How to run PARROT and create POP files

Optimization module for evaluation of thermodynamic and kinetic model parameters

The objective of the PARROT module is to provide unified set of subroutines to be used in critical assessment of various thermochemical and kinetic parameters against avialble experimental and literature data

It makes use of the GES module to handle the models of the various phases

It makes use of POLY module to store and calculate complex heterogeneous equilibria

What is PARROT?

Optimization method The criteria for the best fit derived from the

maximum of the likelyihood principle, will be minimum in the sum of squeare of residuals

Inaccuracies of experimental conditions can be taken in to account in two ways: Independent state variables can be prescribed

in the POLY interface. In this case equilibrium will be calculated with experimental values

The true value can be optimized by using one of the defined variables as the condition.

Commands available in the parrot AMEND_PARAMETER LIST_ALL_VARIABLES RESCALE_VARIABLES AMEND_SYMBOL LIST_CONDITIONS SAVE_PARROT_WORKSPACES BACK LIST_PARAMETER SET_ALTERNATE_MODE COMPILE_EXPERIMENTS LIST_PHASE_DATA SET_EXTERNAL_PARAMETER CONTINUE_OPTIMIZATION LIST_RESULT SET_FIX_VARIABLE CREATE_NEW_STORE_FILE LIST_STORE_FILE SET_INTERACTIVE EDIT_EXPERIMENTS LIST_SYMBOL_IN_GES SET_OPTIMIZING_CONDITION ENTER_PARAMETER MACRO_FILE_OPEN SET_OPTIMIZING_VARIABLE EXIT OPTIMIZE_VARIABLES SET_OUTPUT_LEVELS GOTO_MODULE READ_PARROT_WORKSPACES SET_SCALED_VARIABLE HELP RECOVER_VARIABLES SET_STORE_FILE INFORMATION REINITIATE

Creation of POP filesingle phase equilibria

CREATE_NEW_EQUILIBRIUM 1,1CHANGE_STATUS PHASE LIQUID=FIX 1SET_CONDITION T=1379 P=1E5 X(LIQ,AU)=0.0563SET-REFERENCE-STATE AU LIQ * 1e5SET-REFERENCE-STATTE CU LIQ * 1E5EXPERIMENT HMR=1520:200

CREAT-NEW-EQUILIBRIUM 1 1CHANGE-STATUS PHASE LIQUID FCC=FIX 1SET-CONDITION X(FCC,CU)=0.14 P=1E5EXPERIMENT T=970:2$ H E Bennet, J inst of metals 1962

Creation of POP filesingle phase equilibria

$Congruent melting temperature of Fe2Nb: 1646 C [2011Vos]CREATE_NEW_EQUILIBRIUM 30,1CHANGE_STATUS PHASE LIQUID,C14_LAVES=FIX 1SET_CONDITION P=P0,X(LIQUID,NB)-X(C14_LAVES,NB)=0 SET_START_VALUE T=1919.15,Y(LIQUID,NB)=0.33EXPERIMENT T=1918:2$s-we 1 30

Creation of POP fileInvariant reactions equilibria

Creation of POP fileInvariant reactions equilibria

E.G. 3 Phase equilbrium in binary system:CREAT-NEW-EQUILIBRIUM 1 1CHANGE-STATUS PHASE FCC BCC LIQUID=FIX 1SET-COND P=1E5EXPERIMENT t=912:5

Creation of POP fileTables of experimental data

$ Andersson 1988CREATE_NEW_EQUILIBRIUM 128,1CHANGE_STATUS PHASE FCC_A1, M7C3 M23C6=FIX 1SET_CONDITION P=P0, T=1473EXPERIMENT W(FCC,CR)=0.154:0.005 EXPERIMENT W(M7C3,CR)=0.629:0.05EXPERIMENT W(M23C6,CR)=0.515:0.05

TABLE_HEAD 400CREATE_NEW_EQUILIBRIUM @@,1CHANGE_STATUS PHASE BCC_A2, LIQ, FCC=FIX 1SET_CONDITION P=P0, T=@1EXPERIMENT W(LIQ,C)=@2:10%EXPERIMENT W(LIQ,CR)=@3:0.01EXPERIMENT W(BCC,C)=@4:10%EXPERIMENT W(BCC,CR)=@5:0.01EXPERIMENT W(FCC,C)=@6:10%EXPERIMENT W(FCC,CR)=@7:0.01TABLE_VALUES$ T W(LIQ,C) W(LIQ,CR) W(BCC,C) W(BCC,CR) W(FCC,C) W(FCC,CR) 1761 0.005514 0.025484 0.000977 0.0197 0.002023 0.01757 1751 0.006259 0.044834 0.001139 0.037751 0.002502 0.03483 1741 0.007172 0.06365 0.001351 0.053662 0.003019 0.049626 1731 0.008184 0.082505 0.001538 0.06998 0.003531 0.06333 1721 0.00911 0.100096 0.001824 0.084647 0.0041 0.077481TABLE_END$set-weight 1 400-404

Creation of POP fileTables of experimental data

System definition file: Set up file “TCM” fileBut how do we introduce functions and variables which should be optimized?

System definition file: Set up file “TCM” file

System definition file: Set up file “TCM” file

System definition file: Set up file “TCM” file

The PARROT module reserves 99 variables and can handle 1000 experimental measurements in one single assessment system

Some tricks

• Conflict in experimental data/• Experimental data which gives

strange results• More then one dependent

experimental value exists for one independent experimental value Bo Jannson rule:

If you have two conflicting datasets, then either one or the other may be correct or both are wrong. Both cannot be correct!

Some tricks Use as few experiments as possible to obtain initial fit Use weights! It may be interesting to exclude from optimization intermediate

phases at the initial stage and just optimize liquid and terminal phases. Metastable solubility lines should not contain strange kinks. The values for these phases can be later set fix in order to optimize parameters for intermediate phases

Take care about miscibility gaps Phase with order/disorder require manual start values In case terminal solutions are the same phase (e.g. Cr-Nb) pay

attention that calculated equilibria are on the right side of the system. A final optimization with all variables and all experimental values

with appropriate weighing should be made!!! Check continuously that optimized variables are within reasonable

range of values. When variable starts to change several orders of magnitude during an optimization, one should reconsider weights and if too many parameters are used.

Thank you for your attention!