Neural Network Analysis

1
Neural Network Analysis
www.msm.cam.ac.uk/phase-trans
CSIRO
2
Charpy
fatigue
tensile
critical stress intensity
corrosion
3
Axioms

• All properties can be measured.
• Measurements can be used in safe design.
• Measurements can be used in control.

4
Dogma

• Given a comprehensive description of material,
  process and structure, it is not yet possible to
  predict most properties.

5
hardness
6
Variables

• C, Mn, Si, Ni, Cr, Mo, V, Co, B, N, O..
• Thermomechanical processing of steel
• Welding consumable
• Welding parameters
• Subsequent heat treatment

7
Conclusions

• There are useful ways of expressing hardness
• Limited models relating hardness to
  microstructure
• No method for predicting hardness in general

8
Solution

• non-linear functions
• large numbers of variables
• uncertainties
• exploit large knowledge base

9
Empirical Equations
?y a b (C) c (Mn) d (Ni)
....
1223
10
?y a b (C) c (Mn) ?y a b (C) c
(Mn) d(C x Mn)
11
?y a b (C) c (Mn) ?y a b (C) c
(Mn) d(C x Mn) ?y sin (C)
tanh (Mn)
12
(No Transcript)
13
Hyperbolic Tangents
14
(No Transcript)
15
(No Transcript)
16
non-linear functions
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
y
A
B
x
22
Brun, Robson, Narayan, MacKay Bhadeshia, 1998
23
Components of Creep Strength, 2.25Cr1Mo
iron microstructure
550 C
solid solution
600 C
precipitates
Murugananth Bhadeshia, 2001
24
Cole Bhadeshia, 1999
25
GTA weld at 823 K (data from Nippon Steel)
600
500
400
300
200
100
0
20000
30000
40000
Life / hours
Cole Bhadeshia, 1999
26
Cole Bhadeshia, 1999
27
(No Transcript)
28
Cool, 1996
29
Cool, 1996
30
600 C
As-welded
700 C
650 C
Cool, 1996
31
Siemens Mitsui Babcock Nippon Steel ABB
32
Nickel base alloy FT750dc
wt
33
(No Transcript)
34
Tancret Bhadeshia, 2002
1000
800
600
Yield stress / MPa
400
200
0
0
200
400
600
800
1000
1200
Temperature / C
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)