A Systems Approach to the Computer Aided Design of Reinforced Concrete Structures

Transcription

1 A Systems Approach to the Computer Aided Design of Reinforced Concrete Structures Fátima Farinha 1), João Bento 2) and David Blockley 3) 1) Universidade do Algarve, IPF, Quinta da Penha 8000 Faro, Portugal 2) Instituto Superior Técnico, DECivil, Av. Rovisco Pais 1096, Lisboa Codex, Portugal 3) Univ. of Bristol, Faculty of Eng., Queens Bldg., University Walk, Bristol, BS8 1TR, U.K. ABSTRACT The present paper refers to the initial results of the development of a knowledge based system for the computer aided design of reinforced concrete structural elements. Firstly, a model of the design process specially motivated by structural engineering needs, is presented. The proposed model combines the SAE (Synthesis-Analysis-Evaluation) model, which represents the design process as an exploratory process composed of recursive cycles of synthesis, analysis and evaluation, with the RPL (Reflective Practice Loop) model which represents the design activity as a hierarchically structured set of problem solving processes. The paper follows by describing a developing knowledge based prototype system that implements the most important aspects of the proposed model to the detailing of reinforced concrete columns in buildings. The paper closes with the main conclusions on the work done so far. INTRODUCTION An important objective of structural engineering is to produce artifacts that are safe and reliabil. A model of the design process will be helpful to engineers. Such a model should be consistent at all levels, make use of new knowledge as it evolves and, preferable be able to produce explanations.

2 A systems approach, as an alternative to scientific reductionism, has been used to model complex systems. The scientific method is based on the idea that one can take a complex problem and break it into component parts and tackle each part separately and quasiindependently. The obvious penality of concentrating on a part of a problem is the loss of perspective on the whole, a lessening of awareness of relationship between parts of the problem. Therefore, such reductionist approach may only work when connections are not significant. A 'system' embodies the idea of a set of elements connected together which form a whole and this shows properties which emerge as properties of the whole rather than properties of its components (Checkland, 1981). DESIGN PROCESS MODEL What is a design activity The first difficulty encountered in modeling design processes is that design may involve activities as varied as detailing a reinforced concrete column by engineers, composing a symphony by a musican or even drawing up a political strategy by a politician. Within the scope of the present work, it is essential that distinctive characteristics of design activities be identified in order to avoid limitations arising from a too general treatment. Goel and Pirolli (1989), identified a set of features of design processes, which they called invariants, and these are "good" examples of design activities. For these authors, the complexity of the design activity prevents the drawing of precise borders. They consider that design activities reflect radial properties, that is, that "good", "ideal" or "central" examples can be found in addition to variations of these "central" cases. The characteristics, that identify good examples of design activities, are: I 1 -large and complex problems; I 2 - input as goals and intentions and output as an artifact specification; I 3 - long temporal separation between the design phase and the delivery of the artifact; I 4 - delayed or limited feedback from the world; I 5 - independent functioning of the artifact; I 6 - costs, penalities or benefits, associated with every action; I 7 - predominance of better/worse answers over right/wrong ones; I 8 - many degrees of freedom. Design as a problem solving process Throughout this work design is considered as a problem solving activity. Design as a problem solving process was first presented by Simon (1969). Problem solving may be summarized as the process of finding solutions in a problem space, which represents possible states of the problem (i.e. possible problem descriptions) to be considered in attempting a solution. Such a process consists of searching in that space, for a path connecting an initial state of known facts and conditions to a final one - the goal state. Three hypotheses are adopted: h 1 - design process is a stepwise transformation process from a kernel idea; h 2 - any artifact is satisfactorily described by the goal state; h 3 - design process is limited and driven by constraints. Figure 1 shows this essential idea: design process is a stepwise process of exploration in design space which, from an initial specification of the problem - S o (often incorrectly defined),

3 generates a kernel idea - S k and refines it (by decomposition, generation or transformation) in search of a goal state S g. Design model The proposed model is strongly based on other models, namely on the SAE (Synthesis- Analysis-Evaluation) model put forward by Feijó (1988) and Bento (1992) and on the RPL (Reflective Practice Loop) proposed by Blockley (1992). Figure 1 - Design as a problem solving process It is considered that the design process (Figure 2) begins in a sub-process of PERCEPTION that basically consists of an awareness that the design state is not the goal state; it therefore becomes necessary to change the state - and this sub-process is called REFLECTION. State evolution is then processed though recursive cycles of synthesis (S), analysis (A) and evaluation (E), in which each of these processes is also recursive, that is, each stage of synthesis, analysis or evaluation may set off another SAE cycle. From a computational point of view, the conclusion of a SAE cycle basically corresponds to one of three

4 transformations - entity creation, entity transformation or entity elimination - and leads to a decision-making sub-process - ACTION, which is seen as the creation of a new design state. Each design state (S i ) consists of a set of entities whose properties can be defined in terms of identification attributes (e.g. entity designation), of relation (e.g. relationships), of structure (e.g. physical characteristics) of function (e.g. performance specification). The sequence of states (S o, S k, S 1, S 2,...,S g ) is known as the design history. The evolution of states (i.e. S i S j ) occurs in a space showing the following characteristics: C 1 - extensive problem structuring (an extensive process of finding missing information); C 2 - extensive performance modelling (e.g. cognitive models, scenario immersion, pictorial models); C 3 - personalized evaluation functions (for exemple criteria used to terminate the design process); C 4 - evaluation in three contexts: local context, current context and future context; C 5 - making and propagating decisions (or commitments); C 6 - solution decomposition; C 7 - abstraction hierarchies that is working with entities at various levels of detail; C 8 - use of symbol systems, i.e. bubble diagrams and rough sketches. A detailed analysis of these characteristics may be found in Bento (1992). Figure 2 - Design process model

5 KNOWLEDGE BASED PROTOTYPE SYSTEM The chief objective of the prototype system is to implement the main aspects of the proposed model for sizing and detailing columns of reinforced concrete buildings. The part of the system that had already been implemented and tested involve analysis of columns, of a square or rectangular section, under normal forces, combined bending and double combined bending and detailing of columns, of a square section, under double combined bending. Programming considerations The prototype was implemented using a commercial tool - KAPPA-PC (version 2.3) - from IntelliCorp, Inc. running under MS Windows. Using this tool knowledge was structured hierarchically (Figure 3). The two important aspects that require detailed consideration are the representation of design states and the process by which states progress. The design states are represented by design entities and design entities are represented as KAPPA-PC objects. As an example (figure 3) an entity represented in KAPPA-PC by an object belonging to the Double_Combined_A class inherits all the attributes of Combined_Bending_A class. Some of Double_Combined_A's attributes are shown in figure 4 (e.g. FCD is a function attribute and NOME is a identification attribute). Figure 3 - Hierarchic structure

6 Figure 4 - Design entity and some attributes The second aspect - the evolution of states was solved using KAPPA-PC's features: methods, functions and rules. From a computational point of view the process of promoting changes from one state to another corresponds to one of three entities' modifications: entity creation, entity transformation or entity elimination. Functional and operational considerations Data imput occurs interactively; the user answers the questions put forward by the system and sometimes selects from several options. Units are always indicated and it is possible to change input data without restarting the entire process. Analysis results are displayed in an attractive user interface and are stored on a file only after the designer has accepted them; these files contain all information about the designed element and can be integrated into the technical report of the project. Calculation time is short, so the designer may experiment making of changes in order to improve the design. In the detailing module, the prototype lists a set of possible solutions and the designer is entirely responsible for the selection of the option to be detailed by the system. In some circunstances the system details, for the solution selected by the designer, more than one possible solution. The differences may lie in the positioning of bars or in the type of stirrups (Figure 5).

7 CONCLUSIONS Figure 5 - Knowledge based prototype system - detail of reinforcing bars The work covered by this paper is a preliminary version of a system that is still under development. Nevertheless the following conclusions may already be drawn. (1) The proposed model is based on a cognitive interpretation of the design process, provides a systems approach to the design activity and takes into account the problem solving nature of the design process. (2) The implementation process, of putting into practice the main ideas contained in the model, has achieved the status of a working tool. Two possible applications became immediately apparent: - to help human decision making; - to be used for learning. (3) The system can incorporate the designer's own procedures and "grow" relatively easily. It is sufficient to change rules and/or to add new rules to the knowledge base. This characteristic is deemed essential in any knowledge based system and results from the separation between the knowledge base and the inference mechanism. (4) The use of the commercial tool (KAPPA-PC) in the implementation process led to a reduction of implementation time and of the degree of specification required. ACKNOWLEDGEMENTS The first author wish to express her thanks to Junta Nacional de Investigação Científica e Tecnológica and to the Department of Civil Engineering at EST - Algarve University.

8 REFERENCES Bento, J. (1992), Intelligent CAD in structural steel:a cognitive approach, PhD Thesis, Expert Systems Laboratory, Department of Civil Engineering, Imperial College of Science, Technology and Medicine, University of London, U.K.. Blockley, D. (1992), "Engineering from reflective practice", Research in Engineering Design, pp Checkland, P. (1981), Systems thinking, systems practice, John Wiley & Sons. Eurocode Nº 2 (1988), Design of concrete structures. Part 1: General rules and rules for buildings, European Committee for Standardization. Feijó, B. (1988), Fundamental steps towards an intelligent CAD system in structural steel, PhD Thesis, Expert Systems Laboratory, Department of Civil Engineering, Imperial College of Science, Technology and Medicine, University of London, U.K.. Forsyth, R. (1984), Expert Systems: principles and case studies, Chapman and Hall. Goel, V. and Pirolli, P. (1989), "Design within information-processing theory - the design problem space", The AI Magazine, Spring 89, American Association for Artificial Intelligence, pp KAPPA-PC (1992), Reference manual, IntelliCorp, Inc.. REBAP (1983), Regulamento de estruturas de betão armado e pré-esforçado, Decreto-Lei nº 349-C/83 de 30 de Julho. Simon, H. (1969), The sciences of the artificial, MIT Press, Massachusetts, U.S.A..