Scribd
Upload
47 views8 pages

Exploration of Data Science Techniques To

The paper explores data science techniques to predict the fatigue strength of steel using chemical composition and thermal processing parameters. It utilizes a dataset from the National Institute for Materials Science, achieving high prediction accuracy with models like Multivariate Polynomial Regression and M5 Model Trees. The study demonstrates a significant improvement in prediction accuracy compared to previous research, with R² values exceeding 0.98.

Uploaded by

Moh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
47 views8 pages

Exploration of Data Science Techniques To

The paper explores data science techniques to predict the fatigue strength of steel using chemical composition and thermal processing parameters. It utilizes a dataset from the National Institute for Materials Science, achieving high prediction accuracy with models like Multivariate Polynomial Regression and M5 Model Trees. The study demonstrates a significant improvement in prediction accuracy compared to previous research, with R² values exceeding 0.98.

Uploaded by

Moh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Exploration of data science techniques to

predict fatigue strength of steel from


composition and processing parameters
Objective
The paper aims to use data science techniques-such as machine learning and
regression analysis-to predict the fatigue strength of various types of steel
based on: Chemical composition: Percentages of elements like carbon,
silicon, manganese, etc. Thermal processing parameters: Including processing
temperatures, treatment durations, and cooling rates.
Abstract
Problem Statement:
Fatigue strength is a critical property in the design of mechanical components. However,
measuring fatigue strength is both expensive and time-consuming.
Proposed Solution:
The paper proposes using data science methods (for example, neural networks, decision trees,
and multivariate polynomial regression) to build predictive models that forecast fatigue
strength.
Data Source:
The study uses data from the National Institute for Materials Science (NIMS) database, which
contains 437 samples with 25 features (covering both chemical composition and thermal
processing parameters).
Key Findings:
The developed models achieved very high prediction accuracy (R² > 0.98), representing a
significant improvement over previous studies that reported R² < 0.94.
Dataset
Source : National Institute for Materials Science (NIMS) public
database.
Details :
437 data instances covering carbon steels, low-alloy steels,
carburizing steels, and spring steels.
25 input features :
Chemical composition (e.g., %C, %Mn, %Cr).
Processing parameters (e.g., normalizing temperature, carburization
time, tempering conditions) Microstructural features.
• Predictive Modeling :
Evaluated 12 techniques , including:

-Linear regression, robust regression, and polynomial regression.

-Instance-based methods (k-NN, KStar).

-Decision trees (REPTree, M5 Model Trees).

-Support Vector Machines (SVM).

-Artificial Neural Networks (ANN).


Results
Best Performing Models:
Multivariate Polynomial Regression (MPR) ,Linear
regression achieved the highest accuracy (R² ≈ 0.98).
M5 Model Trees also performed strongly with an R² ≈
0.978.
Artificial Neural Networks (ANN) (R² ≈ 0.97).
Comparison with Previous Studies:
The current models show an improvement of about 66% in
prediction accuracy compared to earlier work (improving
from R² ≈ 0.94 to R² ≈ 0.98

You might also like