affordable. (2) allow designers to make more cost effective designs by providing them with manufacturing cost and feasibility information early in the design process. Unfortunately the processes that one must model in order to automate complex manufacturing tasks is often not well understood. Current models of process planning for tasks such as stamping, turning and milling are at best inexact and incomplete. This thesis presents a model of one expert planning task: forming manufacturing plans for prismatic milled parts. It is implemented in the Machinist program, a rule-based system written in OPS5, which produces correct manufacturing plans judged in a blind test by experts to be equivalent in quality plans produced by human machinists with 7 to 8 years of experience. The model incorporates techniques gleaned from observations of human experts. The domain is highly complex, typically requiring experts with a high degree of training (6 to 10 years) in order to make successful plans for a wide variety of parts. Additionally, an application of the machining planning model is presented in which the planning model is used to produce cost reducing design suggestions. In this application, a manufacturing plan for a part is produced; the program tries to improve the plan by looking for ways in which steps can be combined by making modifications to the original part specification without compromising the function.