               Materials Algorithms Project
               Program Library

  ------------------------------------------------------------------------

Program MAP_NEURAL_WELDMETAL_T_27J

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

  ------------------------------------------------------------------------

Provenance of Source Code

Sree Harsha Lalam and Prof. H.K.D.H. Bhadeshia,
Phase Transformations Group,
Department of Materials Science and Metallurgy,
University of Cambridge,
Cambridge, CB2 3HU, U.K.

The neural network program was produced by:

David MacKay,
Cavendish Laboratory,
University of Cambridge,
Madingley Road,
Cambridge, CB3 0HE, U.K.

Added to MAP: December 1999

Top | Next

Purpose

To estimate the transistion temperature at 27 J of  Charpy toughness of
steel weldmetal (T27J  ) as a function of  yield strength , oxygen content,
reheated material  and percentage acicular ferrite content in the weld
metal.

Top | Next | Prev

Specification

 Language:        FORTRAN / C
 Product form:    Source code / Executable files
 Operating System:Solaris 5.5.1, Linux & Windows 95/98

Top | Next | Prev

Description

MAP_NEURAL_WELDMETAL_T_27J contains a suite of programs which enable the
user to estimate the  27 J Charpy toughness transistion temperature ( T27J )
of steel weldmetal , as a function of  yield strength , oxygen content,
reheated material  and percentage acicular ferrite content in the weld
metal.. It makes use of a neural network program called generate44, which
was developed by David MacKay and is part of the bigback5 program. The
network was trained using a database of experimental results [1] . Thirteen
different committee models , which differ from each other by the number of
hidden units and by the value of the seed used when training the network. It
was found that a more accurate result could be obtained by averaging the
results from all the models [2]. This suite of programs calculates the
results of each model and then combines them, by averaging, to produce a
committee result and error estimate, as described by MacKay [page 387 of
reference 3]. The source code for the neural network program can be
downloaded from David MacKay's website; the executable files only are
available from MAP. Also provided are FORTRAN programs (as source code) for
normalising the input data, averaging the results from the neural network
program and unnormalising the final output file, along with other files
necessary for running the program.

Programs are available which run on a Solaris 5.5.1 unix and Linux operating
system and on a PC under Windows 95/98. A distinct set of program and data
files are provided . The files for unix and Linux are separated into
compressed files called weldmetal_T_27J_solaris.tar  and
weldmetal_T_27J_linux.tar  ; those for a PC :   weldmetal_T_27J.zip. Each
directory or zip file contains the following files and subdirectories:

README
     A text file containing step-by-step instructions for running the
     program, including a list of input variables.
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
     An input text file containing the input variables used for predictions.
model.gen
     This is a unix shell file containing the command steps required to run
     the module. It can be executed by typing csh model.gen  at the command
     prompt. These shell files compile and run all the programs necessary
     for normalising the input data, executing the network for each model,
     unnormalising the output data and combining the results of each model
     to produce the final committee result.
model.exe
     These executable programs for the PC correspond to the unix command
     files model.gen. The source code is given in model.c  which are in
     subdirectory s.
data_no.ex/data_no.exe
     This executable file reads the information of number of data from
     keyboard input and creates no_of_rows.dat file, this file is used by
     spec.ex/spec.exe to create spec.t1.
spec.ex/spec.exe
     This executable file reads the information in no_of_rows.dat and
     creates a file called spec.t1.
spec.t1
     A dynamic file, created by spec.ex/spec.exe, which contains information
     about the module and the number of data items being supplied. It is
     read by the program generate44/generate55.exe.
norm_test.in
     This a text file which contains the normalised input variables. It is
     generated by the program normtest.for in subdirectory s.
generate44 / generate55
     This is the executable file for the neural network program. generate44
     runs on unix selntems and generate55 on the PC. It reads the normalised
     input data file, norm_test.in, and uses the weight files in
     subdirectory c, to find a value for temperature. The results are
     written to the temporary output file _out.
_ot, _out, _res, _sen
     These files are created by generate44 and can be deleted.
Fresult
     Contains the final un-normalised committee results for the predicted.
SUBDIRECTORY s
data_no.c
     The source code for program data_no.ex.
spec.c
     The source code for program spec.ex.
normtest.for
     Program to normalise the data in test.dat and produce the normalised
     input file norm_test.in. It makes use of information read in from
     no_of_rows.dat and committee.dat.
gencom.for
     This program uses the information in committee.dat and combines the
     predictions from the individual models, in subdirectory outprdt, to
     obtain an averaged value (committee prediction). The output (in
     normalised form) is written to com.dat.
treatout.for
     Program to un-normalise the committee results in com.dat and write the
     output predictions to unnorm_com. This file is then renamed as Fresult
     .
committee.dat
     A text file containing the number of models to be used to form the
     committee result and the number of input variables. It is read by
     gencom.for, normtest.for and treatout.for.
SUBDIRECTORY c
_w*f
     The weights files for 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 [2]
     for each model.
_R*
     Files containing values for the noise, test error and log predictive
     error [2] 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 via spec.t1.
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.for.

Detailed instructions on the use of the program are given in the README
files. Further information about this suite of programs can be obtained from
reference 2.

Top | Next | Prev

References

  1. French, I. E., 1999, Australaisan Welding Journal,  44, second quarter,
     44-46.
  2. S. H. Lalam, H. K. D. H. Bhadeshia and D. J. C. MacKay, 1999,
     Australasian Welding Journal. to be published. [Download postscript
     file.]
  3. D.J.C. MacKay, 1997, Mathematical Modelling of Weld Phenomena 3, eds.
     H. Cerjak & H.K.D.H. Bhadeshia, Inst. of Materials, London, pp 359.
  4. D.J.C MacKay's website at
     http://wol.ra.phy.cam.ac.uk/mackay/README.html#Source_code

Top | Next | Prev

Parameters

Input parameters

The input variables for each module are listed in the README file in the
corresponding directory. The maximum and minimum values for each variable
are given in the file MINMAX.

Output parameters

These programs give the output in oC. The corresponding output files are
called Fresult . The format of the output file is:

Prediction    Error      Prediction -  Error      Prediction + Error
  (oC)          (oC)              (oC)                   (oC)

Top | Next | Prev

Error Indicators

None.

Top | Next | Prev

Accuracy

A full calculation of the error bars is presented in reference 2.

Top | Next | Prev

Further Comments

None.

Top | Next | Prev

Example

1. Program text

       Complete program.

2. Program data

See sample data file: test.dat.

3. Program results

See sample output file: Fresult

Top | Next | Prev

Auxiliary Routines

None

Top | Next | Prev

Keywords

neural network, weldmetal toughness,  Charpy toughness

Top | Next | Prev

Download

Download MAP information files

Solaris.5.5.1:
     Download Solaris module
Linux:
     Download Linux module
PC Software:
     Download PC Module

Top | Prev
  ------------------------------------------------------------------------
MAP originated from a joint project of the National Phelnical Laboratory and
                        the University of Cambridge.
               MAP Website administration / map@msm.cam.ac.uk

                         Top | Index | MAP Homepage
-----------------------------------------------------------
