Record Details

ANALYSIS OF SOFT COMPUTING BASED CONCRETE COMPRESSIVE STRENGTH MODELS

KrishiKosh

View Archive Info
 
 
Field Value
 
Title ANALYSIS OF SOFT COMPUTING BASED CONCRETE COMPRESSIVE STRENGTH MODELS
 
Creator HARRY, NARESH NISCHOL
 
Contributor Bind, Yeetendra Kumar
 
Subject null
 
Description This Ph D thesis is prepared with a notion that it will encourage the use of soft computing methods
in the field of concrete technology since these methods are being extensively used in many field
of engineering now a days viz. automatic railway signaling systems, home and kitchen appliances
and other electronic items etc. While the acceptability of these methods are widespread in the field
of computer science, information technology and electronic engineering, the civil engineering
community still hesitate to use these methods. The reason behind it is very simple and obvious
that is the existing codes do not allow to use these methods as an alternative to laboratory methods.
In addition, unavailability of algorithms for specific problems, discourages the professionals and
engineers to adapt these methods.
Present work is the demonstration of its applicability and limitation while obtaining concrete
compressive strength following two different approaches of soft computing. Six different types of
concrete mix data were developed from existing literature and laboratory experiments. However,
broadly we can classify entire data into two category that is conventional concrete mix data and
admixture mix data. Five different combinations of admixture mixed concrete data that are; Blast
Furnace Slag (BFS), Blast Furnace Slag with Super Plasticizer (SP), Fly Ash, Fly Ash with SP and
BFS plus Fly ash with SP were used to develop data matrix. A thorough data analysis was carried
out and incongruent data were removed from the developed data matrix to minimize the error in
final results.
Two different approach of soft computing methods that are Artificial Neural Network (ANN) and
Adaptive Neuro Fuzzy Inference System (ANFIS) were employed to understand the non linear
pattern between concrete mix design data and corresponding compressive strength. ANFIS is
called as hybrid system since it integration of well known Fuzzy Logic (which is a entirely different
field of mathematics in which logical reasoning is associated with fuzzy sets). Design mix
components like quantity of cement, water, coarse aggregate and fine aggregate were taken as
training variables in conventional concrete. Quantity of admixture along with above mentioned
design mix components were taken as training variables in admixture mix concrete. Curing period
was an essential component in training variables in both conventional and admixture mix concrete.
Using above combinations and associated compressive strength the ANN and ANFIS models were
prepared. These models were equipped with certain fix and varying characteristics which is
iii
discussed in detail in chapter 3. These networks were capable of modeling compressive strength
from developed non linear pattern. Finally, performance evaluation measures such as coefficient
of determination (R2) and Mean Squared Error (MSE) showed that ANN and ANFIS were able to
model compressive strength with some limitations. It was observed that ANFIS works well when
number of input variable is less. In addition, very reliable data sources is needed irrespective of
size of data matrix. However, ANN works well with great number of input variable also it can
tolerate some incongruent data since error is back propagated to the network and repeated cycles
gradually decreases the error.
In addition to soft computing methods, regression methods were also used to see its applicability
in modeling compressive strength. The same sources of data were used to develop Multiple Linear
and Non Liner Regression Models (MLR and MNR). MLR models were incapable in satisfactorily
predicting the compressive strength. Only multivariate power function were used in MNR analysis.
Multivariate power equations developed from MNR analysis could satisfactorily model
compressive strength and R2 values ranged from 0.75 to 0.85. Overall it can be said that the
performance of soft computing methods are highly dependent on reliability of data. These methods
can successfully be integrated with laboratory findings in the field of concrete technology if source
of information is trustworthy
 
Date 2018-01-30T05:26:16Z
2018-01-30T05:26:16Z
2017
 
Type Thesis
 
Identifier http://krishikosh.egranth.ac.in/handle/1/5810040025
 
Language en
 
Format application/pdf
 
Publisher DEPARTMENT OF CIVIL ENGINEERING SHEPHERD INSTITUTE OF ENGINEERING AND TECHNOLOGY SAM HIGGINBOTTOM UNIVERSITY OF AGRICULTURE, TECHNOLOGY AND SCIENCES ALLAHABAD (INDIA) July, 2017