Alloy Property Prediction - an introduction
- Jothikrishna, Roakesh
Embedded Files
Predicting Properties of Non- Metallic Alloys
Predicting Properties of Non- Metallic Alloys
Introduction
Introduction
Mechanical Engineering, being the father of all engineering departments has its applications in a lot of fields, still requires a rapid push in research and development of sub-fields like metallurgy, mechanics, Fluid dynamics etc. Therefore to promote Research Activities in these areas we take a step to use Machine Learning which mimics the actual action of the human brain computationally.
Thus, we can have many useful insights which assists us in exploring the field in the intersection of Mechanical Engineering.
Problem Statement
Problem Statement
For any non-metallic alloy to be formed, it is essential and mandatory to determine various properties. The conventional methods of determining this parameter is a time consuming process as this indulges hypothesizing the corresponding values applying other analytical methods.
The desired alloy prepared is also tested physically which might lead to wastage of material. Thereby, applying deep learning techniques, we predict the glass transition temperature parameter without any time complexity.
Deep Learning
Deep Learning
Artificial neural networks are composed of artificially computed neurons which are conceptually derived from biological neurons. They are used to produce desired insights and relationships between attributes of a dataset. They use high computation and mathematical functions to predict the required output and obtain further insights. Their applications include Self-driving cars, Healthcare, Agriculture, Fraudulence detection, Financial services etc.
Alloy Properties
Alloy Properties
The Glass transition temperature Tg is the temperature range where the polymer substrate changes from a rigid glassy material to a soft (not melted) material and generally represented by Tg. The most standard thermal methods for determining transition temperature are Thermomechanical Analysis (TMA), Dynamic Mechanical Analysis (DMA), and Differential Scanning Calorimetry (DSC) .
The shear modulus G describes the material's response to shear stress (like cutting it with dull scissors). These moduli are not independent, and for isotropic materials they are connected via the equations . Normally, to determine this shear modulus of an alloy, we will be needing to find the shear stress and shear strain of the same alloy and then perform some analytical functions which would be a more tedious process.
2G(1+) = E = 3K(1-2)
Pre - Processing Technique
Pre - Processing Technique
Having a dataset which had chemical composition of alloys(X) and their corresponding glass transition temperature value(y), we used the dataset to train our model after pre-processing the chemical composition of metal alloy as follows,
Consider having an alloy with chemical composition Pd2Zr48Cu34Al8Ag8 which would be the primary input of my model. We pre-processed it in such a way that we have the composition as an array which consists of the elements in the alloy individually and labeled each element with an unique numeral. Eg.:[2,4,6,3,7]
Then we consider the composition numerals of the alloy and converted them into another list. Eg.[2,48,34,8,8]. Thereby we had 2 arrays which would be combined as the input to the model to predict the desired output i.e their corresponding glass transition temperature.
@article{mldmmlab,
title={Alloy Property Prediction,},
author={Jothikrishna, Roakesh},
year={2021},
link={https://sites.google.com/smvec.ac.in/mldmmlab/blog/alloy-property-prediction}
}
If interested, join using the below link :
Page updated
Google Sites
Report abuse