نمایه سازی و بازیابی در ماشینکاری برنامه ریزی عملیات با استفاده از استدلال مبتنی بر مورد

To reduce the manufacturing time of a product, one effective way to develop a machining process plan for a new part is to retrieve a relevant case of process planning similar to a new desired part and then adapt the retrieved case to meet the new requirements. This paper proposes a mechanism for retrieval of process planning cases. The core of the retrieval mechanism contains: (1) a feature-based representation of a part and cutting processes; (2) indexing of a part; (3) a feature hierarchical structure based on cutting processes; and (4) a similarity metric used to measure the similarity between a new desired part and any old part in the case base. The application domain here is for axisymmetric part machining. A prototype based on the retrieval mechanism is implemented on a Sun workstation using the ACIS 3D-Toolkit from Spatial Technology Inc.

Process planning is an important activity in CAD/CAM integration. In the past, expert systems have been widely used in automated process planning systems [1], [2], [3] and [4]. However, expert systems are successful only if the domain of applications is well defined and the experts' experiences can be clearly translated into rules. Unfortunately, the domain is difficult to define for most of the manufacturing processes, and the knowledge existing in an experts' mind is not in the form of rules but past successful process plans. Even though one expert system was built, it is static and difficult to change [5]. Besides, the current trend of minimum “time to market” requires that an automated process planning system be able to advance its own knowledge in response to the rapidly changing markets and manufacturing environments. The ability to learn has thus become critical for such a process planning system. In past years, case-based reasoning has been successfully applied in various fields, such as JUDGE [6] for assisting judges in the field of law, CHEF [7] for cooking advisory, and ARCHIE [8] for helping in the design of buildings. The concept of CBR has also been used in design and manufacturing. Yang et al. [9] developed a simple process planning system in machining by employing case-based reasoning. Takahashi et al. [10] combined the concept of CBR and knowledge reuse to solve real, large-scale manufacturing process design problems. Maher and Gómez de Silva Garza [11] used the CBR in structural design. It has been recognized that the retrieval mechanism plays an important role in the case based process planning system. The key factors affecting the performance of the retrieval mechanism are representation, indexing and similarity metric of parts. A good representation, indexing and similarity metric will enable the system to retrieve the most similar case rapidly and correctly. Therefore, the objective of this paper is to develop an appropriate representation and indexing mechanism, and an effective similarity metric. The primary application domain of this study is process planning for axisymmetric parts. Two systems, CAPLAN/CBC [12], [13] and [14] and PROCASE [9] and [15] were developed in this domain. The CAPLAN/CBC assumes that the outlines of a final part can be divided into three areas (two rising areas and a horizontal area) in the axial direction and each area is machined independently. In practice, these areas are not independent of each other in the machining process. Besides, the geometric feature includes only dimensions in their studies. Precision and material were not considered. In fact, precision and material are two important factors affecting the selection of cutting tools and process steps in machining process. On the contrary, PROCASE, though including the information about material and precision ignored them while retrieving the most similar case. In this paper, we extend the concept of the feature-based representation [9] to develop an effective indexing mechanism for retrieving similar parts based on a novel similarity metric. The proposed similarity metric not only covers the geometric shape of a part, but also considers precision and material. In addition, we also develop a bit-representation method to index parts when the number of cases is very large. This method is employed to narrow the searching scope within a large case base. This paper is organized in eight sections. The overview of the retrieval mechanism is described in Section 2. Feature-based representation of parts and processes are discussed in Section 3. Section 4 introduces a case indexing mechanism. Retrieval techniques and similarity measures are presented in Section 5. An algorithm for indexing of massive cases is described in Section 6. In Section 7 three examples implemented in a solid modeling environment are given. Finally, discussions and conclusions are provided in Section 8.

To deal with a machining process planning problem, a conventional rule-based system will accumulate at least hundreds of rules. Extracting those rules from the experts’ experiences is very difficult. Even in the case where rules can be extracted, they are highly coupled. One broken chain will result in the failure of the whole system. Unfortunately, it often occurs, because the machining process planning is so complicated that building a complete rule base is very difficult. In contrast, the case-based system is based on the existing cases. That is, a new plan is built upon an old one. The performance of a case-based system is improved as the number of cases increases. Unlike the rule-based system, each case in the case-based system is often independent. Therefore, no coupling problem exists. In a case-based system for machining process, one of the most important components is its retrieval mechanism. In this paper, a part is represented by an indexing which incorporates geometry, material, and precision. The geometric property of a part is represented by its geometric shape, tolerance, and surface finish, while its material property is represented by its HD and HT. Because the similarity metric developed to measure the similarity of two parts is based on geometric, material, and precision properties, the most similar case can be retrieved from the case base effectively. In addition, a filtering method is introduced to reduce the retrieval time when the size of the case base is large. The empirical results show that the proposed retrieval mechanism can also find out the most relevant case rapidly. A prototype based on the proposed representation, indexing, and similarity metric has been implemented on a Sun workstation using ACIS and 3D-Toolkit from Spatial Technology Inc. Three examples are given to illustrate the proposed mechanism. The empirical results indicate that it performs very well.