SlideShare a Scribd company logo

Improvement of strip thickness control through the process of data analytics

Download as PPTX, PDF
S
Sri Raghavan

This document presents a project update on the improvement of final strip thickness control in high silicon electrical steel using data analytics. It discusses objectives, current progress including feature selection, data preprocessing, model building, and results from various regression models. The findings focus on identifying relationships within the data to reduce thickness and enhance magnetic performance, with an emphasis on addressing data quality issues and employing statistical techniques such as PCA and various regression models.

Engineering
1 of 37
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
PROJECT UPDATE FOR IMPROVEMENT OF ‘FINAL STRIP THICKNESS CONTROL’ IN HIGH ‘SI ELECTRICAL STEEL’ THROUGH THE USE OF ‘DATA ANALYTICS’ Presentation by: Srinivasa Raghavan Ramanujam F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
Objective Current Progress Feature Selection Dimensionality Reduction Model Building Results and Discussion Recommendation Future Work F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
Aim To analyse the cold reduction of high Si steel for reducing the thickness by 90% to achieve sharp magnetic losses which are proportional to the final strip thickness through the use of data analytics. F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
Dissertation Plan F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Topic Weeks 1 2 3 4 5 6 7 8 9 10 11 12 Initial Proposal Feature Selection in Weka Pre-processing Dataset Model Building in Python Analysing Results Current Position Compiling Report Writing
Project Approach F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W •Analyzing the generalized data •Identifying the base of the problem Top down approach •Dividing into test and train sets •Feature Scaling Data pre- processing •Identify the relation between different attribute Feature Selection •Dimensionality ReductionPCA •Build the correct algorithm using python •Predict the results Model Building Visualize Results
Current Update F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Update Check Exporting the IBA data into a csv format Trying to find a relationship between different attribute Develop various mathematical model to predict the snap Working on a GUI to perform data mining using python
Feature Selection O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Problem: Unclean Data O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Fluctuations when velocity is zero
Solution: Removing Zeros O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Feature Selection: Weka O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S It is the process of selecting a subset of relevant features (variables, predictors) for use in model construction. Feature selection techniques are used for four reasons: Simplification of models to make them easier to interpret by researchers/users, Shorter training times, To avoid the curse of dimensionality, Enhanced generalization by reducing overfitting Point where snap happens
Feature Selection: Weka O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Performed for snap and non snap coils 727 Attributes 30 coils Six Different Models BestFirst+ Cfs Subset Eval Greedy StepWise+ Cfs Subset Eval Greedy StepWise + WrapperSubset Eval Ranker + Principal Components Ranker+ ReliefAttribute Eval Ranker + CorrelationAttribute Eval
Feature Selection Result: Entire Coil O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Right gauge coarse deviation Feedback Error Single Shot Gauge Deviation Delayed Raw Mass flow Gauge Error Right Gauge Deviation Component Fit Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back_Capsule_Titt Calc mill speed WR circ
Feature Selection Result: Snap Time Data O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Feedback Error Single Shot Gauge Deviation Right Gauge Deviation Component Fit Mass Flow Exit Gauge Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back_Capsule_Titt Calc mill speed WR circ
Feature Selection Result: Non Snap Selection Coil O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Right gauge coarse deviation Feedback Error Single Shot Gauge Deviation Delayed Raw Mass flow Gauge Error Left Gauge Deviation Component Fit Right Gauge Deviation Component Fit Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back Capsule Titt Calc mill speed WR circ
Data Pre-processing O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Data Pre-processing O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Four steps Importing the Libraries Importing the datasets Splitting into test and train sets Feature Scaling Dimensionality Reduction
Dimensionality Reduction O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Principal Component Analysis Unsupervised Mode of Dimensionality Reduction Statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal componentsSplitting into test and train sets
Model Building O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Model: Multiple Linear Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Multiple linear regression attempts to model the relationship between two or more explanatory variables and a response variable by fitting a linear equation to observed data. Every value of the independent variable x is associated with a value of the dependent variable y
Model: Random Forest Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S It is unexcelled in accuracy among current algorithms. It runs efficiently on large databases. It can deal with a large number of information factors without variable erasure
Model: Decision Tree Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Decision tree builds regression or classification models in the form of a tree structure. It breaks down a dataset into smaller and smaller subsets while at the same time an associated decision tree is incrementally developed. The final result is a tree with decision nodes and leaf nodes
Model Results O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 7.58% 2 37.62% 3 38.54% 4 40.98% 5 75.14% 6 80.61% 7 80.81% 8 82.66% PCA Components Model Accuracy mean 9 82.76% 10 82.76% 11 82.76% 12 82.76% 13 82.76% 14 82.76% 15 82.76% Model: Multiple Linear Regression
Model: Decision Tree Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 92.95% 2 99.42% 3 99.88% 4 99.952% 5 99.96% 6 99.97% 7 99.97% 8 99.97% PCA Components Model Accuracy mean 9 99.98% 10 99.98% 11 99.983% 12 99.985% 13 99.986% 14 99.986% 15 99.986%
Model: Random Forest Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 94.77% 2 96.82% 3 98.42% 4 99.23% 5 99.45% 6 99.54% 7 99.72% 8 99.77% PCA Components Model Accuracy mean 9 99.89% 10 99.98% 11 99.987% 12 99.998% 13 99.998% 14 99.998% 15 99.998%
Snap Identification O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Three primary snap identification parameter Total Force Feedback Force Error RMS Measured Slip
Snap Results O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S 7/8/2017-27/04/2018 2294 coils K-means clustering algorithm Parameter Value Percentage No Snap 1279 56% Snap 1015 44%
Snap Classification O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Year: 2018 837 coils Hierarchical clustering algorithm Parameter Value Percentage No Snap 599 71.5% Edge Split 76 9.08% Pinch 14 1.06% Scab 22 2.6% Stress 10 1.19% Straight Line 12 1.4% Unknown 83 9.91% Mixed Snap 21 2.5%
Snap Relationship Analysis O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Snap Coil Coil ID Speed Total Force Feedback B02595 320 1250 B02776 310 1550 B02825 240 1230 220 1250 B02833 230 1220 B02838 300 1450 405 1250 200 1500 240 1500 B02846 400 1250 370 1260 190 1500 B02856 310 1550 B99514 310 1550 B99659 320 1550 320 1550 300 1500 300 1500 B99680 320 1550 320 1550 B99690 450 1250 250 1500 B99727 350 1250 Non Snap Coil Coil ID Speed Total Force Feedback B02595 200 1200 B02825 210 1300 125 1250 B02786 110 1250 150 1250 B02774 200 1400 220 1250 B02776 170 1250 190 1450 B02786 110 1100 250 1250 B02833 125 1150 175 1250 B02838 190 1300 260 1550 B02843 220 1250 400 1550 B02846 210 1350 175 1150 B02856 125 1250 150 1450 B99514 150 1250 B99727 190 1150
Recommendatio ns O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Snap Improvement Methods: Mechanical System O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Stress Lubrication Increased force value Roll wear Straight Line Snap Rolling Force Bending Force Lubrication Pinch Lubrication Load Values The gauge is too heavy Material defect Scab Material defect Edge Split Guide vane Roll position Lubrication
Thickness Control Improvement Methods: Control System O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Too much noise in the data Control System Kalman Filter Fuzzy logic Convolutional Neural Network Pictures of different snaps and put them into model to predict the snap.
GUI Data Mining O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Version 2.0 O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
Future F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Trying to find more optimization methods Complete the GUI Prepare and compile the report O V E R V I E W
F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Thank you for your time and patience!

More Related Content

PDF
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W4 Autocorrelation and...
J. García - Verdugo
 
PDF
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W3 Sample Size
J. García - Verdugo
 
PDF
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W2 Statistical Process...
J. García - Verdugo
 
PDF
Deep_Learning__INAF_baroncelli.pdf
asdfasdf214078
 
PDF
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
IJCSES Journal
 
PDF
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
IJCSES Journal
 
PDF
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
ijcseit
 
PPTX
Condition Monitoring of Variable State Machinery
Jordan McBain
 
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W4 Autocorrelation and...
J. García - Verdugo
 
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W3 Sample Size
J. García - Verdugo
 
Javier Garcia - Verdugo Sanchez - Six Sigma Training - W2 Statistical Process...
J. García - Verdugo
 
Deep_Learning__INAF_baroncelli.pdf
asdfasdf214078
 
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
IJCSES Journal
 
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
IJCSES Journal
 
THE IMPLICATION OF STATISTICAL ANALYSIS AND FEATURE ENGINEERING FOR MODEL BUI...
ijcseit
 
Condition Monitoring of Variable State Machinery
Jordan McBain
 

Similar to Improvement of strip thickness control through the process of data analytics (20)

PDF
featurers_Machinelearning___________.pdf
AmirMohamedNabilSale
 
PDF
Machine Learning.pdf
BeyaNasr1
 
PPTX
Agile Experiments in Machine Learning
mathias-brandewinder
 
PPTX
Feature Engineering Fundamentals Explained.pptx
shilpamathur13
 
PDF
Machine learning Mind Map
Ashish Patel
 
PDF
Z suzanne van_den_bosch
Hoopeer Hoopeer
 
PDF
Higgs Boson Challenge
Raouf KESKES
 
PDF
A detailed analysis of the supervised machine Learning Algorithms
NIET Journal of Engineering & Technology (NIETJET)
 
PDF
Agile experiments in Machine Learning with F#
J On The Beach
 
PDF
Grasping dataset
gabrielesisinna
 
DOCX
Deep Learning Vocabulary.docx
jaffarbikat
 
PPTX
Data Engineer’s Lunch #67: Machine Learning - Feature Selection
Anant Corporation
 
PDF
Predictive Maintenance - Predict the Unpredictable
Ivo Andreev
 
PDF
Lecture 10 - Model Testing and Evaluation, a lecture in subject module Statis...
Maninda Edirisooriya
 
PPTX
Data Engineer's Lunch #67: Machine Learning - Feature Selection
Anant Corporation
 
PDF
763354.MIPRO_2015_JovicBrkicBogunovic.pdf
srideviramaraj2
 
PDF
Julien vachaudez - projet Autodiag
Synhera
 
PPTX
CBM Variable Speed Machinery
Jordan McBain
 
PPTX
Leveraging Machine Learning for Accelerated Material Design.pptx
ka4768125
 
PDF
Lecture 8 - Feature Engineering and Optimization, a lecture in subject module...
Maninda Edirisooriya
 
featurers_Machinelearning___________.pdf
AmirMohamedNabilSale
 
Machine Learning.pdf
BeyaNasr1
 
Agile Experiments in Machine Learning
mathias-brandewinder
 
Feature Engineering Fundamentals Explained.pptx
shilpamathur13
 
Machine learning Mind Map
Ashish Patel
 
Z suzanne van_den_bosch
Hoopeer Hoopeer
 
Higgs Boson Challenge
Raouf KESKES
 
A detailed analysis of the supervised machine Learning Algorithms
NIET Journal of Engineering & Technology (NIETJET)
 
Agile experiments in Machine Learning with F#
J On The Beach
 
Grasping dataset
gabrielesisinna
 
Deep Learning Vocabulary.docx
jaffarbikat
 
Data Engineer’s Lunch #67: Machine Learning - Feature Selection
Anant Corporation
 
Predictive Maintenance - Predict the Unpredictable
Ivo Andreev
 
Lecture 10 - Model Testing and Evaluation, a lecture in subject module Statis...
Maninda Edirisooriya
 
Data Engineer's Lunch #67: Machine Learning - Feature Selection
Anant Corporation
 
763354.MIPRO_2015_JovicBrkicBogunovic.pdf
srideviramaraj2
 
Julien vachaudez - projet Autodiag
Synhera
 
CBM Variable Speed Machinery
Jordan McBain
 
Leveraging Machine Learning for Accelerated Material Design.pptx
ka4768125
 
Lecture 8 - Feature Engineering and Optimization, a lecture in subject module...
Maninda Edirisooriya
 
Ad

Recently uploaded (20)

PDF
SMART SIGNAL TIMING FOR URBAN INTERSECTIONS USING REAL-TIME VEHICLE DETECTI...
Niraj Aarya
 
PPTX
Artificial Intelligence
dikshendrasarpate2
 
PDF
737-MAX_SRG.pdf student reference guides
ssusere2119a1
 
PDF
Integrating Fractal Dimension and Time Series Analysis for Optimized Hyperspe...
acijjournal
 
PDF
III.4.1.2_The_Space_Environment.p pdffdf
sittiporn2
 
PPTX
CURRICULAM DESIGN engineering FOR CSE 2025.pptx
Amuda3
 
PDF
86236642-Electric-Loco-Shed.pdf jfkduklg
ddeepakbhai790
 
PDF
Automation-in-Manufacturing-Chapter-Introduction.pdf
pcmth867
 
PDF
PPT on Performance Review to get promotions
prashant593122
 
PDF
Categorization of Factors Affecting Classification Algorithms Selection
IJDKP
 
PDF
Enhancing Cyber Defense Against Zero-Day Attacks using Ensemble Neural Networks
IJCNCJournal
 
PDF
Artificial Superintelligence (ASI) Alliance Vision Paper.pdf
SonaliPatil325517
 
PPTX
MET 305 2019 SCHEME MODULE 2 COMPLETE.pptx
VinayB68
 
PDF
Analyzing Impact of Pakistan Economic Corridor on Import and Export in Pakist...
IJAEMSJORNAL
 
PDF
Mitigating Risks through Effective Management for Enhancing Organizational Pe...
IJAEMSJORNAL
 
PPTX
Fundamentals of Mechanical Engineering.pptx
MdMowdudAhmed
 
PPTX
Nature of X-rays, X- Ray Equipment, Fluoroscopy
DJERALDINAUXILLIAECE
 
PDF
Exploratory_Data_Analysis_Fundamentals.pdf
Ashutosh Satapathy
 
PDF
A SYSTEMATIC REVIEW OF APPLICATIONS IN FRAUD DETECTION
IJNSA Journal
 
PDF
Visual Aids for Exploratory Data Analysis.pdf
Ashutosh Satapathy
 
SMART SIGNAL TIMING FOR URBAN INTERSECTIONS USING REAL-TIME VEHICLE DETECTI...
Niraj Aarya
 
Artificial Intelligence
dikshendrasarpate2
 
737-MAX_SRG.pdf student reference guides
ssusere2119a1
 
Integrating Fractal Dimension and Time Series Analysis for Optimized Hyperspe...
acijjournal
 
III.4.1.2_The_Space_Environment.p pdffdf
sittiporn2
 
CURRICULAM DESIGN engineering FOR CSE 2025.pptx
Amuda3
 
86236642-Electric-Loco-Shed.pdf jfkduklg
ddeepakbhai790
 
Automation-in-Manufacturing-Chapter-Introduction.pdf
pcmth867
 
PPT on Performance Review to get promotions
prashant593122
 
Categorization of Factors Affecting Classification Algorithms Selection
IJDKP
 
Enhancing Cyber Defense Against Zero-Day Attacks using Ensemble Neural Networks
IJCNCJournal
 
Artificial Superintelligence (ASI) Alliance Vision Paper.pdf
SonaliPatil325517
 
MET 305 2019 SCHEME MODULE 2 COMPLETE.pptx
VinayB68
 
Analyzing Impact of Pakistan Economic Corridor on Import and Export in Pakist...
IJAEMSJORNAL
 
Mitigating Risks through Effective Management for Enhancing Organizational Pe...
IJAEMSJORNAL
 
Fundamentals of Mechanical Engineering.pptx
MdMowdudAhmed
 
Nature of X-rays, X- Ray Equipment, Fluoroscopy
DJERALDINAUXILLIAECE
 
Exploratory_Data_Analysis_Fundamentals.pdf
Ashutosh Satapathy
 
A SYSTEMATIC REVIEW OF APPLICATIONS IN FRAUD DETECTION
IJNSA Journal
 
Visual Aids for Exploratory Data Analysis.pdf
Ashutosh Satapathy
 
Ad

Improvement of strip thickness control through the process of data analytics

  • 1. PROJECT UPDATE FOR IMPROVEMENT OF ‘FINAL STRIP THICKNESS CONTROL’ IN HIGH ‘SI ELECTRICAL STEEL’ THROUGH THE USE OF ‘DATA ANALYTICS’ Presentation by: Srinivasa Raghavan Ramanujam F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
  • 2. Objective Current Progress Feature Selection Dimensionality Reduction Model Building Results and Discussion Recommendation Future Work F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
  • 3. Aim To analyse the cold reduction of high Si steel for reducing the thickness by 90% to achieve sharp magnetic losses which are proportional to the final strip thickness through the use of data analytics. F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W
  • 4. Dissertation Plan F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Topic Weeks 1 2 3 4 5 6 7 8 9 10 11 12 Initial Proposal Feature Selection in Weka Pre-processing Dataset Model Building in Python Analysing Results Current Position Compiling Report Writing
  • 5. Project Approach F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W •Analyzing the generalized data •Identifying the base of the problem Top down approach •Dividing into test and train sets •Feature Scaling Data pre- processing •Identify the relation between different attribute Feature Selection •Dimensionality ReductionPCA •Build the correct algorithm using python •Predict the results Model Building Visualize Results
  • 6. Current Update F E A T U R E S E L E C T I O N M O D E L B U I L D I N G P R E - P R O C E S S I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Update Check Exporting the IBA data into a csv format Trying to find a relationship between different attribute Develop various mathematical model to predict the snap Working on a GUI to perform data mining using python
  • 7. Feature Selection O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 8. Problem: Unclean Data O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Fluctuations when velocity is zero
  • 9. Solution: Removing Zeros O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 10. Feature Selection: Weka O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S It is the process of selecting a subset of relevant features (variables, predictors) for use in model construction. Feature selection techniques are used for four reasons: Simplification of models to make them easier to interpret by researchers/users, Shorter training times, To avoid the curse of dimensionality, Enhanced generalization by reducing overfitting Point where snap happens
  • 11. Feature Selection: Weka O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Performed for snap and non snap coils 727 Attributes 30 coils Six Different Models BestFirst+ Cfs Subset Eval Greedy StepWise+ Cfs Subset Eval Greedy StepWise + WrapperSubset Eval Ranker + Principal Components Ranker+ ReliefAttribute Eval Ranker + CorrelationAttribute Eval
  • 12. Feature Selection Result: Entire Coil O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Right gauge coarse deviation Feedback Error Single Shot Gauge Deviation Delayed Raw Mass flow Gauge Error Right Gauge Deviation Component Fit Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back_Capsule_Titt Calc mill speed WR circ
  • 13. Feature Selection Result: Snap Time Data O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Feedback Error Single Shot Gauge Deviation Right Gauge Deviation Component Fit Mass Flow Exit Gauge Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back_Capsule_Titt Calc mill speed WR circ
  • 14. Feature Selection Result: Non Snap Selection Coil O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Entry Strip Speed Exit Strip Speed Entry Gauge Deviation Exit Gauge Deviation Total Force Feedback Mill Speed Right gauge coarse deviation Feedback Error Single Shot Gauge Deviation Delayed Raw Mass flow Gauge Error Left Gauge Deviation Component Fit Right Gauge Deviation Component Fit Force Error RMS Back Capsule Force Servo Front Load Servo Back Load Measured Slip Back Capsule Titt Calc mill speed WR circ
  • 15. Data Pre-processing O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 16. Data Pre-processing O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Four steps Importing the Libraries Importing the datasets Splitting into test and train sets Feature Scaling Dimensionality Reduction
  • 17. Dimensionality Reduction O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Principal Component Analysis Unsupervised Mode of Dimensionality Reduction Statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal componentsSplitting into test and train sets
  • 18. Model Building O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 19. Model: Multiple Linear Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Multiple linear regression attempts to model the relationship between two or more explanatory variables and a response variable by fitting a linear equation to observed data. Every value of the independent variable x is associated with a value of the dependent variable y
  • 20. Model: Random Forest Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S It is unexcelled in accuracy among current algorithms. It runs efficiently on large databases. It can deal with a large number of information factors without variable erasure
  • 21. Model: Decision Tree Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Decision tree builds regression or classification models in the form of a tree structure. It breaks down a dataset into smaller and smaller subsets while at the same time an associated decision tree is incrementally developed. The final result is a tree with decision nodes and leaf nodes
  • 22. Model Results O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 23. O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 7.58% 2 37.62% 3 38.54% 4 40.98% 5 75.14% 6 80.61% 7 80.81% 8 82.66% PCA Components Model Accuracy mean 9 82.76% 10 82.76% 11 82.76% 12 82.76% 13 82.76% 14 82.76% 15 82.76% Model: Multiple Linear Regression
  • 24. Model: Decision Tree Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 92.95% 2 99.42% 3 99.88% 4 99.952% 5 99.96% 6 99.97% 7 99.97% 8 99.97% PCA Components Model Accuracy mean 9 99.98% 10 99.98% 11 99.983% 12 99.985% 13 99.986% 14 99.986% 15 99.986%
  • 25. Model: Random Forest Regression O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S PCA Components Model Accuracy mean 1 94.77% 2 96.82% 3 98.42% 4 99.23% 5 99.45% 6 99.54% 7 99.72% 8 99.77% PCA Components Model Accuracy mean 9 99.89% 10 99.98% 11 99.987% 12 99.998% 13 99.998% 14 99.998% 15 99.998%
  • 26. Snap Identification O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Three primary snap identification parameter Total Force Feedback Force Error RMS Measured Slip
  • 27. Snap Results O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S 7/8/2017-27/04/2018 2294 coils K-means clustering algorithm Parameter Value Percentage No Snap 1279 56% Snap 1015 44%
  • 28. Snap Classification O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Year: 2018 837 coils Hierarchical clustering algorithm Parameter Value Percentage No Snap 599 71.5% Edge Split 76 9.08% Pinch 14 1.06% Scab 22 2.6% Stress 10 1.19% Straight Line 12 1.4% Unknown 83 9.91% Mixed Snap 21 2.5%
  • 29. Snap Relationship Analysis O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 30. O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Snap Coil Coil ID Speed Total Force Feedback B02595 320 1250 B02776 310 1550 B02825 240 1230 220 1250 B02833 230 1220 B02838 300 1450 405 1250 200 1500 240 1500 B02846 400 1250 370 1260 190 1500 B02856 310 1550 B99514 310 1550 B99659 320 1550 320 1550 300 1500 300 1500 B99680 320 1550 320 1550 B99690 450 1250 250 1500 B99727 350 1250 Non Snap Coil Coil ID Speed Total Force Feedback B02595 200 1200 B02825 210 1300 125 1250 B02786 110 1250 150 1250 B02774 200 1400 220 1250 B02776 170 1250 190 1450 B02786 110 1100 250 1250 B02833 125 1150 175 1250 B02838 190 1300 260 1550 B02843 220 1250 400 1550 B02846 210 1350 175 1150 B02856 125 1250 150 1450 B99514 150 1250 B99727 190 1150
  • 31. Recommendatio ns O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 32. Snap Improvement Methods: Mechanical System O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Stress Lubrication Increased force value Roll wear Straight Line Snap Rolling Force Bending Force Lubrication Pinch Lubrication Load Values The gauge is too heavy Material defect Scab Material defect Edge Split Guide vane Roll position Lubrication
  • 33. Thickness Control Improvement Methods: Control System O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Too much noise in the data Control System Kalman Filter Fuzzy logic Convolutional Neural Network Pictures of different snaps and put them into model to predict the snap.
  • 34. GUI Data Mining O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 35. Version 2.0 O V E R V I E W F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S
  • 36. Future F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S Trying to find more optimization methods Complete the GUI Prepare and compile the report O V E R V I E W
  • 37. F E A T U R E S E L E C T I O N P R E - P R O C E S S I N G M O D E L B U I L D I N G R E S U L T S R E C O M M E N D A T I O N S O V E R V I E W Thank you for your time and patience!