See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/384246445

AI for Civil Engineers _ Module 4_ Dealing with Uncertainty

Chapter · September 2024

CITATIONS READS

0 223

1 author:

Jayaram M.A
RASTA - Center for Road Technology VOLVO Construction Equipment Campus
196 PUBLICATIONS 1,527 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jayaram M.A on 23 September 2024.

The user has requested enhancement of the downloaded file.

 AI for Civil Engineers
Module 4

Dealing with Uncertainty: Fuzzy Logic

RASTA
Dr.M.A.Jayaram
Centre for Road Technology Professor
VOLVO-Construction Equipment Campus
Bengaluru
 CONTENTS
Introduction
Fuzzy sets and Membership functions
Operations on Fuzzy sets
Differences between fuzziness and
randomness
Differences between Fuzzy sets and crisp
sets
Illustrative examples
Features of MFs
Types of MFs
Building FIS – steps
Recent Applications in Civil Engineering
 Module 4

Learning Under Uncertainty

1. Introduction
In this chapter, AI branch that is devoted to dealing with uncertainty, ambiguity or vagueness
is presented. The intent is to know how AI systems could be enabled to deal with data that is
mired in ambiguity, or vagueness. A few examples of approximate statements are;

• The soil is moderately stable for construction

• The traffic flow is somewhat heavy during peak hours
• The building structure is fairly safe under seismic conditions
• The water quality in the reservoir is acceptable for irrigation purposes
• The concrete mix has a reasonably good compressive strength.
• There is a high chance of moderate flooding in this area
• The road surface is slightly uneven
• The estimated cost of the project is fairly reasonable
• The crew is performing at an acceptable level of productivity
• When water cement ratio is less, strength of concrete is high.
• High intensity rain fall will result in high runoff.
• Silty soil has low shear strength.
• High constriction in the water way beneath the bridge will give rise to high afflux
during floods.
The above statements are made during day-to-day activities as civil engineers. Such statements
are also common in every other walk of life. More often, human beings tend to be approximate
but, they are capable of interpreting the meaning and are able to assess the truth of such
statements for the given context.
Continuing with the examples above, the terms;

• Moderately stable reflects uncertainty or varying degrees of stability.

• Somewhat heavy is a fuzzy descriptor, as the level of congestion may vary and cannot
be quantified precisely.
• Fairly safe is a vague term and reflects the possibility of safety within a certain range
rather than a definitive assessment.
• Acceptable is an uncertain term, as the quality can fluctuate within a range that is neither
perfect nor hazardous.
• Reasonably good reflects an estimate based on experience rather than precise numerical
values.
• Approximately introduces uncertainty, acknowledging that the timeline is not exact.
• High chance and moderate flooding are both fuzzy, describing likelihood and intensity
in non-specific terms.
• Slightly uneven is subjective and introduces a degree of uncertainty regarding the exact
roughness.
1
 • Fairly reasonable is imprecise and allows for variability in the interpretation of the cost.
Acceptable level is an uncertain statement, as the performance may fluctuate within
certain undefined bounds.
Engineering projects often involve significant uncertainties and variability, necessitating
flexibility in both the design and decision-making phases. Fuzzy logic plays a crucial role
in addressing these uncertainties, offering a way for machines, computers, or AI systems to
effectively manage imprecise information. Fuzzy logic is especially useful in two main
scenarios:
• When dealing with highly complex systems whose behaviors are not fully
understood.
• When a quick, approximate solution is needed.
Fuzzy theory is based on fuzzy logic, which handles uncertainties that are non-statistical in
nature. It is important to note that fuzziness is distinct from randomness. Randomness has
long been addressed through probability theory, which underpins the foundation of basic
statistics and the theory of random processes. To recall, stationary random processes exhibit
the following characteristics:
• The sample space remains constant across all experiments, meaning the possible
outcomes do not change.
• The probability or frequency of an event within the sample space is fixed and
remains consistent from one trial to another.
• The results of one trial do not affect the outcomes of previous or subsequent trials,
ensuring replicability.
On the other hand, fuzzy logic tackles uncertainty in information, which can arise in multiple
forms. This uncertainty may be due to complexity, such as the intricate reliability network
of a nuclear reactor, or from other factors like ignorance, various types of randomness,
inadequate measurement tools, lack of knowledge, or the inherent vagueness of natural
language. The nature of the uncertainty in a particular problem is critical, and engineers
must carefully assess it before selecting the most appropriate method to represent it. Fuzzy
sets provide a mathematical framework for addressing vagueness and ambiguity in human-
centered systems. It is important to clearly distinguish between fuzziness and randomness,
as outlined in Table 4.1.

2. Fuzzy Sets
We need fuzzy sets to capture uncertainty in information. Fuzzy sets may be defined as group
of elements, each element attached with a degree of belongingness to the set. The degree of
belongingness is variously called as membership value, degree of truth, degree of possibility.
Unlike conventional or crisp sets that are represented by algebraic notations (such as A,B,C,,,)
fuzzy sets are represented by Linguistic variables.
Table 4.1. Differences between randomness and fuzziness
Randomness Fuzziness
1 Randomness is uncertainty associated Fuzzyness is associated with description of
with happening of the event itself. the event
Ex. The probability of accident in a Ex. Severe accident, fatal accident, gory
particular road is 0.2. accident etc…
Disastrous hurricane, extreme hurricane.

2
 The likelihood of a hurricane is 0.4 from
previous data.
2. Founded on theory of probability Founded on theory of possibility
3. Sum of the probabilities of mutually Sum of the possibilities (degree of truth/
exclusive events should add up to 1. membership value.) need not add up to 1.
4. All possible things are probable All probable things need not be possible.
5. The crisp set theory provides a basis of Fuzzy set theory provides basis for
probability theory possibility theory.

Crisp sets are limited by binary logic – i.e., an element in as set either belongs to the set or
doesn’t belong, nothing in between. Crisp sets are represented by characteristic function
(denoted by  and defined as in equation 4.1:

A = 1 if x  A (4.1)

0 if x A
Here, A is the set. This functional representation of a conventional set indicates that, the degree
of truth is either 1 or 0 depending on an element x, belonging to a set A or not. The
characteristic function may also graphically represent. Suppose A is a crisp set of all concrete
strengths ranging between 50 MPa ≤ f ≤ 80 MPa, is shown in figure 4.1a. A particular concrete
say, C1, has a height of 45 MPa. The membership of this individual sample is equal to 1, or full
membership, given symbolically as χC(c1) = 1. Another sample, say c2, has a strength of 49.9
MPa. The membership of this individual in set C is equal to 0, or no membership, hence χC(c2)
= 0, also seen in Figure 4.1a. In these cases, the membership in a set is binary, either an element
is a member of a set or it is not.

c(c) “High strength Concrete”

C(x) 1 1

0 0
50MPa 80MPa 45 50 80 95
a) Crisp set C b) Fuzzy set
Fig. 4.1 Concrete strength membership functions
Zadeh, often referred to as the father of fuzzy logic, expanded the concept of binary
membership to allow for varying "degrees of membership" within the continuous real interval
[0, 1]. In this system, the endpoints 0 and 1correspond to no membership and full membership,
respectively, similar to how the indicator function operates for crisp sets. However, the values
between 0 and 1 represent different levels of membership for an element xxx within a given set
in the universe. Zadeh coined the term "fuzzy sets" for these sets within the universe XXX that
can accommodate such degrees of membership.
Let’s extend the discussion using the example of concrete strengths. Consider a fuzzy set named
"High Strength Concrete," which includes concrete strengths ranging from 45 MPa to 95 MPa,
as illustrated in Figure 4.1b. A sample with a strength of 45 MPa is part of this set, but its

3
membership value is 0. Meanwhile, for strengths of 45 MPa and 80 MPa, the membership
values might be 0.3 and 1, respectively. A sample with a strength of 85 MPa would have full
membership (a value of 1), while a sample with 95 MPa strength would have a membership
value of 0 in this set. However, the same sample with 95 MPa strength could belong to a
different fuzzy set called "Ultra High Strength."
Fuzzy sets are defined by a membership function, denoted as μA, where the membership values
fall within the interval [0, 1]. This range includes the extreme values of 0 and 1, as well as an
infinite number of intermediate values between them.
A  [0,1]
The membership function provides a mathematical expression of how an element belongs to a
set. The notation for a fuzzy set is typically represented by an uppercase letter with a tilde, such
as Ā. The function μA(x) represents the degree of membership of element x in fuzzy set A. The
differences between fuzzy sets and classical (crisp) sets are outlined in Table 4.2.
Table 4.2. Differences between fuzzy and crisp sets
Crisp Sets/Conventional sets Fuzzy Sets
1.Based on Newtonian set theory Based on Fuzzy set theory
2.The members of the set either belong to the set Members of a fuzzy set will partially belong to a
with full membership value (1) or they do not set, that is their presence may be seen in other sets
belong. also, with different degrees of truth.
3.Crisp sets do not overlap Fuzzy sets will overlap
4.Can not capture vagueness, ambiguity, or Fuzzy sets capture vagueness, ambiguity, or
uncertainty in information. uncertainty in information.
5.They are applicable where precision matters. They are not applicable where precision matters.
6. Axiom of excluded middle and axiom of Axiom of excluded middle and axiom of
contradiction holds good. contradiction does not hold good.
A  A1 = U, and A  A1 =  A  A1 ≠ U, and A  A1 ≠ 
Ex. Set of cars with velocity > 60 km/h Fast cars
Acidic water : 1<pH <7 Acidic Water / not palatable
Distinction grade : 100 ≤Average ≥70 Intelligent students
High strength concrete
Highly workable concrete.

2.1 Representation of fuzzy sets

A notation convention for fuzzy sets when the universe of discourse, X, is discrete and finite,
is as follows for a fuzzy set A. In function theoretic form, they are called membership functions
(MF) . Customarily, MF are represented more often in graphical form.

(4.3)

When the universe, X, is continuous and infinite, the fuzzy set A is denoted as:

(4.4)

In both of these notations, the horizontal bar functions as a delimiter rather than representing
division. The numerator in each term denotes the membership value in set A, while the
4
denominator identifies the corresponding element from the universe. In the first notation, the
summation symbol does not imply standard algebraic addition but instead signifies the
aggregation or collection of elements, meaning that the "+" signs represent an aggregation
operator, not a typical algebraic sum. Similarly, in the second notation, the integral sign is not
meant as an algebraic integral but rather serves as a continuous aggregation operator for
variables. Both of these notational conventions were introduced by Zadeh.

3.Illustrative Examples of MFs

Fuzzy sets form founding stones of a fuzzy mapping systems, fuzzy control systems, or fuzzy
inference systems (FIS). Therefore, identification of appropriate MF is a must before
developing fuzzy models. Following examples will consolidate our idea of fuzzy MFs.
Example 1. To define reasonable membership functions for the fuzzy sets based on the setting
times of epoxies (a) "extra fast," (b) "fast," and (c) "slow" we can use triangular or trapezoidal
membership functions, which are common in fuzzy logic due to their simplicity and intuitive
interpretation.

(t) 1

Extra fast Fast Slow

0
30 60 90 120 150
180 210 240 270 300
Setting time t in mins
Fig. 4.2. Fuzzy MFs for setting times of Epoxy
Example 2. To develop membership functions for the curling of a concrete slab based on
temperature, we can define fuzzy sets corresponding to three states: "curling," "no curling,"
and "some curling." Since the occurrence of curling depends on temperature, the membership
functions will assign different degrees of membership to each temperature range.

Mild Curling
(C) 1 moderate High

0
10 20 30 40 50 60 Temp in 0 C

Fig. 4.3. Fuzzy MFs for curling temperatures of Concrete slab

Example 3. Industry A releases wastewater into a nearby river, which contains high levels of
biological oxygen demand (BOD) and other inorganic pollutants. The discharge rates of both
the river and wastewater remain constant throughout the year. Research indicates that BOD
values up to 250 mg/l have no harmful effects on aquatic ecosystems, but values exceeding 250
mg/l have a significant impact. Create fuzzy membership functions to represent the effects of
BOD levels on aquatic ecosystems

5
 Low effect Severe effect
Safe
(B) 1

0
100 150 200 250 300 350 400 BOD mg/l
Fig. 4.3. Fuzzy MFs effect of BOD

Example 4. To develop membership functions for the three soil types—clay, silt, and sand—
based on grain size in mm , we can use fuzzy logic to represent the gradual transition between
these soil categories, as the boundaries between them are not always distinct..

Clay Silt Sand

(S) 1

0
0.01 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.5 1.0 1.5

Grain size in mm

Fig. 4.3. Fuzzy MFs for kinds of soils

Example 5. circular column subjected to axial loading is considered eccentric if the load is
applied at a distance of 5% of the column's axis, relative to its diameter, ddd, as illustrated in
Figure 4.4. The conditions are as follows: e/d=0.05 is classified as eccentric; e/d<0.05 is
considered not very eccentric; and e/d>0.05 is regarded as highly eccentric. Construct a
membership function for "eccentricity" based on the range of e/de/de/d ratios.

Fig. 4.4. Eccentrically loaded

column

6
 Not very eccentric Excentric Very eccentric
(E) 1

0
0.01 0.05 0.10 0.15 0.20 0.25 e/d ratio➔

Fig. 4.5. Fuzzy MFs for eccentric column loading

Example 6. A bridge pier is considered "safe" if its settlement is less than 25 cm over its lifetime
and "unsafe" if the settlement exceeds 50 cm. Construct the membership function for a "safe"
pier

(C) 1
Safe

0
5 10 15 20 25 30 35 settlements in cms

Fig. 4.5. Fuzzy MF for safe settlement

Example 7. As an example, for discrete fuzzy sets. Consider a scenario where we aim to
compare the strength of two types of concrete. Four concrete masonry units (CMUs) from each
type are subjected to stress until they fail. The lowest stress at failure is assigned a rank of 1,
and the highest is assigned a rank of 4, meaning the CMUs are ranked by failure stress, i.e., X
= {1, 2, 3, 4}. Since the "failure" of CMUs is fuzzy, the membership value for each CMU reflects
the judgment of whether it truly failed. The following fuzzy sets represent the failure estimates
for the two concrete types:

3. Basic Fuzzy Set Operations

Define three fuzzy sets A, B, and C on the universe X. For any element x in X, the following
function-theoretic operations corresponding to the set-theoretic operations of union,
intersection, and complement are defined for A, B, and C on X.
Union: Between two fuzzy sets , the highest value of MF between A and B

7
 (4.4)

Intersection: Between two fuzzy sets, the lowest value of MF between A and B
(4.5)

Complement: The complement of a fuzzy set A is given by;

(4.6)

Union, intersection and complement of fuzzy set A is shown in figure 4.6 a, b, and c for
continuous fuzzy sets.

a) Union b) Intersection c) Complement

Fig.4.6. Operations on continuous fuzzy sets

Two illustrate the operations on discrete fuzzy sets, we have two discrete fuzzy sets namely A,
and B.

In the above discrete fuzzy sets, denominators represent the parametric values, numerator
represents the degree to which this value is a member in the sets A and B respectively.
i. Complement:

ii. Union:

8
iii. Intersection:

iv. Difference:

4. Types of Membership Functions

Fuzzy Membership Functions are a core concept in fuzzy logic, used to represent the degree to
which an element belongs to a fuzzy set. As described in preceding sections, unlike classical
sets where an element either belongs or doesn't (with values of 0 or 1), fuzzy sets allow for
partial membership with values ranging between 0 and 1. The membership function assigns a
membership degree to each element in the set. Some of the widely used MFs are presented
here.
5.1 Triangular MF
The algorithmic representation of triangular MF (Figure 4.7) is provided in equation 4.7. This
is commonly used in most of the civil engineering applications because of its simplicity.

Fig. 4.7 Triangular MF with functional representation

Triangular MF is apt whenever there is only one parametric value that correspond to
membership value of 1. In figure 4.7 when x = b, A(b) = 1.
5.2. Trapezoidal MF
Trapezoidal MF is to be used when there is a range of variable value corresponding to truth
value of 1. (Ex. Teenage, around 4, between 6-8, etc…). The MF and its functional
representation is shown in figure 4.8

9
 Fig. 4.8. Trapezoidal MF and its functional representation
5.3.Gaussian MF
This MF is used whenever the truth value is nonlinear. This function is complex and
computationally intensive (Figure 4.9)

Fig 4.9. Gaussian MF

5. Features of MFs
As the membership function of a fuzzy set encapsulates all its information, it is helpful to
establish a lexicon of terms to describe various key features of this function. For simplicity, the
functions illustrated in the figures will be continuous, though the terminology applies to both
discrete and continuous fuzzy sets. Figure 4.10 aids in this explanation.

Fig. 4.10. Features of fuzzy MF

10
The core of a membership function for a fuzzy set A is defined as the region of the universe
where elements have full membership in A, meaning those elements x for which μA(x)=1.
The support of a membership function for a fuzzy set A is the region of the universe where
elements have non-zero membership in A, meaning those elements x for which μA(x)>0.
The boundaries of a membership function for a fuzzy set AAA are the region of the universe
containing elements with partial membership, meaning those elements xxx for which
0<μA(x)<10. These elements exhibit some degree of fuzziness. Figure 4.10 illustrates the core,
support, and boundaries of a typical fuzzy set.
A normal fuzzy set is one in which at least one element xxx in the universe has a membership
value of 1. When only one element has a membership of 1, it is often called the prototype or
prototypical element of the set. Figure 4.11 shows examples of normal and subnormal fuzzy
sets.

Fig. 4.11 Fuzzy MF types: a) Normal and b) Subnormal

The height of a fuzzy set AAA is the maximum value of its membership function, expressed as
hgt(A)=max{μA(x)}. If hgt(A)<1 , the fuzzy set is considered subnormal. The height of A can
be interpreted as the degree of validity or credibility of the information represented by the set.

6. Developing Fuzzy Inference System (FIS)

FIS is variously called as / Fuzzy Expert System / Fuzzy Decision Support System is a process
that translates a set of fuzzy rules into a decision or control output. To develop a FIS for a given
problem plagued with uncertain, ambiguous, vague or heuristic data, following steps are to be
followed. An overall block diagram of FIS is shown in figure 4.12. Building a FIS involves
following steps:
• Fuzzification
• Formation of Fuzzy Rule base
• Training of the system
• Defuzzification

11
 An example of 3 inputs to 1 output
fuzzy mapping

Fig. 4.12 Block diagram of FIS

6.1 Fuzzification
This is the first step, by fuzzification we mean converting crisp input values (exact numbers)
into fuzzy values (degrees of belonging to certain fuzzy sets). Each input variable, we need to
define MFs that describe how each value fits into a fuzzy set (with gradation of each variable
in to , "low," "medium," "high"). This also means we need to identify appropriate linguistic
qualifiers. Table 4.3 gives some classical fuzzy variables and their gradations.
To make fuzzy gradations, the domain knowledge or heuristic knowledge is a must. The values
of the variables, depends on the context. For example, if we take Temp as the variable for
certain fuzzy system, if it is atmospheric temperature, it may start from 00 to 500 C, on the other
hand if the context is a cement clinker burning kiln, the range of temperature may be 800 0C –
18000C. After deciding the variables involved, we need to make a logical union of all the
variables. We need to fuzzify the output values also in to different gradations.
Table 4.3 Typical fuzzification examples
Linguistic Term set / fuzzy sets/ Names of MFs Numerical measures
Variables
1 Speed {Low, moderate, high, very high } Speed values in Km/h
2 Size {tiny, small, big, huge, very huge} The volume
3 Shape {rounded, flaky, angular, oblong} The aspect ratio, thickness /
largest dimension
4 Temperature {very cool, cool, warm, hot, very hot} Temperature values
5 Strength {low, moderate, high, ultra-high} Stress values
6. W/C ratio {very low, low, moderate, high, very W/C ratio values
high}
7. Contaminant {low, moderate, high, very high} Contaminants expressed in
level ppm , gm/l etc…
8. Earthquake {low, moderate, severe, devastating} Numbers on Richter’s, scale.
intensity

12
6.2 Formation of Fuzzy Rule Base
After having the inputs and outputs fuzzified in to their gradations, we need to develop fuzzy
rules. Rule formation in case of a two input and one output system is elaborated with the help
of figures given in 4.13. The inputs for the system are A and B, while the output is C.

Rule1 Low
low low
C
V A B

IF A is “Low” and B is “ Low” Then ➔ C is “Low”

low medium low high Low high

A B
Rule 2 C

IF A is “medium” and B is “high” Then C is “High”

low medium low high very high Low high very high

A B C
Rule 3

IF A is “Low” and B is “Very High” Then C is “Very High”

Fig. 4.13. Developing 3 rules using two inputs (A & B) and one output (C)

It is important to remember that the rules will emanate through the data base. For example, the
first row of the dataset corresponded to low value of variable A, and low value of variable B,
and the corresponding output C also fell under MF low.
6.3 Training the FIS
A dataset containing input-output pairs is used to train the system. The FIS attempts to learn
the underlying relationship by adjusting the membership functions and rule base based on this
dataset. During the training, the parameters of the fuzzy membership functions (e.g., the shape,
range, or type of function) are adjusted to minimize the error between the predicted output of
the system and the actual target output.
Various optimization techniques like gradient descent, genetic algorithms, or particle swarm
optimization can be used to automate the tuning of the system. This process aims to minimize
some predefined error metric (such as mean squared error) over a set of training data. Training
a FIS is significant because of the following reasons:

13
 • Trained systems are more accurate in predicting or controlling the outputs based on
input data.
• For systems with many inputs or complex dynamics, training ensures that the fuzzy
system can handle subtle variations and interactions between variables.
• Manual tuning of membership functions and rules can be time-consuming. Training
automates this process using data-driven methods.
6.4 Defuzzification
Defuzzification is the final step in a Fuzzy Inference System (FIS), where fuzzy values are
converted into crisp, precise values. In some cases, the output of a fuzzy process needs to be a
single scalar value rather than a fuzzy set. Defuzzification serves this purpose by transforming
a fuzzy quantity into a clear, defined one, similar to how fuzzification converts a precise
quantity into a fuzzy one. The output from a fuzzy process may be the combination (logical
union) of multiple fuzzy membership functions defined over the output variable's universe of
discourse. When new input values for variables A and B are introduced in the trained system,
Mamdani’s method (1966) is used to generate the output. This involves aggregating the rules
through the max-min principle.

(4.7)

This rule can be understood with an example, consider two input one output fuzzy system. Let
the rule 1 be : IF Low A11 and Medium A12 Then Very low B1
Let the rule 2 be : IF Medium A1 and Low A2 Then Low B1
For new inputs, input (i) and input (j) corresponding to A1 and A2 , the projected lines will cut
the input fuzzy MFs, as shown in figure 4.14. The minimum value should be propagated
towards the output MFs and the MFs must be truncated. The union of this truncated output
MFs results in an aggregated output function. The final output must be retrieved by
defuzzification of C1 C2. Defuzzification produces scalar output (crisp) which is most
representative value that encompasses the fuzziness associated with inputs and outputs.
A typical example could be, a fuzzy inference system for predicting the settlement of shallow
foundations can take inputs such as soil type, water table depth, and load magnitude. Each input
is assigned a membership function (a fuzzy number), and through fuzzy rules like
"IF soil type is soft AND load is high, THEN settlement is high"
The system provides an output indicating the settlement category (low, medium, high). This
fuzzy-based approach simplifies the decision-making process and allows engineers to make
more informed judgments even with imprecise or incomplete data.

14
 Truncated output MF for rule 1 C1

Truncated output MF for rule 2 C2

Aggregated output MF C1  C2

Fig. 4.14. Mamdani Max-Min inferencing method.

There are several methods of defuzzification techniques like,

• Centroid method
• Maximum Membership method
• Weighted average method
• Mean max membership method
• Center of sums method
• Center of largest area method
• First (or last of maxima method
Centroid method is the most widely used method as it provides accurate output. Centrod
method is given in algebraic notion as:
C1

(4.8)

1
As per eqn 4.8, the consequent fuzzy output MF should be divided in to elementary areas and
centroid is to be computed. This value is the output of the FIS.

15
7. Illustration of centroid method of defuzzification
In a typical FIS let the rule 1 produce the output as shown in figure 4.13. Also let rule 2 produce
the output MF as shown in 4.15. The union of these two MFs will result in aggregated MF as
shown in figure 4.16.

C1
C2

Fig. 4.15 Truncated output after rule 1 Fig. 4.15 Truncated output after rule 2

The centroid of consequent fuzzy set is the

scalar solution

2 4
1 3

Fig. 4.16 Aggregated output MF

The centrod method is similar to the way in which we find the centroid of a planar figure or
planar cross section of a structural component.

8. Applications of Fuzzy Logic in Civil Engineering

Fuzzy logic and fuzzy inference systems (FIS) have been widely applied in civil engineering
to handle uncertainty and imprecision, which are common in real-world scenarios. The primary
advantage of fuzzy logic is its ability to model human reasoning and make decisions under
conditions of vagueness, which conventional binary logic struggles to address. Some of the key
areas where fuzzy logic and fuzzy inference systems applied in civil engineering are listed.

16
8.1 Structural Engineering
• Structural Health Monitoring (SHM): Fuzzy logic is used to assess the health of
structures like bridges, dams, and buildings. By integrating sensor data and using fuzzy
inference systems, engineers can evaluate damage levels and predict potential failures.
• Structural Design Optimization: In situations where multiple conflicting criteria (like
cost, strength, and durability) need to be optimized, fuzzy logic helps in decision-
making by providing trade-offs between these criteria.
• Seismic Risk Assessment: In seismic analysis, fuzzy logic can model the uncertainty
involved in seismic hazard estimation, building response to earthquakes, and the
decision-making process for retrofitting or evacuation.
8.2 Geotechnical Engineering
• Soil Classification and Slope Stability: Fuzzy logic models are employed for the
classification of soil types where boundaries between classes are not distinct. It is also
used in slope stability analysis where parameters like cohesion, angle of friction, and
water content are not precisely known.
• Settlement Prediction: Fuzzy inference systems can predict settlement of foundations,
particularly where soil conditions are highly variable or uncertain, based on input
parameters such as soil type, load, and consolidation characteristics.
8.3 Water Resources Engineering:
• Flood Risk Management: Fuzzy systems are used to evaluate flood risks by
incorporating uncertain parameters such as rainfall, land use, and drainage conditions.
These models help in decision-making for flood warning systems and urban flood
planning.
• Water Quality Management: Fuzzy logic can be employed to assess the quality of water
in reservoirs, rivers, and treatment plants by accounting for vague and uncertain factors
such as pollutant levels and water chemistry.
8.4 Construction Management:
• Project Risk Assessment: Fuzzy logic helps in assessing risks in construction projects,
where there are uncertainties related to cost, time, and resource availability. Fuzzy
systems can quantify these uncertainties and offer better risk management strategies.
• Resource Allocation and Scheduling: Fuzzy logic-based approaches have been applied
for optimizing resource allocation, particularly when the tasks' durations and resource
availabilities are uncertain. It is also useful in managing workforce, machinery, and
materials more efficiently.
8.5 Pavement Design and Maintenance:
• Pavement Condition Assessment: Fuzzy logic helps in evaluating the condition of
pavements by integrating factors such as surface roughness, cracking, and rutting into
a single, easily interpretable index. This allows for more precise decision-making in
maintenance planning.

17
 • Material Selection: Fuzzy inference systems assist in selecting materials for pavement
layers by taking into account various factors like strength, durability, and cost in
uncertain environments.
8.6 Environmental Engineering:
• Air Pollution Assessment: Fuzzy models are used to assess the impact of construction
activities on air quality by accounting for the uncertain nature of emissions, weather
conditions, and pollutant dispersion.
• Wastewater Treatment: In wastewater treatment plants, fuzzy logic controls can
optimize the process by adjusting variables like flow rate and chemical dosing in
response to uncertain input conditions such as pollutant loads and weather changes.
8.7 Earthquake Engineering:
• Ground Motion Prediction: Fuzzy logic has been used to predict ground motion
parameters like peak ground acceleration (PGA) and spectral acceleration, where the
inputs are often uncertain or incomplete.
• Retrofitting Strategies: Fuzzy inference systems assist engineers in deciding whether to
retrofit a building or structure by considering factors such as cost, risk of damage, and
expected earthquake magnitudes, which are inherently uncertain.
8.7 Traffic and Transportation Engineering:
• Traffic Flow Control: Fuzzy logic is used to model and control traffic flows at
intersections, especially in urban areas where traffic dynamics are highly uncertain. It
helps in reducing congestion and optimizing signal timings.
• Accident Risk Assessment: Fuzzy inference systems can estimate the risk of accidents
in road networks by considering uncertain variables such as weather conditions, road
geometry, and driver behaviour.

Module End Questions

1.What is fuzzy logic? What it captures? Give an example drawn from civil engineering
domain?
2. How fuzzy sets are different from crisp sets? Give examples from civil engineering where
fuzzy sets hold good and crisp sets hold good.
3. Draw fuzzy membership functions using your intuition and the domain knowledge.
• Acidic Water
• Turbid water.
• Soil categories based on grain size.
• Soft clay based on plasticity Index.
• Soft, Medium, Hard ( based on SPT values)
• Load intensity gradations as Light, Moderate, Heavy
• Pavement conditions based on PCI values as Low, Moderate, High
• Seismic risk based on intensity on : Peak Ground Acceleration (PGA) or Magnitude
(Richter Scale)
18
 • Water Quality using terms : Poor, Average, Good, appropriate Measure: Water Quality
Index (WQI) or pH levels
• Traffic congestion as : Low, Medium, High, based on Volume-to-Capacity Ratio (V/C
ratio) or Average Vehicle Delay (seconds).
• Construction Risk as: Low, Medium, High, based on Cost overrun percentage,
Schedule delay percentage, or Risk Index.
• Slope Stability as: Stable, Marginally Stable, Unstable based on factor of safety.
• Crack severity, as Minor, Moderate, Severe, based on Crack Width (mm) or Crack
Length per Unit Area
• Flood risk as , low, moderate, high based on Return period (years), Water Level
(meters)
• Material durability as Low, moderate, High, based on Compressive strength (MPa),
Water absorption (%), or Chloride content.
• Foundation settlement as low, medium , high, based on based on settlement (mm).
4. What are the basic operations on fuzzy sets. Considering two MFs , show how union,
intersection and complement will look like.
5. What are the various types of MFs used while modelling a FIS. Give their functional
representation as well sketches.
6. With neat sketches, explain properties of fuzzy MFs.
7. What is fuzzy inference system? Elaborate on steps involved in building a FIS.
8. List and brief on various applications of fuzzy logic concepts and FIS in Civil Engineering.

References
1. Timoty.J.Ross, Fuzzy Logic with Engineering Applications, III edition, John Wiley Publishing , 2012
2. Michael Gr. Voskoglou, (2024), A Journey from Traditional to Fuzzy Methods of Decision-Making
Transactions on Fuzzy Sets and Systems, Vol.3(1), 136-150.
3. L.A. Zadeh.(1965), Fuzzy sets, Information and Control, 8(3),1965,338-353.
4. M.C.Nataraja, M.A.Jayaram, C.N.Ravikumar, (2006) A Fuzzy-Neuro Model for Normal Concrete Mix
Design, Engineering Letters, 13(2), 1-10.
5. Tan, H., Tarasov, V., Jarfors, A.E.W. et al. (2021) , A design of fuzzy inference systems to predict tensile
properties of as-cast alloy. Int J Adv Manuf Technol 113, 1111–1123.
6. M.C.Nataraja, M.A.Jayaram, C.N.Ravikumar, (2021) Prediction of Early Strength of Concrete, African
Journal of Physics, 8(6), 1-10.
7. Tanyıldızı, H., Ahmet, C., & Kun (2008). fuzzy logic model for prediction of compressive strength of
lightweight concrete made with scoria aggregate and fly ash, Engineering Material Science, 1(2), 1-12.
8. Rong, HJ., Yang, ZX. (2024). Fuzzy Inference Systems. In: Sequential Intelligent Dynamic System
Modeling and Control. Springer, Singapore.
9. Hosseini, S., Gordan, B. & Kalkan, E.(2024) Development of Z number-based fuzzy inference system to
predict bearing capacity of circular foundations. Artif Intell Rev ,57, 146.
10. Hakim, S., Jamaluddin, N., Boon, K., Mokhatar, S., Khalifa, A., Jamellodin, Z. (2024). Development of
Adaptive Neuro-Fuzzy Inference System to Predict Concrete Compressive Strength. Proceedings of
AWAM International Conference on Civil Engineering,Lecture Notes in Civil Engineering, vol 385.
Springer, Singapore.
11. Sadjad Abbasipayam , Natalia V. Mokrova,(2022) Fuzzy logic and intelligent control of engineering
systems of buildings, Management, Computer Science and Informatics, 1, 22-32.
12. Zadeh, L.A. (2019). "Fuzzy logic: A concept of analysis and decision-making in engineering."
Engineering Applications of Artificial Intelligence, 87, 123-135.
19
 13. Kumari, P., & Ahmed, S.M. (2020). "Fuzzy logic in civil engineering: A review of structural and
environmental applications." Journal of Construction Engineering and Management, 146(10), 04020094.
14. Raza, S. & Chaudhry, A. (2021). "Fuzzy inference systems for optimizing geotechnical risk in foundation
engineering." Soil Mechanics and Geotechnical Engineering, 45(2), 112-120.
15. Srinivasan, S., & Gupta, V.K. (2021). "Applications of fuzzy logic in traffic engineering and
transportation planning: A review." International Journal of Traffic and Transportation Engineering,
11(2), 76-84.
16. Jiang, C., & Yan, X. (2022). "Flood risk assessment using fuzzy logic and multi-criteria decision analysis:
A case study in urban areas." Water Resources Management, 36(3), 997-1012.

20

View publication stats