[MAP Logo]

Materials Algorithms Project
Program Library

Program MAP_STEEL_Ms_2004

  1. Provenance of code.
  2. Purpose of code.
  3. Specification.
  4. Description of program's operation.
  5. References.
  6. Parameter descriptions.
  7. Error indicators.
  8. Accuracy estimate.
  9. Any additional information.
  10. Example of code
  11. Auxiliary routines required.
  12. Keywords.
  13. Download source code.
  14. Links.

Provenance of Source Code

Thomas Sourmail and Carlos Garcia Mateo
Phase Transformations Group,
Department of Materials Science and Metallurgy,
University of Cambridge,
Cambridge CB2 3QZ, U.K.

E-mail: ts228athermesdotcamdotacdotuk

Added to MAP: June 2004.

Top | Next


A program for the prediction of the Ms temperature of steels as a function of chemical composition.

Top | Next | Prev


Language: C
Product form: Source Code (all Unix flavours), binary executable (DOS), package for Neuromat Predictor

Operating System : Tested on Solaris, Irix, Linux RH and Windows 98/2000/XP.

Top | Next | Prev


MAP_STEEL_MS_2004 contains the programs which enable the user to estimate the Ms temperature of steels as a function of chemical composition. It was trained using Neuromat's Model Manager, which is based on David Mackay's bayesian neural network program. Previously published databases on Ms temperatures were thoroughly verified and corrected for a large number of mistakes. An additional 700 data points were added, leading to a total of about 1100 data points.

If you have compiled the Unix/Linux version or are using the Windows/DOS program, the only file which you should be concerned with is 'test.dat', which you should edit to make predictions on your own compositions. An example file is provided which contains data for 9 different steels.

Top | Next | Prev


  1. T. Sourmail and C. G. Mateo, Review and assessment of models for predicting the Ms temperature of steels, submitted to Comp. Mater. Sci.
  2. T. Sourmail and C. G. Mateo, A model for predicting the Ms temperature of steels, submitted to Comp. Mater. Sci.
  3. All models and data related to this study are available together on this page.
  4. D. J. C. Mackay, Probable Networks and Plausible Predictions - . A Review of Practical Bayesian Methods For Supervised Neural Networks (1995)

Top | Next | Prev


Input parameters

The input variables are C, Mn, Si, Cr, Ni, Mo, V, Co, Al, W, Cu, Nb, Ti, B and N. The maximum and minimum values for each variable are given in the file MINMAX.

Output parameters

This model predict ln(-ln(Ms/1000)) where Ms is in Kelvin. It is for you to calculate Ms=1000*exp(-exp(output)) to obtain the correct result.
If you use Neuromat's Predictor, this is automatic.

Top | Next | Prev

Error Indicators

A full calculation of the error bars are given in reference [4].

Top | Next | Prev


No information supplied.

Top | Next | Prev

Further Comments

Read README for installation instruction on Linux/Unix

Top | Next | Prev


1. Program text

Complete program.

2. Program data

test.dat file included.

3. Program results

Prediction Error Pred-Err Pred+Err
-0.977297 0.019441 -0.996738 -0.957856
-0.901787 0.015718 -0.917505 -0.886068
-0.762616 0.011469 -0.774084 -0.751150
-0.680312 0.008202 -0.688514 -0.672110
-0.612506 0.009239 -0.621746 -0.603271
-0.508404 0.011323 -0.519727 -0.497080
-0.437531 0.012479 -0.450006 -0.425052
-0.370368 0.012560 -0.382927 -0.357808
-0.286682 0.014232 -0.300914 -0.272453

Top | Next | Prev


Neural networks, Ms temperature, martensite start temperature, steels.

Top | Next | Prev


Download Windows/DOS zip file

Download LINUX/UNIX source code as tar.gz file

Download Neuromat Predictor Package