SM38: Removal of Blends from Boundary Representation Models Sashikumar Venkataraman 1 Geometric Software Solutions Ltd. Plant 14, Pirojshanagar, Vikhroli, Mumbai-400079, India 91-22-5960982 sashiv@geometricsoftware.com Milind Sohoni Dept. of Computer Science and Engg. Indian Institute of Technology, Powai Mumbai-400076, India 91-22-5767729 sohoni@cse.iitb.ac.in Rahul Rajadhyaksha Geometric Software Solutions Ltd. Plant 14, Pirojshanagar, Vikhroli, Mumbai-400079, India 91-22-5960947 rahulr@geometricsoftware.com ABSTRACT This paper reports an algorithm for deletion of blends (or fillets) from Boundary Representation (B-rep) solid models. Blend deletion is usually performed as the first step in feature recognition since it simplifies the model for recognition of volumetric features. The algorithm handles several blend types that include face-face, face-edge and vertex blends. It also handles interactions of blends with other blends and/or volumetric features. Another unique feature of our algorithm is the recreation of new faces in certain situations of blend deletion. Unlike some earlier face-deletion approaches that use geometric heuristics to determine the final topology, our algorithm uses the underlying blend information to directly predict the final topology and geometry. Hence, our algorithm is more efficient and predictable for suppressing large blend networks. Keywords Blend recognition, blend deletion, face-deletion algorithms, euler operators, blend chains, feature recognition. 1. INTRODUCTION Feature recognition is an area of active research in the past two decades [1-5]. Several different approaches have been developed for recognizing features from CAD data. For a comprehensive review of feature recognition, the reader is referred to Refs [4] [5]. These techniques are usually focused towards recognition of volumetric features such as holes, pockets or slots. However, it has been observed that the algorithms presented in the literature do not yet apply to many industrial parts. Most of the examples presented in academic research are usually simple prismatic parts with few feature interactions. It seems a challenge to develop an efficient feature recognition system that works on complex cases containing many curved surfaces and many feature interactions. Another common occurrence in most industrial parts is the presence of complex blend features (or fillets) that are introduced in the design stage to smoothen the sharp edges of the part. This is mainly done to improve its strength, the aesthetics, and also to ensure the manufacturability of the part. Blend features usually manifest as complex networks that can distort the neighboring topology considerably. In presence of such blend features, most current recognition algorithms fail to recognize volumetric features. Even if volumetric features are recognized, their parameterization is difficult in presence of blends. For example, since the edges of the feature faces are also modified by blends, they cannot be directly used as profiles to create feature volumes. To overcome these problems, feature recognition techniques usually assume that blend features are removed prior to the recognition of volumetric features. However, no comprehensive algorithm has yet been presented to remove blends from a part. Apart from feature recognition, blend removal is also useful as a local operator for editing and modification of models. For example in applications involving Finite Element Analysis (FEA), blends below a certain threshold radius are removed prior to analysis. The first step towards blend removal is to recognize the blends in the model. In [7], the authors had proposed an algorithm for recognizing all the blends in a model. The algorithm also deduces the sequence in which the blends were created in the model, and thereby helps in understanding the design intent of the model. The reverse sequence can then be used for blend suppression ensuring valid intermediate models at each step. In that paper, the exact step of blend suppression was performed using a separate function (referred as delete faces) that removed a set of faces from a solid by extending/shrinking neighbor faces of the blends. Such functionality is provided as part of many geometry kernels such as ACIS [12] and Parasolid [13], and also as standalone libraries such as by GSSL [14]. Several algorithms have been presented in the literature to delete a set of faces by extending the adjacent faces [8,9,10]. These algorithms are geometric in nature and use several heuristics to arrive at the final topology. Usually these approaches work for simple and specific kinds of face-sets, and can fail in complex situations. Especially in case of blends, the number of neighbor faces can be large since the blend chain may run across several faces. This increases the likelihood of failure of a particular fixed heuristic in these algorithms. Further, face deletion algorithms usually involve several intersections between surfaces. In case of blend networks, these intersections could be between tangential (or smooth) surfaces that are unstable. Moreover, in many situations, blend deletion needs to be accompanied by recreation of new faces that got removed during making of the blend. This is not within the scope of current face deletion algorithms. Due to these reasons, many kinds of blends cannot be robustly removed with existing face deletion algorithms. 1 Sashikumar V. is also pursuing a Ph.D. in the Dept of Computer Science and Engg. at the Indian Institute of Technology, Powai.