BIOMATH

LABORATORY OF PHARMACEUTICAL PROCESS ANALYTICAL TECHNOLOGY

FACULTY OF PHARMACEUTICAL SCIENCES

BIOMATH, DEPARTMENT OF MATHEMATICAL MODELLING, STATISTICS AND BIOINFORMATICS

FACULTY OF BIOSCIENCE ENGINEERING

Model-based analysis and experimental validation of residence time distribution in twin-screw granulation

Ashish Kumar

3rd European Conference on Process Analytics and Control Technology

Current solid-dosage manufacturing is slow and expensive

Product collected after each unit operation

Actual processing time = days to weeks

2

Continuous manufacturing is better

3

No feed and effluent Concentration is time variant High variability Process control is easy

Constant feed and effluent Concentration are constant Low variability Rigorous control required

Continuous

twin-screw granulator

Granule conditioning

moduleSegmented

Fluid bed dryer

Continuous manufacturing line Consigma™-25 system

4

Twin-Screw Granulation

5

Design of granulator screw, screw speed, material feed rate control granulation

Number of kneading discs and stagger angle

Throughput

Screw Speed

Residence time distribution to know the granulation time and mixing

6

𝑐(𝑡)

𝑡

INLET

Fast particle

(short residence time)

Slow particle, long path (long residence time)

OUTLET

Residence time distribution to know the granulation time and mixing

7

𝑐(𝑡)

𝑡

Screw Configuration- Number of kneading discs- Stagger angle

Process parameters- Material throughput- Screw speed

RTD Measurement by Chemical Imaging

Model Formulation

Outcomes

Analysis of residence time distribution in twin-screw granulation

8

Tracer concentration in granules produced was measured using NIR

chemical imaging

9

API Map was used to measure RTD

time (sec)

10

RTD Measurement by Chemical Imaging

Model Formulation

Outcomes

Analysis of residence time distribution in twin-screw granulation

11

Traceraddition

d

𝑒(𝜃)

Conceptual model to include three main components of RTD

12

Modified Tank-in-Series model used

d d

pn1 n2 nxT

racer

in

t

Tra

cer

Ou

t

t

𝑒 𝜃 =𝑏 𝑏 𝜃 − 𝑝 𝑛−1

𝑛 − 1 !𝑒𝑥𝑝 −𝑏 𝜃−𝑝

where, 𝑏 =𝑛

(1−𝑝)(1−𝑑)

𝑒 𝜃 =𝑏 𝑏 𝜃 − 𝑝 𝑛−1

𝑛 − 1 !𝑒𝑥𝑝 −𝑏 𝜃−𝑝

where, 𝑏 =𝑛

(1−𝑝)(1−𝑑)

𝑒 𝜃 =𝑏 𝑏 𝜃 − 𝑝 𝑛−1

𝑛 − 1 !𝑒𝑥𝑝 −𝑏 𝜃−𝑝

where, 𝑏 =𝑛

(1−𝑝)(1−𝑑)

𝑒 𝜃 =𝑏 𝑏 𝜃 − 𝑝 𝑛−1

𝑛 − 1 !𝑒𝑥𝑝 −𝑏 𝜃−𝑝

where, 𝑏 =𝑛

(1−𝑝)(1−𝑑)

Modified Tank-In-Series model

13

Source: Levenspiel, Chemical Reaction Engineering, 1999

d d d

n1 n2 nxTra

cer

in

t

Tra

cer

Ou

t

t

p

RTD measurement by Chemical imaging

Model Formulation

Outcomes

- Mean residence time- Mean centred variance

- Number of CSTR - Plug flow fraction- Dead volume fraction

Analysis of residence time distribution in twin-screw granulation

14

Measurement based

Model based

API map- quantitative assesment

Variance, σ2

(width of the distribution)

Mean residence time , τ(a measure of the mean of the distribution)

𝜏 = 0∞𝑡∙𝑒 𝑡 𝑑𝑡

0∞𝑒 𝑡 𝑑𝑡

𝜎2 = 0∞(𝑡−𝜏)2∙𝑒 𝑡 𝑑𝑡

0∞𝑒 𝑡 𝑑𝑡

15

Chemical imaging based RTD measurement

Model Formulation

Outcomes

- Mean residence time- Mean centred variance

- Number of CSTR - Plug flow fraction- Dead volume fraction

Analysis of residence time distribution in twin-screw granulation

16

Measurement based

Model based

Residence time reduces with increase in screw speed

17

Measure of the mean of the distribution

Throughput

Residence time reduces with increase in throughput...but not always

18

Residence time increases with increase in number of kneading discs.

19

Residence time reduces with increase in stagger angle.

20

Mean of the residence time distribution

Number of kneading discs

Throughput & stagger angle

Screw Speed

21

Chemical imaging based RTD measurement

Model Formulation

Outcomes

- Mean residence time- Mean centred variance

- Number of CSTR - Plug flow fraction- Dead volume fraction

Analysis of residence time distribution in twin-screw granulation

22

Measurement based

Model based

Width of the distribution shows axial mixing

23

For well-mixed system, NV = 1, For poorly mixed, NV = 0

Axial mixing increases with increase in screw speed

24

Throughput

For well-mixed system, NV = 1, For poorly mixed, NV = 0

Axial mixing increases with increase in throughput…but not always

25

Axial mixing decreases with increase in number of kneading discs

26

Increase in stagger angle caused reduction in axial mixing

For well-mixed system, NV = 1, For poorly mixed, NV = 0 27

Major factors had opposite effects on residence time and axial mixing

28

Factors Residence time Axial Mixing

Number of kneading discs

Increase Decrease

Screw Speed Decrease Increase

Throughput Interaction Interaction

Stagger angle interaction Interaction

Chemical imaging based RTD measurement

Model Formulation

Outcomes

- Mean residence time- Mean centred variance

- Number of CSTR - Plug flow fraction- Dead volume fraction

Analysis of residence time distribution in twin-screw granulation

29

Measurement based

Model based

𝑒 𝜃 =𝑏 𝑏 𝜃 − 𝑝 𝑛−1

𝑛 − 1 !𝑒 −𝑏 𝜃−𝑝

where, 𝑏 =𝑛

(1−𝑝)(1−𝑑)

Dead zone

Parameters of the TIS model estimated using experimental RTD based on least SSE

30

, 0.29 n = 2 , 0.34

n1 n2 nx

Screw speed 900 rpm

p = 0.42 , 4 d = 0.26

Screw speed 500 rpm

Traceraddition

Dead zone

𝑒(𝜃)

Plug flow component of the RTD

31

0

0.2

0.4

0.6

0.8

30

°6

0°

90

°3

0°

60

°9

0°

30

°6

0°

30

°6

0°

90

°3

0°

60

°3

0°

30

°6

0°

90

°6

0°

30

°6

0°

90

°3

0°

60

°3

0°

60

°3

0°

60

°9

0°

30

°3

0°

60

°9

0°

30

°6

0°

30

°

60

°9

0°

30

°6

0°

30

°6

0°

30

°6

0°

90

°6

0°

30

°3

0°

60

°9

0°

30

°6

0°

30

°

2 6 12 2 6 12 2 6 2 6 12 2 12 2 6 12 2 6 12 2 612 2 6 12

10 Kg/h 17.5 Kg/h 25 Kg/h 10 Kg/h 17.5 Kg/h

25 Kg/h 10 Kg/h 17.5 Kg/h 25 Kg/h

500 RPM 700 RPM 900 RPM

Plug flow fraction

Plug flow fraction decreases with increase in screw speed and throughput

32

Traceraddition

Dead zone

𝑒(𝜃)

Mixed flow component of the RTD

33

0

2

4

6

50

07

00

50

07

00

90

05

00

70

09

00

50

07

00

90

05

00

70

09

00

50

05

00

70

09

00

50

07

00

90

0

50

07

00

90

05

00

70

09

00

50

07

00

90

05

00

50

09

00

50

07

00

90

0

50

07

00

90

05

00

70

09

00

50

07

00

90

07

00

90

05

00

70

09

00

70

09

00

30° 60° 90° 30° 60° 90° 30° 60° 30° 60° 90° 30°60° 30° 30 60 90 30 60 30

2 6 12 2 6 12 2 6 12

10 Kg/h 17.5 Kg/h 25 Kg/h

Number of CSTR

34

Material throughput controls mixing which reduces with increase in throughput

Traceraddition

Dead zone

𝑒(𝜃)

Mixed flow component of the RTD

35

0

0.1

0.2

0.3

0.4

0.5

0.6

30

60

90

30

60

90

30

60

90

30

60

90

30

60

60

30

60

30

30

60

90

30

60

90

30

60

90

30

60

30

60

30

60

30

30

60

90

30

60

90

30

60

90

30

60

60

30

60

30

60

30

30

10 17.5 25 10 17.525 10 17.5 10 17.5 25 10 25 1017.525 10 17.5 25 1017.525 1017.525

2 6 12 2 6 12 2 6 12

500 700 900

Dead fraction

Dead zone increases with screw speed, and reduces with increase in kneading discs

36

Model based analysis showed that

Screw speed reduces the conveying fraction

Material throughput dictates mixing.

Number of kneading discs help to reduce the dead volume in TSG

37

Conclusions

NIR-based chemical imaging is fast and robust technique for RTD studies.

Along with experimental study, an improved insight by model based analysis of RTD.

Screw speed controls the residence time, while the material throughput controls mixing.

38

Perspectives

Together with a Granule size distribution study it will be confirmed which mixing regime is most desirable.

In further study we will investigate material properties influence on the RTD and mixing.

Utilise the mixing and residence time information for mechanistic modeling of the TSG.

39

Thomas De BeerIngmar NopensKrist V. Gernaey

Jurgen VercruysseValérie Vanhoorne

Maunu ToiviainenPanouillot Pierre-EmmanuelMikko Juuti

Kris Schoeters

Aknowledgements

40

41

Ashish.Kumar@UGent.be

Model-based analysis and optimization of bioprocesses

Laboratory of Pharmaceutical Process Analytical Technology