Scribd
Upload
24 views15 pages

Updated Machine Intelligence Neuro-Fuzzy Systems

The document discusses fuzzy logic control and neuro-fuzzy systems, highlighting the advantages of PID-type fuzzy logic controllers for achieving fast response and zero steady-state error. It explains the integration of fuzzy logic and neural networks in neuro-fuzzy systems, which enhance learning capabilities and rule creation. Additionally, it introduces the Adaptive Network-based Fuzzy Inference System (ANFIS) as a method for building intelligent controllers that combine fuzzy reasoning with learning from experience.

Uploaded by

lubnamoin65
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
24 views15 pages

Updated Machine Intelligence Neuro-Fuzzy Systems

The document discusses fuzzy logic control and neuro-fuzzy systems, highlighting the advantages of PID-type fuzzy logic controllers for achieving fast response and zero steady-state error. It explains the integration of fuzzy logic and neural networks in neuro-fuzzy systems, which enhance learning capabilities and rule creation. Additionally, it introduces the Adaptive Network-based Fuzzy Inference System (ANFIS) as a method for building intelligent controllers that combine fuzzy reasoning with learning from experience.

Uploaded by

lubnamoin65
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Machine intelligence

Fuzzy logic control and Neuro-fuzzy


systems

`ˆÌi`ÊÜˆÌ ÊÌ iÊ`i“œÊÛiÀȜ˜ÊœvÊ
˜vˆÝÊ*ÀœÊ* Ê `ˆÌœÀÊ

/œÊÀi“œÛiÊÌ ˆÃʘœÌˆVi]ÊۈÈÌ\Ê
ÜÜÜ°ˆVi˜ˆ°Vœ“É՘œVŽ° ̓
PID-like FLC
PID-like FLC
• Generally, a PD-type two-term fuzzy controller cannot
eliminate steady-state error whereas a PI-type two-term
fuzzy controller can eliminate steady-state error but it has
a slower response due to the integral term in the control
variable.
• In order to meet the design criteria of fast rise time, minimum
overshoot, shorter settling time and zero steady-state error, a
further option is to develop a PID-type FLC which enables
fast rise time, smaller overshoot and settling time from the PD
part and minimum steady-state error from the PI part of the
PID controller.
• The generic fuzzy PID controller is a four-dimensional (three-
input single-output) fuzzy system. The basic idea of a PID
controller is to choose the control law by considering the error
e, change of error delta e and integral of error or sum of error.
Modular Fuzzy Controller

A modular structure of FLCs with


minimum number of input/output
variables can reduce the number of
rules.
Fuzzy logic MATLAB Toolbox

https://www.mathworks.com/products/fuzzy-logic.html
Fuzzy logic MATLAB Toolbox
Introduction
• Neuro-fuzzy (NF) systems include fuzzy logic
and neural networks.
• Fuzzy logic systems try to emulate human-like
reasoning using linguistic expression
• Neural networks try to emulate the human
brain-like learning and storing information on a
purely experiential basis.
• A neuro-fuzzy system finds the parameters of a
fuzzy system by means of learning methods
obtained from neural networks. Fuzzy NF Neural
• The most important reason for combining
neural networks with fuzzy systems is their
learning capability. logic networks
• Such a combination should be able to learn
linguistic rules and/or membership functions or
tune existing ones.
Introduction

Learning in NF systems means:


• Creating a rule base,
• Adjusting MFs from scratch and
• Determining other system parameters.
Neuro-fuzzy systems types
• Cooperative neuro-fuzzy systems,
Neuro-fuzzy systems types
• Concurrent neuro-fuzzy systems and

Fuzzy systems and neural networks can work in parallel for a


plant without mutual cooperation among themselves.

• For example, some Japanese air conditioners use an FS to


prevent a compressor from freezing in the winter and use an
NN to estimate index parameters of comfort, known as
predictive mean vote (PMV).
• PMV can be defined as a function of room temperature,
mean radiant temperature, relative air velocity, humidity,
thermal resistance of users’ clothing, metabolic rate.
• Some of the PMV parameters cannot be measured using
sensors, e.g., thermal resistance of clothing and metabolic
rate.
Neuro-fuzzy systems types
• Hybrid neuro-fuzzy systems.

• In any fuzzy system, inferencing using the rule base and


defuzzification using different methods such as centre of
gravity are the most time-consuming part.
• The idea of a hybrid approach is to interpret a fuzzy system
in terms of a neural network.
• The strategy adopted here with a neuro-fuzzy system is,
firstly, to replace the rule base with a neural network so that
the inference processing is simplified and secondly, to find
the parameters of a fuzzy system by means of learning
methods obtained from neural networks.
• A common way to apply a learning algorithm to a fuzzy
system is to represent it in a special neural-network-like
architecture so that a learning algorithm, such as
backpropagation, can be used to train the system.
The ANFIS System
• Adaptive Network-based Fuzzy Inference System
• Neuro-fuzzy system that can identify parameters by using supervised learning
methods
• Sugeno-type fuzzy system with learning capabilities
• First order model

• Nodes have the same function for a given layer but are different from one layer to the
next
ANFIS System
ANFIS System

• Learning algorithm is a hybrid supervised method based on gradient


descent and Least-squares.
• Forward phase: signals travel up to layer 4 and the relevant
parameters are fitted by least squares
• Backward phase: the error signals travel backward and the premise
parameters are updated as in backpropagation
• Neuro-Fuzzy toolbox Matlab
ANFIS System
• Since a wide class of fuzzy controllers can be transformed into
equivalent adaptive networks, ANFIS can be used for building
intelligent controllers that is:
• Controllers that can reason with simple fuzzy inference and that
are able to learn from experience in the ANN style

You might also like