Special Focus Simposium on Collaborative Design and Decision Support System, 9th International Conference on Systems Research Informatics and Cybernetics 19/8/97 Baden-Baden Germany
An Expert System for dating buildings in historical centre of towns
Istituto di Cibenretica
del CNR di Arco Felice
Via Toiano, 6 80072 Arco Felice (Na)
Tel.: (+39 81) 8534204 - Fax: (+39 81) 5267654
One of the hardest problems that
distress architects is the chronological classification
of the buildings in the historical centre of towns.
A First approach to solve such a problem by an Expert System was made, in a previous work, taking into account the architectural features and the historical knowledge about buildings. The results obtained, unlikely, weren't good enough. The buildings, in the course of ages, were often subject to modification and recovering, so the chronological classification become very difficult and the system results weren't reliable.
To improve this approach we studied the walls features and the building technics.
Careful analisys of more than three-hundred buildings whose classification was supported by documents give us the skill to develop a new expert system with a knowledge base and an inference engine based on abductive reasoning. Our system is able to date single pieces of a building on the basis of walls features through a sequence of abductive and deductive inferences performed by a question-answer interaction between the user and the system The finall report, i.e. the chronological classification for the part of the building analysed, is represented by the classification hipothesis better supported by initial data and by the answer provided by the user.
Expert System, knowledge rapresentation, chronological classification of the buildings.
The chronological classification of the
ancient monuments represent, for the historians and
architects, one more complex moment of the learning
process about building art's foundation.
Up to now, the experts used the well proved historical-architectural parameters, i.e. they resort to historical knowledges about architectural features and shape features of some buildings element such as windows, portals, etc., to date an edifice. They gave us a powerful means to make such a classification.
In previous works [Burattini 94a,b], we used these parameters to develop an Expert System that, by an analysis of the buildings architectural features, suggested us some hypotheses about its chronological classification
In the specific case of the historical centre of Naples, such a kindd of classification is too complex.
In fact, in the course of about twentyfive centuries, the buildings were often subject to modification and recovering so, in the same building, we can find some pieces wich can be classified in one age side by side with others classified in another age.
The skill of making a chronological recognition of a building by analysing and classifying its walls, constitute a precius resource not much used till now because of the lack of proper study about walls features and building technics used in the course of age.
To fill up that lack, we have made careful analyses of more than three-hundred buildings whose classification were well supported by documents. In such a manner we built a database that includes, for every analysed stratigrafic unit, several information about wall features ( bricks materials, bricks dimension, workmanship, building technics etc. ), other information about its chronological classification, and others about its history.
A new Expert System
From that database we extracted enough
information to design and to develop a knowledge base,
and the, togheter with the expert, we developd an
inference engine based on an abductive reasoning.
We chosen such a model of reasoning since we believe that, to find the solution, the expert usually doesn't make laborious probabilistic calculus. Instead, in the first analysis he considers the observable wall features, then he makes some hypoteses, amongst the others, that better explain those wall features. Then the expert performs a deeper analysis to get new information in order to confirm or to refuse or to introduce a new hypotesis. He repeats this steps until he presumes he found the solution.
Making a deeper analysis of such a process, we can see that the expert accomplishes a sequence of abductive-deductive inference, until he believes he has found the solution.
Following this model we have built a new Expert System, named WALL, that is able to perform a chronological classification of any stratigrafic unit in Naples and in its outskirts, by analysing the walls.
The Knowledge Base consists of several sets of production rules in order to define the connection between the various pieces of knowledge (bricks material, bricks dimension, bricks workmanship, etc.), to define the way of handle the pieces of knowledge, and to deal with the presence, in the stratigrafic unit being analysed, of particular wall features.
The Inference Engine performs and repeats the following tasks in sequence:
1) it performs an abductive reasoning on the basis of the rules set using the available data; (the system starts with some initial data, and then, during the computation, it adds the data provided by the user and the data yielded by the computation)
2) it handles a question-answer interaction between the user and the system to receive the other informatin needed;
3) it performs a deductive reasoning on the basis of the rules set using the answers provided by the user to decide wich question it will propose to to the user
At the end of the computation the
system yields a group of explicative hypoteses about
chronological classification of the analysed stratigrafic
unit ranked in a plausibility order.
The system, during the computation, gives also several explanations about its work. It explains why it proposes a certain a question instead of others, and it shows the partial computations results, i.e. a first-approach classification hypoteses.
Furthemore WALL includes a context-sensitive help that gives explanations about questions meaning, i.e. it explains the meaning of the ambiguouses terms present in a question, and it explains the correct way to answer that question.
Although the system was developed to emulate the experts behavior, it isnt made to replace the expert in his work. Vice versa, it was designed to assist the expert in analysing a building to get a chronological classification.
The results is very satisfactory, infact in the case in which the expert doesnt have doubt on chronological classification, the system and the expert ansewers were identical. Vice versa in the case in wich the expert was uncertain the system proposed more than one classification hypoteses, helping him in choosing the right one.
A hard chronological classification
In the historical centre of Naples
there is a building, named "Palazzo Carafa di
Montorio" that distresses the Neapolitan architects.
In the course of ages, about five centurys, it was
subject to many modifications, so it is very hard to make
a chronological classification. In several papers many
authors, evaluating its architectural features tried to
reconstruct its history in order accomplish its
Well refer to a paper written by one of the most important Neapolitan hystorian of Naples, Cesare Cundari [Cundari 78].In this work, he Claimed that only a piece of the ground floor of the building was built in the XVI century. The first floor was built in the XVII century and finally in the XVIII century the structure of the ground floor was enlarged.
In figure 1 there is a plan of the building. We can see the peices built in the three different ages.
We tried to get the chronological classification using our system. Analysing the walls of the buildings foundations we found that they were built in XVI century, in opposition to the Cundaris hypotesis. Analysing the rest of the building we found some peices in the ground floor belonging to XVI and some belonging to XVII. Finally, analisyng the top floor we found that the whole top floor was built in XVII century.
Our analisys in part confirms and in part disagree the Cundarys hypotesis.
In conclusion we may claim that the building was subject to several modifiation in its architectural features in the course of age, so Cundari was unable to take into account, in the right way, such modifications, and then his chronological classification was incorrect.
[Burattini 94a] - Burattini E., De Gregorio M. - "Qualitative Abduction and Prediction; Regularities over various Expert Domains", Information and Technologies, 19, 471 - 481, 1994
[Burattini 94b] - Burattini E., Fiengo G., Guerriero L. - "Expert System in the building conservation process", International Symposium "Dealing with Defects in Building", Varenna, 28-30 Sett., 303-312, 1994
[Cundari 78] Cundari, C., "Palazzo Carafa di Montorio", Napoli, 1978