Implemented Innovation on Predictive Analysis Method for...
Transcript of Implemented Innovation on Predictive Analysis Method for...
1SlideTata SteelTata Innovista
ROLLING MILL OPERATING COMMITTEE - 2019
Implemented Innovation on Predictive Analysis Method
for Enhancing Productivity- A Multi Dimensional Approach
Presented By: Pankaj Kumar
New Bar Mil, Tata Steel
Date: 13 February 2019
2SlideTata SteelTata Innovista2
D
Content
Challenges & Overcoming Them
A Trigger for Innovation
B
C Journey of continuous improvementC
D Journey of continuous improvementD
Specific areas of focus – target 750+
Implementation
TBEM 2018
Confirmation of Effects and Way ForwardE
Novelty / Uniqueness
3SlideTata SteelTata Innovista
A. Trigger for Innovation
R Factor = Area(in)
Area(out)
Incorrect R Factor may lead to:
1. Tension or Compression which
can cause cobble in mill
2. Cobble in mill
3. Abnormal load on mill stands
4SlideTata SteelTata Innovista
B. Challenges and Overcoming Them (1/2)
Challenges with R Factor Control:
• Operator has to take decision by monitoring thousands of Data Points every hour.
• Operator adjusts R Factor based on his best judgement.
• While adjusting the potential of a wrong decision is as high as 13%. .
Correct Decision
Incorrect Decision
OEM has no solution available for this problem.
5SlideTata SteelTata Innovista
B. Challenges and Overcoming Them (2/2)
Processing of millions of data Institutionalize Human Knowledge is
key to the problem resolution
Data Analytics was used to
generate meaningful solution
Capturing Operator
experience and
defining Business
Rule
Formation of 18 Dimensional Matrix
through Multivariate Analytics
Determining the regions of high purity using
K Means Clustering
Plane drawing to segregate the good and
bad clusters through Logistic Regression
6SlideTata SteelTata Innovista
C. Implementation
7SlideTata SteelTata Innovista
D. Novelty / Uniqueness
R Factor
Cluster
Operator
independent
Online
visualization
• 1st in world to use R Factor clusters for mill setting.
• Option to choose from 5 different clusters for every section
• Similar R Factor settings in all shifts
• Quantitative capturing of Operator knowledge
• Online R Factor trend available
• Mill settings done based on R Factor
• Data categorized in green-yellow-red based on deviation
from selected cluster
8SlideTata SteelTata Innovista
E. Confirmation of Effects and Way Forward
• Before : on horizontal axis includes delays due to abnormalities in R
factor during the period H2 FY’18 for which campaign wise data is
taken to train the model.
• After : on horizontal axis indicates the delay due to abnormal R
factors over a period H1 Fy’19 which is seen to decline by 34% when
calculated as avg delay in min per day during the period.
• This would lead to a reduction of delay by 7.7 hrs per month which
would lead to incremental savings of Rs 3.22 Cr Annually
1. Inter-stand head tracking time difference
2. Mill Shear (S8 and S16) intelligent monitoring
3. Looper actuation tracking
4. Loop Height abnormality detection
9SlideTata SteelTata Innovista
Thank You
10SlideTata SteelTata Innovista
New Bar Mill: Process Flow
COOLING BED
WALKING BEAMREHEATING FURNACE
ROUGHING MILLSTANDS #1 - #8
INTERMEDIATE MILL #1STANDS #9 - #12 INTERMEDIATE MILL #2
STANDS #13 - #16
CROPPING AND DIVIDING SHEAR
NO TWIST MILLSTANDS #17 - #22
WATER BOX #1, #2
LINE A
LINE B
LINE A
LINE B
BRAKING PINCH ROLLS
COLD SHEAR
STRAPPING MACHINES #1 - #7WEIGHING AND UNLOADING STATION #2
WEIGHING AND UNLOADING STATION #1
BILLET CHARGING BED
11SlideTata SteelTata Innovista
Praveen Thampi
NBM, Analytics Project
Manager
Abhishek Raj
NBM, Business Translator
Pankaj Kumar
NBM, Business
Translator
Sushil Kumar Tripathy
NBM, Operations
Translator
Rahul Anand
Data Scientist
Abhimanyu Kumar Singh
ITS, Data Engineer
Sanjeev Kumar
Automation, Data Architect
Rahul Anand
Visualization Designer
Abhimanyu Kumar Singh
ITS, Visualization Designer