Structural Engineering Laboratory (CE-5255)

EXPERIMENT NO. 01

STUDY ON STRESS-STRAIN CURVE OF DIFFERENT TYPES OF

CONCRETE
Submitted by:
G Mohan Muralidhar Reddy 24CEM1R17
Nagireddy Srinivasu 24CEM1R18
Parasmani Sahu 24CEM1R19
Putta Prasanna 24CEM1R20
Submitted to:
Prof. T. P. Tezeswi
Structures Division
Department of Civil Engineering

Submitted on:
13th September, 2024

STRUCTURES DIVISION
DEPARTMENT OF CIVIL ENGINEERING
NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL
 Group-5
M.Tech 1st Sem.

1.1. Abstract:
Concrete is one of the most important materials in civil engineering, and the stress-strain curve is a
fundamental tool in understanding the mechanical behavior of concrete under load. The shape of the
curve and properties it describes (like modulus of elasticity, strength, and ductility) differ depending
on the type of concrete. Traditional methods of predicting concrete behavior often experimental test
which is costly and timing consuming.

ANN (Artificial neural network) models have demonstrated high accuracy and efficiency in
capturing the full stress-strain curve, even for materials with complex constitutive laws. This study
investigates the stress-strain behavior of different types of concrete using ANN. To understanding
stress-strain behavior using ANNs involves training neural networks to predict material properties,
such as stress responses to given strains, based on a set of input parameters. Various concrete types,
including normal, high strength and fiber reinforced concrete, were tested under compressive loading
to generate stress-stain curves. These experimental data were used to train the ANNs models, which
were designed to predict the stress-strain response for different concrete compositions and curing
conditions. The performance of the ANN models was validated through comparisons with
experimental results.

This approach provides a reliable and rapid method for predicting concrete performance, offering
potential for optimization in engineering applications.

1.2. Introduction:

1.2.1. Background of Artificial Neural Networks (ANNs):

Artificial Neural Networks (ANNs) have become a crucial in engineering, emerging as a key
intersection between computer science and engineering. The origins of ANNs trace back to the 1940s
when Warren McCulloch and Walter Pitts laid the foundational principles of neural computation.
The 1980s saw a resurgence of interest in ANNs due to computational advancements and the
introduction of backpropagation algorithms.

By the 1990s, ANNs were employed for predictive modelling and design optimization, while recent

Structures Division
NIT Warangal 2 Structural Engineering Laboratory
 Group-5
M.Tech 1st Sem.
years have seen their role further augmented by big data and machine learning techniques. In
conclusion, ANNs remain a crucial element in addressing modern engineering challenges.

The use of artificial neural networks (ANNs) to predict concrete strength has undergone a remarkable
evolution since the inception of concrete technology and computational intelligence. This
convergence of materials science and computational methods has led to innovative approaches for
enhancing the reliability and efficiency of concrete structures.

Neural networks have a rich history, tracing back to the mid-20th century, with significant
developments in the 1980s and 1990s leading to their prominence in various scientific fields,
including civil engineering. ANNs emerged as a promising alternative to traditional empirical
methods for predicting concrete strength, showcasing the ability to model nonlinear relationships
and provide more accurate predictions.

Multiple studies have unequivocally demonstrated the superiority of ANNs in predicting concrete
compressive strength, surpassing traditional models and achieving high correlation coefficients with
minimal prediction errors. This success has motivated further exploration into optimizing ANN
architectures, selecting relevant input variables, and integrating novel training techniques.

1.2.2. Introduction to Artificial Neural Networks (ANNs):

Artificial Neural Networks (ANNs) are powerful computational models inspired by the intricate
structure and functionality of the human brain. They are composed of interconnected nodes, or
neurons, organized into layers. Every neuron processes an input signal and then generates an output
signal transmitted to neurons in a subsequent layer.

The fundamental building block of an ANN is the perceptron, which computes a weighted sum of
its inputs and applies an activation function to produce an output. By stacking multiple perceptrons
into layers and hierarchically connecting them, ANNs can effectively learn complex patterns and
relationships from data through a process known as training.

1.2.3. Application of ANN in Civil Engineering:

Structures Division
NIT Warangal 3 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.
The ANN toolbox in MATLAB provides a wealth of tools and functions for creating, training, and
evaluating neural networks, which have become essential in various fields, including Civil
Engineering. In the realm of Civil Engineering, ANNs offer immense potential in addressing
complex problems that involve multiple variables, such as Structural Analysis, Load Forecasting,
and System Control.

By harnessing the capabilities of the ANN toolbox in MATLAB, researchers and practitioners in
civil engineering can develop sophisticated models to accurately predict, classify and optimize
various phenomena, including material behavior, structural responses to different loads, and
infrastructure system condition assessment. For example, ANNs can be trained to predict the
compressive strength of concrete based on mix design parameters or to identify cracks in bridges
using sensor data.

Additionally, the ANN toolbox in MATLAB encompasses a diverse range of algorithms, including
feedforward networks, radial basis networks, and support vector machines, offering versatile
solutions for addressing a wide array of problems in civil engineering. By combining the power of
ANNs with the flexibility of the MATLAB programming environment, civil engineers can craft
tailored solutions that enhance safety, efficiency, and sustainability in the design, construction, and
maintenance of infrastructure systems.

Moreover, recent exploration of ANNs has yielded compelling evidence of their effectiveness in
addressing real-world problems, such as predicting the settlement of shallow foundations,
estimating the loads on bridges, and optimizing structural system designs.
Furthermore, ANNs have been integrated with other data-driven techniques, such as evolutionary
algorithms and fuzzy logic, to create hybrid models capable of addressing complex optimization
problems in civil engineering. By fully exploring the potential of ANNs in civil engineering,
researchers and practitioners can pave the way for innovation and improvement in the field, leading
to the development of safer, more sustainable, and more resilient infrastructure systems.
Overall, the ANN toolbox in MATLAB serves as a robust platform for civil engineers to explore
the intricacies of their field and develop innovative solutions to address the pressing challenges of
the 21st century.

Structures Division
NIT Warangal 4 Structural Engineering Laboratory
 Group-5
M.Tech 1st Sem.

1.3. Methodology Overview of ANN:

• Data collection:

The data collected from the journals which are done by conducting experiments for our analysis
requires a set of stress and strain data for the different grades of concrete by varying the mix
proportions of the constituents of the concrete. These stress-strain values also depend on the curing
conditions degree of supervision and other site conditions these responses to the stress-strain
behavior.

• Preprocessing of data:

The raw data we collected from the different sources may not be compatible with the ANN model
we process the data in a format that is compatible with the ANN model used in MATLAB
including handling missing data and dividing the data for the training, testing, and validation.
• Architecture Design:

Architecture for an ANN is completely based on the type of task we are performing and the data
sets we have. Consider factors such as the number of layers, types of neurons, and connectivity
patterns.

• Model Training:

➢ Set the ANN model's initial parameters such as its weights and biases randomly or by
employing pretraining strategies.
➢ Training data will feed into the model and use algorithms to update the model parameters
iteratively.
➢ Check the training process with testing data to prevent overfitting and to ensure an optimum
performance of the model.

• Model Evaluation:

The model's evaluation can be done by accessing its performance by different values, which
quantify the accuracy of the model. Evaluation is done by finding the mean squared error, or
coefficient of determination, depending on the objective of the study.
Structures Division
NIT Warangal 5 Structural Engineering Laboratory
 Group-5
M.Tech 1st Sem.

• Validation of ANN:

Validation of the ANN model will done by comparing the model output and experimental values
that are done or the commonly used Engineering practices that are reliable we also can compare
empirical equations that are used in engineering practices.

• Simulation and test:

Integrate the neural networks in the models as blocks, which can facilitate integration with a larger
system, testing and deployment to many types of hardware.

WORK FLOW

This methodology helps us develop an ANN model of the stress-strain behaviour of concrete with
different grades, which will further facilitate us to know the stress-strain behaviour of any grade of
concrete.

1.4. Different types of Models used:

Four different models are most commonly used to capture concrete's stress-strain behavior under
uniaxial compression. The most widely used models that apply to wide ranges of concrete grades
are considered. The consideration of such variety is intended to produce general factors and
conclusions, which apply to a wide range of concrete grades ranging from 16 to 120 MPa.

Structures Division
NIT Warangal 6 Structural Engineering Laboratory
 Group-5
M.Tech 1st Sem.

a. Hognestad Model:

The Hognested model is used to represent the stress-strain curve of concrete under uniaxial
compression, capturing both the linear elastic behavior and the nonlinear post-peak softening. It is
particularly useful in structural engineering and materials science for predicting the performance of
concrete under various loading conditions.

a. Kent And Park Model:

The Kent and Park model, developed by Kent and Park in 1971, is an empirical model used to
describe the stress-strain behavior of concrete under compressive loading. It is particularly useful
for analyzing concrete behavior in seismic design and cyclic loading conditions.

b. Popovics Model:

The Popovic model for concrete is specifically designed to describe the stress-strain relationship of
concrete under uniaxial compression. Concrete, being a composite material with complex behavior,
requires a specialized model to accurately represent its mechanical properties, particularly its
nonlinear and brittle characteristics under compression.

c. Rusch Model:

The Rusch model is a constitutive model designed to describe the stress-strain behavior of concrete
under uniaxial compression. It is known for its ability to accurately capture the nonlinear, inelastic
response of concrete, including the post-peak softening behavior

Structures Division
NIT Warangal 7 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.

1.5. Procedure:
2. Assign the training input data and training output/target data in the Matlab workspace.

3. Assign the testing input data in the Matlab workspace:

4. Now assign the following commands in the command window in the Matlab workspace for
transpose of matrix. MATLAB commands:
traininput=traininput’;
traintarget=traintarget’;
testinput=testinput’;

Structures Division
NIT Warangal 8 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.
5. Now develop the network using the following command in the Matlab workspace.

net=feedforwardnet(10);

6. Now by taking the input and target data train the network by using the following
command in the Matlab workspace.

net=train(net,traininput,traintarget);

7. After completion of the training we can check for performance and the regression plot.

8. Till the above steps the training of the data has been completed. Now simulate the test
data by taking the following Commands.
y=net(testinput);

9. After simulating the test data we are going to get the model output data.

Structures Division
NIT Warangal 9 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.

10. Now assign the following commands in the command window for transpose of matrix
and name it as test output.

testoutput=y’;

11. Now copy the model output data and place it with the test target data and check for
correlation.

Structures Division
NIT Warangal 10 Structural Engineering Laboratory
 Group-5
M.Tech 1st Sem.

1.5. Result:

We studied the stress strain curve of the concrete of grade 43 using Artificial Neural Network (ANN)
in MATLAB and EXCEL software by using the known input and output values obtained from the
various research papers and finaly got the test output data.

The majority of the data resulted in Model-3 (i.e. Hognested Model). So, model-3 is best suited for
the M43 grade of concrete.

The Regression value obtained is 0.973420

The Correlation value obtained is 0.989099

1.6. Conclusions:
Through this experiment we learned to approach the Artificial Neural Network (ANN) for predicting
the most suitable model by using the stress-strain behavior of various types of concrete, thereby
improving the understanding of the factors that influence the concrete properties under different
loading conditions.
Using ANN because of its effective tools which are preprocessing, model development, and
evaluation, emphasizes the importance of computational tools in advancing our understanding of the
concrete mechanics under varying loads.
This study seeks to improve the efficiency and accuracy with which the concrete data can be
predicted, hence contributing to the design of structure to the advanced technology that is most
widely used in the other fields.

Structures Division
NIT Warangal 11 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.

1.7. Reference:

1. https://www.mathworks.com/discovery/neural-network.
2. Rajai Z. Al-Rousan, Mohammed A. Alhassan and Moheldeen A. Hejazi (2018), “Novel
nonlinear Stiffness Parameters And Constitutive Curves For Concrete”.
3. Longkang Xu, Yong Yang,Yang Zhang (2023), “Estimation of stress–strain constitutive model
for ultra-high performance concrete after high temperature with an deep neural network based
method
4. Alaa M. Morsy, Predicting mechanical properties of engineering cementitious composite
reinforced with PVA using artificial neural network

Structures Division
NIT Warangal 12 Structural Engineering Laboratory
Group-5
M.Tech 1st Sem.

Structures Division
NIT Warangal 13 Structural Engineering Laboratory