Abstract— A methodology for decomposition of material removed volumes so called Delta Volumes according to the natural sequences of machining is proposed. Tool classification and coding system for automatic characteristics and parameter extraction is presented. Algorithms for identification of major access directions and machining complexity and rules determining operations and tools for roughing and finishing are presented. This article suggests a technique that generates machining operation sequences with appropriate combinations of tool and machine directly from the 3D model of a part and a given block.. I. INTRODUCTION In decades of efforts devoted in computer integrated manufacturing, automatisation of process planning has been a major objective of researches in this field. Till today, the work to generate machine and tool sequence for part machining is still done manually. The main objective of this research is to contribute a robust model that can describe the machining capability of a machine and tool combination. This model analyzes the machinability, accessibility of each sub volume as well as the complexity of the part, chooses suitable machine and tools for the task of machining the part. It thus makes a contribution to the advance in computer aided process planning (CAPP). II. BACKGROUND AND OBJECTIVES CAPP is seen as a communication agent between CAD and CAM. As a widely accepted idea, CAPP has to interpret the part in terms of features. A machining feature is defined as a volume of material that a process planner would consider to machine with the same operation [9]. Feature recognition is considered as a front-end to CAPP and plays a key role in CAD/CAM integration [2]. It is the process of converting the CAD data of a part into sequenced machining activities for producing the part. Hint-based approaches are more successful in recognizing interacting features than the other existing approaches [1,3] but they also have some shortcomings [6]. As it is known that the principle of hint- Manuscript received November 20, 2006. This project is sponsored by the Natural Sciences and Engineering Research Council of Canada (Grant No RGPIN150270-01) rant BS123456 (sponsor and financial support acknowledgment goes here). C. Mascle is with the Mechanical Engineering Department, École Polytechnique, C.P. 6079, Succ. Centre-Ville, Montréal, Québec, H3C 3A7 Canada, (corresponding author to provide phone: 514-340-4711#4398; fax: 514-340-5867, e-mail: christian.mascle@polymtl.ca ) H. Lu is with the Mechanical Engineering Department, École Polytechnique, C.P. 6079, Succ. Centre-Ville, Montréal, Québec, H3C 3A7 Canada, e-mail: hua.lu@polymtl.ca R. Maranzana is with the Mechanical Engineering Department, École Polytechnique, C.P. 6079, Succ. Centre-Ville, Montréal, Québec, H3C 3A7 Canada, e-mail: roland.maranzana@etsmtl.ca . based methods is to match “traces left by the motion of a milling cutter” with predefined features. A feature library does not include all tool traces, and for some complex features it is difficult to find suitable traces. Volumetric decomposition approaches, including convex- hull decomposition [4,5] and cell-based decomposition [8,12] have some natural drawbacks: they cannot directly be used to generate machining features. Graph based feature recognition approaches merely dealt with isolated polyhedral features [10]. Rahmani and Arezoo [6,7] combined graph based and hint based approaches into a hybrid one which brings an advance to graph based approaches in curved feature and interactive feature handling. But the application of their method is still limited to 2.5D features. Venkatadriagaram [11] devoted efforts on facilitating the integration of CAD and CAM for 3-axis milling by combining feature recognition with process planning. Although interfaces were defined for communications between different levels of process planning, machining sequences have to be generated manually. Sridharan and Shah [9] inspected features from a point of view very close by the manufacturing reality. In their taxonomy there are three categories of milling features: cut- through, -around, and -on. The basic feature characteristics considered in their concept are external and internal, blind and through. In their article, no specific methodology was detailed to identify the complexity of machining. Examples given do not show the necessity of using a 4- or 5-axis machining. By investigating all the previous researches that are available to the authors, we could conclude that researches on feature recognition have been halting at the effort of geometry description and generated features are part orientation dependent. Generated features are independent of machining processes. As a result feature recognition creates merely separated geometric volumes which are adequate for tool path and NC code generation. To make a step forward in integration of CAD and CAM, a methodology for machining processes generation together with specification of tool parameters and machine characteristics should be necessities. Our objective is to contribute a robust model to generate possible sequences of machine and tool combination for milling processes planning. It is expected that this model would take the 3D model of an arbitrarily given part and a stock from which the part will be produced, extract the volume ∆V of material to be removed (delta volume), divide ∆V into sub volumes, analyze fitness of tool to the part and Machining process planning using Decomposition of Delta Volume Christian Mascle, Hua Lu, Roland Maranzana Proceedings of the 2007 IEEE International Symposium on Assembly and Manufacturing Ann Arbor, Michigan, USA, July 22-25, 2007 MoC1.2 1-4244-0563-7/07/$20.00 ©2007 IEEE. 106