Ocean Engineering 32 (2005) 1968–1981 www.elsevier.com/locate/oceaneng
Technical note
Recognition of design symbols from midship drawings Ho-Jin Hwanga,*, Soonhung Hanb, Yong-Dae Kima a
Maritime Safety and Pollution Control Division, Korea Research Institute of Ships & Ocean Engineering/KORDI, 171 Jang-dong, Yusong-gu, Daejeon 305-343, South Korea b Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology, ME3080, 373-1, Gusong-dong, Yusong-gu, Daejeon 305-701, South Korea Received 5 October 2004; accepted 19 January 2005 Available online 15 June 2005
Abstract Despite the availability of 3D CAD systems, the designers in shipyards still use 2D CAD systems because of the need to produce drawings rapidly and a shortage of labor. The shipyard spends much time and labor analyzing the design information contained in 2D drawings and creating 3D model for the downstream processes. The design information is represented by symbols that are well known among designers in shipyard. Symbols are expressed by meaningless geometric shapes. These shapes are recognized to have meanings by analysis of experienced and knowledgeable designers. We propose a method for automatic recognition of 2D symbols and the extraction of design information from the midship drawings. The shape and rationale of 2D symbols used in ship design have been analyzed, and symbols have been classified according to the analysis. Based on the classified symbols, the developed system recognizes the symbols expressed in 2D drawings. The extracted design information is assorted by design rationales. The meaningless geometric shape is translated into the design information including designer’s intents. The extracted design data can be applied to the design process in shipyards, and the 3D ship model can be automatically created. q 2005 Elsevier Ltd. All rights reserved. Keywords: Design information extraction; Design symbol; Information model; Midship drawings; Symbol recognition
* Corresponding author. Tel.: C82 42 868 7239; fax: C82 42 868 7229. E-mail address:
[email protected] (H.-J. Hwang).
0029-8018/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.oceaneng.2005.01.008
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1969
1. Introduction 1.1. Motivation Extensive studies have been carried out on the automation of ship design and construction to improve the productivity of shipbuilding. As a system integration for data exchange between related companies and departments, the 3D ship product model is to integrate design systems (Brodda, 1991; Wollert et al., 1992). To realize system integration for data exchange, the 3D ship model containing 3D geometry and topology of ship structures should be generated rapidly from the design information created during the initial structural design process. Korean shipyards are developing CAD systems based on the 3D product model. Shipyards have used 2D CAD systems for 20w30 years. The improvement in CAD systems provides increased functionality to designers including the reusability of data, and shipyards can maximize the design capability through 3D CAD systems (Baum and Ramnkrishnan, 1997; Johansson, 1996). Nevertheless, shipbuilding designers who are accustomed to 2D CAD systems could not easily migrate to 3D CAD systems, and the introduction of new systems is a burden to ship designers. Most shipyards and classification societies use paper drawings or 2D CAD files to exchange design information. The data exchanged through 2D CAD files leads to problems in that the representation of shapes and design information differs according to the tools used in different domains and organizations and the same data is stored again in different forms. The designers in shipyards, however, still use 2D CAD systems because of the need for rapid production of drawings and shortage of engineering staff. Fig. 1 shows the barriers to ship design between the contract design and initial structural design processes which are because of representing design information as 2D data. ‘Barrier 1’ occurs when representing design information in 2D drawings during the contract design process. It will be necessary later, during the outfitting design process, to create 3D models, which creates the problem of potential errors and omissions in design information and also costs additional time and manpower. ‘Barrier 2’ occurs due to the use of different CAD systems in the structure design and outfitting design processes. Much information should be exchanged between these two processes, and an integrated system would enable the active
Fig. 1. Barriers to ship design.
1970
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
exchange of design information. To avoid information breakdowns in the design process, Korean shipyards are recently developing next-generation CAD systems for shipbuilding. This, however, is not a fundamental solution for system integration and data exchange. As the problem occurs due to the exchange of design information with 2D drawings, most shipyards and classification societies spends much time and money for the approval of ship structures. The 2D drawings created during the contract design process are transferred to downstream processes, and the designers in downstream should create 3D models by applying the design rationale represented in 2D drawings. The designers spend time and labor to analyze and interpret the design information described in the drawings. The design information in a midship drawing is represented by symbols that are known and agreed upon by professional ship designers. Seemingly meaningless geometric shapes such as points, lines, curves, and text in the 2D CAD systems are understood as symbols by the designers who analyze and interpret them based on their experience and knowledge. We have proposed a method of extracting design information by recognizing 2D symbols. A 2D midship section drawing, which is the result of the contract design process, contains 2D symbols. The shape and rationale of 2D symbols used in ship design has been analyzed, and the symbols are classified according to their functions. Based on the classified symbols, the developed system recognizes symbols expressed in 2D drawings and extracts the design information. The geometric shape is translated into the design information as intended by the designer. The extracted design data can be applied to the design process in shipyards, and a 3D ship model can be automatically created by a 3D enhancement process using the data (Lovdahl et al., 1994). 1.2. Previous studies Various reports have been published on the extraction and representation of design information from 2D drawings. There have been researches in mechanical domain for the reconstruction of 3D models from 2D drawings. To generate 3D solid models from 2D drawings represented by trigonometry, Shin and Shin (1998) used the geometric properties and the topology of geometric primitives. Shum et al. (2001) proposed a 2-stage extrusion method that reconstructed a solid by modifying the incremental extrusion. Gorti et al. (1996) mapped an evolving symbolic description of design into a geometric description. Bidarra and Bronsvoort (2000) defined semantic features and proposed a modeling method based on semantic features. All of the above research used the method of assigning functional meanings to user’s event using predefined features, but does not deal with the reusability of legacy drawings. Lee and Han (2000) defined the 2D patterns of feature shapes and reconstructed 3D models from 2D drawings of mechanical parts based on defined patterns. Shin and Shin (1998) created a 3D shape using the feature recognition method of 2D midship drawings. Lee and Han (2002); Shin and Han (1998) had dealt with the recognition of features from 2D drawings. They had only focused on the reconstruction of the 3D shape of a product, however, and did not consider the configurations and properties of a product model. Most reconstruction methods have been focused on solid shape models and machining features of mechanical parts represented by trigonometry or orthogonal views. They cannot be directly applied to the shipbuilding industry, which represents design information with symbols, and are it is difficult to extract design
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1971
information from design symbols. Commercial CAD systems such as Solidworks, Pro/E, and I-DEAS currently provide a module which extracts and utilizes information from 2D drawings. These systems have adopted a sketch method that uses contour profiles in 2D drawings, not a method of recognizing design symbols to create 3D models. Some CAD systems like SolidEdge (EDS, 2003) provides an automatic reconstruction from 2D drawing, but it can only recognize a drawing represented by trigonometry and limited to mechanical parts. The reconstruction and recognition modules in these systems are not fully operational for representing design rationale with symbols in the shipbuilding industry. There have been studies regarding the product modeling of early ship structural information. Kim et al. (1991) introduced an object-oriented concept for the integration of CAD information for plates of ship structures. Yum (1995) proposed a methodology for building ship hull model with an object-oriented method. Aoyama and Nomoto (1997) modeled ship structural geometry using the product model concept and the object-oriented concept. The integration of the ship design process and the production process is possible by defining the data structure of the product model. The model can be used to analyze the hull and to implement the computerized model. Lee et al. (2002) proposed a data structure that defines the semantic data model of a ship for the early design stage. They use UML (unified modeling language) for the class modeling. Lee also developed and verified a CAD system that supports the proposed data structure. Lee analyzed 2D data that is generated in the initial ship design stage and proposed a semantic data structure to create a 3D ship model. His approach is different from the proposed information model in this paper in that we extract design information with symbols in midship drawings and store the midship data into the 2D information model in the same manner as represented in the drawings. 2. Information model of midship drawings The design information represented in midship drawings shows conventional or similar patterns for the same ship type. To define the shapes of plates, parameters that depend on the properties of plates are used (Kobayashi et al., 2002). An engineer designs a ship structure by setting values of the parameters of the plates. Because the parameters of ship structures represented in 2D drawings have been analyzed and structured, the recognition system can store the design information acquired by recognition of design symbols. The user can also add insufficient information through the developed user interface that could not be acquired by drawing recognition process. We have developed the information model of midship drawings that can hold design rationale obtained from 2D drawings of ship structures (Hwang et al., 2002). Fig. 2 shows a partial representation of the information model with EXPRESS-G (ISO, 1994). We have used VisualExpress of EPM’s EDM (Express Data Management) system (EPM Technology, 2002). The information model is not a part of ship STEP of ISO standards, and is specialized for the design information of 2D drawings. The root entity of the midship section is defined as midship_model. It has three subtypes such as general_ information to describe general data of a ship, configurations to describe the shape and outline of the ship structure, and piece_definitions to describe the detail shapes and properties of each structural piece.
1972
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
Fig. 2. Information model for 2D drawing using EXPRESS-G notation.
Fig. 3(a) shows the entity configuration items of an inner section. The breadth and depth of the ship are defined in the entity general_information, and the information of the inner bottom plate can be calculated from the double bottom height. The information of the hopper tank plate and side shell plate can be obtained using the x, y coordinates and width of the hopper tank. In this fashion, the entity configurations of midship section can be defined by parameter representation in 2D drawings. Fig. 3(b) shows the entity piece_definitions of the bottom plate. The layout of bottom plates can be generated by the breadth and depth of a ship from the entity general_information and the bilge radius from the entity configurations. The location and seam information of plates for the bottom plates are defined based on this layout. The entity piece_definitions in Fig. 3(b) is defined as follows: center is the start point offset from the centerline; plate width is the width of each plate; bilge plate height is the width of the bilge plate.
3. Recognition of midship drawings 3.1. Symbol representation in midship drawings The midship drawing contains geometric information, which expresses shapes of structural pieces, and the non-geometric information, which describes properties of pieces
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1973
Fig. 3. Entity definitions of configurations and piece_definitions.
such as dimensions and materials. A data model that can represent the geometric and nongeometric information is needed to convert midship drawings into digital version of design The ‘data model’ should have not only geometric shape representations but also semantic representations of ship structures (Kim et al., 1991). The semantic representation is implicitly described in 2D drawings. ‘Implicit’ representation means that designers describe the design rationale with symbols as a kind of series of rules, and the design information of ship structures is expressed using design symbols. Fig. 4 shows an example of symbols represented in midship drawings. The centerline symbol of the midship section is shown in Fig. 4(a). The seam line symbol in ship plates is described in Fig. 4(b) and it is made of one line and two arcs. Designers use a seam line symbol to divide one plate into two or more pieces. The seam line is utilized for welding data in downstream processes. The longitudinal symbol is represented in Fig. 4(c) and it
Fig. 4. Example of symbols in 2D drawing.
1974
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
Fig. 5. Different semantics by representation of symbol.
constitutes two lines to represent a shape, numerical characters to express the longitudinal number, and text to describe the properties of the members. Design symbols are described simply by points, lines, curves, and texts in 2D drawings and do not have the functional meanings of ship structures. In the case of the seam line symbol, 2D CAD systems can only represent shapes formed by one line and two arcs but cannot represent meanings of the seam line. Fig. 5 shows the different semantics according to the representations of symbols for the same piece. The stiffener expressed by a solid line as shown in Fig. 5(a) is located in front of the reference plate, and the stiffener expressed by a dotted line as shown in Fig. 5(b) is located to the back of the reference plate. Texts and numerical values placed on a stiffener symbol mean properties of the stiffener such as web length, flange length, thickness, and materials. The description of properties should be expressed with structural symbols and includes the data that cannot be represented by symbols of structural members. In the case of the longitudinal symbol, as shown in Fig. 5(c), the properties such as the longitudinal number and specifications are described together. As described above, the form of a symbol is 2D geometric shapes, but the contained meaning in the symbol is 3D structural piece. The 3D design information expressed by symbols can be recognized by designers. If symbols are systematically comprehended and the system extracts design information from the symbols, the design information described in a 2D drawing can easily be translated into a 3D model and downstream applications in the design process can utilize the model. 3.2. Recognition of design symbols Fig. 6 shows the representation of a longitudinal symbol on a midship drawing and an example file based on the DXF specification. The longitudinal symbol constitutes two LINE entities and one TEXT entity in a DXF file. Two LINE entities are joined by the start point of one entity and the end point of the other entity. The entities are connected within the defined error bounds and the two LINE entities are at right angles to each other. A TEXT entity is placed within an error bound determined by a range between flange and web. The TEXT entity describes the longitudinal number as an integer. The recognition system interprets the group of entities as a longitudinal of ship structure based on the relationships of entities.
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1975
Fig. 6. Representation of longi. Symbol.
The system parses DXF file into LINE entities, ARC entities, CIRCLE entities, and TEXT entities. The locations and properties of parsed entities are stored and classified based on the specification of the entity. The relationships between entities are analyzed. The connected and related entities are stored in the candidate list of symbols. The system compares the candidate list with the pre-defined feature symbols of the midship section and finds the meanings of entities. There are no direct relationships among geometric entities of DXF file. Human designers can recognize a symbol by combining several geometric shapes, but the system cannot recognize a symbol based only on a DXF file. The relationships between the entities that represent geometric shapes are needed to recognize a symbol. Fig. 7 shows the process of analyzing symbols in the system. DXF entities are parsed, and the parsed entities are classifies based on contents of the entities. The system searches for connected entities and predefined ranges of the entities and analyzes the relationships among entities. The candidate symbol list is prepared by assigning the entity connections and by grouping the related entities. The system can finally recognize a symbol by comparing the candidate symbol list with the pre-defined symbol lists. The shaping rules of symbols represented in 2D drawings are defined in the symbol lists. Fig. 8 shows an example of a shaping rule for symbols in drawings. The rule for a longitudinal symbol is shown in Fig. 8(a). The longitudinal symbol is represented by two LINE entities for a flange and a web, respectively. The flange and web entities are perpendicular each other. The LINE entity, which represents the reference plane, is searched at the end point of the web plate. The system analyzes the web plate whether it is perpendicular to the reference plane, parallel to the z direction, or y direction. If the group of entities satisfies the above conditions and the numeric text of longitudinal exists within
1976
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
Fig. 7. Process of symbol recognition.
the predefined range, the system can recognize the group of entities as the longitudinal symbol. Fig. 8(b) shows a recognition rule for the feature symbol of hopper tank web or topside tank web. A closed contour is searched for around of hopper tank or topside tank. As shown in Fig. 8(b) the geometric shape is a closed contour. The feature symbol of a web can be recognized by the loop contour and inside fillet information and outside offset information. These rules define the symbol representations in 2D midship drawings. The system compares the symbols list with groups of entities and recognizes a symbol from midship drawings. To recognize 2D midship section drawings, the system firstly interprets the centerline and the baseline of the midship section. As the symbols which represent centerline and baseline are recognized, the datum point is established by the recognized lines. In certain cases there are two or more centerlines and baselines on a drawing. The system recognizes them as multiple sections of ship structures. Based on the recognized datum information, the layout information of ship structures such as the inner bottom height, tank position, and the camber height is extracted by recognizing the design values from dimension lines.
Fig. 8. Shaping rules of symbols in midship drawings.
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1977
Fig. 9. Sequence of symbol recognition.
The locations of functional structures are classified using the extracted information and the system generates the basic layout of midship structures. The detailed information for plates is extracted from the seam line symbol and dimension values. The dimension value described in midship drawings is represented by parameters and is located between two seam line symbols. Previous researches cannot deal with the seam symbol and parameters of structural plates. The information of stiffeners, such as longitudinal stiffeners and stringer, is finally extracted. The stiffener symbol is expressed by properties in the form of texts together with a symbol. The properties of stiffeners are sometimes described as a table. Fig. 9 shows the sequence of symbol recognition. As symbols can be systematically interpreted and automatically recognized, the system can finally construct 3D ship models from 2D midship drawings.
4. Implementation and test The graphical user interface provides popup windows to express the design information with parameters in a similar manner to the way they are represented in 2D drawings. Fig. 10 shows the GUI expressing bottom and bilge plates with seam information. The user can provide the properties of the plates such as the starting point from the centerline, the width, the thickness, the material of the bottom plates, and the height of the bilge plate. The user can provide the parameter values for the plates in the same manner as represented in the drawings in addition to the recognized parameters. The design information prepared through the feature parameter input method transforms design rationale represented in drawings into the 2D data structure of Fig. 3. The recognition system has been implemented in a Windows NT environment. The DIME (DXF Import, Manipulation, and Export) library has been utilized to parse DXF files. We have used the EDM system (EPM Technology, 2002) of EPM to prepare EXPRESS-G diagrams. We have also used MS Visual Basic 6.0 for the development of GUI, and MS Visual CCC6.0 for the development of the module to store the information model of midship structure. Fig. 11 shows the architecture of symbol recognition system. The system has two modules, the 2D drawing recognizer and the midship data completion module. A 2D DXF drawing of the midship section is used as the input data. The system
1978
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
Fig. 10. GUI for piece definitions.
parses the DXF entities and finds entity relationships among disconnected by analyzing the forms of symbol entities. The inter-related entities are compared with the symbols list that store patterns of symbols which appear in midship drawings, and the compared entities can capture the meaning of symbols. The captured design information is stored in the data structure for 2D drawing information model of Fig. 8. The user can additionally provide insufficient design information through the parameter input GUI. The system stores design
Fig. 11. Architecture of the symbol recognition system.
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1979
Fig. 12. Data Specification of a ship model using symbol capture and GUI.
information captured by the symbol recognition method and the additional information provided by the GUI in the data structure based on the information model of midship drawing. The developed system has been tested and verified with a real ship data from a Korean shipyard. The midship of a bulk carrier has been used in the experiment. Fig. 12 shows an example of data specification of the midship model. It contains the design information captured from symbols in midship drawings and the additional information through GUI provided by users. The design information can be used in the structure analysis system and the ship CAD system for detail design. It also can be enhanced into a 3D ship model through the mapping of the STEP AP218 schema for ship structures (Hwang et al., 2002, 2003), and used to exchange information between design departments in a shipyard as well as between classification societies and shipyards. The 3D ship model in Fig. 13 shows a 3D ship model transformed from a midship drawing for a bulk carrier. The 3D ship model in Fig. 13 is generated from the design information that has been captured by symbol recognition of the 2D midship drawing and added by the parameter GUI. We can see the shape of midship sections that are composed of many plates.
5. Conclusion Despite improvements in 3D CAD systems and the effectiveness of 3D models, most shipyards and classification societies still use 2D paper drawings and 2D CAD files to exchange design information, which is both time-consuming and costly. We have proposed a method of extracting design information from symbols contained in midship drawings and
1980
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
Fig. 13. Visualization of 3D ship model.
storing the data to an information model for a midship. The symbols in drawings are analyzed and the design information can be captured by representing the symbols as features. The information model of midship drawings, which is represented by the same manner as midship drawings, has been utilized to store captured design information. To supplement the insufficient recognized information, a GUI that supports the information model of midship drawings has been developed. The developed system, which is based on the symbol recognition using symbols and information model of midship drawing, has been tested and verified with real ship data. The design information captured from the drawing can be transformed into 3D ship models through 3D enhancement process. The enhanced ship model can resolve the delay of the design period that occurs due to the barriers described in Fig. 1 and be exchanged between the engineers of structural design and outfitting design. The proposed method of design capture using symbol recognition of 2D drawings and storage of data in a similar manner to the way designs are represented in 2D drawings can be extended to the building and construction industries, which also describe design information using symbols. This method can maximize the design capability of experienced designers accustomed to 2D CAD systems, and provides an effective design environment for the transfer of design information to downstream processes.
H.-J. Hwang et al. / Ocean Engineering 32 (2005) 1968–1981
1981
Acknowledgements This research has been partially supported by the Development of Maritime STEP Technology (M1-0104-00-0265) project from the National Research Laboratory (NRL) Program of the Ministry of Science and Technology of Korea.
References Aoyama, K., Nomoto, T., 1997. Information models and functions for CIM in shipbuilding. Journal of Marine Science and Technology 2 (3), 148–162. Baum, S., Ramnkrishinan, R., 1997. Applying 3D product modeling technology to shipbuilding. Marine Technology 43 (1), 56–65. Bidarra, R., Bronsvoort, W.F., 2000. Semantic feature modeling. Computer-Aided Design 32 (3), 201–225. Brodda, J., 1991. Shipyard modeling—an approach to a comprehensive understanding of functions and activities. Journal of Ship Production 7 (2), 79–93. EDS, 2003. SolidEdge Drafting tools. http://www.solidedge.com/prodinfo/drafting.htm. EPM Technology, 2002. EXPRESS Data Manager, http://www.epmtech.jotne.com/products/index.html. Gorti, S.R., Ram, D., Sriram, R.D., 1996. From symbol to form: a framework for conceptual design. ComputerAided Design 28 (11), 853–870. Hwang, H., Han, S., Kim, Y., 2002. Mapping 2D midship drawings into 3D structural models based on STEP AP218, Proceedings of the Geometric Modeling and Processing (GMP 2002) 2002 (pp. 226–232). Hwang, H.J., Han, S., Kim, Y.D., 2003. Generation of 3D STEP model from 2D drawings using feature definition of ship structure. Transactions of the Society of CAD/CAM Engineers 8 (2), 122–132 (in Korean). ISO TC184/SC4/WG11, 1994. ISO/IS 10303 Industrial automation systems and integration—Product data representation and exchange Part 11: Description methods—The EXPRESS language reference manual (only available in hardcopy, http://www.tc184-sc4.org: WG11/N149 http://www.tc184-sc4.org/SC4_Open/ SC4_and_Working_Groups/WG11/N-DOCS/Files/wg11n145.pdf). Johansson, K., 1996. The product model as a central information source in a shipbuilding environment. Journal of Ship Production 12 (3), 106–119. Kim, Y., Lee, K., Seo, S., Ryu, H., Hong, C., 1991. A study on the relational data model for the representation of ship structure. Journal of the Society of Naval Architects of Korea 28 (2), 12–20 (in Korean). Kobayashi, H., Sasaki, Y., Suzuki, T., Yoneya, T., 2002. Development of a new direct strength assessment system for double hull tankers, Proceedings of the 11th International Conference on Computer Applications in Shipbuilding (ICCAS 2002) 2002 (pp. 35–41). Lee, H., Han, S., 2000. Reconstruction of 3D CAD models from 3D drawing files, Proceedings of the ‘2000 ASME Computers and Information in Engineering (CIE) Conferences 2000. Lee, W., Lee, K., Roh, M., Kwon, O., 2002. Development of an initial hull structural CAD system for computeraided process planning (CAPP), Proceedings of the 11th International Conference on Computer Applications in Shipbuilding (ICCAS 2002) 2002 (pp. 113–123). Lovdahl Jr.., R.H., Martin, D.J., Polini, M.A., Wood, R.W., Geradi, M.L., Lazo, P.L., Wooley, D., 1994. The NIDDESC ship product model: the STEP solution. Journal of Ship Production 10 (1), 39–50. Shin, Y., Han, S.-H., 1998. Data enhancement for sharing of ship design models. Computer-Aided Design 30 (12) , 931–941. Shin, B.S., Shin, Y.G., 1998. Fast 3D solid model reconstruction from orthographic views. Computer-Aided Design 30 (1), 63–76. Shum, S.S.P., Lau, W.S., Yuen, M.M.F., Yu, K.M., 2001. Solid reconstruction from orthographic views us ing 2-stage extrusion. Computer-Aided Design 33 (1), 91–102. Wollert, J., Lehne, M., Hirsch, B.E., 1992. Modeling for ship design and production. Journal of Ship Production 9 (1), 48–58. Yum, J., 1995. A study on the object-oriented structural modeling of double hull tankers and the automatic generation of structural analysis model. PhD Thesis, Seoul National University, The South Korea. (in Korean).