Transcript Bez tytułu slajdu

COMPUTER-AIDED PROJECTING OF GAS NITRRIDING
PROCESSES WITH UTILIZATION OF SIMULATION AND
METHODS OF ARTIFICIAL INTELLIGENCE
Jerzy DOBRODZIEJ, Jacek WOJUTYŃSKI,
Jerzy RATAJSKI, Tomasz SUSZKO,
Jerzy MICHALSKI
INSTITUTE FOR SUSTAINABLE TECHNOLOGIES – NATIONAL RESEARCH INSTITUTE
POLAND
PRESENTATION PLAN
PROBLEMS TO SOLVE
METHODS OF SOLVING
MODULE OF DATABASES
EXPERT SYSTEM
MODULE DESIGN OF DYNACMICS CHARACTERISTIC
OF NITIRIDING PROCESSES
MODULE OF NEURAL NETWORK
MODULE OF OPTIMISATION –
EVOLUTIONARY ALGORITHMS
PROBLEMS TO SOLVE
layer
thickness=2.8mm
Substrat material
Material with a layer
Substrate
material
Material selection
Process milieu
Material with
a layer
Selection and inspection
of control parameters
Selection of the
layer’s properties
Classical approach – empirical methods of trial and error
Computer-aided processes of layers creation – How it to do ?
METHODS OF SOLVING
Substrate
material
Process milieu
Measurements on-line
Archival data
Material with
a layer
Measurements off-line
DATABASE
APPLIED MODELS
Input parameters
Fuzzy logic
(expert systems)
Output parameters
Artificial neural networks
Evolutionary algorithms
Forecasted properties
of a layer
Data mining models –
detection of similarities and
differences in processes
Analytical models:
thermodynamic, statistical,
heuristic
Computer-aided design
of layers creation
MODULE OF DATABASES - INFORMATION STRUCTURE
Archival process
In-situ process
Process
Parameter name
Value
Parameter type
Parameters for
the whole
process
Devices
Device 1
...
Parameter name
Value
Parameter type
Device m
Parameter name
Value
Parameter type
Materials
Substrate
(before the
process)
Parameter name
Value
Parameter type
Parameter name
Value
Parameter type
Effects of the process
(economical, ecological,
innovative, etc.)
Stages of the
process
Materials with
layers
(after the process)
Parameter name
Value
Parameter type
Stage 1
...
Parameter name
Value
Parameter type
Stage n
Parameter name
Value
Parameter type
MODULE OF DATABASES - APPLICATION
Local database
Collection of data in local databases
Operational tasks
Registration of a new process by defining
process structure and saving the created
structure into the database
Data modification
•parameters set which describes process,
•data of technological stages,
•device data,
•material or layer data,
•dynamic characteristics of the process
(or stage),
•graphical data concerning results of layer
structures tests,
Assuring accomplishing transactions such as
adding, removing, modyfing and
selecting/searching data
Transaction synchronisation with the
concurrent access and creation of
appropriate blocades while simultaneous
modyfing the same data by many users
Data coherence, that is inviolability of data
integrity rules
Removing data from database
Replicationality (data repetitiveness, reverse
copy)
Data coping
Concurrent access for many users
Aggregating dispersed data from local
databases
Making access to data via the Internet
according to users rights
Data search
•SQL queries,
•ranking search,
•fuzzy search for data mining requirements
and artificial intelligence models.
Providing multi-level security systems
against access to data:
•setting accounts for users
•setting system rights
•assigning access rights to objects
in database
•guaranting access to tables and atributes
in tables
EXPERT SYSTEM
User interaction module
Selection of
input and output
parameters
set
Formulation
of database
query
- STRUCTURE OF EXPERT SYSTEM
Database integration module
DATABASE
Creation of the fuzzy logic function
Knowledge bases generation
Inference module
Set of processes
Optimisation module
Knowledge bases optimisation
Fuzzification
of input parameters
values
Rules
congregation
Defuzzification
INFERENCE RESULTS:
LAYER PARAMETERS VALUES
(output parameters)
12/16
EXPERT SYSTEM
- APPLICATION
TASK
Prediction of layers properties manufactured
in nitriding and PVD processes.
Support for designing the nitriding processes
technologies on the basis of substrate
and process milieu parameters.
System properties
Inference versatility
Inferencing with diverse parameters.
Flexibility and coherence of inferencing
Inferencing on the basis of different
domains parameters: continue (e.g. temperature
in time function), discrete (e.g. value of layer
resistance to corrosion), nominaly ordered
(e.g. type of mechanical treatment used
for substrate surface).
Inference adaptation and self-learning
Using data referring to new and completed processes
as well as created layers in order to improve inference
quality.
IFHTSE 2007 Congress
Adam Mazurkiewicz, 31.10.2007
13/16
EXPERT SYSTEM
- VALIDATION IN THE FIELD OF NITRIDING PROCESSES
Process 1
Process 3
570
60
480
4.75
3.25
6
Process duration [min]
Mean nitride potential [atm½]
Process 2
530
Dependence of the530
error value570
on the number
of rules in knowledge base
Temperature [°C]
60
20
60
12 atmosphere [%]
Amount of NH3 in the
40
10,6710,37 10,49
9,57
Substrate material 10
40
HMJ
9,36
60
40
Amount of N2 in the atmosphere [%]
triangle
Error value [%]
Fuzzification method
8
40 HMJ
38 HMJ
triangle
triangle
8,95
7,1 7,37
7,81
Process 1
6
Process 2
Parameter name
4
Obtained
Predicted
Obtained
Process 3
4,49 4,62 4,34
Predicted
Obtained
Predicted
Effective thickness g400 [mµ]
0.185
0.1767
0.17
0.16210
0.2
0.1913
Effective thickness g500 [mµ]
0.09
0.086
0.07
0.06680
0.1
0.0957
0.345
0.3295
0.24
0.22890
0.3
0.2870
4.5
4.3047
2
0Effective thickness gr+50 [mµ]
22
23
Grey area thickness [mµ]
24
Nitride layer thickness [mµ]
Number
of rules
10
10.4500
Maximum hardness HV
Process 1
5
25
4.7750
26
27
4
4.18480
12.5
13.0775
10.5
10.0443
551
575.795
538
562.8556
552
575.9568
659
629.345
692
723.9704
644
671.9496
Surface hardness HV10
642
670.890
630
600.8940
625
652.125
Surface hardness HV0.5
519
495.645
512
535.6544
532
555.0888
Surface hardness HV5
544
519.520
559
533.1742
562
537.6092
Surface hardness HV1
Results
Process milieu
and substrate
Process 2
Process 3
MODULE DESIGN OF DYNACMICS CHARACTERISTIC OF
NITIRIDING PROCESS
Designing of process
environment characteristics
Purpose
Designing of atmospheres for gas nitriding process.
Module properties
Two- and tree-component atmospheres:
Nitriding potential model on the basis of isoconcentrative
characteristics or established by the designer.
Model of dissociation level.
MODULE DESIGN OF DYNACMICS CHARACTERISTIC OF
NITIRIDING PROCESS
Purpose
nitrides area thickness
Simulation of layer growth kinetics.
Simulation of nitrogen concentration profiles on phases borders.
System properties
Short timechanges
of calculations.
temperature
Additional software for mathematical calculations not required.
Possibility of layer growth in time animation.
Possibility of concentrations on phase border animation.
Possibility of concentration profiles on phase border animation.
concentration
on phase borders
potential changes
nitrogen
concentration
profiles
MODULE OF NEURAL NETWORK
Purpose
Prediction of micro hardness distribution in the function of:
Process duration
Temperature
Nitridning potential
Result
Module properties
Optimal structure of neuron network.
Generalization option.
Possibility of adapting for diverse materials substrates.
MODULE OF OPTIMISATION – EVOLUTIONARY ALGORITHMS
Purpose
Temperature and nitriding potential prediction in order to obtain the projected
micro hardness distribution
System
Result: process properties
parameters
Determining optimal average values of temperature and potential
in successive gas nitriding process.
Possibility of adapting for diverse materials substrates.
CONCLUSIONS
Designed system enables:
Modification and development of technologies, particulary working out new technological
solutions.
Reduction in energy and material consumption, as a result of processes duration shortening.
Precise planning of processes and obtaining surface layers described by set parameters
The system might be used for:
Competitiveness’ enhancement of SMEs operating in surface treatment area by improving en end
product quality
Designing of new properties profiles, for instance, toward development of extremely hard layers
with high adhesion in aim to increase their life by surface hardness enhancement, wear resistance
(pitting, micro-pitting and scuffing) and endurance of machine and tools’ elements
Creating new SMEs which are consultants in the area of surface treatment, i.e. selection of single
treatment or joint treatment and their parameters for certain applications