Computational methods to support sketching, reverse engineering, and optimization of shapes

The shape has a unique role in product design. On the one hand, it allows to realize mechanical functions, on the other hand, it supports achieving positive aesthetic impressions and use experiences. For engineering designers, the shape is defined by technical requirements, functionality, physical principles, structure, ways of materialization, and manufacturing technologies, while for industrial designers it is determined by aesthetics, ergonomics, usability of and experiences with products. This duality gives rise to many challenges when it comes to computer support of shape design and optimization. There is an intensive research into efficient computer-oriented methods and tools that would be able to concurrently support conceptualization, detailing, and optimization of shapes from the aspects of both engineering design and industrial design. Various solutions have been proposed in the past years, the majority of which addresses specific problems of either engineering design or industrial design. For this special issue, we have selected papers that consider both human aspects and engineering aspects, to some extent, in the definition of the shape of products. These papers show not only the efficiency, but also the flexibility of these methods in terms of processing deterministic or stochastic, formal or tacit, static or dynamic, and incomplete or complete information and knowledge.

What is common in the proposed solutions is that they rely on a computational basis. Though the computational principles and target applications vary in a wide range, a common feature is that they tend to adopt computational approaches to engineering design of shapes that have been previously used to support industrial design, and vice versa. It tells us that future computational tools for shape design will integrate the aspects of engineering design and industrial design more intensively, and will provide a multifunctional toolkit for shape designers. Actually, the selected papers belong to three topic groups which deal with computational issues of sketching shape in industrial design, reverse engineering of shapes, and shape optimization, respectively. We believe that these novel computer oriented methods and tools are not only important from an academic point of view, but also valuable for the industrial practice. From the pool of papers submitted to the Sixth International Symposium on Tools and Methods of Competitive Engineering (TMCE 2006), we picked up those which report on mature research and results that had been tested through practical application.

The first paper titled “Semantic-based operators to support car sketching”, co-authored by Cheutet, V., Catalano, C.E., Giannini, F., Monti, M., Falcidieno B. and Leon, J.C., addresses the problem of sketch interpretation with the goal to support higher level interaction with shape concepts when they are represented in digital forms. The authors propose an ontology of the aesthetic curves for automotive applications and a method for curve manipulation based on a shape grammar. This shape grammar deals with only the intrinsic characteristics of shapes, which lend themselves to a concise and reusable vocabulary to describe and compare shapes. The major benefits of this approach for aesthetic design of automotive products are that it establishes a formal basis for shape conceptualization involving high-level constraints and enables a more efficient communication between designers. Development of shape ontologies seems to be straightforward for automotive applications, but it may be more challenging in other application fields, where the aesthetic curves of the product may have arbitrary topological structures. Hence, generalization of this method holds hidden challenges and requires further research.

The second paper titled “Reverse engineering of aesthetic products: Use of hand made sketches for the design intent formalization”, by Mengoni, M., Germani, M. and Mandorli, F., proposes a method for capturing design intent based on reverse engineering of heterogeneous design representations such as points clouds, sketch lines, and textual notes. The proposed approach enables to combine implicit information (lines/curves) and explicit information (textual notes, 3D points) and knowledge extracted from design representations produced by different design techniques, e.g. by sketching and 3D scanning. To connect the explicit and implicit information contained in the sketches to the shape elements of a CAD model, characteristic lines are identified on the point cloud representing the 3D physical object. The point cloud is segmented by characteristic lines, sectioning curves are extracted to help fitting surfaces, and geometrical constraints are identified, in order to govern the subsequent shape modifications. The explicit shape information is used to define a set of rules which support the identification of the styling lines on the point cloud. This multimodal shape design method significantly reduces the reconstruction time of CAD models, and helps to recognize and preserve both explicit and implicit aesthetic information during the entire design process.

Titled “Principal component and Voronoi skeleton alternatives for curve reconstruction from noisy point sets”, the paper of Ruiz, O., Vanegas, C. and Cadavid, C. also involves reverse engineering. This proposes a new surface reconstruction method that takes into account the ‘noise’ of data in scanned point clouds. Noisy data may pose challenges for reconstruction of the transitional region of neighbouring surfaces of local shapes, especially in the case of branching shapes—a case difficult to handle by deterministic algorithms. The proposed method considers the stochastic nature of the point clouds and reconstructs the connecting surface through the, in the statistical sense, best possible geometric locus that represents the point clouds. One of the proposed methods involves principal component analysis (statistical) and Voronoi-Delaunay (deterministic) algorithms for the reconstruction of the transition surface. The other method applies principal component analysis to find a direct piecewise linear approximation of curves. A comparison of the complexity of the two methods is presented together with a qualitative comparison with some previous solutions. An application is also presented, in which transition surface between surface meshes derived from range images of an art piece are generated.

The paper titled “Repairing triangle meshes built from scanned point cloud”, co-authored by Pernot, J-P., Moraru, G. and Véron P., presents a method for filling in holes in triangular meshes reconstructed by reverse engineering. This problem appears in both engineering design and industrial design when geometric data are incompletely measured for certain regions of a physical object. Therefore, the incomplete representation of the surface needs to be completed and the continuity of the surface in the concerned region needs to be reconstructed. A complete toolbox is proposed to fill in the unwanted holes. In a two-step procedure, first the contour of the hole is cleaned from badly formed triangles. Then, a topological grid is inserted and its shape is adjusted, in order to match the shape of the neighbouring region. Additional geometric constraints are specified to control the formation of the shape of the inserted mesh. A prototype software has been developed to demonstrate the application of the proposed method to practical examples.

The fifth paper, titled “Case study - surface reconstruction from point clouds for prosthesis production”, written by Vukašinoviæ, N., Kolšek, T. and Duhovnik, J., presents a high end application of reverse engineering to manufacture orthopaedic prosthesis. The goal of this case study was to investigate the possibilities of computer aided surface reconstruction and supplementation to improve the quality of the prosthesis and shorten the manufacturing time. To this end, a model of a human finger, made of plaster, was scanned with a high fidelity laser scanner. From the scanned data, a watertight, high resolution 3D computer model of the human finger was created, which can be scaled, mirrored or stretched in arbitrary dimensions. The final result is a digital model of the human finger with all surface details such as fingerprints and wrinkles. By additional measurements, the authors showed that the arithmetical average of the deviation of the digital model from the physical model was below a threshold value. They also found possibilities for accelerating the scan alignment while increasing the accuracy, and various solutions for avoiding the influences of geometric errors caused by the 3D laser scanner and the triangulation method. Fast reverse engineering of CAD files are of interest both for industrial design and engineering design, thus the results presented in this paper can be directly adopted by both disciplines.

The paper of Langerak, T. R. and Vergeest, J.S.M., titled “A new framework for the definition and recognition of free form features”, presents a framework which has been conceptualized based on a critical analysis of two existing free form features recognition methods. These two recognition methods have their roots in techniques used for recognition engineering features. The first method, template matching, was originally used to recognize manufacturing features based on matching the graph representation of features to existing templates. The second method, called feature line detection, resembles a hint-based search for regular manufacturing features. The framework is supposed to lead to a better understanding of free form features and to the development of more efficient and uniform methods. However, some problems have also been identified. The assumption that the representations of the entities in the feature library and that of the target shape match is not always the case in the practice. The framework does not offer means for handling multiple representations and needs to be extended to be able to translate representations.

The seventh paper, titled “Shape optimisation of parts in dynamic mechanical systems regarding fatigue”, by Häussler, P. and Albers, A., contributes an integrated structural optimisation process for dynamic mechanical systems based on a new shape optimisation approach. Shape optimisation of dynamic mechanical systems is a very complex task, as it is not possible to reduce the number of load cases to a limited set, due to the need for loading histories. This new approach allows a multi-aspect investigation of dynamically loaded parts in complex mechanical systems and a straightforward optimisation with regard to fatigue. Finite element analysis (FEA), multibody system simulation (MBS), fatigue analysis, and shape optimisation are integrated into a fully automated process. In addition, a method has been developed which enables the coupling between optimisation operations and system dynamics. A test-bed application example is discussed in depth with the goal to outline and illustrate the potentials of this method. The results show that there is a close relationship between the dynamic properties of the parts and the overall dynamic system. The implications of this relationship to the optimisation process are also demonstrated. A more complex application example, namely, optimisation of a passenger car suspension arm, is presented which demonstrates the applicability and feasibility of the optimisation process for real world problems.

The last paper, “Optimisation of a bow riser using the autogenetic design theory”, written by Vajna, S., Edelmann-Nusser, J., Kittel, K. and Jordan, A., presents a case study about the application of the Autogenetic Design Theory (ADT). This was developed based on the idea that new solutions can be produced by using evolutionary methodologies which employ a ’mixed‘ process of searching, adopting of existing knowledge, learning, evaluating, selecting, and combining. In the genetic selection process, good properties of the preceding solutions are passed on to the succeeding solutions. The theory was applied to redesign a riser of a recurve bow with the aim to reduce its weight while achieving the highest stiffness. With the help of ADT, the designer can consider a large number of possible designs, which raises the probability to find the best possible solution. In this case, the ADT helped to reduce the mass by 22% while keeping the stiffness on the same level.

We hope that with this composition of papers we managed to provide good examples of new computational methods and tools that can be used in product design in the industry. We also hope that we could demonstrate the best research practices for fellow researchers by means of these papers. We are grateful to all authors for their contribution to this special issue, and do appreciate their efforts and nice collaboration. We are also indebted to our reviewers who helped us to increase the scientific/professional value and quality of the papers.

   Dr Imre Horváth

   Dr Zoltán Rusák

   Faculty of Industrial Design Engineering

   Delft University of Technology

   Dr Jože Duhovnik

   Faculty of Mechanical Engineering

   University of Ljubljana