Sie sind auf Seite 1von 17

ELSEVIER

Automation

in Construction

6 (1997) 121-137

KBES for evaluating R.C. framed buildings using fuzzy sets


Xilin Lu a,*, Sidney H. Simmonds b
a Research Institute of Engineering Structures, Ton@ Uniuersity, Shanghai, 200092, China b Department of CiGi Engineering, The University of Alberta, Edmonton, Alberta, T6G 2G7, Canada

Abstract A me~odology for developing a knowledge-based expert system for assessing the structural condition of existing reinforced concrete buildings is proposed in this paper. The procedure incorporates the results of field observations and, if necessary, strength computations of individual members and combines these using weighting factors to obtain the overall structural condition of the building. The lack of precision associated with verbal descriptions used in recording some field observations is expressed using fuzzy set notation. Elements of fuzzy set theory are also used in the reasoning process. A KBES incorporating this meth~olo~ and using expertise based primarily on the experience of a team of investigating engineers at Tongji University, China, is described. 0 1997 Elsevier Science B.V.
Keywor& Building structure; Expert system; Fuzzy set theory; Reinforced concrete; Safety; Structural evaluation

1. Int~~u~ion

The reason for evaluating the structural safety and overall serviceability of existing reinforced concrete buildings is diverse and includes such cases as general aging of the structure, change in occupancy, proposed major renovations or the result of natural disasters such as fires, wind storms or earthquakes. The traditional procedure for evaluating buildings consists of a number of steps, each of which is time consuming and requires considerable experience and judgement on the part of the investigating engineers. Therefore, especially after a major natural disaster such as an earthquake when there may be many buildings requiring immediate evaluation, it is desir-

* Corresponding author. Discussion is open until December 1997 (please submit your discussion paper to the Editor on Construction Technologies and Engineering, M.J. Skibniewski). 0926-5805/97/$17.00 0 1997 Elsevier Science B.V. All rights reserved. PII SO926-5805(97)00012-5

able to have a simple but reliable method of assisting the structural engineer in the assessment process. This paper describes a framework for the development of an expert system to assist in the evaluation of reinforced concrete framed buildings. While the procedures used to evaluate the structural safety and condition of existing buildings may vary somewhat depending on the behavior of the structure and the reason for the evaluation, most procedures incorporate the following steps. The first is to study the original design and construction documents and to ensure that the structure was built in accordance with them. If such documents are not available, it may be necessary to survey the building to obtain measurement and details of the structural framing. The second step is to examine visually different members of the structure for their physical condition. The third step is to obtain an overall evaluation of the structural condition of the building. This may involve an analysis of the structure to determine the internal forces required in each mem-

122

X. Lu, S.H. Simmonds/Automation

in Construction

6 (19971 121-137

ber and a judgement of the ability of each member to resist these forces. Should the investigator have confidence in the original design and there are no signs of distress in the individual members, this third step may be omitted. In designing an expert system for evaluating reinforced concrete framed buildings using the above steps, it is necessary to develop a methodology for combining field observations, numerical calculations and structural expertise. In many instances it is desirable to give a mathematical significance to the results of visual observations that are often expressed in linguistic terms. The lack of precision associated with words used to describe field observations and variations between observers can be expressed by using the fuzzy set theory. In the evaluation of any structure, decisions must be made on the weighting to be given to the different observations and calculations relating to the strength and serviceability of individual members and to their effect on the overall structure. Such decisions must be based on the experience and knowledge of engineers who have expertise in this domain. In the program described in this paper such decisions are based primarily on the observations and accumulated data of a team of investigators at the Research Institute of Engineering Structures, Tongji University, Shanghai, China. During the past five years this team has evaluated the structural condition of more than 45 reinforced concrete buildings with different types of structural framing.

make precise statements about that problem decreases. This motivated him to develop the theory of fuzzy sets. In fuzzy set theory an observation may be assigned membership to more than one element in the set. To demonstrate this concept, consider a set that represents the quality of construction. The elements to define all possibilities are selected as GOOD (above average), AVERAGE, POOR and BAD (unsatisfactory). Classification of an observation is made by the investigating engineer. It is certain that there will be cases when the proper classification of construction quality for a given member is not obvious and, indeed, there is an inherent vagueness in the terms defining the elements of the set. These uncertainties can be incorporated using fuzzy set theory. Let Q denote the fuzzy set defining quality of construction and G, A, P and B the elements in that set. The case where the engineer describes the construction quality as GOOD may be expressed as: Q = (1.0/G + 0.5/A + 0.1/P + 0.0/B}. (1)

2. Fuzzy set theory A set is a collection of elements that define all possibilities in a probabilistic problem. These elements may be finite and discrete or infinite and continuous. Many problems in engineering are probabilistic in nature and decisions are based on data collected. Set theory provides a convenient means of grouping this data and a framework for mathematical interpretation. In crisp set theory an observation (sample point) either belongs (membership value of 1.O) or does not belong (membership value of 0.0) to a particular element in the set. Zadeh [l] recognized that as the complexity of a problem increases our ability to

Eq. (1) indicates that the most likely membership is in the element GOOD but that there is also a possibility that it has membership in an element denoting lesser quality of construction. The value assigned to the degree of membership in the other elements when a particular element has been selected depends on the particular characteristics of that problem represented by that set. In some instances, it is convenient to normalize the membership values in a set, that is, to have the degrees of membership add to 1.0. In this way the degrees of membership may be considered as the probabilities of belonging to an element of a set. Thus, Eq. (1) expressed in normalized form is: Q = {0.625/G + 0.3125/A + 0.0625/P + 0.0/B]. (2) The concept of fuzzy set theory has important applications in the field of knowledge-based expert systems as it permits not only a mathematical treatment of transition properties (i.e., safe to unsafe; allowable stress (deformation) to unsatisfactory stress (deformation); etc.) and linguistic-terms (see above example), but also operations simulating human inferences about complex interactions between variables that are not functionally related. Therefore, it is

X. Lu, S.H. Simmonds/Automation

in Construction

6 (1997) 121-137

123

not surprising that it has been used by many authors to develop KBES in the field of structural engineering (see, e.g., Blockley [2]; Brown and Yao [3]; Ogawa et al. [4]; Shiraishi et al. [5]; Furuta et al. [6]; Castaueda et al. [7]; Chao et al. [8]). Since its introduction by Zadeh [l], fuzzy set theory has undergone a substantial theoretical development (see, e.g., Zadeh [9]; Zimmermann [lo]; Kitainik [ 1I]). In this paper only the basic concept and notation of fuzzy set theory are used.

3.2. Factor sets All factors that are to be considered in the solution of the problem represented by the goal set are selected. These are then expressed as fuzzy sets which are referred to as factor sets. Since the final goal set is obtained from these factor sets, all factor sets must have the same number of elements as there are states in the goal set. The degree of membership either computed or assigned to an element in the factor set represents the degree of support that the element has for the corresponding state in the goal set. 3.3, Goal sub-sets In obtaining the solution for complex problems, there may be levels where a factor represents the state or condition of a sub-problem at that level. Such a set is referred to as a goal sub-set. For example, for the case where the goal set represents the solution of the structural condition of a building frame, there will be at some level a factor set that represents the state or condition of a particular column. This factor set is also the goal sub-set for that column. A goal sub-set is designated as V to distinguish it from a goal set U. 3.4. Weighting factor sets At any given level there will be several factor sets that contribute to the solution of the next higher level set. All con~buting factors may not have the same importance in the decision making process and this can be reflected by assigning different weights to the different contributing factor sets. The number of terms in the weighting factor set must obviously correspond to the number of con~buting factor sets. The choice of values in the weighting factor set is part of the expertise built into the system. 3.5. Selection categories
The number of levels used in the solution of a problem will depend on the complexity of the problem. At the lowest level, the factor set is divided into a number of categories from which the user selects. It is this selection of category that constitutes much

3. Multilevel fuzzy composition The concept of multilevel fuzzy composition is to divide a complex problem into a level of simpler problems which, in turn, are subdivided into even simpler problems at corresponding lower levels. This process is repeated until a level is reached in which the problems or questions posed can be answered by the user based on either his expertise or on an observed or computed value. These answers are expressed as fuzzy sets with appropriate membership values and are then combined using weighting factors and fuzzy composition. In this step by step way, answers are obtained for each successive level until the answer to the highest level or originally posed problem is obtained. To facilitate later discussion, the following terms are defined. 3.1. Goal set
The set of elements that define the solution to the problem to be solved is referred to as the goal set. Elements in the goal set are generally referred to as states since they represent the possible states or conditions of the solution. The number of states in the goal set is arbitrary and depends on the nature of the problem being considered. The case of a goal set, U, with IZ possible states can be expressed mathematically as

(3) Note that U is a fuzzy set and so can have membership in any or all of the defined states since it is the composition of many fuzzy sets used in the solution. The state with largest mem~rship is the most likely solution to the problem.

U={u,,

U2 . . . . u,}.

124 Table I Degrees of membership Categories quality G A P B

X. Lu, S.H. Simmonds/Automation

in Construction

6 (19971 121-137

u2
for construction quality States of the goal sub-set L: 1.0 0.5 0.1 0.0
12 L!3 14

for construction

u3
0.0 0.1 0.5 1.0

Good (above average) Average Poor Bad (unsatisfactory)

0.5 1.0 0.5 0.1

0.1 0.5 1.0 0.5

u4

the building as a whole is sound but some structural elements need repair, (Average). the building as a whole does not meet the requirements of the building code and needs to be strengthened; some elements may need to be streng~ened immediately, (Poor). the building is unsafe and must be vacated, (Bad).

of the input to the expert system. Associated with each category is a set of degrees of membership which correspond to the level of support that category has to the corresponding states in the goal set. The degrees of mem~rship assigned to the factor set are those for the category selected. This concept is demonstrated in Table 1. Here the quality of construction is a factor set at the lowest level in the problem of evaluating a goal sub-set. The factor set is divided into the categories G to B shown in Table 1 and for each category there is a set of degrees of membership corresponding to the states in the goal sub-set. The degrees of membership assigned to the quality of construction factor set for later use are those corresponding to the category selected by the user.

The stmctural condition of the building will correspond to the state that has the highest computed degree of membership. Since four states are defined in the goal set, all factor sets used in determining the goal set must also have four states that correspond in meaning to the states in the goal set. That is, the degree of membership in each state of the factor sets is a measure of the degree of support that factor has to the corresponding state in the goal set. The first step in dete~ining which state best describes the structural condition of the building is to

Tbird

Ftw&

La!.&

GfXl.X81 _ mConditions

Use Environment Service Years

4. Evaluating the structural condition of a building


A possible strategy for evaluating the structural condition of a buil~ng is given in Fig. 1. At the highest level is the goal set which contains the states that are to be used to define all possible conditions for the building. The number of such states selected will depend on the degree of refinement in the distinction between states that the user is prepared to make. For the goal set for the evaluation of a reinforced concrete framed building, a document by the Ministry of Construction [12,13] suggests that four states are adequate. Linguistically, these states are: the building as a whole is good; possibly some elements that do not effect the structural behavior may need repair, (Good).

FootinSs

-/
i

--

DiEerentisl Settlement Slip from Original Position Visual lnsptction Strength Visual Inspection Crack Width / ~- Story Drift L strength

/ structurs1 Condition of the Building iVertical Members

walls

i-

I
-4

I
Columns -.---

Visual Inspection ~ Crack Width Story DriA L Strength i !Visual lospsctioa Crack Width ~ Denectioo L Streogth

iBeams Horizontal Members / _- i Slabs -

-.

.-.

.-

im Visual hspectioo tmm~Crack Width

Quality

of Connections

u1

Fig. 1. Factors ing. for evaluating

Horizontal

Offsets

the structural

condition

of a build-

X. Lu, S.H. Simnwnds/Automtion

in Construction

6 (1997) 121-137

125

select the factors that will indicate the structural condition. As seen from Fig. I, the factors selected refer to problems associated with the original condition and history of the structure, the current condition of the footings, vertical elements (walls and columns), horizontal elements (beams and slabs) and the structural connectivity. Each of these factors are evaluated by subdividing into further factors at a lower level until the question posed by the factor can be answered by the user. The reason that the different structural components are grouped as separate factors is to permit the program to assign greater weight to the structural condition of the vertical elements. To demonstrate the procedure, we will consider in detail the evaluation of the columns as part of the factor set related to vertical elements. The other factor sets are evaluated in a similar manner and will only be discussed briefly. 4.1. Vertical elements The current condition of the columns must be inferred from an evaluation of each individual column. This is accomplished by considering for each column the factors to be used to determine the condition of that column. The factors selected are the surface condition, the maximum crack width, the story drift and the column strength. Each of these factors have categories that are used by the investigating engineer to establish the degree of membership for the states that correspond to the four defined states in the goal sub-set for the column. Consider the factor related to the column surface condition as determined by visual inspection. Four categories are suggested as being adequate to describe the surface condition.
G good or shrinkage

judgement since the boundaries between categories are not defined precisely and there is a likelihood that a different investigator or the same investigator at a different point in time may select another category. Since only one category can be selected, this vagueness is represented by assigning degrees of membership to the different states in the goal sub-set. As shown in Table 2, the selection of category G shows strong support for goal sub-set state u,, moderate support for uZ, little support for us and no support for uq. Similarly, selecting a different category does not preclude support for adjacent states in goal sub-set. The degrees of membership for the category selected are assigned to the factor set for surface condition. Since the factor sets at this level will be used to evaluate a single column the states in goal sub-set at this level pertain to the condition of that column but must have a correspondence to the states in the original goal set for the structure. This means that u1 for the column must indicate that the column is sound to correspond to the meaning of u1 for the structure. Consider now the maximum crack width. Since this is a measured quantity, it can be recorded and entered as input to the system. However, even if the maximum crack width could be determined precisely, its effect on the overall condition of column is fuzzy and degrees of membership to the different states in the goal sub-set must be assigned. This is done internally in the KBES using the ranges shown in Fig. 2. In Fig. 2, the transition point from ut to v2 of a crack width of 0.3 mm corresponds to the serviceability limit state and the transition to uq of 1 mm corresponds to the limit for dangerous state given in documents supplied by the Ministry of Construction [12]. The other points are proportioned between these two transition points. From the values

cracking only; A microcracks (maximum width less than 0.3 mm); P moderate cracking with possible minor spalling; B severe cracking or spalling with possible exposure of reinforcement.

Table 2 Degrees of membership Categories inspection


G Good

for visual surface inspection States of the goal sub-set I 1.0 0.6 0.1 0.0
u2 03 u4

for visual surface

In the program these four categories are listed as a menu from which the engineer chooses. Obviously, choosing a specific category involves considerable

A P B

or shrinkage cracking Microcracks Moderate cracking Severe cracking or spalling

0.6 1.0 0.6 0.1

0.1 0.6 1.0 0.6

0.0 0.1 0.6 1.0

126

X. Lu, S.H. Simmonds/Automation

in Construction 6 (1997) 121-137

~~xli:isIl
0.1 0.3 0.4
Maximum

0.6
Crack

0.7
Wtdth

0.9 (mm)

1.0

1.3

Fig. 2. Membership

function for maximum

crack width.

in Fig. 2, a maximum crack width between 0.4 and 0.6 mm would assign a degree of membership supporting state v2 of 1.0 and 0.0 for the other states whereas a maximum crack width of 0.35 mm would be interpreted as having degrees of membership supporting states vi and v2 of 0.5 and 0.0 for states us and vq. In a similar manner the story drift for a given column is also a measured quantity that can be input. Again the system assigns degrees of membership in relation to the level of support for the different states in the goal sub-set. This is done using the rules implied in Fig. 3, which are also obtained from limits contained in documents prepared by the Ministry of Construction [12]. From Fig. 1, it is seen that the fourth factor in evaluating a column is strength, which is defined as the ratio of the provided capacity of the member to the required capacity. This differs from the first three factors which are all based on physical observations by the investigating engineer and can be entered as input data. However, experience indicates that if there are no signs of distress in a member of an existing structure then there is little reason to undertake a full evaluation of the strength for that member. This is considered by the system using production rules.
P

For reinforced concrete columns or walls and denoting the visual surface inspection, crack width and ratio of interstory drift to story height as VSI, CRW and SD, respectively, the following are samples of the production rules invoked. IF {(VSI = G) OR (VSI = A) AND (CRW < = 0.4 mm) AND (SD < = 0.002)) THEN (The strength of the column or wall is good and a degree of membership of 1.O is assigned to state v, and 0.0 to the other states) OTHERWISE (a calculation of strength is required). IF {(VSI = P) OR (VSI = B)} THEN (The wall or column must be repaired). And similar rules are used for the other factor sets. Should the conclusion be that a calculation of strength is needed, additional input is required. An evaluation of the provided and required capacities can be made separately by the engineer and the ratio input. In the expert system described the required capacity of the member must be input but, instead of entering the capacity required, the engineer has the option of entering the material properties, concrete dimensions and amount and position of the reinforcement. The system then determined the capacity provided and computes the strength ratio. For columns only a single strength ratio is used that is based on equating the moments and taking the ratio of corresponding provided to required axial capacities. The degrees of membership for the different states in the goal sub-set are assigned in accordance with Fig. 4. Thus, it is seen that regardless of the condition of the column, membership values are assigned to each of the factors considered. This results in a matrix of membership values which will have as many rows as there are factors considered for the goal sub-set and as many columns as there are states in the goal set.

rfKLXiKL__
0.2 0.4
Story

i; f

1.5r

I~M_xLKy_;
0.7 0.78 Prwldsd

0.5

0.7

0.8

1.0 (pmmt)

1.1

1A

0.80

0.66

0.90

0.98 capacit,

1.0

1.1

Drift/Column Height

Copaelty/R.qulrsd

Fig. 3. Membership

function for story drift ratio.

Fig. 4. Membership

function for strength ratio.

X. Lu, S.H. Simmonds/Automation

in Construction 6 (1997) 121-137

127

To arrive at the goal sub-set r~p~senting the condition of that column it is necessary to combine these factors taking into account their relative importance. This is done by pre-multiplying the matrix of the degrees of membership by a vector representing the different weighting factors. For a column these are taken to be W, = (0.2 0.2 0.2 0.4).
(4)

0.30

0.40

0.42

0.52 Oaflection/Span

0.54

0.6 (percent)

0.64

0.80

MaxImum

Fig. 5. Membership

function for beam deflections

ratio.

This matrix product is the goal sub-set for a particular column. Similar fuzzy sets are obtained for each column in the structure. Unless there is some reason for weighting the columns differently, all columns are assumed to have the same influence on the overall condition of the structure. Thus a single fuzzy set representing the condition of all columns is obtained by averaging the degrees of membership for each state in the column goal sub-sets. The process for evaluating wah elements is identical to that described for evaluating columns. To obtain a single fuzzy set representative of the vertical members the factor sets representing the condition of the columns and walls must be combined. Again, the weighting to each factor will depend on the nature of the structure, for example, there may be no walls. Unless otherwise entered the two element types are weighted equally. 4.2. Horizontal elements A process similar to that used to evaluate the columns is used to evaluate the beams and slabs. For these components the same four factors are used at the lowest level except that for the beams and slabs the out-of-alignment (story drift) is replaced with the maximum deflection. The limits for defining the different categories are again obtained from interpolating between values recommended by the Ministry of Construction [12]. For example, the limits for assigning categories for beam and slab deflections are given in Fig. 5. Again there are production rules that are similar to those for the column to determine whether strength calculations are required. For each structural element, the factors are combined using the same weighting factors to arrive at a fuzzy set indicating the structural condition of that element. Those elements deemed defective are noted for later repair.

As for the columns, a single fuzzy set representing the condition of all beams or all slabs is obtained by averaging for each element type the degrees of membership for each state in the goal sub-sets. Again, these two factor sets are combined using a weighting factor set to obtain the factor set representing the structural condition of all ho~zontal members. 4.3. Footings The single fuzzy set that represents the condition of all footings is again obtained by considering the individual footings. Obtaining the degrees of membership for the four states in the goal sub-sets for each footing follows the same methodology as used for the columns. From Fig. 1 it is seen that four factors are selected to evaluate the footing so that the matrix of degrees of membership for each footing will have four rows and the degrees of membership in each row wiI1 be those corresponding to the category selected for that factor. The first two factors, differential settlement and the amount of displacement of the footing from the original position, can be estimated from a field survey of the building. However, depending on the structure, a visual examination of the surface condition of the footing may not be easily accomplished. The decision as to whether such a visual examination is required is determined from production rules. For example, if the differential settlement and displacement categories selected support state ur , then a visual inspection of the footing is not required and a membership value of 1.0 is assigned to state u, and lesser degrees of membership to the other states. Should there be signi~cant differential settlement or slip of the footing a visual examination of the foot-

128

X. Lu, S.H. Simmonds/Automation

in Construction 6 f1997) 121-137


CL

ing is required and membership values assigned corresponding to the category selected. After the matrix of degrees of membership for a footing has been determined, it is pre-multiplied by the weighting factor set which for a footing is the same as for the column. The fuzzy set for the footings is obtained by averaging the footing sub-sets. 4.4. General conditions The first two factors listed in Fig. 1, design quality and construction quality have categories and corresponding degrees of membership as given in Table 1. The same categories are used for the factor, use environment, where POOR is used to indicate the conditions of alternatively wet and dry or exposure to strong wet winds and BAD to exposure to a corrosive atmosphere such as from salt spray or severe freeze-thaw conditions. The factor, service years, distinguishes between masonry and reinforced concrete construction. The transition points between the states in the goal sub-set are 20, 40 and 60 years for masonry construction and 25, 50 and 75 for reinforced concrete construction. Again for each category there is a set of degrees of membership and the factor set representing general conditions is obtained by applying appropriate weighting factors. The weighting factor set for general conditions is wg = (0.30 0.30 0.25 0.15). 4.5. Structural connectivity (5)

f;

1.5,

Maximum

Vertical

Offset

(cm)

Fig. 6. Membership

function for maximum vertical offset.

for the different states in the goal sub-set for offsets for floor levels are given in Fig. 6. 4.6. Structural condition

Once the five factors sets have been determined they are combined to obtain the goal set or the single fuzzy set that describes the structural condition of the building. The weighting factor set used when the factors are listed in the order given in Fig. 1 is v, = (0.1 0.2 0.4 0.2 0.1). (6)

5. Example
5.1. Main symbols and their definitions example used in this

The fifth factor in evaluating the structural condition of the building is structural connectivity and relates not only to the quality of the connections themselves but also the amount of offset in the connection geometry. The first two factors for evaluating structural connectivity, quality of connections and structural integrity, and divided into categories GOOD, AVERAGE, POOR and BAD. Structural integrity evaluates how well the structure is tied together and considers such items as the ties between intersecting walls, the degree of reinforcing and the condition of bond beams in masonry walls, etc. The latter two factors have categories of NONE (negligible), SMALL, INTERMEDIATE and MAJOR. The transition points

VSI visual inspection; CRW maximum crack width; SD ratio of interstory drift/story height; ratio of provided capacity/required capacity; ; ratio of maximum deflection/span; IDS footing differential settlement; FSP footing slip from original position.

5.2. Brief description

of the structure

This building was built in early 1920s and is a two-story R.C. framed building used as a chemical plant. Because of overloading and corrosion conditions, some of the R.C. columns are cracked very severely, and the deflections of some beams are very large, and the concrete of some beams and most slabs are spalling. The simplified profile of the structure is shown in Fig. 7 in which the footings, beams

X. Lu, S.H. Simmonds/Automation in Construction 6 (1997) 121-137

129

and columns are represented by f,b and c respectively. 5.3. The reason for its evaluation With the development of the production and the updating of the equipment, the owners of the building plan to install new machines whose weights are to be larger than the previous ones. Therefore, the owners want us to evaluate the overall structural condition and to propose some remedial action if necessary.

f;
Fig. 7. Simplified

f2

f,

f,

profile of the structure of the example building.

5.4. The detailed evaluation

5.4.1. General condition Design quality Observations Average Representation by fuzzy sets f0.238, 0.476,0.238, 0.048] (These data are normalized from C0.5, 1.0, 0.5,O.l)) (0.048, 0,238, 0.476, 0.238) (0.048, 0.238, 0.476, 0.238) {O,0.062,0.313,0.625)

Construction quality Use environment Service years

Poor Poor 68 years

Weighting factor set: ws = {0.30,0.30,0.25,0.15}. Fuzzy relation matrix: 0.476 0.238 0.062 0.238 0.238 0.476 0.313 0.476 0.048 0.238 0.625 0.238

Evaluation sub-set for general conditions E,: E, = Wg. R, = {0.0978,0.2830,0.3802,0.2390}.

130

X. Lu, S.H. Simmorrds/Automation

in Construction

6 (1997) 121-137

5.4.2. Footings footings .f, FDS FSP VSI Y .f2 FDS FSP VSI Y f, FDS FSP VSI Y Observations 12 mm 19 mm Poor 0.75 or calculation Representation by fuzzy sets

iO.210, 0.527, 0.210, 0.053) IO.210, 0.527, 0.210, 0.053) {0.043, 0.261, 0.435, 0.261) {O, 0, 0, 11 IO.526, 0.421, 0.053, 0) {0.210, 0.527, 0.210, 0.053) {0.043, 0.261, 0.435, 0.261) {O,O,O, 11 (0.526, 0.42 1, 0.053, O} {0.210, 0.527, 0.210, 0.053) (0.043, 0.261, 0.435, 0.261) (0, 0, 0, 11 {0.210, 0.527, 0.210, 0.053) IO.210, 0.527, 0.210, 0.053) IO.043, 0.261, 0.435, 0.261) {O,O,O, 11

Omm 15 mm Poor 0.63

Omm 15 mm Poor 0.63

f4
FDS FSP VSI Y 11 mm 16mm Poor 0.77

Weighting

factor set:

Wr= {0.20,0.20,0.20,0.40}.

Averaging

all footings to form fuzzy relation matrix

R,:

0.368

0.527 0.261 0 0.474

0.210 0.435 0 0.132

0.053 0.261 1 0.026

Evaluating

sub-set for footings

E2:

E,=

W,~R,={0.124,0.252,0.155,0.468}

(8)

X. Lu, S.H. Simmonds/Autotnation

in Construction

6 (1997) 121-137

131

5.4.3. Vertical members Columns Cl, VSI CRW SD Y VSI CRW SD Y C,, VSI CRW SD Y CM VSI CCRW SD Y Observations Bad > 1.0 mm 0.0035 0.80 Poor 0.95 mm 0.0032 0.76 Poor 0.90 mm 0.0033 0.78 Bad > 1.0 mm 0.0038 0.85 or calculation Representation
by fuzzy sets

(0, 0.059, 0.353,0.588) (0,0, 0, 11 (1, 0, 0, 01 (0, 0, 0, 11 (0.043, 0.261, 0.435, 0.261) (0, 0, 0.5, 0.5) (LO, 0, 01 (0, 0, 0, 11 (0.043, 0.261, 0.435, 0.261) (o,o, LO1 (1, 0, 0, 01 (0, 0, 0, 11 (0, 0.059, 0.353, 0.588) 1) (1, 0, 0, 0) (0, 0, LO1

(0,0, 0,

Averaging

all columns

in the first story to form fuzzy relation matrix

R,,:

r0.021

0.160 0 0 0

0.394 0.375 0 0.5

0.425 0.625 0 0.5

Weighting

factor set for columns:

w, = {0.2,0.2,0.2,0.4}.

Evaluation

sub-set for the columns

in the first story, Es,:

Es, = W, . R,, = (0.204,0.032,

0.354, 0.410).

132

X. Lu, S.H. Simmonds/Automation

in Construction

6 (19971 121-137

Colunlns

Obseruations Bad > l.Omm 0.006 0.85

or calculation

Representation

by fuzzy sets

G
VSI CRW SD Y C,, VSI CRW SD Y C,, VSI CRW SD Y (0, 0.059, 0.353, 0.588) (0, 0, 0, 11 IO, l,O,OI {O,O, 1,OI 10.043, 0.261, 0.435, 0.261) mo,o, 11 (0, l,O,OI (0, 0, 0, 1) (0, 0.059, 0.353, 0.588) (0, 0, 0, 1) IO, l,O,OI {O,O, l,O>

Poor > 1.0 mm 0.005 0.78

Bad > 1.0 mm 0.006 0.83

Averaging

all columns

in the second story to form fuzzy relation matrix,

R,,:

R,, =

I
0 0

0.014

0.126 0 I 0

0.380 0 0.667

0.479 0 1 I 0.333

Evaluation

sub-set for the columns

in the second story, E,,:

E,, = WC. R,, = {0.003,0.225,0.343,0.362}.

Weighting

all columns

in different

stories by weighting

factor set IV,,

W,, = {0.6, 0.4},

where 0.6 and 0.4 are for the first story columns Evaluation sub-set for vertical members, E,:

and second story columns,

respectively.

E3 = wS,

E31 =
E
32

{0.124,0.109,0.350,0.391)

(9)

X. Lu, S.H. Simmonds/Automation

in Construction

6 (1997) 121-137

133

5.4.4. Horizontal members

Observations or calculation

Representation by fuzzy sets


(0, 0.059, 0.353, 0.588) 10, 0, 0, 1) 10,0,1,0) 10, 0, 1) 0, (0, 0.059, 0.353, 0.588) 1) 10, 0, 0, 1) 10, 0, 0, 1)

4,
VSI CRW SD Y B,, VSI CRW SD Y B,, VSI CRW SD Y Bad > 1.0 mm 0.0063 0.76

Bad > 1.0 mm 0.0081 0.70

10, 0, 0,

Bad > 1.0 mm 0.006 0.78

(0, 0,059, 0.353, 0.588)

(0, 0, 0, 1)
(O,O, LO) (0,0, 0, 1)

Averaging 0

all the beams in the first story to form fuzzy relation matrix 0.059 0.353 : 667 0 0.588 :, 333 1 I members:

Rj,

Weighting

factor set for horizontal

w, = {0.2,0.2,0.2,0.4). Evaluation sub-set for the beams in the first story E:,

Ej, = W,, . R& = {0,0.012,0.204,0.784} Beams B,,


VSI CRW SD Y B,, VSI CRW SD Y

Obseruations or calculation
Bad > 1.0 mm 0.0073 0.81

Representation by fuzzy sets


(0, 0.059, 0.353, 0.588) (0, 0, 0, 1) IO, 0, 0, 1) (o,o, LO) (0, 0.059, 0.353, 0.588) 10, 0, 0, 1) IO, 0, 0, 11 IO,O, 1,O)

Bad > 1.0 mm 0.070 0.85

134

X. Lu, S.H. Simmonds/Automation

in Construction

6 (1997) 121-137

Averaging f-0

all beams in the second story to form fuzzy relation matrix 0.059 0.353 0.588 1

R&:

Evaluation

sub-set for the beams in the second story E&2:

E& = W,, . Ri2 = {0,0.012,0.471,0.518}, Weighting factors for the beams in different stories W,,:

Ws, = {0.6,0.4}, Evaluation sub-set for beams Ei:

E, = IV& . [ Slabs
311

E44= E4h2

(0,0.012,0.311,0.678}. Observations Bad > 1.0 mm 0.006 0.78 or calculation Representation (0,0.059,
by fizzy sets

VSI CRW SD Y
312

0.353, 0.588)

IO,& 0, 1) (O,O, l,O) (O,O, l,OJ (0, 0.059, 0.353,0.588) (O,O,O, 1) (O,O, 1,01 (0, 0, 0, 11 (0, 0,059, 0.353, 0.588) (0,0, 0, 11 (O,O, 1,01 (O,O,O, 11

VSI CRW SD Y VSI CRW SD Y

Bad > 1.0 mm 0.0058 0.73 Bad > 1.0 mm 0.0053 0.75

Averaging 0 R;,=;; [ 0 Evaluation 0

all slabs in the first story to form fuzzy relation matrix 0.059 0.353 ; 0 0.588 0 1

1
.

Ri,:

sub-set for the slabs in the first story Ei,:

E;, = W,. R;, = {0,0.012,0.271,0.718}.

X. Lu, S.H. Simmonds/Automation

in Construction 6 (1997) 121-137

135

Slabs S &I CRW SD Y S Vi1 CRW SD Y

Observations Bad > 1.0 mm 0.0053 0.87

or calculation

Representation

by fuzzy sets

(0, 0.059, 0.353, 0.588)

(0,0, 0,

11 (O,O, LOI {O,O, l,OI

Poor > 1.0 mm 0.0057 0.83

(0.043, 0,261, 0.435, 0.261) 1) (O,O, LOI (O,O, LO1

(0,0, 0,

Averaging 0.022 R&= [ ; 0

all slabs in the second story to form fuzzy relation matrix 0.160 0 0 0.394 0 1 1 0.424 0 1 0 I.

R$:

Evaluation

sub-set for the slabs in the second story E&:

E& = W,, . Rig = (0.004,0.032,0.679,0.285}. Averaging all the slabs in different stories, W,, = (0.6, 0.4):

Ej = W,, .

= {0.002,0.020,0.434,0.545}.

Weighting

Ed = W,,, . [

E, 1
Es
4

beams and slabs, W, = (0.6, 0.4):

= {0.001,0.015,0.360,0.625}.

5.4.5. Structural connectivity Quality of connection Structural integrity Vertical offsets Horizontal offsets Observations Bad Bad 0 > 1500 mm or calculation Representation by fuzzy sets (0, 0.063, 0.313, 0.625) (0, 0.063, 0.313, 0.625) 11,0,0,0~ (0, 0, 0, 11

136

X. Lu, S.H. Simmonds/Automation

in Construction

6 (1997) 121-137

Fuzzy relation matrix:

0.063 0.313 0 0 0.063 0.313

0.625 0. 1 0.625 I

Weighting factors for structural connectivity: w, = {0.4,0.3,0.2,0.1]. Evaluation sub-set for structural connectivity, Es: Es = W, . R, = {0.2OO, 0.044,0.219,0.538}.

(11)

5.4.6. Overall structural conditions Fuzzy relation matrix for overall structural conditions, R, is composed of E, , E, , E,, E4 and Es in Eq. (7) to Eq. (11): 0.098 0.124 R= i 0.124 0.001 0.200 0.283 0.252 0.109 0.015 0.044 0.380 0.155 0.350 0.360 0.219 0.239 0.468 0.391 . I 0.625 0.538

Weighting factors for overall structural conditions, W,: ys= {0.1,0.2,0.4,0.2,0~1}. The final evaluation set for the overall structural condition, E: E = W, . R = {0.104,0.130,0.303,0.453}. The final evaluation set E, E = {e,, e2, e3, e4} is a fuzzy sub-set in the goal set U, U = {u,, uZ, us, uq). The numerical value in set E are the degrees of membership of the building corresponding to the different states, respectively. In this study, the final result is determined according to the maximum value in the evaluation set E. Because e4 = 0.453 is the maximum value within the set ZZ, overall structural condition of the building is the uq, which means the building is unsafe and must be vacated. This result has been adopted by the owner of the building and a new building has been completed recently.

X. Lu, S.H. Simmonds/Automation

in Construction

6 (1997) 121-137

137

6. Summary

and conclusions

A methodology is described that permits combining field observations, numerical calculations and engineering expertise to be incorporated into the writing of a knowledge-based expert system. The procedure is demonstrated for a KBES developed to obtain an evaluation of the structural condition of a reinforced concrete framed building and a detailed example is given to show the whole procedure of structural evaluation. It is concluded that the use of fuzzy sets permits adequate representation of the inherent vagueness in much of the data collected in the evaluation of buildings and the method of multilevel fuzzy composition provides a practical means of manipulating this data to obtain meaningful conclusions.

References
[l] L.A. Zadeh, Fuzzy sets, Information and Control 8 (1965) 338-353. [2] D.I. Blockley, The rule of fuzzy sets in civil engineering, Fuzzy Sets and Systems 2 (1979) 267-278. [3] C.B. Brown and J.T.P. Yao, Fuzzy sets and structural engi-

neering, Joumal of Structural Engineering 109(5) (1983) 1211-1225. [4] M. Ogawa, K.S. Fu and J.T.P. Yao, SPERIL-II: An expert system for damage assessment of existing structures, in: M.M. Gupta et al., eds., Approximate Reasoning in Expert Systems (Elsevier Science Publishers, 1985) 731-734. [5] N. Shiraishi, et al., Fuzzy expert system for assessment of structural reliability, Journal of the Society of Material Science Japan 39(436) (1990) 19-25. [6] H. Furuta, et al., Knowledge-based expert system for damage assessment based on fuzzy reasoning, Computers and Structures 40(l) (1991) 137-142. [7] D. Castaneda and C. Brown, Methodology for forensic investigations of seismic damage, Journal of Structural Engineering 120(12) (1994) 3506-3522. [81C.J Chao and F.P. Cheng, Fuzzy diagnostic model for diagnosing cracks in reinforced concrete structures, Microcomputers in Civil Engineering 1 l(2) (1996) 115-122. [91 L.A. Zadeh, Outline of a new approach to the analysis of complex systems and decision process, IEEE Transaction on Systems, Man and Cybernetics 3(l) (1973) 28-34. Fuzzy Set Theory and Its Application DOI H.J. Zimmermann, (Kluwer-Nijhoff Publishing, 1985). illI L. Kitainik, Fuzzy Decision Procedures with Binary Relations: Towards a Unified Theory (Kluwer Academic Publisher, 1993). El21 Ministry of Construction, China, Assessment criteria for dangerous buildings, Beijing, 1986, (in Chinese). [I31 Ministry of Construction, China, Reliability appraisal of industrial buildings, Beijing, 1992, (in Chinese).

Das könnte Ihnen auch gefallen