SlideShare a Scribd company logo
1 of 30
Utilizing DeltaV to Perform
PAT Calculations Real Time
Real-Time Non-Linear Regression
Chromatography Endpoint Detection
Photography & Video Recording Policy
Photography and audio/video recording is not permitted in any sessions or in the
exhibition areas without press credentials or written permission from the Emerson
Exchange Board of Directors. Inquiries should be directed to:
EmersonExchange@Emerson.com
Thank you.
Presenters
 Michael Hausladen
Director of Manufacturing
Technology, BMS
 Paul Brodbeck
Engineering Manager,
Control Associates, LBP
Introduction
 Chromatography Project
 Functional Requirements
 Chromatography Elution
 Modeling
 Prototyping
 Implementation
 Results
 Summary
Chromatography Project
 BMS Syracuse
 Scientist: Mike Hausladen
 Pilot Plant Chromatography Skid Modification
 Purpose:
– Scale-up laboratory endpoint detection
– Ensure process robustness
• Minimize incorrectly determined collection end-point
– Pilot scale model of full scale production system
– Demonstrate capability for full scale production
 Focus of this presentation: Robust chromatography
elution end-point determination
Chromatography Basics
The basics of bind and elute chromatography:
– Modify the conditions of the mobile phase to cause binding
or elution of the product (protein)
– Aqueous systems: pH and conductivity
Stationary
Phase
Mobile phase:
1. Flush and prep column
2. Load protein on column
3. Wash
4. Elute protein
5. Clean and sanitize column
6. Store column
Monitor elution for product fraction
• UV adsorbance @ 280nm
• Collect the desired portion
of the elution peak
Elution Curve
0
10
20
30
40
50
60
70
80
90
100
mAU
Worse case chromatography
generated in the lab. Data is
not smooth., has multiple peaks.
z
w
h
The whole point – determine h (peak height) real time – to
calculate percent of peak maximum collection end-point
Percent of peak max (h)
stop collecting product
BMS Requirements
 Robust endpoint of Elution Determination - real time
 Peak Maximum Calculation
– Endpoint = Percent of Peak Maximum
• Model Predict Peak Max. of absolute optical density.
• Percent of Peak Max lookup table of Sialic Acid vs. load material.
 3 Models – in DeltaV Controller
– Smoothing (1st Order Filter) Model
– Linear Regression Algorithm - Polynomial Fit
– Non-Linear Model
• Extreme Value Function fit with Real-time Data
– Column Volume (mobile phase volume) vs. UV absorbance
 Alarming/Auto-Switching
– Limits to ensure that the Peak Maximum is not determined early.
– Ensure data is fitting the model with a sufficient level of accuracy.
– Limits end Model if the algorithm is not converging.
 Focus of this presentation is on the non-linear model real-time fit of the Extreme
Value Function
Implemented
 3 Models
– Smoothing Model
– Polynomial Fit
– Non-Linear Model
 1 Algorithm
– Newton-Raphson
System Integrator Process
 Select Model Equation
 Basic Curve Fitting
– Least Squares Error (LSE)
 Select Numerical Method Algorithm
– Gradient Descent
– Newton Raphson
– Levenberg-Marquardt
 Machine Learning Parallel
 Grey Box Modeling
 Math
– Linear Algebra
– Vector Based Programming Languages
(MATLAB)
– Solving of Partial Differential Equations
 Research
– Internet Literature
– Example Programs
– Textbook – Numerical Methods by Dahlquest &
Bjorck
 Program Construction
– Flowchart
 DeltaV CALC Block Limitation
– 2000 lines per scan
– 64 If-Then loops per block
– 256 field arrays
– Dynamic Reference delays
 Excel Solver
– Troubleshooting
– Math Checking
– Convergence Issues
 Switch Gradient to Newton
– Bad Hardware
– Convergence Issues
 Final Program
 Model Statistics
 IQ/OQ
 Levenberg-Marquart Future
 Numerical Integration
 Non-linear Models
 Controller vs. Application Station
 Other Approaches
– synTQ, MATLAB, SoftPhase
 Summary
Non-Linear Model
 Extreme Value Function: is an equation that
approximates a chromatography peak
 Using a dataset of x’s (volume of mobile phase) and y’s
(elution response – UV adsorbance )
– Determine h, w, and z that give the best approximation of the peak
– From h, determine endpoint
 There are many mathematical functions for the representation
of chromatographic peaks, extreme value function chosen for
simplicity (Journal of Chromatography)
Elution Curve Fitted
0
10
20
30
40
50
60
70
80
90
100
mAU
fit
z
w
h
The whole point – determine h (peak height) real time – to
calculate percent of peak maximum collection end-point
Percent of peak max –
stop collecting product
Numerical Method – Basic Curve Fitting
 Minimize the Squares of the Error
– Least Squares Error
 Curve fitting, minimizing error, finding best solution/ best fit
 Analytical solutions for simple fit
 Iterative numerical solution required for complex equations
– Initial guess required
– Convergence satisfied?
 Solving in multidimensional space
Numerical Methods – Solving Least Squares
 Gradient Descent
– Make Initial Guess Xi-1
– New Guess: Xi = Xi-1 – e * F’(Xi-1)
– e is tuning constant
• Too low = slow convergence, Too high = unstable
 Newton - Raphson
– Make Initial Guess Xi-1
– New Guess: Xi = Xi-1 - F’(xi-1)/2*(F’’(xi-1))
– Can be unstable with a poor initial guess
 Levenberg-Marquardt
– Start with Gradient, end with Newton
 Iteratively repeat, check for
convergence
SquareofError
Math
 In order to mathematically determine the slopes of a
multi-parameter system need to calculate the partial
differentials.
– Partials are slopes in n-dimensional space
 Linear Algebra
– When solving multivariate we need vectors, arrays or
matrices.
– Solutions become complex when dividing matrices
 Vector Based Programming
– Matlab, Python, Octave, …
Partial Differentials – 3D Slopes
Solving the Extreme Value Function
 Need to solve for h, z, w.
– h = maximum peak height.
– w = peak width.
– z = retention time of column.
 Vectorization Required
 Challenge here is dividing by a
Matrix
 Complex Linear Algebra
 Determinants
 Easy in Matlab
 Not so Easy Otherwise
Research
 Internet Literature
– Methods for Non-Linear Least Squares Problems1
• Nomenclature difficult
• Theoretical
– Wikipedia
– Dr. Math
• Least Squares Regression for Quadratic Curve Fitting2
– Example Programs
• L-A Algorithm by Pradit Mittrapiyanuruck
– MATLAB
• L-M Method for Non-Linear Least Squares by Henri Gavin
– MATLAB, extra code… weighting, CHI, R^2, lamda
 Textbook – Numerical Methods by Dahlquist & Bjorck
– 5.2.1 Numerical Linear Algebra
– 5.6 Iterative Methods
– 6.9.2 Newton-Raphson’s Method…
– 10.5.1 Non-Linear Optimization Problems (Hessian)
Project Challenges
DeltaV Limitations Other Obstacles
 No matrix math functions
 2000 lines per scan
 64 loops per scan
 Dynamic reference time
delay
 256 cell arrays max
 Hardware problem
 Literature for Matlab
 Solving Diff. Eq.’s
 Notation theory
 Matrix Algebra
 Numerical Methods
 Least Squared Error
Program Flowchart
Program – DeltaV CALC Block
Math – Solving PDE’s
 Solving PDE’s
Prototyping in MS Excel
h = 15
z = 0.24
w = 270
Model Performance
 Trust Zone
 5 Second scan time
 1000 pts each scan
 Model converges to solution each scan
 40% MX Controller Usage @ 5 sec scan
 90% MX Controller Usage @ 1 sec scan
IQ/OQ Validation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
6 6.5 7 7.5 8 8.5 9 9.5 10
Raw Data
Excel Prediction
DeltaV NonLinear
DeltaV Linear
Levenberg-Marqaudt Future Enhancement
 Prototyped in Matlab
 Tested in DeltaV
 Ready to be upgraded
 Faster, more robust
Implemented Models
 All 3 running simultaneously
 Auto Switch to best Model
 Non-Linear Model is primary
 Polynomial Model is secondary
 Smoothing Model last
 All 3 models tested within X% of each
 Monitor all 3 and Alarm if > X% Error
Other Options for Future
 MATLAB
– OPC communication
– Runs in App station
 synTQ
– $$$
– MATLAB or other math package
 Soft Phase
– Requires C, C++, C#, or VB
 Controller vs. Application Station
– Controller slightly more robust
• If App Station goes down Model still runs in controller
– Application Station
• Run at higher speeds
• Reduced controller memory & capacity
Summary DeltaV
 Non-linear was successfully implemented to predict
elution endpoint
 Model updates 1000pt every 5 seconds
 Model is capable of handling model non-convergence
 Model passed IQ/OQ
Thank You for Attending!
Enjoy the rest of the conference.

More Related Content

What's hot

Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...
Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...
Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...Brati Sundar Nanda
 
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...sipij
 
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...Youness Lahdili
 
Low power vlsi implementation adaptive noise cancellor based on least means s...
Low power vlsi implementation adaptive noise cancellor based on least means s...Low power vlsi implementation adaptive noise cancellor based on least means s...
Low power vlsi implementation adaptive noise cancellor based on least means s...shaik chand basha
 
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...Raj Kumar Thenua
 
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...An Efficient DSP Based Implementation of a Fast Convolution Approach with non...
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...a3labdsp
 
Using FME for Topographical Data Generalization at Natural Resources Canada
Using FME for Topographical Data Generalization at Natural Resources CanadaUsing FME for Topographical Data Generalization at Natural Resources Canada
Using FME for Topographical Data Generalization at Natural Resources CanadaSafe Software
 
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...Databricks
 
Using capability assessment during product design
Using capability assessment during product designUsing capability assessment during product design
Using capability assessment during product designMark Turner CRP
 
Implementation Adaptive Noise Canceler
Implementation Adaptive Noise Canceler Implementation Adaptive Noise Canceler
Implementation Adaptive Noise Canceler Akshatha suresh
 

What's hot (20)

Dct and adaptive filters
Dct and adaptive filtersDct and adaptive filters
Dct and adaptive filters
 
Adaptive filter
Adaptive filterAdaptive filter
Adaptive filter
 
Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...
Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...
Noice canclellation using adaptive filters with adpative algorithms(LMS,NLMS,...
 
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...
PERFORMANCE ANALYIS OF LMS ADAPTIVE FIR FILTER AND RLS ADAPTIVE FIR FILTER FO...
 
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...
4 - Simulation and analysis of different DCT techniques on MATLAB (presented ...
 
Low power vlsi implementation adaptive noise cancellor based on least means s...
Low power vlsi implementation adaptive noise cancellor based on least means s...Low power vlsi implementation adaptive noise cancellor based on least means s...
Low power vlsi implementation adaptive noise cancellor based on least means s...
 
Aca11 bk2 ch9
Aca11 bk2 ch9Aca11 bk2 ch9
Aca11 bk2 ch9
 
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...
Simulation and hardware implementation of Adaptive algorithms on tms320 c6713...
 
Introduction to Adaptive filters
Introduction to Adaptive filtersIntroduction to Adaptive filters
Introduction to Adaptive filters
 
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...An Efficient DSP Based Implementation of a Fast Convolution Approach with non...
An Efficient DSP Based Implementation of a Fast Convolution Approach with non...
 
Adaptive filter
Adaptive filterAdaptive filter
Adaptive filter
 
Using FME for Topographical Data Generalization at Natural Resources Canada
Using FME for Topographical Data Generalization at Natural Resources CanadaUsing FME for Topographical Data Generalization at Natural Resources Canada
Using FME for Topographical Data Generalization at Natural Resources Canada
 
ADAPTIVE NOISE CANCELLATION
ADAPTIVE NOISE CANCELLATIONADAPTIVE NOISE CANCELLATION
ADAPTIVE NOISE CANCELLATION
 
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...
Time Series Forecasting Using Recurrent Neural Network and Vector Autoregress...
 
Ch1
Ch1Ch1
Ch1
 
Av 738 - Adaptive Filtering Lecture 1 - Introduction
Av 738 - Adaptive Filtering Lecture 1 - IntroductionAv 738 - Adaptive Filtering Lecture 1 - Introduction
Av 738 - Adaptive Filtering Lecture 1 - Introduction
 
Lecture 09: SLAM
Lecture 09: SLAMLecture 09: SLAM
Lecture 09: SLAM
 
ANCLMS
ANCLMSANCLMS
ANCLMS
 
Using capability assessment during product design
Using capability assessment during product designUsing capability assessment during product design
Using capability assessment during product design
 
Implementation Adaptive Noise Canceler
Implementation Adaptive Noise Canceler Implementation Adaptive Noise Canceler
Implementation Adaptive Noise Canceler
 

Similar to EGUE Technikrom Final_8_12_13

Heuristic design of experiments w meta gradient search
Heuristic design of experiments w meta gradient searchHeuristic design of experiments w meta gradient search
Heuristic design of experiments w meta gradient searchGreg Makowski
 
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...Tom Hubregtsen
 
PAT Process Control IFPAC 2013
PAT Process Control IFPAC 2013PAT Process Control IFPAC 2013
PAT Process Control IFPAC 2013Paul Brodbeck
 
My Postdoctoral Research
My Postdoctoral ResearchMy Postdoctoral Research
My Postdoctoral ResearchPo-Ting Wu
 
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...maikelcorleoni
 
Nafems15 systeme
Nafems15 systemeNafems15 systeme
Nafems15 systemeSDTools
 
Nafems15 Technical meeting on system modeling
Nafems15 Technical meeting on system modelingNafems15 Technical meeting on system modeling
Nafems15 Technical meeting on system modelingSDTools
 
An Approach to Overcome Modeling Inaccuracies for Performance Simulation Sig...
An Approach to Overcome Modeling  Inaccuracies for Performance Simulation Sig...An Approach to Overcome Modeling  Inaccuracies for Performance Simulation Sig...
An Approach to Overcome Modeling Inaccuracies for Performance Simulation Sig...Pankaj Singh
 
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...A Framework for Scene Recognition Using Convolutional Neural Network as Featu...
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...Tahmid Abtahi
 
Making Custom Oscilloscope Measurements
Making Custom Oscilloscope MeasurementsMaking Custom Oscilloscope Measurements
Making Custom Oscilloscope Measurementsteledynelecroy
 
Fast optimization intevacoct6_3final
Fast optimization intevacoct6_3finalFast optimization intevacoct6_3final
Fast optimization intevacoct6_3finaleArtius, Inc.
 
Biomedical Signal and Image Analytics using MATLAB
Biomedical Signal and Image Analytics using MATLABBiomedical Signal and Image Analytics using MATLAB
Biomedical Signal and Image Analytics using MATLABCodeOps Technologies LLP
 
Intro to LV in 3 Hours for Control and Sim 8_5.pptx
Intro to LV in 3 Hours for Control and Sim 8_5.pptxIntro to LV in 3 Hours for Control and Sim 8_5.pptx
Intro to LV in 3 Hours for Control and Sim 8_5.pptxDeepakJangid87
 
Universal Adiabatic Quantum Computer v1.0
Universal Adiabatic Quantum Computer v1.0Universal Adiabatic Quantum Computer v1.0
Universal Adiabatic Quantum Computer v1.0Aditya Yadav
 
Virus, Vaccines, Genes and Quantum - 2020-06-18
Virus, Vaccines, Genes and Quantum - 2020-06-18Virus, Vaccines, Genes and Quantum - 2020-06-18
Virus, Vaccines, Genes and Quantum - 2020-06-18Aritra Sarkar
 
Probability and random processes project based learning template.pdf
Probability and random processes project based learning template.pdfProbability and random processes project based learning template.pdf
Probability and random processes project based learning template.pdfVedant Srivastava
 
Two methods for optimising cognitive model parameters
Two methods for optimising cognitive model parametersTwo methods for optimising cognitive model parameters
Two methods for optimising cognitive model parametersUniversity of Huddersfield
 

Similar to EGUE Technikrom Final_8_12_13 (20)

Heuristic design of experiments w meta gradient search
Heuristic design of experiments w meta gradient searchHeuristic design of experiments w meta gradient search
Heuristic design of experiments w meta gradient search
 
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...
Aritra Sarkar - Search and Optimisation Algorithms for Genomics on Quantum Ac...
 
PAT Process Control IFPAC 2013
PAT Process Control IFPAC 2013PAT Process Control IFPAC 2013
PAT Process Control IFPAC 2013
 
My Postdoctoral Research
My Postdoctoral ResearchMy Postdoctoral Research
My Postdoctoral Research
 
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...
Multi-Period Integer Portfolio Optimization Using a Quantum Annealer (Present...
 
Nafems15 systeme
Nafems15 systemeNafems15 systeme
Nafems15 systeme
 
QuantumFuzzylogic
QuantumFuzzylogicQuantumFuzzylogic
QuantumFuzzylogic
 
Quantum Fuzzy Logic
Quantum Fuzzy LogicQuantum Fuzzy Logic
Quantum Fuzzy Logic
 
Nafems15 Technical meeting on system modeling
Nafems15 Technical meeting on system modelingNafems15 Technical meeting on system modeling
Nafems15 Technical meeting on system modeling
 
An Approach to Overcome Modeling Inaccuracies for Performance Simulation Sig...
An Approach to Overcome Modeling  Inaccuracies for Performance Simulation Sig...An Approach to Overcome Modeling  Inaccuracies for Performance Simulation Sig...
An Approach to Overcome Modeling Inaccuracies for Performance Simulation Sig...
 
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...A Framework for Scene Recognition Using Convolutional Neural Network as Featu...
A Framework for Scene Recognition Using Convolutional Neural Network as Featu...
 
Making Custom Oscilloscope Measurements
Making Custom Oscilloscope MeasurementsMaking Custom Oscilloscope Measurements
Making Custom Oscilloscope Measurements
 
Fast optimization intevacoct6_3final
Fast optimization intevacoct6_3finalFast optimization intevacoct6_3final
Fast optimization intevacoct6_3final
 
Biomedical Signal and Image Analytics using MATLAB
Biomedical Signal and Image Analytics using MATLABBiomedical Signal and Image Analytics using MATLAB
Biomedical Signal and Image Analytics using MATLAB
 
Intro to LV in 3 Hours for Control and Sim 8_5.pptx
Intro to LV in 3 Hours for Control and Sim 8_5.pptxIntro to LV in 3 Hours for Control and Sim 8_5.pptx
Intro to LV in 3 Hours for Control and Sim 8_5.pptx
 
Universal Adiabatic Quantum Computer v1.0
Universal Adiabatic Quantum Computer v1.0Universal Adiabatic Quantum Computer v1.0
Universal Adiabatic Quantum Computer v1.0
 
Quantum programming
Quantum programmingQuantum programming
Quantum programming
 
Virus, Vaccines, Genes and Quantum - 2020-06-18
Virus, Vaccines, Genes and Quantum - 2020-06-18Virus, Vaccines, Genes and Quantum - 2020-06-18
Virus, Vaccines, Genes and Quantum - 2020-06-18
 
Probability and random processes project based learning template.pdf
Probability and random processes project based learning template.pdfProbability and random processes project based learning template.pdf
Probability and random processes project based learning template.pdf
 
Two methods for optimising cognitive model parameters
Two methods for optimising cognitive model parametersTwo methods for optimising cognitive model parameters
Two methods for optimising cognitive model parameters
 

More from Paul Brodbeck

Rutgers CM Seminar - QbD Process Tech 051716-R1
Rutgers CM Seminar - QbD Process Tech 051716-R1Rutgers CM Seminar - QbD Process Tech 051716-R1
Rutgers CM Seminar - QbD Process Tech 051716-R1Paul Brodbeck
 
Applying Batch Data Principles to Continuous Manufacturing AIChE Final
Applying Batch Data Principles to Continuous Manufacturing AIChE FinalApplying Batch Data Principles to Continuous Manufacturing AIChE Final
Applying Batch Data Principles to Continuous Manufacturing AIChE FinalPaul Brodbeck
 
AAPS Advanced Controls Uploaded 2
AAPS Advanced Controls Uploaded 2AAPS Advanced Controls Uploaded 2
AAPS Advanced Controls Uploaded 2Paul Brodbeck
 
RS_AIChE presentation 2013
RS_AIChE presentation 2013RS_AIChE presentation 2013
RS_AIChE presentation 2013Paul Brodbeck
 
ERC_EGUE_FINAL_Aug 12_PJB
ERC_EGUE_FINAL_Aug 12_PJBERC_EGUE_FINAL_Aug 12_PJB
ERC_EGUE_FINAL_Aug 12_PJBPaul Brodbeck
 
ISA FPID Presentation Final 3
ISA FPID Presentation Final 3ISA FPID Presentation Final 3
ISA FPID Presentation Final 3Paul Brodbeck
 

More from Paul Brodbeck (6)

Rutgers CM Seminar - QbD Process Tech 051716-R1
Rutgers CM Seminar - QbD Process Tech 051716-R1Rutgers CM Seminar - QbD Process Tech 051716-R1
Rutgers CM Seminar - QbD Process Tech 051716-R1
 
Applying Batch Data Principles to Continuous Manufacturing AIChE Final
Applying Batch Data Principles to Continuous Manufacturing AIChE FinalApplying Batch Data Principles to Continuous Manufacturing AIChE Final
Applying Batch Data Principles to Continuous Manufacturing AIChE Final
 
AAPS Advanced Controls Uploaded 2
AAPS Advanced Controls Uploaded 2AAPS Advanced Controls Uploaded 2
AAPS Advanced Controls Uploaded 2
 
RS_AIChE presentation 2013
RS_AIChE presentation 2013RS_AIChE presentation 2013
RS_AIChE presentation 2013
 
ERC_EGUE_FINAL_Aug 12_PJB
ERC_EGUE_FINAL_Aug 12_PJBERC_EGUE_FINAL_Aug 12_PJB
ERC_EGUE_FINAL_Aug 12_PJB
 
ISA FPID Presentation Final 3
ISA FPID Presentation Final 3ISA FPID Presentation Final 3
ISA FPID Presentation Final 3
 

EGUE Technikrom Final_8_12_13

  • 1. Utilizing DeltaV to Perform PAT Calculations Real Time Real-Time Non-Linear Regression Chromatography Endpoint Detection
  • 2. Photography & Video Recording Policy Photography and audio/video recording is not permitted in any sessions or in the exhibition areas without press credentials or written permission from the Emerson Exchange Board of Directors. Inquiries should be directed to: EmersonExchange@Emerson.com Thank you.
  • 3. Presenters  Michael Hausladen Director of Manufacturing Technology, BMS  Paul Brodbeck Engineering Manager, Control Associates, LBP
  • 4. Introduction  Chromatography Project  Functional Requirements  Chromatography Elution  Modeling  Prototyping  Implementation  Results  Summary
  • 5. Chromatography Project  BMS Syracuse  Scientist: Mike Hausladen  Pilot Plant Chromatography Skid Modification  Purpose: – Scale-up laboratory endpoint detection – Ensure process robustness • Minimize incorrectly determined collection end-point – Pilot scale model of full scale production system – Demonstrate capability for full scale production  Focus of this presentation: Robust chromatography elution end-point determination
  • 6. Chromatography Basics The basics of bind and elute chromatography: – Modify the conditions of the mobile phase to cause binding or elution of the product (protein) – Aqueous systems: pH and conductivity Stationary Phase Mobile phase: 1. Flush and prep column 2. Load protein on column 3. Wash 4. Elute protein 5. Clean and sanitize column 6. Store column Monitor elution for product fraction • UV adsorbance @ 280nm • Collect the desired portion of the elution peak
  • 7. Elution Curve 0 10 20 30 40 50 60 70 80 90 100 mAU Worse case chromatography generated in the lab. Data is not smooth., has multiple peaks. z w h The whole point – determine h (peak height) real time – to calculate percent of peak maximum collection end-point Percent of peak max (h) stop collecting product
  • 8. BMS Requirements  Robust endpoint of Elution Determination - real time  Peak Maximum Calculation – Endpoint = Percent of Peak Maximum • Model Predict Peak Max. of absolute optical density. • Percent of Peak Max lookup table of Sialic Acid vs. load material.  3 Models – in DeltaV Controller – Smoothing (1st Order Filter) Model – Linear Regression Algorithm - Polynomial Fit – Non-Linear Model • Extreme Value Function fit with Real-time Data – Column Volume (mobile phase volume) vs. UV absorbance  Alarming/Auto-Switching – Limits to ensure that the Peak Maximum is not determined early. – Ensure data is fitting the model with a sufficient level of accuracy. – Limits end Model if the algorithm is not converging.  Focus of this presentation is on the non-linear model real-time fit of the Extreme Value Function
  • 9. Implemented  3 Models – Smoothing Model – Polynomial Fit – Non-Linear Model  1 Algorithm – Newton-Raphson
  • 10. System Integrator Process  Select Model Equation  Basic Curve Fitting – Least Squares Error (LSE)  Select Numerical Method Algorithm – Gradient Descent – Newton Raphson – Levenberg-Marquardt  Machine Learning Parallel  Grey Box Modeling  Math – Linear Algebra – Vector Based Programming Languages (MATLAB) – Solving of Partial Differential Equations  Research – Internet Literature – Example Programs – Textbook – Numerical Methods by Dahlquest & Bjorck  Program Construction – Flowchart  DeltaV CALC Block Limitation – 2000 lines per scan – 64 If-Then loops per block – 256 field arrays – Dynamic Reference delays  Excel Solver – Troubleshooting – Math Checking – Convergence Issues  Switch Gradient to Newton – Bad Hardware – Convergence Issues  Final Program  Model Statistics  IQ/OQ  Levenberg-Marquart Future  Numerical Integration  Non-linear Models  Controller vs. Application Station  Other Approaches – synTQ, MATLAB, SoftPhase  Summary
  • 11. Non-Linear Model  Extreme Value Function: is an equation that approximates a chromatography peak  Using a dataset of x’s (volume of mobile phase) and y’s (elution response – UV adsorbance ) – Determine h, w, and z that give the best approximation of the peak – From h, determine endpoint  There are many mathematical functions for the representation of chromatographic peaks, extreme value function chosen for simplicity (Journal of Chromatography)
  • 12. Elution Curve Fitted 0 10 20 30 40 50 60 70 80 90 100 mAU fit z w h The whole point – determine h (peak height) real time – to calculate percent of peak maximum collection end-point Percent of peak max – stop collecting product
  • 13. Numerical Method – Basic Curve Fitting  Minimize the Squares of the Error – Least Squares Error  Curve fitting, minimizing error, finding best solution/ best fit  Analytical solutions for simple fit  Iterative numerical solution required for complex equations – Initial guess required – Convergence satisfied?  Solving in multidimensional space
  • 14. Numerical Methods – Solving Least Squares  Gradient Descent – Make Initial Guess Xi-1 – New Guess: Xi = Xi-1 – e * F’(Xi-1) – e is tuning constant • Too low = slow convergence, Too high = unstable  Newton - Raphson – Make Initial Guess Xi-1 – New Guess: Xi = Xi-1 - F’(xi-1)/2*(F’’(xi-1)) – Can be unstable with a poor initial guess  Levenberg-Marquardt – Start with Gradient, end with Newton  Iteratively repeat, check for convergence SquareofError
  • 15. Math  In order to mathematically determine the slopes of a multi-parameter system need to calculate the partial differentials. – Partials are slopes in n-dimensional space  Linear Algebra – When solving multivariate we need vectors, arrays or matrices. – Solutions become complex when dividing matrices  Vector Based Programming – Matlab, Python, Octave, …
  • 17. Solving the Extreme Value Function  Need to solve for h, z, w. – h = maximum peak height. – w = peak width. – z = retention time of column.  Vectorization Required  Challenge here is dividing by a Matrix  Complex Linear Algebra  Determinants  Easy in Matlab  Not so Easy Otherwise
  • 18. Research  Internet Literature – Methods for Non-Linear Least Squares Problems1 • Nomenclature difficult • Theoretical – Wikipedia – Dr. Math • Least Squares Regression for Quadratic Curve Fitting2 – Example Programs • L-A Algorithm by Pradit Mittrapiyanuruck – MATLAB • L-M Method for Non-Linear Least Squares by Henri Gavin – MATLAB, extra code… weighting, CHI, R^2, lamda  Textbook – Numerical Methods by Dahlquist & Bjorck – 5.2.1 Numerical Linear Algebra – 5.6 Iterative Methods – 6.9.2 Newton-Raphson’s Method… – 10.5.1 Non-Linear Optimization Problems (Hessian)
  • 19. Project Challenges DeltaV Limitations Other Obstacles  No matrix math functions  2000 lines per scan  64 loops per scan  Dynamic reference time delay  256 cell arrays max  Hardware problem  Literature for Matlab  Solving Diff. Eq.’s  Notation theory  Matrix Algebra  Numerical Methods  Least Squared Error
  • 21. Program – DeltaV CALC Block
  • 22. Math – Solving PDE’s  Solving PDE’s
  • 23. Prototyping in MS Excel h = 15 z = 0.24 w = 270
  • 24. Model Performance  Trust Zone  5 Second scan time  1000 pts each scan  Model converges to solution each scan  40% MX Controller Usage @ 5 sec scan  90% MX Controller Usage @ 1 sec scan
  • 25. IQ/OQ Validation 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 6 6.5 7 7.5 8 8.5 9 9.5 10 Raw Data Excel Prediction DeltaV NonLinear DeltaV Linear
  • 26. Levenberg-Marqaudt Future Enhancement  Prototyped in Matlab  Tested in DeltaV  Ready to be upgraded  Faster, more robust
  • 27. Implemented Models  All 3 running simultaneously  Auto Switch to best Model  Non-Linear Model is primary  Polynomial Model is secondary  Smoothing Model last  All 3 models tested within X% of each  Monitor all 3 and Alarm if > X% Error
  • 28. Other Options for Future  MATLAB – OPC communication – Runs in App station  synTQ – $$$ – MATLAB or other math package  Soft Phase – Requires C, C++, C#, or VB  Controller vs. Application Station – Controller slightly more robust • If App Station goes down Model still runs in controller – Application Station • Run at higher speeds • Reduced controller memory & capacity
  • 29. Summary DeltaV  Non-linear was successfully implemented to predict elution endpoint  Model updates 1000pt every 5 seconds  Model is capable of handling model non-convergence  Model passed IQ/OQ
  • 30. Thank You for Attending! Enjoy the rest of the conference.