Professional Documents
Culture Documents
com
ISSN : 2248-9622, Vol. 4, Issue 12( Part 3), December 2014, pp.12-14
ABSTRACT
Ready mixed concrete (RMC) is an essential material in contemporary construction and engineering projects.
Compressive strength of concrete is a major and perhaps the most important mechanical property, which is
usually measured after a standard curing of 28 days. In this research work, 28-day compressive strength of
Ready Mix Concrete has been estimated by using feed forward back propagation neural network, Fuzzy Logic
and Adaptive Neuro Fuzzy Inference System (ANFIS) modeling. The data for the ready mixed concretes (RMC)
were collected from RMC batching plant. Various models has been has been developed for different input
scenarios. The compressive strength was modeled as a function of five variables, the effects of each parameter
on networks were studied for Artificial Neural Network (ANN), Fuzzy Logic and ANFIS models.
Keywords- neural network, Fuzzy Logic, compressive strength, ANFIS models
www.ijera.com 12 | P a g e
Mr. Vipin V. Munot Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 12( Part 3), December 2014, pp.12-14
RMC is preferred to on-site concrete mixing because methods and, hence, where they would be
of the precision of the mixture and reduced worksite particularly useful. The data for the ready mixed
confusion. It facilitates speedy construction through concretes RMC were collected from RMC batching
programmed delivery at site and mechanized plant. Of these records were used for training and for
operation with consequent economy. It also testing.
decreases labor, site supervising cost and project
time, resulting in savings. Proper control and IV. SOFT COMPUTING TECHNIQUE
economy in use of raw material results in saving of ARTIFICIAL NEURAL NETWORK
natural resources, it assures consistent quality FUZZY MODELS
through accurate computerized control of aggregates ANFIS MODEL
and water as per mix designs. It minimizes cement Work on artificial neural networks, commonly
wastage due to bulk handling and there is no dust referred to as neural networks, has been motivated
problem and therefore, pollution-free. right from its inception by the recognition that the
brain computes in an entirely different way from the
II. SOFT COMPUTING TECHNIQUE conventional digital computer The term “fuzzy set”
The definition of soft computing is not precise. first appeared in 1965 when professor Lotfi A.
Lotfi A. Zadeh, the inventor of the term soft Zadeh from the university of Berkeley, USA,
computing, describes it as follows (Zadeh, published a paper entitled “Fuzzy sets”. Since then
1994):„„Soft computing is a collection of he has achieved many major theoretical
methodologies that aim to exploit the tolerance for breakthroughs in this field and has been quickly
imprecision and uncertainty to achieve tractability, joined by numerous research workers developing
robustness, and low solution cost. Its principal theoretical works. ANFIS is the famous hybrid
constituents are fuzzy logic, neuro computing, and neuro-fuzzy network for modeling the complex
probabilistic reasoning, soft computing is likely to systems. ANFIS incorporates the human-like
play an increasingly important role in many reasoning style of fuzzy systems through the use of
application areas, including software engineering. fuzzy sets and a linguistic model consisting of a set
The role model for soft computing is the human of IF–THEN fuzzy rules. The main strength of
mind.‟‟ ANFIS models is that they are universal
Soft Computing is an emerging field that approximators with the ability to solicit interpretable
consists of complementary elements of fuzzy logic, IF– THEN rules. The Adaptive Network Based
neural computing, evolutionary computation, Fuzzy Inference System (ANFIS) which integrates
machine learning and probabilistic reasoning. Due to the best features of fuzzy systems and neural
their strong learning, cognitive ability and good networks is defined by Jang (1992).
tolerance of uncertainty and imprecision, soft
computing techniques have found wide applications. V. PROPOSED WORK
Generally speaking soft computing techniques It is observed from the above literature that, the
resemble human reasoning more closely than number of individual studies using Artificial Neural
traditional techniques which are largely based on Network (ANN), Fuzzy Logic (FL) and Adaptive
conventional logical systems, such as sentential Network based Fuzzy Inference System (ANFIS)
logic and predicate logic, or rely heavily on the have been carried out for predicting compressive
mathematical capabilities of a computer. strength of concrete. But, the number of comparative
studies of above techniques (ANN, FL and ANFIS)
III. OBJECTIVES is very less in literature. So, it is significant to
The main objective was to explore the conduct a comparative study on common platform
feasibility of ANN, Fuzzy Logic and ANFIS models using various soft computing techniques.
for predicting the 28 day compression strength of So the objective will,
Ready Mix Concrete. The objective was to compare To explore the feasibility of ANN model
and to know the best of ANN, Fuzzy Logic and for predicting the 28 day Compression
ANFIS models for predicting the 28 day Strength of Ready Mix Concrete.
compression strength of Ready Mix Concrete. To explore the feasibility of Fuzzy Logic
Concrete were considered, namely normal strength model for predicting the 28 day
ready mixed plant mixes where only retarders with Compression Strength of Ready Mix
moderate water reducing properties were used. The Concrete.
data were obtained from industry. Furthermore, To explore the feasibility of ANFIS model
types of mixes involved admixtures and/or additives for predicting the 28 day Compression
over and above the basic concrete constituents, thus Strength of Ready Mix Concrete.
testing the soft computing in applications which are
not very well covered by conventional mix design
www.ijera.com 13 | P a g e
Mr. Vipin V. Munot Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 12( Part 3), December 2014, pp.12-14
VI. APPLICATIONS USING SOFT these networks will help in solving more
COMPUTING complex problems.
• Application of soft computing to handwriting Models having all five parameters show good
recognition performance in all models than changing one
• Application of soft computing to automotive parameter, so it proves that all five parameters
systems and manufacturing are essential for estimating of compressive
• Application of soft computing to image processing strength of RMC.
and data compression
• Application of soft computing to architecture REFERENCES
• Application of soft computing to decision-support Journal Papers:
systems [1.] Paresh Chandra Deka, Somanath N Diwate,
• Application of soft computing to power systems “Modeling Compressive Strength of Ready
Mix Concrete Using Soft Computing
VII. CONCLUSIONS Techniques”, International Journal of Earth
Neural network model trained by feed forward Sciences and Engineering ISSN 0974-5904,
back-propagation algorithm have the Logsig Volume 04, No 06 SPL, October 2011, pp.
transfer function with 1 hidden layers having 3 793-796.
numbers of neurons in hidden layer could [2.] Abdulhamit Subasiwas, AhmetSerdarYilmaz,
predict the 28- day compression strength of HanifiBinici, 2009, “Prediction of early heat
ready mix concrete with satisfactory of hydration of plain and blended cements
performance. using neuro-fuzzy modelling techniques”,
The ANFIS which have the Triangular Expert Systems with Applications 36 (2009)
membership function could predict the 28-day 4940–4950.
compression strength of ready mix concrete in a Books:
best manner and the minimum required error in 1. Dias,W.P.S, & Pooliyadda S.P. (2001).
comparison with the fuzzy models. “Neural networks for predicting properties of
The ANN models, which can easily incorporate concretes with admixtures”, J. of Construction
additional model parameters, give less scattered Building Material, 2001. 15:371–9.
estimated value results than those given by the 2. Jafar Sobhani, Meysam Najimi, 2010,
other models. “Prediction of the compressive strength of no-
slump concrete: A comparative study of
From the results obtained it can be concluded
regression, neural network and ANFIS
that the ANN models are more suitable and
models”, Construction and Building Materials
feasible in modeling of complex problems and
24 (2010) 709–718
save a lot of computational effort compared to
conventional methods significantly. The use of
www.ijera.com 14 | P a g e