Sunteți pe pagina 1din 10

Int. J.

Man-Machine Studies (1983) 19, 9-18

Potential applications of fuzzy sets in Civil Engineering


J. L. A. CHAMEAU, A. ALTESCtIAEFFL,H. L. MICHAEL AND J. T. P. YAO

Purdue University, West Lafayette, Indiana 47907, U.S.A.


The authors are all civil engineers but with various specialties in soil dynamics, foundation engineering, transportation and structural engineering. Because they are colleagues on the same faculty, there have been many occasions to discuss potential applications of the theory of fuzzy sets in their respective fields. The different disciplines that are involved appear to have complemented each other in the development of ideas to solve problems in the civil engineering profession using the fuzzy sets mathematics. In this paper their ideas ate summarized and selected applications and future developments are discussed.

Introduction
Empirical or theoretical techniques are generally available to assess art engineering problem and make technical decisions. However, even the most sophisticated techniques cannot take into account all the issues of any given problem. This is the case especially in Civil Engineering, not only for our lack of understanding of the p h e n o m e n a involved and factors affecting them, but also because of the limited quantity and inexact quality of information usually available. Consequently, information, experience and engineering judgment must supplement scientific knowledge and become inherent parts of the evaluation. Engineering decisions are based on a combination "objective" scientific knowledge and "subjective" information and engineering judgment. The theory of fuzzy sets can be used to perform this combination in a logical manner. Recent developments of this theory in the fields of medicine and economics show that it can interpret and manipulate information from different sources, evaluate uncertainty, and incorporate judgment to improve the decisionmaking process. A civil engineer must exercise engineering judgment in making scientific calculations and in making decisions based on imprecise information. Experienced engineers may be capable of reaching a suitable decision on the basis of their professional judgment. The making of such decisions is difficult for younger engineers who lack the decision-making skills accrued from experience. The theory of fuzzy sets may provide both a rational and systematic approach to the transfer of experience and decision-making practice, as well as a means of improving the assembly of data for the experienced engineer. An informal discussion group was organized among faculty members of the Department of Civil Engineering at Purdue University to discuss the potential applications of the theory of fuzzy sets to various sub-disciplines within Civil Engineering. This led to the development of ideas to solve problems in the fields of geotechnical, structural and transportation engineering, and in the initiation of research work in such areas. In this paper, some of these ideas are summarized and selected potential applications are presented. Although this work is in the early stages of development, 9 0020-7373/83/070009 + 10503.00/0 O 1983 Academic Press Inc. (London) Limited

l0

L. A. C H A M E A U E T A L .

it is hoped that this presentation may encourage other engineers to consider the theory of fuzzy sets a useful tool to express engineering judgment and make complex engineering decisions.

Potential applications in geotechnical engineering


Most of the analytical procedures used in geotechnical engineering are deterministic techniques developed within the framework of soil mechanics in accordance with previous experience. Even the most sophisticated tools are only steps which force the engineer to consider the essential aspects of the problem and excercise his best judgment, on the basis of his experience, observations, and those of others. There is a growing tendency to base engineering decisions upon hazard and risk analysis procedures, instead of upon classical factors of safety or margins of conservatism. Deterministic procedures have been embodied in probabilistic approaches which are based on reliability theory and deal with the uncertainty due to random parameters. When assessing hazards in geotechnical engineering, random uncertainty is not necessarily the most important uncertainty, and human and system uncertainties can be the most significant. These uncertainties are related to the lack of understanding of the problem, or to the difficulty of precisely defining the phenomena involved. They can also involve our inability to describe clearly how the different elements of the phenomenon are-related. An example of the above situation is present in the assessment of liquefaction potential in earthquake-prone regions. The term liquefaction describes the loss of strength of cohesionless soil during an earthquake due to an increase in pore water pressure. This can subsequently lead to large ground movements which damage structures and cause loss of life. Such damage occurred in Niigata, Japan, during the 1964 earthquake, where many structures settled more than 3 ft and suffered up to 80 ~ of tilting, and in Valdez, during the 1964 Alaskan earthquake, where extensive flow slides washed entire sections of the waterfront into the sea. The liquefaction susceptibility of a cohesionless soil at a given site can be evaluated by comparing the shear stress generated by the earthquake loading with the shear strength of the soil, as determined by laboratory experiments, or by using empirical correlations between the stress causing liquefaction and an in situ soil characteristic. These theoretical or empirical techniques are based upon our best scientific knowledge of the phenomenon at present and should be an essentiaI part of our evaluation. It must, however, be recognized that our understanding of the problem is limited and the answer given by any method of analysis is consequently uncertain. Furthermore, it is impossible for any of the available techniques to take into account all of the factors which may influence the potential for liquefaction. There are three major factors which are conducive to liquefaction: intensity of ground shaking, low relative density of sand materials, and shallow water-table. These are taken into account in the analytical and empirical solutions, but there are many other factors which may influence the resistance of a site to liquefaction; these are the presence of a slope, geologic age of deposit, pretressing, grain size distribution of the soil, amount of cementation, lateral confinement, presence of clay layers, etc. It is believed that the theory of fuzzy sets can be a tool to improve on the present state-of-the-art by: (1) taking into consideration the random and system uncertainty

FUZZY SETS AND CIVIL ENGINEERING

11

inherent to the liquefaction evaluation techniques; and (2) incorporating the different factors influencing the potential for liquefaction as part of a diagnosis. As an example, fuzzy statistics can be applied to the correlation chart between cyclic stress ratio and penetration resistance commonly used for the evaluation of liquefaction potential (Fig. 1). Any combination of penetration resistance and cyclic shear stress ratio which
0'5

0-4

IO

I-0 1.0

1.0
f

I'0 13r 0.3 I'0 1.0 ._O 0.2


io

0.95 /

1.0 07 ~0.6

,,, ~0.75 0.4


0-55 0.2 0-3 0.1 0.1 0

/9 O-Ii 0.6 O.2

o/
0

I0

20

30

40

Modified penetration

resistance N I ( b l o w / f t )

FIt3.1. Chart for evaluation of liquefaction potential (after Seed & Idriss, 1981).

lies in the region above the line represents circumstances when liquefaction is presumed to occur. As an example, liquefaction is predicted for a corrected penetration resistance of 15 and a stress ratio of 0.20 (Fig. 1). If the stress ratio is reduced to 0.10 the site is considered non-liquefiable. In this deterministic approach the S-N1 curve is treated as a one-to-one relation and the penetration resistance is assumed to be accurate in spite of a known large scatter in data. The same calculation can be performed when the relation between the (stress, penetration resistance) pairs leading to liquefaction is treated as a fuzzy relation R (Fig. 1), and the penetration resistance is a fuzzy variable. For example, let the uncertainty in the number of blows to be expressed by N1 = 0.6/15 + 1/20 + 0.5/25. The arithmetic of fuzzy compositions (Zadeh, 1975) generates S = N~ o R with membership function /.ts(S):
S = u/Zs(S) = 0.2/0.05 + 0-6/0.1 + 0 . 6 / 0 . 1 5 + 0 . 7 5 / 0 . 2 0 + 0.95/0-25

+ 1 / 0 . 3 0 + 1/0.35 + 1/0.40.

12

1.. A .

CHAMEAU

ETAL.

Now assume that the applied stress ratio S is a random variable taking the values 0-15, 0.20 and 0.25 with the following probabilities: P(S = 0.15) = 0.30; P(S = 0-20) = 0.40; P(S = 0.25) = 0.30.

Then, the probability of liquefaction can be calculated as (Zadeh, 1968) P(liquefaction) = ~tZs(S) 9P(S = s),
$

which gives a probability of 0-77. To keep this example very simple, a limited number of discrete values were arbitrarily assigned to S and N1, but fuzzy statistics can be extended to actual problems with large number of discrete values or continuous variables. As mentioned earlier, there are many important factors which are not taken into account in any of the liquefaction evaluation techniques. The next logical step in the fuzzy sets approach is to introduce these factors as part of a diagnosis. One possible way to express the diagnosis will be as a grade of the damage state after the earthquake. As an example, it could be a numerical value between 0 and 1, where 0 and 1 correspond to no damage and total failure, respectively: 0 No 0.1 0.2 0-3 Slight 0.4 0.5 0.6 Moderate 0.7 0.8 Severe 0-9 1.0 Total failure

Similarly, a factor affecting liquefaction can be characterized by its conduciveness to damage. Conduciveness may range from 0 to 1: 0 No 0.1 0.2 Weakly conducive 0.3 0.4 0.5 0.6 0.7 Conducive Strongly conducive 0.8 0.9 Almost definitely conducive 1.0 Definitely conducive

Each of the influencing factors will have to be examined and a methodology developed to measure their conduciveness to damage due to liquefaction. This measure will reflect the uncertainty and judgement involved in the assessment. As examples, the presence of a slope and the age of a deposit at a given site might be regarded as strongly conducive and weakly conducive, respectively, and expressed by the fuzzy sets strongly conducive (slope) = 0 . 3 / 0 . 4 + 0.8/0.5 + 1/0.6 + 0 . 8 / 0 . 7 + 0.2/0.8, weakly conducive (age) = 0 . 2 / 0 + 0.8/0.1 + 1/0.2 + 0.4/0.3. Similarly the state of damage might be specified by a fuzzy set. A moderate state of damage may be represented by moderate state of damage = 0.1/0.2 + 0 . 5 / 0 . 3 + 0 . 8 / 0 . 4 + 1/0.5 + 0 . 8 / 0 . 6 + 0 - 4 / 0 . 7 . Future research will provide the means to generate fuzzy sets representation of these factors given the information at a site, define the relation between these factors and the damage state, and provide a global assessment of the state of damage using fuzzy sets statistics. The goal of this methodology is to provide the logical means to differentiate between those situations where liquefaction can cause damage and those situations where liquefaction, per se, may not cause significant damage. Other

FUZZY

SETS AND

CIVIl. ENGINEERING

13

approaches than diagnosis theory can be considered to solve the liquefaction problem. In particular the approach proposed by Blockley (1975, 1977) and Brown (1977) to fuzzify a risk probability appears promising. Subjective assessment is clearly necessary in real situations, not only in the making of assumptions but also in the examination of the analytical results. One attraction of the fuzzy methodology is that subjective assessments obtained from different sources can be combined. The authors believe that the approach outlined for liquefaction potential can be applied to other areas of geotechnical engineering. Geotechnical engineers use a variety of scientific theory to help their design decisions; however, this scientific knowledge is very often inadequate, and rules-of-thumb and engineering judgment are used extensively. Typical examples are: evaluation of the results of field investigation; tolerability of predicted settlements for a given structure; evaIuation of landslide potential and slope stability; and assessment of the performance of a geotechnical engineering structure such as an earth dam. As a simple application, consider the evaluation of the performance of an embankment. Let A be the compaction control,of an embankment and B the engineering judgment on the good performance of the embankment. Assume that A is good, and B is likely. A and B could be expressed as A = 1/1 + 0 . 8 / 0 . 9 + 0 . 7 / 0 . 8 + 0.1/0.7 B = 1/1 + 0 . 7 / 0 . 9 + 0.6/0.8. The membership function of the Cartesian product R = A x B (Zadeh, 1975) can be expressed by the following table: Blikelytohaveagoodperformance
1

0-9 0.7 0.7 0.7 0.1

0.8 0-6 0-6 0-6 0-1

A good compaction control

1 0.9 0.8 0.7

1 0.8 0'7 0.1

This fuzzy relation expresses the idea that when the compaction control is good, it is likely that the performance will be good. If our opinion of the quality control of the compaction of a given embankment is A = 0.6/1 + 0.8/0.9 + 1/0.8 + 0 . 6 / 0 . 7 + 0.4/0.6. Then the "likelihood" of a good performance of the embankment is B = A . R = 0.8/1 + 0 . 7 / 0 . 9 + 0.6/0.8. In this example, subjective engineering judgment (quality control) is combined with experience (fuzzy relation R) to express a judgment on the performance of the embankment.

14

L. A.

CHAMEAU

ETAL.

Potential applications in structural engineering


Since the introduction of the fuzzy sets to the civil engineering profession by Brown & Leonard (1971), several attempts have been made to solve problems in structural engineering. Just like other engineers, structural engineers are concerned with input (or loads or excitation or disturbance), the system (or structure or structural system), and the output (response of the structure in the form of internal forces or displacements). Generally, problems in structural engineering can be classified into analysis, design, identification, and reliability as listed in Table 1. Frequently, practical solutions TABLE 1

Classification of structural engineering problems


Input (forcing function, distrubance or excitation) Known or specified Known or specified Measured Information System (material, configuration, capacity) Known or specified Unknown, to be determined Actual and existing structure with unknown characteristics (to be determined) Statictics of capacity known or given Output (response of the structure) Unknown, to be determined Limits specified Measured Classification of the problem

Structural analysis Structural design Structural identification

Statistics known or given

Statistics of response and/or probability of survival to be estimated

Structural reliability

require the combined use of approaches in two or more of these classifications. As an example, to design a structure, it is necessary to assume initially the material and configuration, with which an analysis can be made. On the basis of results of such an analysis, the configuration (and/or material) may be altered to " i m p r o v e " the initial design. Then the improved design is further analyzed, etc. Therefore, most structural designs involve an iterative procedure of design as well as analysis techniques. In the development of design codes, fuzzy information must be processed to produce concise design rules (Sozen, 1980). As a simple example, the interstory drift of a tall building is an important consideration in the design of earthquake-resistant structures (Sozen, 1980). Therefore, a simple and precise formula should be given to prevent any excessive inter-story drifts in building design codes. On the other hand, the building structure

F U Z Z Y SETS A N D CIVIL E N G I N E E R I N G

15

is an extremely complex system and the earthquake ground motions are difficult to predict (or even to measure, due to noise problems). It seems that fuzzy sets can be useful in the development of code provisions, which usually require a lot of subjective evaluation and judgment. Attempts have been made to apply fuzzy sets in the identification of structural damage for existing structures (Yao, 1980; Ishizuka, Fu & Yao, 1981). Other applications include those for the ready-mixed concrete industry (Munro & Jowitt, 1978), analysis of structural safety (Brown, 1979, 1980, Ditlevsen, 1980), the analysis of accidents (Blockley, 1975), and the behavior of building columns and metals fatigue (Blockley, 1979). Recently, fault trees and fragility curves have been used in probabilistic risk analysis of nuclear power plant structures (CorneIl, 1982). There seem to exist some arbitrary and subjective decisions concerning the simplification of some fault trees and the determination of certain fragility curves. In summary, with the possible exception of structural analysis, many problems can be solved with the use of fuzzy sets in structural design, structural identification, and structural reliability.

Potential applications in transportation and highway engineering


Deterministic approaches to transportation and highway engineering problems appear to have inherent shortcomings because: (1) these problems involve the complex socio-economic environment; (2) the variables are not completely defined and cannot be precisely measured; and (3) personal bias and subjectiveness enter in the analysis. These characteristics of complexity, indefiniteness, and subjectiveness make attractive the use of the theory of fuzzy sets to divide these complex problems into simpler questions and to relate verbal statements and subjectiveness to quantitative statements. Discussions with practicing engineers of the Indiana Department of Highways suggest that this approach may improve the decision-making process for highway investment. Many of the phenomena which control the performance of highways are not precisely defined. The quality of performance is also not clearly defined because the ultimate judge is the user; different users have different criteria and the criteria are often defined only in a vague, or fuzzy, manner. This creates uncertainty for the engineer decision-makers. As a result, experience and engineering judgment are used to supplement scientific knowledge, and they have become inherent components of any performance evaluation. The system performance of a highway is a key element entering decisions regarding allocation of resources to highway improvement and the type of improvement needed (reconstruction, resurfacing, widening, etc.). It is a "gross measure" which can not be precisely defined and is currently expressed by descriptive words such as "good" or "fair", based upon the subjective assessment of a review team. However, the system performance can be represented as a combination of safety, physical condition, and service performance. Each of the three determinants of system performance (i.e. safety, physical condition, and service) are themselves combinations of simpler determinants. As an example, among the factors governing the physical condition of a highway are structural adequacy of the pavement, skid resistance, manifestation of distress, and bridge rating. Structural adequacy is the ability of the pavement to carry the load. The manifestation

16

L.A.

CHAMEAU ETAL.

of distress concerns blow-ups and cracking. The bridge rating is itself a combination of bridge deck and structural adequacy. The variables which control each factor and their assembly into some "level" of quality correlated with performance are subjective matters. The evaluation of highway system performance is a problem similar to the damage assessment of structures. The rule-inference techniques recently proposed by Ishizuka et al. (1981) can certainly be applied to this performance evaluation. These techniques follow along two lines: (1) organize information (experts' knowledge) in a logical manner; and (2) use an inference technique to manipulate the information to obtain an answer to the original problem, i.e. highway system performance. The information can be organized in a hierarchical tree. The first level of the tree will be system performance, while the second level will consist of safety performance, physical condition performance, and service performance. Other levels will be introduced as needed. A third level for physical condition performance would be composed of structural adequacy of the pavement, manifestation of distress, skid resistance and bridge rating. A fourth level would be necessary for bridge rating, etc. Levels will be connected by a set of logical rules which represent the necessary knowledge to obtain an intermediate diagnostic state. A simple 7ule may be IF the structural adequacy is good, AND the skid resistance is good, THEN, it is almost definitively true that the physical condition is good. The rules will be defined using expertise from highway engineers. Any complex rule can be expressed with the clauses IF and THEN, and the the relations AND, OR, and COMB (combination). Each of the pieces of information present in the inference tree will be represented by a fuzzy set. The performance level will be represented by a numerical grade between 0 and 1 with a verbal interpretation attached to this grade: 0 Very poor 0.1 0.2 Poor 0.3 0.4 0.5 Fair 0.6 0.7 0.8 Good 0.9 1.0 Excellent

Therefore, the rules will be written with fuzzy subsets, such as excellent, good, fair, etc. The fuzzy subsets can be characterized with membership functions such as excellent performance = 0-2/0.8 + 0.8/0.9 + 1/1, fair performance = 0.4/0.3 + 1/0.4 + 0.8/0.5 + 0.3/0.6. The use of descriptive words or phrases to evaluate performance is not foreign to highway engineers. From the first inception of highway performance, the ultimate judge has been considered the road user. In particular, rating panels have been used to make the judgment of riding quality (service performance). They have been developed using cress-sections of the road-user population. The AASHO Test Road used panels of 30-35 people; others have used different numbers. Each rater is asked to rate rideability between very poor to very good, using a number scale such as 1-5. There is usually a wide variation in opinions of panel members.

FUZZY

SETS A N D C I V I l , E N G I N E E R I N G

17

An inference technique is then necessary to process the rules and fuzzy subsets in an effective manner. The Dempster and Shafer theory (Dempster, 1967; Shafer, 1976) and its extension to fuzzy sets have been used successfully for damage assessment of structures (Ishizuka et al., 1981) and a similar approach may be developed for highway performance. The Dempster and Sharer probability has been found to be an effective alternative to the Bayesian probability in expressing subjective judgment. The application of these ideas are still of an exploratory nature and future work is needed to define the logical rules and a satisfactory inference technique. It is hoped that they will stimulate interest among highway engineers. Their collaboration is essential to generate the needed expert knowledge data base.

Concluding remarks
Potential applications of the theory of fuzzy sets to various sub-disciplines within Civil Engineering have been presented. Civil Engineering is an art, the practice of which requires not only scientific knowledge but also engineering judgment and experience. The theory of fuzzy sets offers us a new tool to express subjective information and engineering judgment in the making of engineering decisions. It can also provide a rational approach to the transfer of experience to younger engineers. The authors believe that, with further studies, the approach can be most useful in solving practical problems in the Civil Engineering profession.

References
BI.OCKLEY, D. I. (1975). Predicting the likelihood of structural accidents. Proceedings of the Institution of Civil Engineers, 2 (59), 659-668. BLOCKLEY, D. I. (i977). Analysis of structural failures. Proceedings of the Institution of Civil Engineers, 1 (62), 51-74. BLOCKI.EY, D. I. (1979). The role of fuzzy sets in civil engineering. Fuzzy Sets and Systems 2, pp. 267-278, Amsterdam: North-Holland Publishlng Company. BROWN', C. B. (1977). Analytical methods for environmental assessment and decision making. Regional Environmental Systems, Assessment of RANN Projects, Department of Civil Engineering, University of Washington, Seattle. BROWN, C. B. (1979). A fuzzy safety measure. Journal of the Engineering Mechanics Division, American Society of Civil Engineers, 105 (EM5), 855-872. BROWN, C. B. (1980). The merging of fuzzy and crisp information. Journal of the Engineering Mechanics Division, American Society of Civil Engineers, 106 (EM1), 123-133. BROWN, C. B. & LEONARD, R. S. (1971). Subjective uncertainty analysis. Presented at the American Society of Civil Engineers Meeting, Baltimore, Maryland, Preprint No. 1388. CORNEI.I., C. A. Private Communication (27 April 1982). DEMPSTER, A. P. (1967). Upper and Lower probabilities induced by a multi-valued mapping. Annals of Mathematical Statistics, 38. DITLEVSEN, O. (1980). Formal and real structural safety, influence of gross errors. In THOFT CHRISTENSEN, P., Ed., Lectures on Structural Reliability, 121-147. Aalborg, Denmark: Aalborg University. DuBOIS, D. & PRADE, H. (1980). Fuzzy Sets and Systems: Theory and Applications. New York: Academic Press. ELMS, D. G. (1981). Use of fuzzy sets in developing code risk factors American Society of Civil Engineers National Convention, St Louis ['reprint No. ,91.509. GUPTA, M. M., SARIDIS, G. N. & GAINES, B. R. (1977). Fuzzy Automata and Decision Processes. New York: Elsevier/North-Holland.

18

L. A. CHAMEAU ETAL.

ISHIZUKA, M. (1982). Extension of Dempster and Shafer's theory to fuzzy set for constructing expert systems. Summary of Papers on General Fuzzy Problems, Report No. Z The working Group on Fuzzy Systems, Tokyo, Japan. ISHIZUKA, M., FU, K. S. & YAO, J. T. P. (1981). A rule-inference method for damage assessment. American Society of Civil Engineers Convention St. Louis. Paper presented at Session No. 81-EM, Civil Engineering Applications of Bayesian Statistics and Fuzzy Sets. ISHIZUKA, M., FU, K. S. & YAO, J. T. P. (1981). Inference procedure with uncertainty for problem reduction method. Structural Engineering Report, Purdue University, CE-STR-8124. MUNRO, J. & JOWITT, P. W. (1978). Decision analysis in the ready-mixed concrete industry. Proceedings of the Institution of Civil Engineers, 2 (65), 41-52. SEED, H. B. & IDRISS, I. M. (1981). Evaluation of liquefaction potential of sand deposits based on observations of performance in previous earthquakes. American Society of Civil Engineers National Convention, St Louis, Preprint No. 81.544. SHAFER, G. (1976). A Mathematical Theory of Evidence. Princeton: Princeton University Press. SOZEN, M. A. (1980). Review of earthquake response of reinforced concrete buildings with a view to drift control. In ERGUNAY, O. & ERDIK, M., Eds, State-of-the-Art in Earthquake Engineering, pp. 383-418. Turkish National Committee on Earthquake Engineering. YAO, J. T. P. (1980). Damage assessment of existing structures. Journal of the Engineering Mechanics Division, American Society of Civil Engineers, 106 (EM4), 785-799. YAO, J. T. P., Fu, K. S. & ISHIZUKA, M. (1981). Fuzzy statistics and its potential applications in Civil Engineering. American Society of Civil Engineers National Convention, St Louis, Preprint 81.501. ZADEH, L. A. (1965~. Fuzzy sets. Information and Control, 8, 338-353. ZADErt, L. A. (1968). Probability measures of fuzzy events. Journal of Mathematical Analysis and Applications, 23, 421-427. ZADEH, L. A. (1973). Outline of a new approach to the analysis of complex systems and decision processes. Institute of Electrical and Electronics Engineers, Transactions on Systems, Man and Cybernetics, SMC-3(1), 28-44. ZADEH, L. A. (1975). The concept of linguistic variable and its application to approximate reasoning, parts I, II and III. Information Science, 8, 199-249, 301-357, 9, 43-80.

S-ar putea să vă placă și