AI, IoT and Blockchain tech briefing to the industry to showcase our research at NUS.
by Dr. Xiaonan Wang
Assistant Professor
NUS Department of Chemical & Biomolecular Engineering
1. How can AI and IoT power
the Chemical Industry?
Department of Chemical and Biomolecular Engineering
National University of Singapore
NUS Faculty of Engineering, Block E5, Unit #03-04
+65 6601 6221 (Tel) chewxia@nus.edu.sg
2.
3. __________________________________________________________
Motivation: what and how
AI: Machine Learning/Deep Learning
http://blogs.teradata.com/data-points/tree-machine-learning-algorithms/
A computer program is said
to learn from experience E
with respect to some task T
and some performance
measure P, if its performance
on T, as measured by P,
improves with experience E.
“If you invent a breakthrough in
artificial intelligence, so machines can
learn, that is worth 10 Microsofts”
(Bill Gates, Chairman, Microsoft)
4. The new context in Industry
Artificial Intelligence &
Potential Applications of AI and IoT
• AI automates the monotonous tasks.
• AI recognizes subtle patterns in industry data and helps predict
when each node will need servicing long before a problem occurs.
• AI systems gradually understand each operational element better,
allowing it to quickly identify patterns
Credits to Exxon Mobil and Prof Marco Seabra dos Reis
6. Overview of Smart Systems
Engineering (SSE) Research
Smart Energy
• Multiple sustainable energy systems
• Supplying demand with resource efficiency
Smart Healthcare
• Targeted healthcare data mining and system
• Improving pharmaceutical manufacturing
Optimal
Design
and
Operation
in multi-
scales
Smart Materials and Accelerated Manufacturing
• Advanced data and optimization strategies
• Expediting materials and industry innovations
Smart Nation/ Smart City
• Network of data and technologies
• Providing sustainable and resilient progress
7. Model
Application A
Features
Power (kW)
Discharge Duration (h)
Temperature range (oC)
etc…
Technology Probability
T1 P1
T2 P2
…
…
Tn Pn
Requirements or
characteristics of
the applications
Technical suitability of
various technologies
for the application
Research Project: Use data driven methods to model technical suitability of various energy
storage technologies for different application
Supervised classification task using various machine learning algorithms:
Logistic regression, Random forest, Neural networks
AI for Smart Energy: Technology and System Design
Data source: DOE Global Energy Storage Database
More than 1000 installations in various parts of the world
Features selected (X)
Power (kW)
Discharge Time (hr)
Label (y)
Technology
L. Li, P. Liu, Z. Li and X. Wang. "A Multi-Objective Optimization Approach for Selection of Energy Storage
Systems", Computers & Chemical Engineering 115 (2018): 213-225.
8. 8
AI for Smart Energy: blockchain enhanced micro-grid
Decentralized Peer-to-Peer mode
Promotion of trust
Reduction of cost
Enhancement of security
S. Noor, W. Yang, M. Guo, K. H. van Dam, and X. Wang*. “Energy Demand Side Management within
micro-grid networks enhanced by blockchain”, Applied Energy 228 (2018): 1385-1398.
9. 9
1. Storage for digital
records
2. Exchanging digital
assets (called tokens)
3. Executing smart
contracts
AI for Smart Energy: blockchain enhanced micro-grid
https://baas.zhigui.com
10. AI for healthcare: Protein engineering
Protein
sequence
Machine
learning
algorithms
Predict
protein
solubility
Guide
experiments
A series of machine learning models
Accurate prediction which reduces cost of experiments in vivo
Data augmentation for small training dataset
Generative Adversarial Networks (GANs)
Continuous values of solubility
More applicable values compared with binary values
Experiments in vivo Machine learning in silico
https://arxiv.org/abs/1806.11369
11. 1) Demand Forecasting: Machine Learning Based techniques such as
Artificial Neural Networks, Support Vector Machines etc.
Any function within the time series data can be learned.
2)
Allowing connectivity of
processes, products
and people
Reducing time to
market for final
products
Real-time monitoring of storage conditions
for sensitive products with shorter shelf-lives
helps improving drug safety.
AI for healthcare: Precision therapies
To Ensure right Cell and Gene therapies delivered to right patient:
Chain of Custody, Chain of Identity, Real time tracking.
12. Mission
Time
reduction 10x
Economic
Value
Quantitative
Relationship
AMD (Accelerated
Materials Development)
aims combining AI and
Materials.
Data-driven features
have already replaced
hand-crafted features in
speech recognition,
machine vision, and
natural-language
processing.
Carrying out the same
task for virtual
screening, drug design,
and materials design is
a natural next step.
Apply ML/AI to accelerate
material process-
optimization and discovery
stage.
Taking into account both the
initial conditions for
experiment as well as the
whole synthesis process
parameter. Enabling better
reasoning in material world
with the aid of data science.
Serve data mining,
data management, data
analyzing and data
interpretation role.
Labor
reduction
Technical
Excellence
AI for Smart Materials: Accelerated manufacturing
http://www.accelerated
materials.org
14. AI and IoT funding opportunities for SMEs
MATCHING THE INDUSTRY TO THE TOP AI MINDS IN SINGAPORE TO SOLVE THEIR AI PROBLEM STATEMENTS.
100 Experiments (100E) consists of significant industry-surfaced problem statements, brought forward by the project
sponsor, for which no existing commodity-off-the-shelf (COTS) solution exists, but for which existing AI technologies can be
quickly built with limited research. 100E funds the assembled academics and researchers in the IHLs and RIs up to
$250,000 to work on the project sponsor’s problem statement.
To develop industry-ready capabilities
towards deepening alignment of public
sector research, and to develop
multidisciplinary and integrated programmes
with early industry involvement.
INDUSTRY ALIGNMENT FUND - PRE-POSITIONING PROGRAMME (IAF-PP)
https://www.aisingapore.org/100e/
https://www.nrf.gov.sg/rie2020 Advanced Manufacturing and Engineering | Health and Biomedical Sciences |
Urban Solutions and Sustainability | Services and Digital Economy
15. AI and IoT tools and learning materials
ML and DL
https://www.tensorflow.org/ – TensorFlow™ is an open source software library for
numerical computation using data flow graphs.
https://pytorch.org/ – Tensors and Dynamic neural networks in python with strong
GPU acceleration.
https://keras.io/ an open source neural network library written in Python. It is capable
of running on top of TensorFlow, Microsoft Cognitive Toolkit or Theano.
Matlab Deep Learning – Matlab Deep Learning Tools
Microsoft Cognitive Toolkit – a unified deep-learning toolkit by Microsoft Research.
Andrew Ng Machine Learning https://www.coursera.org/learn/machine-learning
Courses by Udacity https://www.udacity.com/courses/georgia-tech-masters-in-cs
Jeremy Howards Practical Deep Learning Course http://course.fast.ai/
Advanced deep learning: deeplearning.ai Deep Learning Specialization
IoT and blockchain platforms
https://www.ibm.com/internet-of-things/spotlight/blockchain
https://www.ibm.com/blockchain/hyperledger
https://baas.zhigui.com/login
AutoML systems: Throughout recent years several off-the-shelf packages have been
developed which provide automated machine learning http://www.ml4aad.org/automl/
16. Thank you!
Department of Chemical and Biomolecular Engineering
National University of Singapore
chewxia@nus.edu.sg
http://sse-wang.strikingly.com/