Materials Algorithms Project
![[A logo showing the University of Cambridge Crest]](https://www.phase-trans.msm.cam.ac.uk/Univ.jpg)
|
Materials Algorithms Project
|
|
 |
Dominique Carrouge,
Phase Transformations Group,
Department of Materials Science and Metallurgy,
University of Cambridge,
Cambridge CB2 3QZ, U.K.
E-mail: dc249@cus.cam.ac.uk
Added to MAP: September 2002.A program for the prediction of the Ac1 temperature of steels, including martensitic stainless steels, as a function of chemical composition and heating rate.
|
Language: |
C |
|
Product form: |
Source Code (Unix only) |
Operating System : Tested on Solaris, Linux and PC.
MAP_STEEL_AC1TEMPcontains the programs which enable the user to estimate the Ac1 temperature of steels, including martensitic stainless steels, as a function of chemical composition and heating rate. It makes use of a neural network program called generate44, which was developed by David Mackay and is part of the bigback5 program. Eighty three experimental data relevant to 9-13 wt% Cr steels were added to the original database from Gavard et al work [1] and the updated database "database_ac1_martensitic.txt " is also provided . Following a Gaussian process analysis by Bailer-Jones et al [2], the elements considered to affect the Ac1 temperature are C, Si, Mn, Cu, Ni, Cr, Mo, Nb, V, W, Co. The model presented here combines the output of two different models which had five hidden units and different initial weights. The programs automatically calculate the result of each model and average those results to produce a prediction of Ac1 temperature and an error estimate.
A Linux/Unix and a PC version of the program are provided. The source code is accompanied by a program to install the program, which should run on most version of UNIX. Once uncompressed, and once the 'install' program runs (sh install), the directory contains
MINMAX
A text file containing the minimum and maximum limits of each input and output variable. This file is used to normalise and unnormalise the input and output data.
test.dat
model.gen
This is a UNIX shell file containing the commands required to run the model. It can be executed by typing `sh model.gen' at the command prompt. This shell file normalises the input data, executes the neural network program, unnormalises the results and combine them to produce the final committee result.
.normalise
Hidden executable file, to normalise the input data.
.generate44
Hidden executable file, for the neural network program. It reads the normalised input file and also uses a configuration file spec.t generated by .generate\_spec and the weight files, located in the subdirectory c.
.generate\_spec
Hidden executable file, generates the configuration file read by .generate44.
.gencom
Hidden executable file, it combines the output of the different models in a committee result.
.treatout
Unnormalise the results.
SUBDIRECTORY c
\_w*f
The weight files of the different models.
*.lu
Files containing information for calculating the size of the error bars for the different models.
\_c*
Files containing information about the perceived significance value [3] for each model.
\_R*
Files containing information about the noise, test error and log predictive error [3] for each model.
SUBDIRECTORY d
outran.x
A normalised output file which was created during the building of the model. It is accessed by .generate44.
SUBDIRECTORY outprdt
out1, out2, etc.
The normalised output files for each model.
com.dat
The normalised output file containing the committee results. It is generated by .gencom .
To make predictions with the PC version: alter the test.dat file and double click on model.exe. The results are in the file Model_result.dat.
The input variables are C, Si, Mn, Cu, Ni, Cr, Mo, Nb, V, W, Co, Heating Rate. The maximum and minimum values for each variable are given in the file MINMAX.
The output is written in the file result in the Unix/Linux version and in Model_RESULT.dat for the PC version. This file contains the predicted Ac1 temperature (in degree Celsius), the error bar value, the predicted value - the error and the predicted value + the error.
A full calculation of the error bars are given in reference [3].
No information supplied.
None.
Complete program.
"C" "Si" "Mn" "Cu" "Ni" "Cr" "Mo" "Nb" "V" "W" "Co" "Heating Rate" 0.005 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1 0.01 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1 0.015 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1 0.02 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1 0.025 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1 0.03 0.26 0.46 0 6.46 12.2 2.48 0 0 0 0 1
599.620728 27.860588 571.760193 627.481323 600.653503 28.078754 572.574768 628.732239 601.681885 28.325672 573.356201 630.007568 602.705444 28.600571 574.104858 631.306030 603.724976 28.901979 574.822998 632.626953 604.739319 29.228968 575.510376 633.968262
Neural networks, Ac1 temperature, heating rate, martensitic stainless steels, carbon steels.
Download PC source code as zip file