Recent Research Activities in Laboratory

Recent Research Activities in Laboratory

Intelligent Mechanics LaboratoryIntelligent Mechanics Laboratory

School of Mechanical Engineering College of Engineering Pukyong National University

San 100 Yong dang-dong Nam-gu, Pusan 608-739, Korea

Tel) 82-51-620-1604, 625-1604

Fax)+ 82-51-620-1405

E-mail) dmlab@dolphin.pknu.ac.kr

Home page: http://vibration.pknu.ac.kr

School of Mechanical Engineering College of Engineering Pukyong National University

San 100 Yong dang-dong Nam-gu, Pusan 608-739, Korea

Tel) 82-51-620-1604, 625-1604

Fax)+ 82-51-620-1405

E-mail) dmlab@dolphin.pknu.ac.kr

Home page: http://vibration.pknu.ac.kr

Faculty :

Prof. Bo Suk Yang, Ph.D.

Prof. Soo-Jong Lee, Ph.D.

Prof. Dong-Jo Kim, Ph.D.

Faculty :

Prof. Bo Suk Yang, Ph.D.

Prof. Soo-Jong Lee, Ph.D.

Prof. Dong-Jo Kim, Ph.D.

Students :

Ph.D. : 8

Sung-Pil Choi, Jang-Woo Lee, Dong-Soo Lim, Young-Chan Kim, Yong-Han Kim, Jing-Long An, Jun-Ho Park, Woo-Kyo Jang

M.S. : 6

Yong-Min-Oh, Soo-Mok Lee, Jin-Dae Song, Bum-Jung Han, Young-Ho Choi, Tian Han

Undergraduate: 6

Students :

Ph.D. : 8

Sung-Pil Choi, Jang-Woo Lee, Dong-Soo Lim, Young-Chan Kim, Yong-Han Kim, Jing-Long An, Jun-Ho Park, Woo-Kyo Jang

M.S. : 6

Yong-Min-Oh, Soo-Mok Lee, Jin-Dae Song, Bum-Jung Han, Young-Ho Choi, Tian Han

Undergraduate: 6

Alumnus(1986 ~2000) Alumnus(1986 ~2000)

Ph.D. M.S. B. Eng.

7 50 82

Ph.D. M.S. B. Eng.

7 50 82

Faculty & Student

Research Area

Rotating Machinery(Pump,Motor,Turbine Generator, Compressor, etc.)

Rotating Machinery(Pump,Motor,Turbine Generator, Compressor, etc.)

Dynamic Optimum DesignDynamic Optimum Design

Monitoring & DiagnosticsMonitoring & DiagnosticsVibration AnalysisVibration Analysis

Area of Research:

Rotordynamics & Vibration Analysis

Intelligent Optimum Design & System Identification

Intelligent Condition Monitoring & Diagnostics

Area of Research:

Rotordynamics & Vibration Analysis

Intelligent Optimum Design & System Identification

Intelligent Condition Monitoring & Diagnostics

List of Research Applications:

Integrated Classification Techniques for Vibration Diagnostics

Development of Case-Based Reasoning Algorithm

Vibration Diagnostics by Petri-Net Algorithm

Model Updating Using Artificial Neural Networks

Development of Enhanced Genetic Algorithm for Optimum Design

Development of Optimization Algorithm Using Artificial Life

List of Research Applications:

Integrated Classification Techniques for Vibration Diagnostics

Development of Case-Based Reasoning Algorithm

Vibration Diagnostics by Petri-Net Algorithm

Model Updating Using Artificial Neural Networks

Development of Enhanced Genetic Algorithm for Optimum Design

Development of Optimization Algorithm Using Artificial Life

Research Area

Methods & Tools:

Artificial Neural Network (SOFM, LVQ, RBF, etc)

Random Tabu Search Method

Genetic Algorithm, Immune-Genetic Algorithm

Artificial Life

Methods & Tools:

Artificial Neural Network (SOFM, LVQ, RBF, etc)

Random Tabu Search Method

Genetic Algorithm, Immune-Genetic Algorithm

Artificial Life

Applications :

Optimum Shape Design of Rotor Shaft

Optimum Design for Bearing & Seal Geometry

Optimum Layout of Damping Material

Optimum Allocation of Piping System

Sensitivity Analysis

Applications :

Optimum Shape Design of Rotor Shaft

Optimum Design for Bearing & Seal Geometry

Optimum Layout of Damping Material

Optimum Allocation of Piping System

Sensitivity Analysis

Dynamic Optimum DesignDynamic Optimum Design

Vibration Analysis

Software Development for Vibration Analysis

Horizontal Pumps and Vertical Pumps

(General Centrifugal Pump, Boiler Feedwater Pump)

Hydraulic Turbine-Generator Rotor System for Hydro-Power Plant

Steam Turbine/Generator System for Thermal & Nuclear Power Plant

Rotary Compressor Rotor System for Small Refrigerator

Software Development for Vibration Analysis

Horizontal Pumps and Vertical Pumps

(General Centrifugal Pump, Boiler Feedwater Pump)

Hydraulic Turbine-Generator Rotor System for Hydro-Power Plant

Steam Turbine/Generator System for Thermal & Nuclear Power Plant

Rotary Compressor Rotor System for Small Refrigerator

Vibration Analysis

Motor/Generator Rotor System with Electromagnetic Pull

Geared & Coupled System (Bending & Torsional Vibration)

Vibration Analysis

Motor/Generator Rotor System with Electromagnetic Pull

Geared & Coupled System (Bending & Torsional Vibration)

Development of Vibration Diagnostics Algorithms

Neural Network & Fuzzy Theory

Decision Tree & Decision Table

Expert System

Petri Net Technique

Development of Condition Monitoring System

Case-Based Reasoning & Diagnosis

Construction of Case Base by Vibration Troubleshooting

Web Site of Case Base Search(http://vibration.pknu.ac.kr)

Wavelet Analysis & Feature Extraction

Ball Bearing Defect, Rubbing

Development of Vibration Diagnostics Algorithms

Neural Network & Fuzzy Theory

Decision Tree & Decision Table

Expert System

Petri Net Technique

Development of Condition Monitoring System

Case-Based Reasoning & Diagnosis

Construction of Case Base by Vibration Troubleshooting

Web Site of Case Base Search(http://vibration.pknu.ac.kr)

Wavelet Analysis & Feature Extraction

Ball Bearing Defect, Rubbing

Condition Monitoring & DiagnosticsCondition Monitoring & Diagnostics

Vibration Analysis of Turbine-Generator System for Nuclear Power Plant

HP Rotor LP Rotor

Turbine/Generator Rotor System

Steam Turbine-Generator Shaft Model (Kori #3 & 4, 1007MW)

Vibration Analysis : Campbell Diagram

Vibration Analysis : Damping Ratio & Root Locus

Vibration Analysis : Mode Shape

Damping and Stiffness Coefficients of No. 2 Bearing

Vibration Analysis : Bearing Dynamic Coefficients

Vibration Analysis : Comparison of Natural Frequency & Error

0 500 1000 1500 20001E-5

1E-4

1E-3

0.01

0.1

1

#5 bearing

#9 bearing

Am

plitu

de (m

m)

Rotating speed (rpm)

Comparison of unbalance response at bearing No.5, 9

Vibration Analysis : Unbalance Response

Vibration Analysis of Turbine-Generator System for Pumped-Storage Power Plant

Pump\Turbine-Generator\Motor Shaft Model (Muju #1, 2, 336MW)

0 100 200 300 400 500 600 700 8000

500

1000

1500

2000

2500 Runaway speed

Operating speed

1X

9XN

atu

ral

freq

uen

cy (

cpm

)

Rotating speed (rpm)

Vibration Analysis : Campbell Diagram

100 200 300 400 500 600 700 800

1

10

Operating speed

Generator/motor

Runner

without add mass/magnetic force

with add mass/magnetic force

Am

pli

tud

e (

m)

Rotating speed (rpm)

100 200 300 400 500 600 700 800

0.01

0.1

1

10

Operating speed

Generator/motor

Runner

Am

pli

tud

e (

m)

Rotating speed (rpm)

100 200 300 400 500 600 700 800

0

500

1000

1500

2000

2500

3000

Operating speed

without add mass of water

with add mass of water

Nat

ura

l fr

equ

ency

(cp

m)

Rotating speed (rpm)

Hydraulic unbalance

Mechanical unbalanceMode shape

Add-mass effect of water

Vibration Analysis : Mode Shape & Unbalance Response

Analytical Model & Earthquake Wave

0 5 10 15 20 25-600

-400

-200

0

200

400

600

Acc

eler

atio

n (g

al)

Time (sec)0 2 4 6 8 10

0

2

4

6

8

10

12

Frequency (Hz)

Am

plit

ude

Turbine-generator model and KOBE earthquake wave(EW)

Seismic Response Analysis

-0.4

-0.2

0.0

0.2

0.4

No.2 bearing

-0.4

-0.2

0.0

0.2

0.4

No.5 bearing

Dis

pla

cem

ent

(mm

)

0 5 10 15 20

-0.4

-0.2

0.0

0.2

0.4

No.9 bearing

Time (sec)

0 5 10 15 20

-0.6

-0.4

-0.2

0.0

0.2

0.4

No.9 bearing

Time (sec)

-0.4

-0.2

0.0

0.2

0.4

Dis

pla

cem

ent

(mm

)

-0.4

-0.2

0.0

0.2

0.4

0.6

No.5 bearing

No.2 bearing

Modal superposition method

Direct integration method

Seismic Response : Comparison of Analysis Methods

0 50 100 150 200 250

2

4

6

8

10

12

14

16

2

Max. allowable stress14.614

GENLPCLPBLPAHPB

end

ing

str

ess

(MN

/m )

Station number

Distribution of bending stress at maximum displacement position

Seismic Response Analysis : Bending Stress

0 10 20 30 400.0000

0.0002

0.0004

0.0006

0.0008

10

Level

6

7

8

9

Am

plit

ude

Time(sec)

Wavelet transform of seismic response wave at bearing No.9

0 10 20 30 400.000

0.001

0.002

10

Level

6

7

8

9

Am

plitu

de

Time(sec)

EW component UD component

Seismic Response Analysis: Wavelet Transform

Flow Chart for Immune Genetic Algorithm (IGA) Flow Chart for Immune Genetic Algorithm (IGA)

StartStart

Production of the initial chromosome calculation of the fitness

Production of the initial chromosome calculation of the fitness

Calculation of the fitness and affinity of individuals

Calculation of the fitness and affinity of individuals

Production of the individuals

Production of the individuals

SelectionSelection

Crossover and Mutation

Crossover and Mutation

Change old population with new population

Change old population with new population

Gen Max.GenGen Max.Gen

EndEnd

No

Yes

The differentiation of memory and suppressor cell

The differentiation of memory and suppressor cell

Calculation of the affinity between individuals and suppres

sor cellsAffinity Tacl

Calculation of the affinity between individuals and suppres

sor cellsAffinity Tacl

Proliferation andSuppression of

individuals

Proliferation andSuppression of

individuals

Optimization Result of IGAOptimization Result of IGA

IGAInitial

value Weight, W Bearing force, bF3

Critical speed, cn (rad/s)

c2 685.84 538.34573 549.86c3 2646.14 2309.7938 2301.05

Shaft weight, W (kg)

W 10.235 5.59691 9.82036

Transmitted force, bjF (N)

bF1 13.562 9.808 15bF2 475.599 339.90555 543.1bF3 461.366 183.49775 114.71

Maximum bending stress, (MPa)

MaxBS 45.29 22.056 35.75

Maximum amplitude, (m)

MaxAmp 78.9 65.87 68.9(b) Original model

(a) Optimum model

Flow Chart for Enhanced Genetic Algorithm Flow Chart for Enhanced Genetic Algorithm

StartStart

Production of the initial chromosome calculation of the fitness

Production of the initial chromosome calculation of the fitness

Calculation of the fitness Calculation of the fitness

Production of the individualsProduction of the individuals

SelectionSelection

Crossover and MutationCrossover and Mutation

Calculation of the affinity between saved candidacy solution set

Calculation of the affinity between saved candidacy solution set

FAC = 1 FAC = 1

Yes

No

Affinity < 0.1Affinity < 0.1

Decision and reallocation of candidacy solution

Decision and reallocation of candidacy solution

Erasion candidacy solution

Erasion candidacy solution

No

Selection and CrossoverSelection and Crossover

Production of the individualsProduction of the individuals

Fmin = Fmax Fmin = Fmax

Yes

Change old population with new population

Change old population with new population

No

Yes

EndEndGlobal search Local search

Change old population with new population

Change old population with new population

Solution num. = NSolution num. = NNo

Yes

Yes

Affinity < 0.1

Affinity < 0.1

Selection next solutionSelection next solution

No

),( yxf

x y

)5.3cos3cos1.2()1.25.2cos2(cos),( yyxxyxf

Optimum values

SGA 16.091155

IGA

16.09171316.09171316.09105116.091051

EGA

16.09172016.09172016.09172016.091720

Comparison of Optimization ResultsComparison of Optimization Results

Characteristics of Artificial life Algorithm

Circular food chain Dynamic interaction in the environment

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

xs

Artificial worldx(x

1

max, x2

max )

x(x1

min, x2

min )

x2

x1

D

Neighborhood region(acting radius)

: Resource: Artificial organism

Step 1 Initialization

Step 2 Search resource

Step 3 Movement using elite reservation strategy

Step 4 Metabolism

Step 5 Increasing age

Step 6 Reproduction

Step 7 Reducing energy

Step 8 Increasing generation

Flow chart for Artificial Life Algorithm

)0.2,0.2()0.1()(0.100),( 2122

122

1221 xxxxxxxf

Emergent Colonization produced at the optimumpoint

= (1.0, 1.0) optx )( optf x = 0

Contour line and emergent colonization for banana function

Optimization result of Artificial life Algorithm

Emergent Colonization produced at the optimumpoint

Contour line and emergent colonization for camel function

)0.2,0.2(

)0.40.4()3/1.20.4(),(

21

22

22

21

41

2121

xx

xxxxxxxf

= {(0.0898, -0.7126), (-0.0898, 0.7126)} optx )( optf x = -1.0316

Optimization result of Artificial life Algorithm

Structure of Classification System for Diagnostics

Database Storage Module Database Storage Module

Data Training Module Data Training Module

Classification Module Classification Module

1. Experimental Configuration

2. A/D Converting System

3. Database Management

1. Experimental Configuration

2. A/D Converting System

3. Database Management

1. Wavelet Transform

2. Statistical Feature Extraction

3. Training using Neural Network

1. Wavelet Transform

2. Statistical Feature Extraction

3. Training using Neural Network

1. Untrained New Data

2. Store to Database

1. Untrained New Data

2. Store to Database

A/D ConversionA/D Conversion

Hardware

Transient Stable

Software

Signal Data, Condition, Specification,Date, Sensor Information

Database

W/T TransformFeature Extraction

W/T TransformFeature Extraction

Training (Neural Network) :SOFM, LVQ

Training (Neural Network) :SOFM, LVQ

W/T TransformFeature Extraction

W/T TransformFeature Extraction

Trained DataCondition ClassificationCondition Classification

Software Process

Software

Integrated Classification System for Diagnostics

Normal Abnormal

Wavelet Transform

Statistical Evaluation Value : Mean, Standard deviation, Skewness, Kurtosis

Statistical Evaluation Value : Mean, Standard deviation, Skewness, Kurtosis

More & Robust Features than Time-waveform

Neural Network Classification

Training Data

Trained Data

Self-Organizing Feature Map & Learning Vector Quantization

Technique

Re-organized into CODEBOOK Vectors

k-NN TechniqueWhich Class?

Untrained Data

Main Window for Diagnosis System

Database Storage Module : Data Input

Database Storage Module : Management

Diagnosis Module : Classification

Introduction of Case-Based Reasoning System

- Memorize previous situation and case-history - Reuse to for solving new problem - Previous problem solving current problem solving

- Memorize previous situation and case-history - Reuse to for solving new problem - Previous problem solving current problem solving

CBR System for Vibration Diagnosis

• Categories and Details : Can be Added through CBR Cycle

• Keywords and Weights : Extracted from the Case-Base and Stored to Library

Input and Output of CBR System

http://vibration.pknu.ac.kr

Petri Net Algorithm for Abnormal Diagnosis

BackgroundBackground

Cause Diagnosisby Symptom

Frequency

Cause Diagnosisby Symptom

Frequency

Occurrence of Abnormal Vibrationat Rotating Machine

Occurrence of Abnormal Vibrationat Rotating Machine

Expression ofSymptom Freq.by Target Transition

Expression ofSymptom Freq.by Target Transition

Addition of Source Transition

at All SourcePlaces

Addition of Source Transition

at All SourcePlaces

Minimal Support

T-invariantCalculation

Minimal Support

T-invariantCalculation

Petri NetCarl A. Petri

1962

Petri NetCarl A. Petri

1962

PUKYONG NATIONAL UNIV. Intelligent Mechanics Lab.

Diagnosis of Rotating Machine by Petri Net

ModelingModelingModelingModeling

Place Transition

P1 축상에서 헛도는 베어링의 기계적 느슨함

P2 하우징내에서 베어링 느슨함

P3 과도틈새

P4 회전속도의 많은 회전성분

P5 구별되는 회전속도의 4X 성분

P6 기본열주파수

Transition Rule

T1 P1 P4

T2 P2 P4

T3 P2 P5

T4 P3 P4

T5 P3 P6

Source Transition

Source Place

T6 P1

T7 P2

T8 P3

PUKYONG NATIONAL UNIV. Intelligent Mechanics Lab.

Diagnosis of Rotating Machine by Petri Net

Diagnosis ResultsDiagnosis Results

Symptoms Causes Minimal support

T-invariant

회전속도의 많은 조화성분 (P4)과 구별되는 회전속도의 4X성분 (P5)

축상에서 헛도는 베어링의 기계적인 느슨함 (P1)과 하우징 내에서 베어링 느슨함 (P2)

[1010011011]

하우징내에서 베어링 느슨함 (P2) 과 과도틈새 (P3) [0011001111]

원인 추정 하우징 내에서 베어링 느슨함 (P2)

회전속도의 많은 조화성분 (P4)과 기본열주파수 (P6)

축상에서 헛도는 베어링의 기계적인 느슨함 (P1)과 과도틈새 (P3)

[1000110111]

하우징 내에서 베어링 느슨함 (P2) 과 과도틈새 (P

3)[0100101111]

원인 추정 과도틈새 (P3)

PUKYONG NATIONAL UNIV. Intelligent Mechanics Lab.