How To Deliver Step Changes in Manufacturing Operations with Predictive Insights using AI/ML and Pharma 4.0 Principles

Copyright © 2019, Bigfinite
www.bigfinite.com
How to Deliver Step Changes
in Manufacturing Operations
with Predictive Insights
Using AI/ML and Pharma 4.0
Principles

Toni Manzano
2
■ CSO & co-founder of Bigfinite
■ Physics Degree, Master in Information and Knowledge Society, Post-graduated in
quality systems for manufacturing and research pharmaceutical processes
■ 7 US Patents: encryption, transmission, storage and processing big data for
regulated environments in the cloud
■ Articles & white papers based on cloud, pharmaceutical industry, and data science
■ AI Health Xavier University of Cincinnati: AI Core Team and AI Manufacturing Team
Lead
■ PDA, Scientific Committee & Europe co-chair
■ AI & Big Data SME for Sciences in the Spanish Parliament
■ Professor at the University Autonomous of Barcelona
■ Bioinformatics of Barcelona, Project Leader of the Data Integrity team

What is Artificial Intelligence?
3
“AI can be thought of as simulating the capacity for
abstract, creative, deductive thought - and particularly
the ability to learn - using the digital, binary logic of
computers.”
“Artificial Intelligence (AI) is no longer some bleeding
technology that is hyped by its proponents and mistrusted by
the mainstream. In the 21st century, AI is not necessarily
amazing. Rather, it is often routine. Evidence for the routine and
dependable nature of AI technology is everywhere.”
“Verification and Validation and Artificial Intelligence,” Tim Menzies, Portland State
University, Charles Pecheur, NASA Ames Research Center. July 2004.

Supervised and Unsupervised Algorithms
4
Learning Supervised Learning Unsupervised
➔ Linear Regressor
➔ Support Vector Machines
➔ Random Forests
➔ Neural Networks
➔ K-Nearest Neighbours
➔ Gradient Boosted Trees
➔ ...
➔ K-Means Clustering
➔ Hierarchical Clustering
➔ Isolation Forests
➔ Graphical Lasso
➔ Bayesian Networks
➔ Markov Hidden Models
➔ ...
AI Supervised Vision Example AI Unsupervised Vision Example

Using AI, Believing in Data
5
Get it right
the first time!

Bulk #1
Bulk #2
Bulk ‘n’
(...)
Dissolution (S4) Sterile filtration
Concentration by
Ultrafiltration
UF-3
Lot formulation:
24,5% bulk #1
12,3% bulk #2
(...)
2% bulk ‘n’
Bulk Polarimetry (g/L)
Final Product
Polarimetry (g/L)
Recirculation #1
Recirculation #2
(...)
Recirculation ‘n’
~ f (bulk polarimetry,
filtrate pressure)
Which is the CPP set point
to obtain the optimal polarimetry avoiding recirculation?
Biz Case: Increase Process Robustness → Right the First Time
6

Vessel pressure
Membranes outlet pressure
Membranes pressure control
Membranes inlet pressure
Filtrate pressure
Temperature
Retentate flow
Tank level
System weight
Pump speed
Average pump speed
Trans-Membrane pressure
Plastic bag weight
AVG of pH
Average of Osmolality
Mean Viscosity at Shear Rate 0.1/sec
Mean Viscosity at Shear Rate 1000/sec
%RSD for Shear Rate 0.1/sec
Mean Original Concentration
CPP
CQA
7

Biz Case: Process Robustness → Right the First Time
8
Reducing Process
Step Iterations
“CQA”
[g/L]
# Recirculations0 1 2 3
10.2%
10.2
9.9
CPP ~ pump speed9.8% 9.5% 9.2%
9.1g/L
8.2g/L
9.8g/L
10.1g/L
“CQA”
[g/L]
# Recirculations0
10.2
9.9
CPP ~ pump speed9.2%
10.1g/L
Which is the CPP set point
to obtain the optimal CQA
avoiding recirculation?
Process Optimization with AI
3
Breaking Data Silos
1
ERP
MES
LIMS
(...)
Discovery
2
Support
Vector
Regressor

In 58 batches over 2 years, 90 recirculations were performed
With the Bigfinite platform, no recirculations would be necessary
61% reduction of runs
From the 63 samples used for modeling, only 58 had all data needed
There was valuable knowledge lost in 5 batches
From 8% of knowledge lost to 0%
In 58 batches over 2 years, only 27 batches were right the first time
Increase by 53% process effectiveness
9

Using AI, Believing in Data
10
Learning the
rules by playing
the game!

Discover Root Cause Analysis in OEE
11
“We were able to visualize patterns that we
couldn’t understand before because we
started tracking and identifying the
communication delays between the IT
systems surrounding the line!”
- Kasper Malthe Larsen
Chief Technology Architect
BUSINESS CHALLENGE
Packaging line was always behind and not performing well
● Very complex set-up, find a way to retrieve data loggings from their site
SOLUTIONS
Take all of the data from the packaging line (temp/hum, PLC, etc.),
put it into the platform, then identify patterns with AI/ML
● Built a ML model (by a random forest regressor) that can predict the
number of pens produced per minute
● Able to quickly get access to real-time data and historical data
RESULTS
IT systems are producing more effectively and in the right sequence
after learning there were common delays between systems
● Learned the majority of the best batches of pens were produced during
the day shift
● Identified 10 batches that outperformed all other batches; able to
investigate baseline of the ‘golden batch’

Improving Packaging Line OEE
12
Classic OEE
Actual Output
AI Prediction
● 2,500 batches of
experience
● 130 factors
● Critical variables:
Temperature
Pens/Carton
Batch size
Day of the year
● One single
variable to look
at!

When the Statistical Results have Huge Impact on the Users
13

AI: More than Just Multivariable Models...
14

How to start?
Next Steps
15

Thank you!
info@bigfinite.com | www.bigfinite.com
Providing simple solutions to complex needs in
biotech and pharma.
Visit us in Booth #3

How To Deliver Step Changes in Manufacturing Operations with Predictive Insights using AI/ML and Pharma 4.0 Principles

More Related Content

What's hot (10)

Similar to How To Deliver Step Changes in Manufacturing Operations with Predictive Insights using AI/ML and Pharma 4.0 Principles (20)

Recently uploaded (20)

How To Deliver Step Changes in Manufacturing Operations with Predictive Insights using AI/ML and Pharma 4.0 Principles

Editor's Notes