LECTURE NOTES FOR NEURAL NETWORKS AND FUZZY LOGIC (20APE0203)

UNIT-5
PATTERN RECOGNITION USING A FUZZY CONTROLLER
Pattern recognition using a fuzzy controller is a technology that involves identifying and categorizing
objects or data patterns based on their characteristics and features. Fuzzy controllers are particularly
useful in handling imprecise or uncertain information, making them suitable for various pattern
recognition applications. Let's explore the components and working of pattern recognition using a
fuzzy controller:
Components of Pattern Recognition using a Fuzzy Controller:
1. Input Data: This is the raw data or information that is fed into the pattern recognition system for
analysis. It can be in the form of images, signals, text, or any other data type relevant to the
application.
2. Feature Extraction: In many pattern recognition tasks, raw data needs to be transformed into a set
of relevant features that capture the essential characteristics of the objects or patterns. Feature
extraction methods help in reducing the dimensionality of the data and enhancing its discriminatory
power.
3. Fuzzy Controller: The fuzzy controller is the heart of the pattern recognition system. It consists of
several key components:

- Fuzzy Sets: Fuzzy sets are defined to represent linguistic variables, such as "small," "medium,"
and "large." These sets describe the degrees to which input data belong to certain categories.
- Fuzzy Rules: Fuzzy rules are defined to establish relationships between input features and output
categories. They are expressed in terms of "if-then" statements, where the "if" part involves fuzzy sets
and the "then" part involves fuzzy categories.

- Fuzzy Inference System: This component computes the degree to which input data satisfy the fuzzy
rules. It uses fuzzy logic to make inferences and assign membership values to different output
categories.
- Defuzzification: The output of the fuzzy inference system is typically a fuzzy set representing the
categories' membership degrees. Defuzzification converts these fuzzy values into crisp output
categories or decisions.
4. Output Categories: These are the predefined classes or categories into which the input patterns are
classified. The fuzzy controller assigns membership degrees to each category, indicating the degree
of similarity between the input data and the categories.
Working of Pattern Recognition using a Fuzzy Controller:
1. Data Preprocessing: The process begins with the collection and preprocessing of input data.
Preprocessing may involve data cleaning, normalization, and feature extraction to prepare the data
for pattern recognition.
2. Fuzzy Rule Design: Experts or domain specialists define fuzzy rules that capture the relationships
between input features and output categories. These rules are usually based on domain knowledge
and experience.

1
3. Fuzzification: The input data is transformed into fuzzy sets using membership functions. Each
feature's value is assigned a degree of membership to various fuzzy sets based on its similarity to
those sets.
4. Fuzzy Inference: The fuzzy inference system evaluates the fuzzy rules and calculates the degree to
which the input data satisfy each rule. This results in a set of membership values for each output
category.
5. Aggregation: If multiple rules apply to the same output category, their membership values are
aggregated to determine the overall membership of the input data to that category.
6. Defuzzification: The final step involves converting the fuzzy output into a crisp decision. This is
done by considering the aggregated membership values and selecting the category with the highest
membership as the recognized pattern or class.
7. Decision and Output: The pattern recognition system outputs the recognized category, which
corresponds to the input data's classification based on the fuzzy controller's inference.
Pattern recognition using a fuzzy controller is beneficial in scenarios where the input data is
imprecise, uncertain, or involves linguistic variables. It can be applied in various fields, including
image recognition, speech recognition, medical diagnosis, and industrial automation, to make
intelligent decisions based on complex and uncertain data.

2
 FUZZY CONTROL SYSTEM
A fuzzy control system, also known as a fuzzy logic controller (FLC), is a type of control system that
uses fuzzy logic to model and control complex, nonlinear systems. Fuzzy control systems are
particularly useful when dealing with systems that are difficult to describe using conventional
mathematical models or when there is uncertainty in the system's behavior. Here's an overview of
the components and working of a fuzzy control system:
Components of a Fuzzy Control System:
1. Fuzzification Interface: This component is responsible for converting crisp (precise) input data or
variables into fuzzy sets. Fuzzy sets represent linguistic terms like "low," "medium," and "high" to
describe the degree to which an input belongs to a specific category. Fuzzification involves defining
membership functions that assign degrees of membership to these linguistic terms.
2. Fuzzy Rule Base: The rule base contains a set of if-then rules that describe the relationship between
the fuzzy input variables and the fuzzy output variables. Each rule combines one or more fuzzy input
variables using logical operators like "and" or "or" to determine the fuzzy output variable's value.
These rules are typically based on expert knowledge or domain-specific information.
3. Inference Engine: The inference engine is responsible for executing the fuzzy rules and determining
the degree to which each rule is satisfied by the current input values. It combines these rule
activations to compute the fuzzy output values for each rule using operators like "min" (for "and"
operations) and "max" (for "or" operations).
4. Fuzzy Output Interface: This component converts the fuzzy output values generated by the
inference engine into crisp output values. This process is called defuzzification. There are various
defuzzification methods, such as centroid, weighted average, and maximum membership, to obtain a
single, precise output value.
5. Control Actions: The crisp output values obtained through defuzzification are used to control the
system. These values represent control signals or adjustments that need to be applied to the system's
actuators to achieve the desired system behavior.

3
Working of a Fuzzy Control System:
1. Fuzzification: The process begins by measuring or obtaining input data from sensors or other
sources. These input data are then fuzzified using membership functions, which assign degrees of
membership to linguistic terms describing the input variables. For example, if the input is
temperature, fuzzy sets like "cold," "warm," and "hot" might be defined.
2. Fuzzy Rule Evaluation: The fuzzified input values are used to evaluate the fuzzy rules in the rule
base. Each rule is assessed to determine how much it applies to the current input conditions. This is
done by computing the degree of membership of each input variable in the rule's antecedent (if-part)
using the membership functions.
3. Inference Engine: The inference engine combines the results of the rule evaluations to determine
the fuzzy output values for each rule. This often involves using logical operators (e.g., "and" and "or")
and aggregating the rule activations to create fuzzy output sets for each rule.
4. Defuzzification: The fuzzy output sets from each rule are defuzzified to obtain crisp output values.
Various methods can be used for defuzzification, such as finding the centroid (center of mass) of the
fuzzy output sets.
5. Control Actions: The crisp output values obtained from defuzzification are used as control signals
to adjust the system's actuators. These control actions are applied to the system to achieve the desired
system behavior.
6. Feedback Loop: In many applications, the control system operates in a closed-loop fashion,
continuously measuring the system's response to the control actions and adjusting the inputs
accordingly. This feedback loop ensures that the system maintains the desired state or setpoint.

Fuzzy control systems are widely used in various applications, including industrial control,
automotive systems, home appliances, and robotics, where they can effectively handle complex and
uncertain control tasks by modeling human-like reasoning and decision-making processes.

4
 EXPLAIN ABOUT COMPONENTS AND WORKING OF AIRCRAFT LANDING SYSTEM USING
NEURO FUZZY CONTROLLER
The aircraft landing system is a critical component of an aircraft's avionics, which is responsible for
guiding the aircraft safely during the landing phase. One approach to designing the landing system is
by using a neuro fuzzy controller. An aircraft landing system, often referred to as an Instrument
Landing System (ILS), is a crucial component of aviation that helps pilots safely land aircraft in
various weather conditions, including low visibility. While traditional ILS systems use basic control
algorithms and feedback loops, incorporating a neuro-fuzzy controller can enhance the system's
performance and adaptability. Let's break down the components and workings of an aircraft landing
system using a neuro-fuzzy controller:
Components of an Aircraft Landing System:
1. Localizer (LLZ): The localizer is responsible for guiding the aircraft horizontally along the
centerline of the runway. It emits a radio signal that the aircraft's receiver uses to maintain the correct
lateral alignment during the approach.
2. Glide Slope (GS): The glide slope provides vertical guidance to the aircraft, helping it maintain the
correct descent path to the runway. It emits a radio signal that helps the aircraft stay on the proper
glide path.
3. Marker Beacons: These are ground-based transmitters that provide distance and altitude
information to the aircraft during its approach. There are typically three marker beacons: Outer
Marker (OM), Middle Marker (MM), and Inner Marker (IM).
4. Neuro-Fuzzy Controller: The neuro-fuzzy controller is the advanced component that manages the
aircraft's guidance and control during the landing approach. It combines neural network elements
and fuzzy logic to adaptively control the aircraft's movements based on various input parameters and
environmental conditions.

5
Working of an Aircraft Landing System with a Neuro-Fuzzy Controller:
1. Data Collection: The landing system continuously collects data from various sources, including the
localizer, glide slope, marker beacons, altimeters, airspeed sensors, and other aircraft instruments.
These data sources provide information about the aircraft's position, altitude, speed, and alignment.
2. Data Processing: The neuro-fuzzy controller processes the incoming data to make real-time
decisions about the aircraft's course corrections. It considers factors such as wind speed and
direction, visibility, aircraft weight, and runway conditions to optimize the landing approach.
3. Fuzzy Logic: Fuzzy logic allows the controller to handle imprecise and uncertain data effectively. It
uses linguistic variables and membership functions to interpret input data and generate appropriate
control commands. For example, it can assess the "degree of alignment" or "degree of descent" and
make decisions accordingly.
4. Neural Networks: Neural networks within the controller can learn from previous landing
experiences and adapt to changing conditions. They help the system recognize patterns and make
intelligent decisions to improve landing accuracy.
5. Control Commands: Based on the processed data, fuzzy logic rules, and neural network predictions,
the neuro-fuzzy controller generates control commands for the aircraft's control surfaces, such as
ailerons, elevators, and rudders. These commands ensure that the aircraft maintains the correct
approach path and alignment.
6. Feedback Loop: The system continuously monitors the aircraft's response to its control commands
and adjusts them as needed to maintain the desired approach parameters. It ensures that the aircraft
remains on the glide slope and localizer centerline.

7. Decision Altitude: As the aircraft approaches the decision altitude, the neuro-fuzzy controller
assesses whether the required visual cues for a safe landing are visible. If they are, the pilot can
proceed with the landing manually. If not, the controller can initiate a missed approach or go-around
procedure.
In the context of an aircraft landing system, a neuro fuzzy controller can be used to analyze various
sensor inputs, such as altitude, airspeed, runway conditions, wind speed, and others, to determine
the appropriate control actions for a safe landing. The controller learns from historical data to
establish relationships between the sensor inputs and suitable control outputs.
The fuzzy logic component enables the controller to handle uncertainties and imprecise inputs. Fuzzy
logic allows for flexible decision-making based on linguistic rules, which are typically expressed as
"if-then" statements. These rules define the relationships between the input variables and the desired
control actions.
The neuro fuzzy controller uses the input sensor readings to calculate a degree of membership for
each linguistic term in the fuzzy rule base. This degree of membership represents the degree of
influence of each linguistic term on the control action. The actual control action is then determined
by aggregating and defuzzifying the outputs of the fuzzy rules.

By using a neuro fuzzy controller, the aircraft landing system can adapt to changing conditions and
make intelligent decisions based on the inputs received from various sensors. This helps to enhance
the accuracy and safety of the landing process.

6
 FUZZY-BASED STATISTICAL PROCESS CONTROL (SPC)
Fuzzy-Based Statistical Process Control (SPC) is an approach that combines statistical techniques
with fuzzy logic to monitor and control industrial processes. It helps in maintaining consistent
product quality by detecting and addressing variations and abnormalities in manufacturing
processes. Here's an explanation of the components and working of a fuzzy-based SPC application:

Components of a Fuzzy-Based SPC Application:

1. Data Collection: This is the initial step where data related to the manufacturing process is collected.
Data can include measurements of various process parameters, such as temperature, pressure,
dimensions, or chemical compositions. It may also include data related to product quality, such as
defect counts or variations in product characteristics.
2. Fuzzification: In this stage, the collected data is converted into fuzzy sets. Fuzzy sets represent
linguistic terms or categories to describe the degree to which data points belong to a specific category.
For example, "low," "medium," and "high" are linguistic terms often used in fuzzification.
3. Fuzzy Logic Rules: Fuzzy logic rules are created based on domain knowledge and experience to
describe the relationships between the fuzzy input variables (e.g., process parameters) and the fuzzy
output variables (e.g., quality indicators or control actions). These rules are typically expressed in the
form of "if-then" statements.
4. Fuzzy Inference Engine: The fuzzy inference engine evaluates the fuzzy rules using the fuzzified
input data. It calculates the degree to which each rule is satisfied for the current set of input values.
This is done using operators such as "and," "or," and "not" to combine the fuzzy membership values
of the input variables.
5. Defuzzification: After executing the fuzzy rules, the defuzzification process converts the fuzzy
output into a crisp, actionable decision or control signal. Various defuzzification methods can be used,
such as centroid defuzzification or weighted average defuzzification.
6. Control Actions: The crisp output obtained from defuzzification represents control actions that
need to be taken to maintain or improve the manufacturing process. These control actions can include
adjustments to process parameters, alarms, or notifications to operators.
Working of a Fuzzy-Based SPC Application:
1. Data Collection: The process begins by collecting data from sensors, instruments, or quality control
inspections within the manufacturing process. These data points are recorded and monitored over
time.
2. Fuzzification: The collected data is fuzzified by converting numerical values into linguistic terms
using membership functions. For example, temperature measurements might be fuzzified into terms
like "low," "normal," or "high."
3. Fuzzy Logic Rules: Expert knowledge or historical data is used to define fuzzy rules that relate the
fuzzified input variables to quality indicators or control actions. For instance, if the temperature is
"high" and the pressure is "low," it might trigger a rule to increase cooling.
4. Fuzzy Inference Engine: The fuzzy inference engine evaluates the fuzzy rules for the current set of
fuzzified input data. It calculates the degree of membership for each rule and combines them to
determine the overall fuzzy output.

7
5. Defuzzification: The fuzzy output, which represents the recommended control action, is defuzzified
to obtain a crisp decision. This can be a specific numerical value or a qualitative instruction,
depending on the application.
6. Control Actions: The final output is used to adjust process parameters, initiate alarms, or take other
actions to maintain or improve the quality of the manufacturing process. These actions are aimed at
minimizing variations and deviations from the desired process state.
7. Continuous Monitoring: The SPC application continuously collects new data, re-evaluates the fuzzy
rules, and adapts control actions as needed to ensure the process remains within acceptable quality
limits.
Fuzzy-based SPC applications are particularly valuable in situations where traditional statistical
methods may not capture the complexity or uncertainty of a manufacturing process adequately. They
provide a flexible and adaptive approach to process control, allowing for real-time adjustments to
maintain product quality.

8
 FUZZY COGNITIVE MAPPING (FCM)
Fuzzy Cognitive Mapping (FCM) is a knowledge representation and decision-making tool that
combines concepts from fuzzy logic and cognitive science to model complex systems and analyze
their behavior. It is commonly used in fields like systems analysis, policy modeling, and decision
support systems. Here's an explanation of the components and working of Fuzzy Cognitive Mapping:

Components of Fuzzy Cognitive Mapping:

1. Concepts (Nodes): FCMs consist of a set of interconnected concepts or nodes. Each node represents
a relevant variable, factor, or concept within the system being modeled. These nodes can be related
to physical entities, abstract concepts, or decision variables.
2. Causal Links (Edges): Causal links are the connections or edges between nodes in the FCM. They
represent the cause-and-effect relationships or influences between concepts. These links can be
positive (indicating a direct positive influence) or negative (indicating a direct negative influence).
The strength of these links is often represented using fuzzy values, allowing for gradual and imprecise
relationships.
3. Fuzzy Values: Fuzzy logic is used to represent the degree of membership or strength of causal links
between nodes. Instead of binary (0 or 1) connections, fuzzy values allow for a range of values
between 0 and 1, expressing the degree of influence or certainty in a more flexible manner.
4. Activation Levels: Each node in the FCM has an associated activation level, which represents the
current state or value of that concept within the system. Activation levels can also be represented
using fuzzy sets, allowing for imprecise and gradual values.

9
Working of Fuzzy Cognitive Mapping:
1. Model Initialization: The first step in creating an FCM is to define the set of concepts (nodes) that
are relevant to the system under consideration. These concepts can be identified through domain
expertise or data analysis.
2. Establishing Causal Links: For each pair of concepts in the FCM, you establish causal links that
represent the influence of one concept on another. These links are often determined through expert
knowledge, literature review, or empirical data. The strength and direction (positive or negative) of
these links are represented using fuzzy values.
3. Initializing Activation Levels: Initially, each concept's activation level is set to a specific value or a
fuzzy set to represent the initial state of the system. This represents the starting conditions for the
system.
4. Activation Propagation: The FCM is then used to simulate the behavior of the system over time.
Activation levels propagate through the network based on the causal links and fuzzy values. This
propagation process involves calculating the effect of each concept on its connected concepts
according to the strength of the causal links.
5. Aggregation of Influence: As activation levels propagate through the network, they aggregate at
each concept, considering both positive and negative influences from other concepts. This
aggregation often involves using fuzzy logic operators like minimum, maximum, or weighted
aggregation.
6. Iterative Simulation: FCMs are typically simulated iteratively, with activation levels updating in
each iteration based on the current state of the network. The simulation continues until a stopping
condition is met, such as reaching a predefined number of iterations or achieving a stable state.
7. Analysis and Decision-Making: Once the FCM has been simulated, the results are analyzed to draw
insights about the system's behavior. FCMs can help identify critical concepts, feedback loops, and the
impact of various interventions or policy changes on the system.
Fuzzy Cognitive Mapping is a powerful tool for modeling and analyzing complex systems that involve
uncertainty, imprecision, and nonlinear relationships between variables. It can support decision-
making by providing insights into the dynamic behavior of a system and the potential consequences
of different actions or policies.

10
 FUZZY PROBABILITY MEASURES
Fuzzy probability measures are a mathematical concept that extends classical probability theory to
handle uncertainty and imprecision using fuzzy sets. These measures are used in various applications
where traditional probability theory may not adequately capture the uncertainty inherent in real-
world problems. Here's an explanation of the components and working of fuzzy probability
measures:
Components of Fuzzy Probability Measures:
1. Fuzzy Events: In classical probability theory, events are defined as subsets of a sample space. In
fuzzy probability theory, events are represented as fuzzy sets. These fuzzy events express the degree
of membership of elements in the sample space to the event. Fuzzy events allow for the modeling of
imprecise or uncertain events.
2. Fuzzy Probability Distribution: Instead of a traditional probability distribution that assigns precise
probabilities to events, a fuzzy probability distribution assigns degrees of membership to events. This
distribution represents the uncertainty associated with each event. It can be represented using fuzzy
numbers, membership functions, or other fuzzy set representations.
3. Fuzzy Random Variables: Fuzzy random variables are variables whose values are fuzzy numbers
or fuzzy sets. These variables capture the uncertainty or imprecision associated with real-world data.
Fuzzy random variables are used to model data that cannot be precisely measured or classified.

11
Working of Fuzzy Probability Measures:
1. Definition of Fuzzy Events: The first step is to define fuzzy events, which represent uncertain or
imprecise outcomes. Fuzzy events are often expressed using membership functions, which describe
the degree of membership of each element in the sample space to the event.
2. Fuzzy Probability Distribution: A fuzzy probability distribution is defined to assign degrees of
membership to each fuzzy event. This distribution captures the uncertainty associated with the
likelihood of each event occurring. It can be based on expert opinions, data, or a combination of both.
3. Operations on Fuzzy Events: Like classical probability theory, fuzzy probability theory defines
operations on fuzzy events, such as union, intersection, and complement. These operations allow you
to combine and manipulate fuzzy events to calculate the likelihood of complex events.

4. Fuzzy Random Variables: Fuzzy random variables are introduced to model data with uncertainty.
These variables can take on fuzzy values, representing the imprecision in measurements or
observations. Fuzzy random variables can be used to describe real-world phenomena where data
may vary within a range of values.
5. Fuzzy Probability Distributions for Fuzzy Random Variables: When working with fuzzy random
variables, you can define fuzzy probability distributions to represent the likelihood of different fuzzy
values for these variables. These distributions capture the uncertainty in the variable's values.
6. Probabilistic Inference: Just as in classical probability theory, fuzzy probability theory allows for
probabilistic inference. You can calculate conditional probabilities, expectations, and other statistical
measures based on fuzzy probability distributions and fuzzy events.

7. Applications: Fuzzy probability measures find applications in decision making, risk analysis, fuzzy
modeling, and uncertainty quantification in various fields, including engineering, finance, and
artificial intelligence.
Fuzzy probability measures provide a framework to deal with uncertainty and imprecision in a more
flexible and intuitive manner than classical probability theory. They are particularly useful when
dealing with real-world problems where precise numerical probabilities may be challenging to
determine or when dealing with data that inherently contains uncertainty.

12
 FUZZY PROBABILITY & NECESSITY MEASURES
Fuzzy probability and necessity measures are concepts that extend classical probability theory to
deal with uncertainty and imprecision using fuzzy sets. They are often used in decision-making, risk
analysis, and modeling scenarios where traditional probability theory may not adequately capture
the vagueness and ambiguity inherent in real-world problems. Here's an explanation of their
components and how they work:
Components of Fuzzy Probability and Necessity Measures:
1. Fuzzy Events: Similar to classical probability theory, fuzzy probability and necessity measures
work with events. In this context, events are represented as fuzzy sets. These fuzzy events express
the degree to which an outcome belongs to an event, allowing for the modeling of uncertain or
imprecise events.
2. Fuzzy Probability Measures: Fuzzy probability measures quantify the degree to which a fuzzy event
is likely to occur. Instead of assigning precise probabilities, they assign degrees of membership to
events, indicating the likelihood of occurrence in a more flexible manner. Fuzzy probability measures
are typically represented as membership functions or fuzzy numbers.
3. Fuzzy Necessity Measures: Fuzzy necessity measures provide the complementary view to fuzzy
probability measures. While fuzzy probability measures indicate the likelihood of an event's
occurrence, fuzzy necessity measures indicate the necessity or certainty of its occurrence. High
necessity values suggest that the event is almost certain to occur, while low necessity values indicate
a more uncertain event.

Working of Fuzzy Probability and Necessity Measures:

1. Definition of Fuzzy Events: The process begins by defining fuzzy events, which represent uncertain
or imprecise outcomes. These events are often described using linguistic terms (e.g., "likely,"
"unlikely") or membership functions that assign degrees of membership to different outcomes.
2. Fuzzy Probability Measures: Fuzzy probability measures are defined to assign degrees of
membership to each fuzzy event, indicating the likelihood of the event's occurrence. These measures
capture the uncertainty associated with the event's likelihood. They can be based on expert opinions,
data, or a combination of both.

13
3. Fuzzy Necessity Measures: In parallel with fuzzy probability measures, fuzzy necessity measures
are defined to assess the necessity or certainty of each fuzzy event's occurrence. These measures
provide complementary information to fuzzy probability measures and help in understanding the
certainty associated with each event.
4. Operations on Fuzzy Events: Fuzzy probability and necessity measures support operations on fuzzy
events, such as union, intersection, and complement. These operations allow for the combination and
manipulation of fuzzy events to calculate the likelihood and necessity of complex events.
5. Probabilistic Inference: Just as in classical probability theory, fuzzy probability and necessity
measures allow for probabilistic inference. You can calculate conditional probabilities, expectations,
and other statistical measures based on these measures and fuzzy events.
6. Applications: Fuzzy probability and necessity measures are applied in decision-making, risk
assessment, and modeling scenarios where uncertainty and imprecision need to be taken into
account. They are particularly valuable when dealing with subjective judgments, vague data, or
situations where traditional probability theory may not be appropriate.

The combination of fuzzy probability and necessity measures provides a comprehensive framework
for handling uncertainty in a more flexible and intuitive manner. These measures are especially useful
when addressing real-world problems that involve imprecise information and where it is essential to
consider both the likelihood and necessity of events.

14