dynamic process control: implementation of feedback/feedforward
TRANSCRIPT
Dynamic Process Control: Implementation
of Feedback/Feedforward Control for
Continuous Systems
Mojgan Moshgbar (PhD)
Director of Advanced Manufacturing Technology Group Director of Pfizer Intellicentic® Partnership
Overview
• Introduction to Static & Dynamic System Analysis
• Continuous Processes – Examples of Process Dynamics
• Advanced Process Control in Continuous Pharmaceutical Processes
• Example Results from Continuous SOD processing
• New Platform Technology
Statics
• “Steady State” Conditions
• Initial vs Final States
Dynamics • Transition Between States
• Change over Timer
Reality • Superimposed dynamics & Statics
• Analysis must consider both
System Behavior
Introduction & Terminology
Static & Dynamic Response
0 10 20 30 40 50 60 20.8
21
21.2
21.4
21.6
Time
Te
mp
era
ture
0 10 20 30 40 50 60 20.8
21
21.2
21.4
21.6
Te
mp
era
ture
Response to Setpoint Change
Dynamic Response
0 50 100 15018
20
22
24
26
Ro
om
Te
mp
era
ture
0 50 100 1501.5
2
2.5
3
Time (mins)
Po
we
r (k
W)
time delay time constant
Process Gain = output
input
1)(
Ts
KesG
std
First Order System
Dynamic Response
First Order Systems Response to Input Change
0 10 20 30 40 50 60 70 80 90 1000
2
4
6
8
10
12
14
16
18
20
Time
Out
put
Oscillatory response for under-damped systems
Over-damped Response
Many continuous processes exhibit first order system dynamics
Stopping Starting
Continuous Processes Dynamics
Startup API Concentration (NIR) 28.7% 30.2% 31.7%
Stopping Starting
85
90
95
100
105
110
115
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Pred
icted
Pote
ncy %
LC (N
omina
l Weig
ht)
Sample NumberSIMCA-P+ 11.5 - 6/13/2008 12:53:18 PM
88%LC
100%LC
Typical Dynamic Responses of Continuous Solid Formulation Processes Response Time: dependent on mixing/throughput parameters
Steady state fluctuations partially due to measurement errors caused by moving powder
APC in Continuous Processing
Continuous processes are time dependent & may experience transient process upsets
Fast PAT systems are essential to monitor real-time performance, and divert OOS material
Simulated Transient Disturbance
A transient ½ minute process
upset, in a high throughput
continuous Process can result in
high volume of rejected OOS (eg.
5000 OOS units with throughput
of 120 kg/hour
(assumes 200mg dosage weight)
APC Architecture for
Continuous Blending Applications
Model Predictive
APC
Raw Materials Concentration
Setpoints
Feeders Ratios
Impeller Speed
Retention Mass
Continuous Formulation
System
Blend Composition Fast NIR System
Retained Mass
Blend
Initial Process Setpoints
Throughput Setpoint
Static Setpoints
Dynamic Setpoints
Multiple Feeders
Load Cells
Load Cells Readings
Feedback Loops
Feedforward Loops
Feedback vs Feedforward
Feedback Loops provide dynamic information on the process outputs error
Feedback Loop Measurement Point (location) is critical for stability of control loops
Large separation between measurement & Actuation points increases the response delay (dead-time)
Feedforward loops are used to preemptively reduce the impact of dead-time and delay in system response to feeder excursions
Pfizer’s New Continuous SOD Portable, Continuous, Miniature & Modular
Continuous Mixing Direct Compression
ConsigmaTM -25 TSWG
Raw Material Dispensing
Slide Curtsey: Dan Blackwood et al. AIChE 2013
APC Turned ON 2. Feeder performance adjusted to achieve target API
Concentration
3. API Concentration by on-line NIR is centered at Target
1. API Concentration is above target
Slide Curtsey: Dan Blackwood et al. AIChE 2013
5. Potency setpoint change from 14 to 10%.
2. Production has been
increased from 3 to 3.5 kg/hr
3. Feeder Performance is adjusted
6. The APC implements the
potency change while controlling
production and hold-up mass to setpoint.
1. API Potency is at target
4. Hold Up Mass is held
Constant
Slide Curtsey: Dan Blackwood et al. AIChE 2013
Real-time Material Characterization Platform Capturing Process Signature
Real-time NIR Imaging System
Dynamic Process Signature
Hidden Process Fingerprint
Beginning Middle End
Visualization of MgSt dispersion and lubrication film in SOD Matrix
Advanced Image Processing Algorithm for Visualization of Blend Lubrication, Micro Temp Variations & Environmental Impacts on
Blend Performance
Color Key: Relative Lubricant Concentration Distribution
APC in Continuous Processing
APC system can reduce the frequency and severity of transient excursions and associated cost impact
In Pfizer, APC is rapidly gaining acceptance as a key enabling and transformational technology for both continuous and batch processes
New platform technologies for real-time material characterization provide new capabilities for validation
Acknowledgement
Pfizer PCMM Team
Pfizer GTS Colleagues
Pfizer AMT Colleagues