Anda di halaman 1dari 19

Neural Based Fault Tolerance System Using FPAA

Basic Concepts of FTS


Process resilience Reliable client-server communication Reliable group communication Distributed commit Recovery

Introduction
What is fault tolerance? What are the ways in which systems can fail? What is the most basic way of achieving fault tolerance?

Dependability
Basics: A component provides services to clients. To provide services, the component may require the services from other components a component may depend on some other component. Specifically: A component C depends on C* if the correctness of Cs behavior depends on the correctness of C*s behavior. Some properties of dependability: Availability: Readiness for usage

Reliability: Continuity of service delivery Safety: Very low probability of catastrophes Maintainability: How easy can a failed system be repaired Note: For distributed systems, components can be either processes or channels

Neural Networks
Biological approach to AI. Developed in 1943. Comprised of one or more layers of neurons Several types, well focus on feed-forward networks.

http://research.yale.edu/ysm/images/78.2/articles-neural-neuron.jpg

Back Propagation Algorithm


function BACK-PROP-LEARNING(examples, network) returns a neural network inputs: examples, a set of examples, each with input vector x and output vector y network, a multilayer network with L layers, weights Wj,i , activation function g repeat for each e in examples do for each node j in the input layer do aj xj[e] for l = 2 to M do ini j Wj,i aj ai g(ini) [Russell,] for each node i in the output layer do Dj g(inj) i Wji Di for l = M 1 to 1 do for each node j in layer l do Dj g(inj) i Wj,i Di for each node i in layer l + 1 do Wj,i Wj,i + a x aj x Di until some stopping criterion is satisfied return NEURAL-NET-HYPOTHESIS(network)

Back-Propagation Algorithm:

Fig. 20.25 Pg. 746

Back-Propagation Illustration

ARTIFICIAL NEURAL NETWORKS Colin Fahey's Guide (Book CD)

Training
Datasets from gaming sessions of human vs. human are best Must decide whether training will occur in-game, during development, or both Learning during play provides for adaptations against individual players

Training Patterns of Neural Network :

Fuzzy Logic Controller


FUZZY SETS FOR LOAD AND SPEED To choose membership functions, first of all one needs to consider the universe of distance for all the linguistic variables, applied to the rules formulation. To specify the universe of discourse, one must firstly determine the applicable range for a characteristic variable in the context of the system designed. The range you select should be carefully considered.

Example : For example, if you specify a range, which is too small, regularly occurring data will be off the scale, that may impact on an overall system performance. Conversely, if the universe for the input is too large, a temptation will often be to have wide membership functions on the right or left to capture the extreme input values. Because of this situation it is usually desirable and often necessary to scale or normalize the universe of discourse of an input/output variable. Normalization means applying the standard range of [1, +1] for the universe of discourse both for inputs and outputs. The universe of discourse for percentage load, percentage speed and percentage control voltage is chosen as {0% to 100%). Error, change in error and control voltage are taken as fuzzy variables and are assigned the membership functions as shown in the following figure.

Fuzzy calculation:
NL NM NS AZ

Available in Knowledge base.


PS PM PL

-30

-20

-10

10

20

30 Error

NL

NM

NS

AZ

PS

PM

PL

-3

-2

-1

3 Change in Error

In the membership function plot NL Negative Large NM Negative Medium NS Negative Small AZ Absolutely Zero PS Positive small PM Positive Medium PL Positive Large

NL

NM

NS

AZ

PS

PM

PL

-80

-55

-30

30

55

80 Control voltage in %

Rule-Base for the proposed FLC The Fuzzy rule base relating the variables error, change in error and control voltage is shown in Table below.

ce e

NL

NM

NS

AZ

PS

PM

PL

NL NM NS AZ PS PM PL

NL NL NL NL NM NS AZ

NM NL NL NL NS AZ PS

NS NL NM NS AZ PS PM

AZ NM NS AZ PS PM PL

PS NS PS PS PM PL PL

PM AZ PS PM PL PL PL

PL PS PM PL PL PL PL

FPAA and its Features


FPAA is Field Programmable Analog Array. Its architecture is shown in the next slide. It is a dynamically reconfigurable device. It is quickly accessible.

Special features of FPAA

In our project
Neural network is used as an Estimator. FPAA is used as a reconfigurable device. FLC is used as the controller.

Block diagram of our project

Anda mungkin juga menyukai