Structural Engineering Laboratory (CE-5255)

EXPERIMENT NO. 01

STUDY ON STRESS-STRAIN CURVE OF DIFFERENT TYPES OF

CONCRETE
Submitted by:
Yadagiri Manoj Kumar 23CEM1R30
Rittul Mamta 23CEM1R31
Vadthyavath Shiva 23CEM1R32
Bellamkonda Yashwanth 23CEM1S01
Submitted to:
Prof. T. P. Tezeswi
Division Head – Structures Division
Department of Civil Engineering

Submitted on:
09th February, 2024

STRUCTURES DIVISION
DEPARTMENT OF CIVIL ENGINEERING
NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL
Group-8
M.Tech 2nd Sem.

Abstract:

Concrete, as one of the most fundamental materials in civil engineering, exhibits complex
mechanical behavior under varying loading conditions. Understanding the stress-strain
relationship of concrete is crucial for designing safe and durable structures. Traditional
methods of predicting concrete behavior often rely on empirical equations and experimental
tests, which can be time-consuming and costly.

In recent years, artificial neural network (ANN) models have emerged as promising tools for
predicting concrete properties with improved accuracy and efficiency. By leveraging the
computational power of neural networks, these models can capture the nonlinear and intricate
nature of concrete behavior under different loading regimes.

This study aims to investigate the stress-strain curve of various types of concrete using an ANN
model developed specifically for this purpose. The ANN model has been trained on a
comprehensive dataset encompassing different concrete mixes and loading conditions,
allowing it to learn and generalize the complex relationships between input parameters and
concrete behavior.

By utilizing the ANN model, this research seeks to enhance our understanding of how different
factors such as mix proportions, aggregate properties, and curing conditions influence the
stress-strain response of concrete. Additionally, the study aims to demonstrate the potential of
ANN models as efficient tools for predicting concrete properties under increased loading,
thereby contributing to advancements in structural engineering design and construction
practices.

1.1 Introduction:

1.1.1. Background of Artificial Neural Networks (ANNs):

Artificial Neural Networks (ANNs) trace their origins back to the pioneering work of
Warren McCulloch and Walter Pitts in the 1940s. They proposed a computational model
inspired by the structure and function of the human brain, where interconnected nodes
(neurons) process information. However, it wasn't until the latter half of the 20th century,
with the advent of computational technology and advancements in machine learning
algorithms, that ANNs gained practical significance.

Structures Division
NIT Warangal 1 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

The field of ANN experienced periods of both enthusiasm and skepticism over the years.
In the 1960s and 1970s, research in neural networks flourished, but progress was limited
by computational constraints and theoretical challenges. During the 1980s and 1990s,
interest waned due to the dominance of other machine learning techniques and the
perception of ANNs as black-box models lacking interpretability.

However, the resurgence of interest in deep learning in the early 21st century, fueled by the
availability of large datasets, powerful computing hardware, and novel training algorithms,
revitalized the field of ANNs. Deep learning, a subfield of machine learning based on
ANNs with multiple layers, has achieved remarkable success in various applications,
including image recognition, natural language processing, and robotics.

Today, ANNs have become ubiquitous in modern technology, powering a wide range of
intelligent systems and applications. From virtual assistants to autonomous vehicles, ANNs
continue to push the boundaries of what is possible in artificial intelligence and machine
learning. In the context of engineering and science, ANNs offer powerful tools for
modeling complex systems, predicting behavior, and optimizing processes, as evidenced
by their application in fields such as structural engineering, materials science, and
environmental monitoring.

1.1.2. Introduction to Artificial Neural Networks (ANNs):

Artificial Neural Networks (ANNs) are computational models inspired by the structure and
function of the human brain. They consist of interconnected nodes, known as neurons,
organized into layers. Each neuron receives input signals, processes them, and generates an
output signal that is transmitted to neurons in the next layer.

The basic 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 connecting them in a hierarchical fashion, ANNs can learn
complex patterns and relationships from data through a process known as training.

Structures Division
NIT Warangal 2 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

Figure 1 Adopted Stress v/s Strain Curve for M65

Training an ANN involves adjusting the weights of connections between neurons based on
input-output pairs from a training dataset. This process is typically done using optimization
algorithms such as gradient descent, which minimize a predefined loss function that
quantifies the difference between predicted and actual outputs.

One of the key advantages of ANNs is their ability to learn and generalize from data,
making them suitable for a wide range of tasks, including classification, regression, and
pattern recognition. Moreover, ANNs can automatically extract relevant features from raw
input data, reducing the need for manual feature engineering.

ANNs come in various architectures, ranging from simple feedforward networks to more
complex architectures such as convolutional neural networks (CNNs) and recurrent neural
networks (RNNs). Each architecture is tailored to specific types of data and tasks, with
CNNs being well-suited for image processing tasks and RNNs being effective for sequence
modeling and time-series prediction.

In recent years, ANNs have seen widespread adoption across diverse domains, including
computer vision, natural language processing, robotics, and healthcare. Their versatility,
scalability, and ability to learn complex patterns from data have made them indispensable
tools in the era of big data and artificial intelligence. In the context of engineering and
science, ANNs offer powerful capabilities for modeling, prediction, and optimization,
enabling advancements in fields such as materials science, structural engineering, and
environmental monitoring.

Structures Division
NIT Warangal 3 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

1.1.3. Methodology Overview of ANN:

Data Collection:
• Gather a comprehensive dataset containing input-output pairs relevant to the task at
hand. In the context of studying stress-strain curves of different types of concrete, this
dataset would include variables such as mix proportions, aggregate properties, curing
conditions, and corresponding stress-strain responses.

Data Pre-processing:

• Clean and preprocess the dataset to ensure consistency and compatibility with the ANN
model. This may involve normalization of input features, handling missing values, and
splitting the dataset into training, validation, and testing sets.

Architecture Design:

• Select an appropriate architecture for the ANN model based on the nature of the task
and characteristics of the dataset. Consider factors such as the number of layers, types
of neurons, and connectivity patterns.

Figure 2 Architecture of ANN

Model Training:
• Initialize the parameters of the ANN model (e.g., weights and biases) randomly or using
pretraining techniques.
• Feed the training dataset into the model and use optimization algorithms such as
gradient descent to update the model parameters iteratively.
• Monitor the training process using validation data to prevent overfitting and ensure
generalization performance.

Structures Division
NIT Warangal 4 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

Model Evaluation:
• Assess the performance of the trained ANN model using the testing dataset. Evaluate
metrics such as accuracy, mean squared error, or coefficient of determination,
depending on the specific objective of the study.

Analysis and Interpretation:

• Analyze the learned patterns and relationships captured by the ANN model to gain
insights into the underlying behavior of concrete under different loading conditions.
• Interpret the model predictions in the context of engineering principles and domain
knowledge, identifying factors that influence the stress-strain curve of concrete.

Validation and Verification:

• Validate the ANN model against experimental data or theoretical predictions to verify
its accuracy and reliability.
• Compare the performance of the ANN model with other prediction methods or
empirical equations commonly used in concrete engineering practice.

Sensitivity Analysis:
• Conduct sensitivity analysis to understand the impact of input parameters on the
predicted stress-strain curve of concrete. Identify influential factors and their relative
contributions to the overall response.

By following these steps, the methodology enables the development and utilization of an
ANN model to capture the stress-strain behavior of different types of concrete, facilitating
prediction and analysis in structural engineering applications.

1.1.4. OriginLab

In this experiment, we leverage the capabilities of ScanIt to extract data points from stress-
strain curves obtained through experimental testing of various types of concrete. These data
serve as the foundation for developing an artificial neural network (ANN) model to predict
concrete behavior under increased loading. ANNs are computational models inspired by

Structures Division
NIT Warangal 5 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

the structure and function of the human brain, capable of learning complex patterns and
relationships from dataSoftware Used

MATLAB can play a crucial role in several aspects of the experiment:

1. Data Preprocessing:

MATLAB provides powerful tools for data preprocessing, including importing, cleaning, and
organizing experimental data collected from stress-strain tests on various concrete specimens.
This may involve removing outliers, normalizing data, and preparing it for use in training the
ANN model.

2. ANN Model Development:

MATLAB offers extensive support for developing ANN models through its Neural
Network Toolbox. Researchers can use MATLAB to design, train, and validate ANN
architectures tailored to predict concrete properties based on input parameters such as material
composition, loading conditions, and environmental factors.

2. Feature Selection and Engineering:

MATLAB's computational capabilities allow for advanced feature selection and

engineering techniques to enhance the performance of the ANN model. Researchers
can leverage MATLAB's statistical and optimization functions to identify relevant
features and optimize model parameters for improved accuracy and generalization.25

3. Model Training and Evaluation:

MATLAB facilitates the training and evaluation of ANN models using various
algorithms and techniques. Researchers can utilize MATLAB's built-in functions for
training algorithms such as backpropagation, radial basis function networks, and deep
learning architectures. Furthermore, MATLAB provides tools for cross-validation,
hyperparameter tuning, and performance evaluation to ensure robust model
performance.

4. Visualization and Interpretation:

MATLAB offers rich visualization capabilities for analyzing and interpreting the

Structures Division
NIT Warangal 6 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

results of the ANN model. Researchers can generate plots, graphs, and interactive
visualizations to explore the relationship between input parameters and predicted
concrete properties. This enables a deeper understanding of the model's behavior and
insights into concrete behavior under different loading conditions.

5. Integration with Other Tools:

MATLAB can seamlessly integrate with other software tools and datasets, allowing
researchers to incorporate additional data sources or analysis techniques into their
experiment. This interoperability enhances the versatility and extensibility of the
research workflow, enabling comprehensive analysis and interpretation of experimental
results.

Overall, MATLAB serves as a valuable computational platform for conducting

experiments involving ANN modeling of concrete behavior, offering a wide range of tools
and capabilities for data analysis, model development, and visualization. Its flexibility and
versatility make it well-suited for addressing the complex challenges inherent in studying
and predicting the mechanical properties of concrete under various loading conditions.

1.1.5. Different types of Models used

There are many constitutive models that have been proposed to capture the stress-strain
behaviour of concrete under uniaxial compression. We considered, the most widely used
models that are applicable to different ranges of concrete grades. 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.

a. Hognestad Model:

In 1951, Hognestad introduced one of the first constitutive models for concrete under
uniaxial compression. In His model, the concrete stress-strain curve has been
represented by two branches. The first being an ascending branch that follows a second-
degree parabola till reaching the peak strength represented by 0.85 fc'; maximum
reliable strength. The second branch is a linear descending branch. Leading to the
following equations

Structures Division
NIT Warangal 7 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

b. Kent And Park Model:

In 1971, Kent and Park (1971) introduced a modification to Hognestad model (1951),
by eliminating the strength reduction factor of 0.85, and by fixing the strain ε0
corresponding to fc ' at a value of 0.002 and defining the descending branch with respect
to a strain corresponding to 0.5 fc '.

where ε0 =0.002 the strain corresponding to the peak stress, Z is the slope of the
descending branch the strain in the descending branch corresponding to 0.5 the peak
stress.

c. Popovics Model:

In 1973, Popovics (1970) represented both ascending and descending branches using
the following equation

where n= (0.4×10-3 fc+1), ε0=2.7×10-4×(fc') 1/4 defined as strain at peak stress. In

1985, Based on Popovics model, Carriera and Chu (1985) introduced a modified model
capable of capturing the constitutive behaviour of high strength concrete up to 80 MPa.

d. Rusch Model

Structures Division
NIT Warangal 8 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

1.2 Procedure:
Assign the training input data and training output/target data in the Matlab workspace

Figure 3

Assign the testing input data in the Matlab workspace

Now assign the following commands in the command window in the Matlab workspace for
transpose of matrix.

Structures Division
NIT Warangal 9 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

MATLAB commands:

traininput=traininput’;
traintarget=traintarget’;
testinput=testinput’;

Now develop the network using the following command in the Matlab workspace.

net=feedforwardnet(10);

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);

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

Till the above process the training of data has completed.

Now simulate the test data by taking the following command.

y=net(testinput);

Structures Division
NIT Warangal 10 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

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

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

testoutput=y’;

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

Structures Division
NIT Warangal 11 Structural Engineering Laboratory
Group-8
M.Tech 2nd Sem.

1.3 Results:

1.4 Conclusions:

Reference:

[1] Hognestad, E. (1951), “Study of combined bending and axial load in reinforced concrete
members”, University of Illinois at Urbana Champaign, College of Engineering, Engineering
Experiment Station
[1] Kent, D.C. and Park, R. (1971), “Flexural members with confined concrete”, J. Struct. Div.,
97, 1969-1990
[1] Popovics, S. (1970), “A review of stress-strain relationships for concrete”, J. Proc., 67(3),
243-248.
[1] Rusch, H. (1960), “Researches toward a general flexural theory for structural concrete”, J.
Am. Concrete Inst., 57, 1-28

Structures Division
NIT Warangal 12 Structural Engineering Laboratory