Prediction of compressive strength of concrete by machine learning

Compressive strength is the resistance of a material to break under compression. The compression test is usually performed in a universal testing machine. This varies from tabletop to large machines. To avoid large machines, we have introduced machine learning and automation in construction.

Machine learning and automation are both a subset of Artificial Intelligence (AI). Automation has a very vast future in construction.

Looking to build projects on Machine Learning?:

Machine Learning Kit will be shipped to you and you can learn and build using tutorials. You can start for free today!

1. Machine Learning (Career Building Course)

2. Fraud Detection using Machine Learning

3. Machine Learning using Python

4. Movie Recommendation using ML

5. Handwritten Digits Recognition using ML

6. Machine Learning Training & Internship

7. Brain Tumor Detection using Deep Learning

Machine learning is the way of studying algorithms and statistical models on computer systems that performs a specific task without the use of explicit instructions. It uses inferences instead of relying on patterns.

Machine learning includes all the stages of construction life from starting of plan and design to construct the facility, its maintenance, and operation to its failure to its rebuilding the engineered structures. The recent advances in the branch of computer science and robotics have developed new technologies for the growth of the construction industry.

Another important part that we require for a successful machine learning model is the data-set for training. A data-set is a set of data that we have collected to perform the experiment. This data-set should be of accurate data and it is better when there is a large amount of data available to get the results accurately. Now as discussed there are two types of data which are training data and test data. For the machine to recognize patterns in the data, training data is used whereas to predict new answers test data is used.

There may be many variations between the training and the test data. To measure the accuracy RMSE (root mean square error) method is used for more accurate results. The ideal value of RMSE is 0.

Methodologies

There are three methodologies i.e. Decision tree learning, multivariate adaptive regression splines (MARS) and neural network.

1) Decision tree learning

Decision tree learning uses a tree-type structure called the decision tree, which is a predictive model to understand the observations of an item to conclude the item's target value. This method is used in the analysis of visual and detailed decision making. In data mining, a decision gives detailed data, but the outcome of this classification can be an input to decision making.

Did you know

Skyfi Labs helps students learn practical skills by building real-world projects.

You can enrol with friends and receive kits at your doorstep

You can learn from experts, build working projects, showcase skills to the world and grab the best jobs.
Get started today!

2) Multivariate adaptive regression splines (MARS)

The non-linearity and reciprocals between variables are modeled in this regression technique where parameters have no value. It is an algorithm that creates a piece-wise linear model and which provides an intuitive stepping block into non-linearity after grasping the concept of linear regression and other intrinsically linear models. In MARS a model can be built in two phases i.e the forward and backward pass. This two-stage approach is the same as that used by recursive partitioning trees.

3) Neural network

A neural network is a broad network of neuron circuits in a modern sense composed of artificially made neurons or nodes. It can be made of real biological neurons or an artificially made neural network. The biological neuron is modeled as weights. The excitatory and inhibitory connections are shown by positive and negative weights.

Programming language: Python

Software requirements:

Coding in anaconda software

Coding in Jupiter notebook

Basic knowledge in machine learning and automation.

Objectives:

1. To collect enough data required for the prediction of compressive strength.
2. To learn the machine learning models.
3. To gain the highest possible training speed for the model.
4. To achieve the greatest accuracy in predicting the Compressive strength and to come up with the best model for prediction.

Conclusion:

The NEURAL NETWORKS model performs the best among the three models. It achieves the greatest accuracy compared to other models for the performed variation. Several other tests can be performed for accurate results.

• Universal Testing Machine

• Construction Management
• Strength of Concrete

Prediction of compressive strength of concrete by machine learning