34
Continuous Manufacturing & Automation Solutions 2017 Engineering, GMP & Validation Forum Anand Sunit Kulkarni Life Sciences Industry Leader_ Asia Pacific Emerson Automation Solutions

Continuous Manufacturing & Automation Solutions · Konstantinov and Cooney, Journal of Pharmaceutical Sciences 104:813-820, 2015 Buffer / Media Prep Perfusion Cell Culture Continuous

  • Upload
    others

  • View
    9

  • Download
    0

Embed Size (px)

Citation preview

Continuous Manufacturing &

Automation Solutions

2017 Engineering, GMP & Validation Forum

Anand Sunit KulkarniLife Sciences Industry Leader_ Asia PacificEmerson Automation Solutions

Agenda

2Emerson Confidential

Continuous Manufacturing: The Path Ahead

Pharma Manufacturing: Batch Vs Continuous

Dynamics of Continuous Manufacturing

Automation Management and Control Strategies

Closing

Pharmaceutical Manufacturing: The Path Ahead..

Emerson Confidential

Pharmaceuticals Need to Change3

“ Right now, manufacturing experts from the 1950s would easily recognize the pharmaceutical

manufacturing processes of today. It is predicted that manufacturing will change in the next 25 years as current manufacturing practices are abandoned

in favour of cleaner, flexible, more efficient continuous manufacturing.”

Dr.Janet Woodcock, AAPS Annual meeting, October 2011

Continuous Production is Coming

• Example of a modular manufacturing capable of supporting process development, clinical, or full commercialized production.

4Emerson Confidential

Strong Financial Incentives to Change

5Emerson Confidential

Improved Quality Capital Savings

Higher Throughput Cost Savings

New facility

makes 100,000

tablets / hour

vs >4 weeks at

a batch plant 1

Operational Flexibility

Low residence time

Real-Time

monitoring, PAT,

& Advanced Control 2

Shorter production times

Waste reduction / don’t

discard entire batch

>30% operating cost

reduction—Novartis / MIT 1

Vertex facility is only

4,000 sq. ft. vs 100,000

sq. ft in a traditional

deployment 1

Scales based on time

and parallelization vs

size

Portability on site

Use across stages 2

1 WSJ Feb. 8, 2015 Drug Making Breaks Away from It’s Old Ways2 Biotechnology and Bioengineering Vol. 112, Issue 4, April 2015, Pages 648-651

6Emerson Confidential

Pharmaceutical Manufacturing (Synthetic Drugs)

• Different sites and large equipment

• Larger Supply Chain management

• Collectively for Months

• Single Site Operations

• Modular Equipment/Shorter Supply Chain

• Shortened to Days

Pharmaceutical Manufacturing (Biotech / Biosimilars)

7Emerson Confidential

• Product transfers between each unit

operation in batches

• Finished product is tested at off-line

laboratories

• Actual processing time = days to

weeks

Buffer /

Media

Prep

Feb Batch

Cell

Culture

Harvest Capture Chromatograph

y

Filtratio

n

Drug

Substance

Konstantinov and Cooney, Journal of Pharmaceutical Sciences 104:813-820, 2015

Buffer / Media

Prep

Perfusion

Cell CultureContinuous Capture

(Chrome 1)

Continuous

Chromatography

(Chrome 2 and 3) Filtration

Drug Substance

Continuous

• Product flows between each unit

operation

• Product is adjusted based upon in-

process measurements

• Actual processing time = hours to days

Batch

Continuous

8Emerson Confidential

Secondary Manufacturing – Batch to Continuous

Batch Continuous

9

Emerson Confidential

Residence Time Distribution

• Variation in material attributes and process parameters

• Optimization of the Equipment Design

• Tracking of material and disturbances

Real Time Release Testing and Process Control

• On-line / In-line / At-line Measurements

• Process Control and Multivariate Models

• Adequate Models and Validation

Production Capacity

• Demonstrate manufacturability within expected range

• Equipment performance and Process Dynamics

• Uniformity of Quality

Dynamics of Continuous Manufacturing

Comprehensive Control Solutions Required

10

• Continuous Regulatory Control for minimum Variability

– PID Control is the workhorse

• Advanced Control for Constraint Optimization

– Multivariable Control

– Control of Difficult Dynamics

– Inferential Control

– Optimization Objectives

• Statistical Process Control

– Multivariate

• Principle Component Analysis

• Partial Least Squares

Emerson Confidential

Real-time Optimization (RTO)

Target tracking MPC Layer

Regulatory PID Layer

Process

Continuous Automation Implementation Challenges

11Emerson Confidential

Automation is Crucial for Continuous Manufacturing

Automation Management Strategies Control Strategies

• Understand the unit operation(s) / how far does “continuous manufacturing” extend?

• Modular automation design

• Data integrity from source through use

• Equipment states

• Start- Up and shutdown activities

• Basic PID Control

• Unit Operation Modeling

• Advanced Process Control

• Batch / Lot definition

• Batch context

• Standardized, modular automation

Automation Management Strategy :Modular Automation Designs

12

• Units need to be autonomous

• Support a complete scope of solutions for optimized unit production

– Measurement with integrity

– Local control

– Start-up / shutdown / product change-over sequencing for on-the-fly transitions

– Process characterization to determine optimized end result

– On-line analyzers are critical in many cases

– Alarm Management

– Process History

Emerson Confidential

Unique ID in System

Sequences

PID Control

Advanced

Control

Local

Alarms

Trips

Integrity

Status

Running

Data

GMP

Status

Service

Time

Process

Capabilities

Flow Rate

Pressure

Temperature

Dynamic

Characteristics

Recip

e

Req

uire

men

tsData

Hours Run

Local Control

Compliance

Paperwork

Characterization Data

Control Strategy: Start with “Basic” PID Control

13Emerson Confidential

• Temperature profile, T(t), is main design variable for cooling crystallization

– Cannot set T(t) directly; T(t) specified from valve positions of cold and hot water streams

Hot Water

Cooling Water

( )cF t

( )cx t

( )hx t

hT

cT

( )hF t

Physical

actuation

Jacket

Solution

Crystallizer

energy

balance, T

Jacket energy

balance, Tj

Flow-valve

relationship, F=f(x)

Real-time Optimization (RTO)

Target tracking MPC Layer

Regulatory PID Layer

Process

Control Strategy: Then Add on Optimization

• There are two main classes of models used in Multivariate Process Control (MPC) formulations

Real-time Optimization (RTO)

Target tracking MPC Layer

Regulatory PID Layer

Process

• Objective: Optimizes economics

• Strategy: Steady state optimization

• Objective: Drive system to target setpoints

• Strategy: Dynamic constrained optimization

Can improve by combining

these two layers

Drawbacks of layered approach

• RTO targets may not be reachable

• True optimal situation may not be

achieved at a constant setpoint

14Emerson Confidential

Automation Approach Combining Regulatory and Optimization

Data Integrity is critical for

pharmaceutical quality

measurements and becomes

even more important with

continuous for Real Time

Release and Continuous

Process Verification.

Doug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 201615

Tablet Press

Blender

Feeders

mill

API M

M

M

M

M

M

M

M

Content/DensityBlend Uniformity

NIR, Raman

LT

Thickness

Density

US

Hardness

NIR

Dissolution (check)

Feed forward control

Force

Weight

Compression Gap

Cross-check

Content (check)

Emerson Confidential

Key FDA Issues for Continuous Manufacturing

16Emerson Confidential

Regulatory Definitions – Applicable to continuous processes

“Lots”

21 CFR 210.3

• Lot – a batch, or a specific identified portion of a batch, having uniform character and quality within specified limits; or, in the case of a drug product produced by a continuous process, it is a specific identified amount produced in a unit of time or quantity in a manner that assures its having uniform character and quality within specified limits.

Need to track Lots

“Batch”

21 CFR 210.3

• Batch – a specific quantity of a drug or other material that is intended to have uniform character and quality, within specified limits, and is produced according to a single manufacturing order during the same cycle of manufacture.

• Batch refers to the quantity of material and does not specify the mode of manufacture

Need to track Batches

MIT-CMAC International Symposium on Continuous Manufacturing of Pharmaceuticals, May 20, 2014

Janet Woodcock, M.D. Center for Drug Evaluation and Research, FDA

Managing Batch / Lots:

Residence Time Distribution (RTD)

Diagnostic tool used to determine the time (mean and distribution) a particle stays in a unit operation

RTD models are used to determine:

1. Disturbance dissipation

2. Raw material traceability

3. Mixing patterns in unit operations

Detector(i.e., PAT tools)

Pulse of Tracer(i.e., API)

Time

[Tra

cer]

Time

[Tra

cer]

t0

τ

0

( )(t)

( )out

C tE

C t dt

0

( )MRT t E t dt

Inlet

Outlet

Doug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 2016 17Emerson Confidential

Integration of RTDs in Continuous Direct Compression

Case Study

Overlap of the curves represents

the mixing of the lots

M

M

EXM

M

API

M

M

MgSt

Lot A

Lot B

Lot AB

How much of AB?

Using RTD

API

Given the relationship of batch API and continuous drug product manufacturing, multiple lots

of API will inherently need to be mixed in order to make product

Regulatory question: How can we determine the API lot composition of the drug product?

Managing Batch / Lots:

Residence Time Distribution (RTD)

Doug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 2016 18Emerson Confidential

Unlocking the Value of Continuous Manufacturing Requires Planning

Continuous

Manufacturing in Life

Science is progressing.

Assess people,

processes and

technologies within

your organization to

determine how to get the

benefits

The number of approvals is growing

The technology barriers are being addressed

Secondary manufacturing is moving the fastest

but primary is also progressing

Approach requires re-thinking your

manufacturing approach

Requires people with a combination of process

and automation / data technology skills

19Emerson Confidential

Back Up Slides

21Emerson Confidential

New Data Integrity Requirements

Data Integrity• Completeness, consistency, and accuracy of data.

Managing Data Integrity and Change Control

• Attributable

• Legible

• Contemporaneous

• Original

• Accurate

• Enduring

• Complete

• Consistent

• Retrievable

ALCOA +

(or true copy)

22Emerson Confidential

It Starts with the Sensor:

Data with Status / Unit Operations Using Status

Data

Applications

Security

Insight

Tools

Secure

First Mile

Standalone

Rich Tools

PERFORMANCE ENERGYHEALTH

Advisory

Tools

Traditional Data Pathways IIoT Application Gateways

Pervasive

Sensing

Radar

Level

Electrical

Monitoring

Analyzers Steam

TrapsVibration Valve Position

Monitor

Single Use

Pressure

Single Use

pH

Single Use

Pinch Valve

23Emerson Confidential

Out of Service Available

CalibratedCalibrateCleanCIP’d

Inactivated Unavailable

In Use

YesNo

In-Use

/ Start

In-Use /

Complete

Inactivate

Equipment

Out of

Service

Equipment

State Model

Procedural Automation for Continuous Process (ISA 106) Growing in Importance

Process State Operational States

Not Ready (Out

of Service

Preparing

Ready (Idle)

Filling

Heating

Running

Abnormal

(Off Spec)

Shutting Down

Not Ready

Preparing

Ready

Start-up

Running

Shutting

DownISA-TR106.00.01-2013 Procedure Automation for Continuous Process Operations

Manufacturing Savings via Procedural Automation

• Dow is a major contributor to development of the ISA106 standard.

– State Based Control has been applied in around 200 Dow plants globally

– 75% of Dow Assets are run using state based control*

• In one recent example, Dow reported the reduction in shutdown time achieved by automating cleaning procedures resulted in a 16% increase in plant profit.

Shorter Maintenance

Faster Shutdown

Faster Startup

25Emerson Confidential

Procedural Automation State Based Control, Yahya Nazer, Janette Brightwell, ARC Forum, 2015

Account for Likely Sources of Disturbances

Hot Water

Cooling Water

( )cF t

( )cx t

( )hx t

Pressure

drop

hT

cT

Temperature

fluctuations

Heat of

crystallization +

other non-idealities( )hF t

Mitigate disturbances by

design of regulatory

control layer

26Emerson Confidential

Design of Regulatory Layer: Cascade Control

CW ValvePID JacketcFcx

Crystallizer

PIDjT,c SPF

,j SPT

cP

HW ValvePIDhFhx,h SPF

PID

hP

cT hT

T

Crystallizer

temperature

PIDSPT

jT

27Emerson Confidential

When Combined, There Are Two Main Approaches

• There are two main classes of models used in MPC formulations

1. Empirical / black-box models

• Based completely on process data

• Majority of methods generate linear models

2. First principles models

• Derived from physics of problem (mass/energy

balances, kinetic rate expressions, thermodynamics)

28Emerson Confidential

Empirical vs. First Principles

Characteristics Empirical Model First Principles Model

Model derivation Only from process data Based on physics; need some

modeling expertise

Number of parameters Usually high due to black-box nature Smaller; based on physical

parameters

Identification Must perturb process for relatively

long time for each manipulated

variable

Smaller perturbations (if past data is

informative, no experiments needed)

Accuracy Only around a steady state, where

experiment conducted

Accurate in wider range than a

steady state

Capability Mostly target tracking to keep model

accuracy reasonable

General objectives (e.g., profit,

energy usage)

Versatility Hard to deal with unmeasured states Can handle unmeasured states and

disturbances explicitly

Resultant optimization in MPC LP or QP: typically fast to solve

online

NLP: need more sophisticated

solvers and expertise to speed up for

real time usage

29Emerson Confidential

First Principles Models are Accurate over Larger State-space Region

x1

x2

Steady-state where

step response model

was developed

Accuracy region for first

principles model

Ex: State trajectory for startup

endpoint

Only guaranteed

to be valid locally!

30Emerson Confidential

Add on Modeling per Unit Operation Requirements:

Modeling Dissolution

• When C < Csat, crystals will dissolve in solution

• Dissolution can be modeled as a first-order process

in terms of relative supersaturation

( )satC C T

, 0

, 0

gg

d

k S SG

k S S

Usually large fast dynamics

Need to calculate rate of crystal loss

(equivalent B for dissolution in μ0 ODE)

Moment equations same for j > 0

or crystals dissolve

completely…

31Emerson Confidential

More Research / Proofs Exist in Secondary Process Modeling

Feedback

control

Feedforward

control

Flowsheet model

Doug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 2016 32Emerson Confidential

FEEDERS:Model: Delay Differential

Equation, Transfer Function

rpmd50,ρFset

d50,ρFout

MIXER:Model: Population

Balance model, RTD

model, surrogate model

rpmd50,ρi

Fin,Ci

d50,ρFout, Ci,

CO-MILL:Model: RTD model, Mass

balance equation

rpmd50,ρFin,Ci,

d50,ρFout, Ci,

TABLET PRESS:Model: Kawakita equation and

feed frame RTD model

P, rpmd50,ρFin,Ci,

ε, σ, Ci, Fout

Process Modeling for Control Development:

Basic Regulatory

Doug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 2016 33Emerson Confidential

Flowrate

Fill level

Flowrate

Mass Hold-up

Flowrate

Mass Hold-up

Co-mill

Feeder

Blender

Process Modeling for Control Development:

Add in Optimization for the Unit

34Emerson ConfidentialDoug Hausner, Ph.D., Associate Director, ERC SOPS, Rutgers University

Emerson Exchange Life Science Forum Oct. 2016