Thermomechanical- Modelling Types

Question 1

Name 3 types of model for thermomechanical processing

Answer 1

Data-based
Physically-based: internal state variable approach.
Spatially-based: Monte Carlo, cellular automata, phase field, vertex

Question 2

Name 5 modelling techniques for thermomechanical processing

Answer 2

Spreadsheet
Finite element method
Artificial neural networks
Fuzzy logic
Hybrid modelling

Question 3

For thermomechanical processing what do you derive maximum benefit from?

Answer 3

Control of dimensional accuracy.
Microstructure evolution for product properties

Question 4

What are the two broad drivers for modelling TMP?

Answer 4

Insight
Accuracy

Question 5

Types of modelling on an accuracy vs insight schematic

Answer 5

From top left to bottom right have data-based (black box), probabilistic (grey box), knowledge-based (grey box), knowledge-based (white box). Then up to top right which is knowledge and data-based (hybrid)

Question 6

Why is black box modelling at the top left of the accuracy vs insight schematic?

Answer 6

This is where you use lots of data and get an accurate answer but you don’t really know how that answer came about and you gain little metallurgical understanding. An example of this approach is the use of artificial neural networks

Question 7

What is used in black box modelling to deduce the product properties?

Answer 7

Chemical composition and processing parameters

Question 8

For the example of heat treatment (austenitising, quenching, tempering then cooling), what are process variations due to?

Answer 8

Chemical composition of steel.
Prior thermomechanical treatment.
Heat treatment temperatures and quenches.
Geometry of bars and batch distribution.
Measurement, control and operator errors.

Question 9

Describe the layers in artificial neural networks (ANNs)

Answer 9

Input layer
Hidden layer (may be several)
Output layer

Question 10

What are ANNs good and bad for and how to use them?

Answer 10

Good for non-linear data and learning from data.
Mainly functional and not structural so poor at explanation/human interaction.
Should be used for interpolation only and also in populated zones

Question 11

What is an example of a graph that ANNs can produce?

Answer 11

Continuous cooling transformation diagrams. Predictions based on concentration of alloying elements.