This presentation on batch process analytics was given at Emerson Exchange, 2010. A overview of batch data analytics is presented and information provided on a field trail of on-line batch data analytics at the Lubrizol, Rouen, France plant.
1. Batch Process Analytics
- update -
Robert Wojewodka, Technology Manager and Statistician
Terry Blevins, Principal Technologist
Willy Wojsznis, Senior Technologist
3. 3
Introduction
Lubrizol and Emerson Process Management
have worked together over the last three years
to develop and install a beta version of
Emerson’s on-line batch analytics
This new functionality is currently in operation
after successful field trials at the Lubrizol,
Rouen, France plant
In this session we will present the lessons learned in implementation of
this technology in a running plant
We will also summarize the results achieved by the process operators
and operations management using this new capability
We outline the basic principles and objectives of analytic application
We sketch some innovative analytic concepts which were validated at
the field trial
Discuss current activities
4. 4
• Operators and engineers work in a highly complex, highly
correlated and dynamic environment each day
• Operators and engineers manage a large amount of data
and information on a running unit
• Operators and engineers need to avoid undesirable
operating conditions
• Operators an engineers need to reduce variation, improve
throughput and improve quality yet maintain safety
The Setting
5. 5
Jointly develop viable on-line multivariate batch
process data analytics
The primary objectives of the field trial were:
– Demonstrate on-line prediction of product quality
– Evaluate different means of on-line process fault
detection and identification; abnormal situations
Document the benefits of this technology
Learn from the field trial to update and improve these
new and evolving modules
Objectives of the Beta TestObjectives of the Beta Test
6. 6
Process holdups
Access to lab data
Variations in feedstock
Varying operating conditions
Concurrent batches
Assembly and organization of the data
Challenges in Applying Online Data
Analytics to Batch Processes
7. 7
Functionality of the Analytics Application
Take all inputs and process variables associated with a batch
process and characterize “acceptable variation” and process
relationships associated with “good” batches
Identify how these variables relate to each other and to end of
batch product quality characteristics
Use the analytic techniques to identify typical process
relationships and faults as current and future batches are running
Use the analytic techniques to predict end of batch quality
characteristics at any point in time as a batch is evolving
Identify and diagnose faults and provide recommendations to
operations personnel how to improve batch operation and product
quality
8. 8
The “Golden Batch” comparison approach is
plagued with problems
What is the “best” batch?
Only refers to ONE batch; but there are many
“good” batches
Does not address fact that variation exists nor does
it address defining an acceptable level of variability
May significantly miss direct resources
May significantly miss direct control emphasis
Does not promote process understanding nor does
it promote identifying important process
relationships; nothing is learned
The economics may be completely wrong
Does not promote identification and control of
critical parameters and relationships with quality
parameters (analytical, physicals, time cycle, yield,
waste, economics, etc.)
…and the list goes on…
Golden
Batch
Comparison
9. 9
Analytics Drive the Power of Information
The Power of
Information
Raw Data
Standard
Reports
Descriptive
Modeling
Predictive
Modeling
Data Information Knowledge Intelligence
Optimization
What happened?
Why did it happen?
What will happen?
What is the
Best that
could happen?
$$$
ROI
$$$
ROI
Adapted by Bob Wojewodka from slide courtesy of SAS Inst.
Ad hoc Reports
& OLAP
10. 10
services
SAP Process Order
& Recipe
Consumption
Data
Firewall
Resource
Optimization and
Planning Application
Batch Exec &
Campaign Mgr
Historian &
Recipe Exchange
PRO+
Operator Interface
Recipe Transfer
via XML
Consumption from
Batch Historian
event file via XML
Control Network
LZ Domain
SAP Analysis server(s)
Analysts
Embedded
analytics
Device level analysis /
diagnostics
Device level analysis/
diagnostics
Embedded
analysis &
diagnostic apps.
Embedded
analysis and
diagnostic apps.
Business &
Process Analytics
Business and
process analytics
Data Transfer
via XML
Pro+
.net
Web services
Batch exec.
Consumption
SAP®
Data historian
Operator interface
Data transfer
Analysis servers
Analysts
chemists
engineers
Embedded
analysis
XML
Recipe +
schedule
DeltaV and SAP Integration With Data Analytics
Statgraphics®
11. 11
Summary of Actual Field Trial Analyses
2 units / products
18 input variables
38 process
variables
4 output variables
(2 initially for the online)
All data at 1-minute time intervals for the analysis
Total of 172 historical batches used for analysis and
model development across these two processes
12. 12
PCA – Principal Components Analysis
– Provides a concise overview of a data set. It is powerful for
recognizing patterns in data: outliers, trends, groups,
relationships, etc.
PLS – Projections to Latent Structures
– The aim is to establish relationships between input and
output variables and developing predictive models of a
process.
PLS-DA – PLS with Discriminant Analysis
– When coupled, is powerful for classification. The aim is to
create predictive models of the process but where one can
accurately classify the material into a category.
The Primary Multivariate Methods
19. 19
Web-based Interface - There’s an App for that
Since the user interface is
web-based it can be
accessed from multiple sites
over the intranet (or internet)
As will be demonstrated at
the Rouen beta site, access
is also available through an
iPod Touch or iPhone.
20. 20
Summary of field trial results
Operators and engineers at Rouen are using these new tools for
faults detection and quality parameter prediction.
The impact of the on-line analytic tools installed at Rouen on the
plant operation have been evaluated over a 6 month period and
since then the installation is in use beyond the initially planned
period.
Examples of faults detected using this capability are provided in the
presentation given at Emerson Exchange 2009 – see Benefits
Achieved Using On-Line Data Analytics by Robert Wojewodka
and Terry Blevins.
21. 21
Lessons Learned - Key concepts / approaches
that have evolved from the beta work
Use of Stage in data analytics to define the major
manufacturing steps
Selection and pre-processing of data used for
model development and on-line analytics
On-line interface designed to meet operator’s
requirements
Web based architecture for operator interface and
data exchange
Development of a web based dynamic process
simulation to enable effective operator training
22. 22
Current Activities
Emerson progressing with the commercialization of the batch
analytics modules
– Will be part of the DeltaV Version 12 release
Following process improvement design changes on the field
trial units, models will be updated and redeployed
Completion of a Design of Experiments to further
characterize the modeling process relative to differing
process relationships
23. 23
Design Of Experiments
Examining more process relationships
and impacts on the analysis methods
Results will be used to further refine the
modeling approach
Results will be used for pre-assessment
of candidate units for use of the analysis
modules
24. 24
Added Work Prior to Commercial
Release
Off-line modeling tool set with enhanced diagnostics to
aid the process engineers during model development
steps
Ability to simultaneously predict multiple “Y” output
variables while on-line
On-line diagnostics of the “health” of the running
models; alert when model errors deviate beyond initial
levels when deployed
Additional functionality for being able to update and
redeploy models quickly following processing changes
25. 25
Where to Get More Information
Interactive demonstration of data analytics applied to the saline process
http://207.71.50.196/AnalyticsOverview.aspx
Robert Wojewodka and Terry Blevins, “Data Analytics in Batch Operations,” Control, May 2008
Video: Robert Wojewodka, Philippe Moro, Terry Blevins Emerson - Lubrizol Beta:
http://www.controlglobal.com/articles/2007/321.html
Emerson Exchange 2010 Workshop – SAP to DeltaV integration using the DeltaV SOA Gateway and SAP Web
Services – Philippe Moro, Joe Edwards, Chris Felts
Emerson Exchange 2009 Workshop – Benefits Achieved Using On-line Data Analytics - Robert Wojewodka,
Terry Blevins
Emerson Exchange 2008 Short Course: 366 – The Application of Data Analytics in Batch Operations - Robert
Wojewodka, Terry Blevins
Emerson Exchange 2008 Short Course: 364 – Process Analytics In Depth - Robert Wojewodka, Willy Wojsznis
Emerson Exchange 2008 Workshop: 367 – Tools for Online Analytics - Michel Lefrancois, Randy Reiss
Emerson Exchange 2008 Workshop: 412 – Integration of SAP®
Software into DeltaV - Philippe Moro, Chris
Worek
Emerson Exchange 2007 Workshop: 686 – Coupling Process Control Systems and Process Analytics to
Improve Batch Operations – Bob Wojewodka, Philippe Moro, Terry Blevins
26. 26
Thank You
Q & A
Vision without action is merely a dream.
Action without vision just passes the time.
Vision with action can change the world.
--- Joel Barker, Futurist
Hinweis der Redaktion
For beginning with I will let Bob explain to you the power of Information.
What is the purpose of my work, here and in France ?
For having a great business all company need to analyze what’s happened inside.
For that, Process system provide Data. Mixing data.
The company can after edit Standard Report, for modeling this data. Next step for a better presentation of data is Ad hoc Reports & OLAP.
At this state we know what happened ? Data are become information.
But for data become efficient for increasing the profitability of the company, we need to continue analysis.
For knowing what did it happen by descriptive modeling. What will happen with Predictive modeling and to finish what is the best that could happen ? This step is the Optimization. As this state, data are become from information to knowledge to Intelligence. We have able to find key for increasing the potential of the company.
Lubrizol want to improve this part.
OMS Phase I and II are working on this : How data can become Intelligence. How is the best way to use Lubrizol Data.
OMS Phase II enables us to move to expand our data analysis capabilities. Therefore OMS Phase II is an enabler.
This slide is a bit more busy…
…but it depicts some of the next steps we are starting to transition to into 2006.
Phase II of our integration work is to bridge the various data streams and to truly analyze and optimize our manufacturing processes.
There are 3 layers of data and data analytics that we see… …describe these… The newer buzz term out there is PAT.
In Phase II of our work activities, we will be working with vendors and bridging what ever gaps exist ourselves to automate the extraction and organization of data.
We will be moving the organization further from a reporting and trending mindset to a process and business data analysis mindset.