/ The Yingzao fashi
project / Teaching materials / Coursework
/ Reaction / Papers
/ Who's who / Related
/ About the materials / Design and construction / Examples / Animations / Interactive models / Student work / Glossary / Readings and bibliography /
In this paper we discuss a method of using virtual modelling to teach and study aspects of traditional Chinese wood construction as prescribed in the building manual Yingzao fashi, published in 1103. There are two types of argument for using virtual modelling. First, there are important conceptual commonalities, the most compelling of which is the distinction between primitive and instance. The Chinese construction system, in its treatment of building components and their relations, makes precisely the same distinction.
Second, the precision of the virtual model and its ease of duplication and modification - its very virtuality - seem to open up new areas of investigation through comparative models. The amount of labor required to make a real model is a serious impediment to comparison studies, which involve multiple models. The meaning of the Chinese construction system lies not in the manufacture of repeated components, but in their combination.
The first phase of this project culminates in the demonstration of a prototype system. We found that users had many operational difficulties, which tended to distract them from more important issues. Based on this demonstration, we propose directions for expansion and refinement in the second phase, which will include use of the system in an advanced undergraduate elective course in architectural history in the fall of 1994-95.
The overall goal of this research is to develop a new approach to teaching and studying traditional Chinese wood construction. We have developed a prototype system in order to investigate the validity of this approach. Based on the results of a demonstration workshop, we propose possibilities for future expansion and refinement.
The Yingzao fashi is the principal source for official Song dynasty (960-1127) construction methods. Indeed, it is one of only two books dealing with the construction of Chinese official buildings; the other is the Gongbu gongcheng zuofa zeli, published in 1733. Although the Yingzao fashi was written for bureaucrats and not for builders, its specificity and prescriptive nature provide detailed information about the forms and dimensions of almost all structural elements in the wooden frame and their organization into the structural frame of a building. Twentieth-century studies of the Yingzao fashi (by Liang, Chen, and others) have made it possible to build architectural models (from wood, for example) and even buildings in the official Song style.
The construction system prescribed in the Yingzao fashi has several significant characteristics. First, it separates structure and enclosure. Like contemporary steel frame and curtain wall construction - and unlike the load-bearing masonry construction of western Europe - the system in the Yingzao fashi consists of a structural wood frame and a non-load-bearing enclosure.
Second, the system is modular. Virtually all dimensions in the Yingzao fashi are given in fen. This is a modular unit, which has eight grades ranging in size from 0.60 to 0.30 cun, or about 19.2 to 9.6 mm. The rank of the building determines both the size of the fen and the size of the building (in bays). (See ill. 1.)
Third, the system employs a small number of what we may call component-types. Each of these has a specific role and position and is repeated individually and in groups throughout the structural frame. Each has only modular dimensions (in fen), and has an actual size (in cun) only in a specific example with an assigned grade. (See ill. 2.)
Virtual modelling seems to have two areas of advantage in the teaching and study of traditional Chinese construction: conceptual commonalities and what perhaps may be called operational characteristics.
There are two important conceptual commonalities between virtual modelling and Chinese construction. The first and most compelling is that both Chinese construction and virtual modelling distinguish the primitive and the instance. The Chinese component-type, such as the cap block (lu dou, ill. 2), is precisely a primitive: it is instantiated as necessary to combine with other instances to form a building. The component-type's modularity and lack of absolute dimensions emphasize that it is not an individual component (or instance), but a type (or primitive). In short, the building (or the model) is constructed of instances and not of component-types or primitives themselves.
In fact, the hierarchy is more complex than simply primitives and instances; there are intermediate levels of what might be called group primitives. For example, a set of blocks (dou) and brackets (gong) combine to form a bracket set (puzuo), which is then manipulated as a unit. (See ill. 4.)
The second commonality is that a virtual model can be made without reference to actual dimensions, mirroring the Chinese modular approach. A virtual model can be reproduced at different scales (different grades of fen) for comparison. As we discuss below, this would be impractical with a real model.
The first operational advantage of the virtual model is that, like texts, spreadsheets, and other virtual artifacts, it can be duplicated and modified effortlessly. This makes possible comparative, "what-if" studies, which would be prohibitively laborious using real models. Virtual modelling opens new areas of investigation.
One such area is departures from orthogonality, for which the Yingzao fashi includes explicit instructions. An example is the curved roof, so often mentioned as one of the characteristic features of Chinese architecture. By using comparative models to study the various rules which contribute to the definition of the multiple curves - the eaves are curved in both plan and in elevation, and the slope is curved in section - it should be possible to understand more deeply the nature of the curved roof and its construction. With real models, this would be so labor-intensive as to be impossible.
Other examples in this area are rules whose purpose seems to be optical, such as cejiao, or the slant of columns away from vertical. Again, through comparative models, it should be possible to identify the individual and combined effects of the rules and thereby to evaluate their possible roles as optical devices.
The second operational advantage of the virtual model, its precision, also figures in the study of cejiao. The Yingzao fashi specifies a slant of 8/1000. Only a virtual model can attain this degree of precision and make it possible to investigate the implications of such a specification.
As a first step in investigating the feasibility of the proposed system, we ran a three-hour workshop for students to do an exercise with a prototype system which we had developed over the previous ten months. (See ill. 3-6.) To introduce the research project and the exercise, we had a brief discussion and gave a computer-based slide show.
The exercise was to construct a model of the structural frame of a small building. We gave the students a set of component-types (primitives) from which they were to assemble the frame. We simplified this task in the following ways:
One student complained that she had no sense of gravity or structure, that the model stood up no matter how she built it. All but one of the students would have preferred a real model at this stage. Clearly, the intangibility, the incorporeality of the virtual model has an important drawback, at least in the prototype system, namely its lack of tactility and kinesthetic feedback.
This does not invalidate our approach; on the contrary, knowing its limits helps us understand its strengths. We believe that tactility is most important in the early stages of acquaintance with the construction system. A wood model reinforces structural ideas because it is structural. The feel of two components locking together, for example, demonstrates how a rigid joint resists moment. Indeed, we would expect that the real model's ability to engage all the senses is an important factor in developing intuition. Turning a bracket set over in one's hands, viewing it from above and below, taking it apart and putting it back together are all ways of learning.
Effective use of the virtual model seems to presuppose an intuition that it cannot teach. We believe, however, that once students have developed that intuition by using a real model, they will be able to operate at a second remove from reality: using the virtual model to pursue more sophisticated inquiries, such as we have mentioned earlier.
We developed the prototype system on the PC platform, using the Solid Modeller of AutoCAD Release 12 in Windows. AutoCAD provides enough functions for us to create detailed building components, but the modelling process is time-consuming and tedious. For example, it took two whole nights to generate the slide show used in the workshop. Manipulating several complex components simultaneously is similarly unpleasant.
Our student users were also confused by the complex functions and tedious procedures involved in manipulating objects. An example is the assembly of a complete bracket set, which is composed of numerous individual objects. First, the user selects the required objects and moves them to their approximate location. Then, he or she aligns them, using their predefined reference points. Finally, he verifies their positions with a three-dimensional coordinate system. Preoccupied by CAD operations, the user can easily overlook the larger issues which the system is supposed to elucidate.
We have initiated three approaches to improve the user interface. The first is to apply object constraints (or relationships) to the primitives. For example, StrataVision 3D and Swivel 3D both provide functions to define the relationship among building components. Defining these relationships not only would help students to build the model efficiently, but would also encourage them to think in terms of functional groups (e.g., a bracket set) rather than of individual objects.
The second is to provide menu templates or graphic-oriented tools for manipulating objects. This would free the student from having to memorize text-based commands. Drafting-oriented CAD systems like AutoCAD typically contain large numbers of such commands, which however do not enhance the studentís understanding above the operational level.
The third is to provide self-aligning reference points on architectural elements. Even with a 3-D mouse, manually aligning the elements of an architectural model in virtual space is time-consuming, difficult, and frustrating.
What types of domain information should our teaching tool address? For the "virtual corporation" of the future, Davidow has proposed four categories of information: content, form, behavior, and activity [Davidow 92]. In the domain of architecture, we propose three categories of information: text-based, graphic, and non-graphic domain-specific.
In the first phase of this research, the prototype system provides only sequential text-based general information (i.e., the user manual) and graphic information (i.e., the 3-D CAD model). One resulting shortcoming is that students identified and manipulated the architectural components only visually. They found it difficult to understand the architectural meaning or structural requirements of an individual component.
We propose to expand the current system by using an information management system (e.g., ADE of AutoCAD) to link non-graphic data to the 3-D primitives in the virtual model. This would enable students, for example, to click on a primitive to inquire about its properties, such as its name, its function, or its dimensions. Still images of the component in a real building can similarly be provided to assist learning.
Students found it challenging and exciting to manipulate and transform objects in order to construct a virtual model of the wood frame of a traditional Chinese building. From the completed base model, they can generate photo-realistic 3-D images and other views.
However, the process of constructing the model is at least as important as the final product. In architecture, an understanding of the construction process is essential to proper design. The architect must consider the structural characteristics of building components and how they are combined. This is unlike virtual modelling in other disciplines. A television advertisement, for example, requires only well rendered photo-realistic models; how these models are created is not important.
Our prototype system provides a computer-based slide show to illustrate the building process of a traditional Chinese building. This is to help students establish a conceptual framework for both design and the construction process. This framework is necessary, because without the constraint of gravity - to mention it once again - acting on the virtual objects, there is no penalty for flouting the logic of the building construction process.
One possible expansion of the prototype system is to implement limitations on the interaction among components. Following the principles of object-oriented programming (i.e., encapsulation and message passing), this mechanism could be embedded in the primitives to follow the constraints of the construction process.
In order to investigate the issues discussed above, we intend to continue with a second phase of research. Over the summer of 1994 we will refine and expand the system along the lines discussed above and will use it in an advanced undergraduate elective course in architectural history in the fall of 1994-95.
We are pleased to thank our research assistant, Mr. Eric Ngai Lik Tsang; our student volunteers, Ms. Alice Cheung Shuk Han, Mr. Bingram Lai Tung Ming, Mr. August Ma Ho Wai, and Mr. William Shum Wai Lap; and our colleague, Mr. Jeff Kan Wai Tak.
This project was funded by a UPGC Direct Grant for Research 1993/94 (small project 220 201 810) from the Chinese University of Hong Kong, which we acknowledge with thanks.
/ Top of papers / Top of reading /