Combining Chemical Robotics and Statistical Methods to Discover Complex Functional Products
Lead Research Organisation:
University of Cambridge
Department Name: Chemical Engineering and Biotechnology
Abstract
Robotics and statistical machine learning have revolutionised manufacturing of mechanical devices and our ability to deal with large amounts of data in many areas of human activities. Chemical manufacturing remains one area where both robotics and statistics have seen very limited uptake. However, both are the likely solutions to many challenges facing the chemicals manufacturing industries. The challenge of sustainable manufacturing requires a rapid switch to locally available renewable feedstocks, new sources of energy and use of rapidly re-configurable intensive reactor technologies. However, conventional methods of process development are slow: this stems from the inherent complexity of chemical processes, including multiple interactions of many components over a very broad range of length and timescales, from behaviour of single molecules to the behaviour of cubic-meter-scale manufacturing reactors. This is where machine learning algorithms could provide the solution, with the ability to rapidly identify the underlying interactions and to design the most useful experiments to perform. To use such algorithms effectively we require a new type of a chemical experiment - highly automated and 'intelligent', equipped with sensors and the ability to link with mathematics, the data handling and the machine learning. The main advance on knowledge that this proposal will realise, is the translation of discovery of new chemical products to their manufacture. For this in this project the chemical robot will learn to identify key process parameters that will impact on scaling the process and will help to develop a scale-up model of a process. This ability will have a tremendous impact in all sectors of chemical products manufacturing, with the main societal impact of better process safety, guaranteed product quality and reduced impact of manufacturing on climate change through reduced emissions and feedstocks waste.
Planned Impact
Automation of the processes of discovery and of development of manufacturing protocols is a breakthrough technology, which will revolutionize high-value manufacturing in the chemical industries. Robotics is already used in several chemical sectors, specifically in agrochemical and consumer products industries, for high-throughput development of formulations; catalysts development companies are using robots for combinatorial high-throughput material synthesis. However, the link of robotics with machine learning and its use to guide molecular and product discovery is yet at a nascent state. The project will solve the identified key fundamental challenges in developing the future industrial platforms based on robotics and AI. The subsequent enhancement of the UK industrial base in high-value manufacturing will contribute to a range of important societal impacts, including: more cost-effective development of medicines, faster access to complex and customised functional products, delocalization of manufacturing, leading to demand for skilled work-force away from large cities, and many others, yet difficult to predict positive societal outcomes. The project will achieve further impact through its contribution to the people-pipeline through training and mentoring of a group of multi-disciplinary scientists with strong core skills in chemistry, chemical engineering and statistics.
Impact deliverables: In summary, to achieve impact we will deliver
- Three open-innovation workshops, facilitating interactions with academic community and potential end-users.
- Interactions with centres of excellence and networks: DaM, CPI and EPSRC funded CMAC, targeting self-selected communities of early adopters.
- Papers, presentations in industrially accessible outlets, and open source research software, promoting rapid uptake of methods and tools.
- Exemplars to demonstrate the methodology and its advantages over current state-of-the-art, providing numerical justification for industrial adoption of robotics and AI in chemical industries.
- A website, social media platforms including videos show casing the various aspects e.g. discovery, DoE, scale up, promoting SET subjects in the wider society.
- Industrial, as well as academic, PDRA secondments and identification of exploitation and commercialisation routes.
- Public engagement via science and engineering events.
Impact deliverables: In summary, to achieve impact we will deliver
- Three open-innovation workshops, facilitating interactions with academic community and potential end-users.
- Interactions with centres of excellence and networks: DaM, CPI and EPSRC funded CMAC, targeting self-selected communities of early adopters.
- Papers, presentations in industrially accessible outlets, and open source research software, promoting rapid uptake of methods and tools.
- Exemplars to demonstrate the methodology and its advantages over current state-of-the-art, providing numerical justification for industrial adoption of robotics and AI in chemical industries.
- A website, social media platforms including videos show casing the various aspects e.g. discovery, DoE, scale up, promoting SET subjects in the wider society.
- Industrial, as well as academic, PDRA secondments and identification of exploitation and commercialisation routes.
- Public engagement via science and engineering events.
Publications
Bai J
(2022)
From Platform to Knowledge Graph: Evolution of Laboratory Automation.
in JACS Au
Cao L
(2021)
Optimization of Formulations Using Robotic Experiments Driven by Machine Learning DoE
in Cell Reports Physical Science
Cao L
(2021)
Automated robotic platforms in design and development of formulations
in AIChE Journal
Felton K
(2021)
Summit: Benchmarking Machine Learning Methods for Reaction Optimisation
in Chemistry-Methods
Janusson E
(2019)
Synthesis of polyoxometalate clusters using carbohydrates as reducing agents leads to isomer-selection.
in Chemical communications (Cambridge, England)
Jorayev P
(2022)
Multi-objective Bayesian optimisation of a two-step synthesis of p-cymene from crude sulphate turpentine
in Chemical Engineering Science
Khan A
(2022)
Designing the process designer: Hierarchical reinforcement learning for optimisation-based process design
in Chemical Engineering and Processing - Process Intensification
Neumann P
(2020)
A new formulation for symbolic regression to identify physico-chemical laws from experimental data
in Chemical Engineering Journal
Porwol L
(2020)
An Autonomous Chemical Robot Discovers the Rules of Inorganic Coordination Chemistry without Prior Knowledge.
in Angewandte Chemie (International ed. in English)
Tibbetts JD
(2021)
Efficient Syntheses of Biobased Terephthalic Acid, p-Toluic Acid, and p-Methylacetophenone via One-Pot Catalytic Aerobic Oxidation of Monoterpene Derived Bio-p-cymene.
in ACS sustainable chemistry & engineering
Zhan CH
(2019)
Controlling the Reactivity of the [P8 W48 O184 ]40- Inorganic Ring and Its Assembly into POMZite Inorganic Frameworks with Silver Ions.
in Angewandte Chemie (International ed. in English)
Description | The project developed a generic methodology for uncovering physical laws from robotically-generated data, using the method of symbolic regression. This has been widely appreciated by the research community and will lead to further work that is required to improve the methodology. The project has developed a generic methodology for developing machine learning models for formulated products. This has led to the set-up of a new project with an industrial partner (BASF) and a new academic partner - Singapore A*STAR IMRE. |
Exploitation Route | The algorithms and code are available through GitHub. Outputs of the ongoing new collaboration will be both publications, equipment designs and codes released publicly through GitHub. |
Sectors | Chemicals |
Description | Industrial interest to this project resulted in several industrial studentships awards to the project team. University of Cambridge has set-up a facility to support SMEs in UK to enter into the digital chemical technologies - iDMT (https://idmt.online). |
First Year Of Impact | 2020 |
Sector | Chemicals |
Impact Types | Economic |
Description | Chemical Robotic Workflows for a High Throughput Digital Laboratory |
Amount | £100,000 (GBP) |
Organisation | BASF |
Sector | Private |
Country | Germany |
Start | 01/2020 |
End | 12/2024 |
Description | Development of Multistep Reactions in Pharma |
Amount | $1,800,000 (SGD) |
Funding ID | C4 |
Organisation | A*STAR Graduate Academy |
Department | PIPS |
Sector | Academic/University |
Country | Singapore |
Start | 05/2019 |
End | 05/2022 |
Description | Innovation Centre in Digital Molecular Technologies |
Amount | £4,994,642 (GBP) |
Organisation | European Commission |
Department | European Regional Development Fund (ERDF) |
Sector | Public |
Country | Belgium |
Start | 01/2021 |
End | 12/2023 |
Description | Pharma Innovation Platform Singapore (PIPS) |
Amount | $1,800,000 (SGD) |
Organisation | A*STAR Graduate Academy |
Department | PIPS |
Sector | Academic/University |
Country | Singapore |
Start | 11/2020 |
End | 12/2022 |
Description | Towards plug and play rectification columns |
Amount | £100,000 (GBP) |
Organisation | BASF |
Sector | Private |
Country | Germany |
Start | 09/2019 |
End | 09/2023 |
Description | UCB Biopharma SPRL PhD studentship |
Amount | £80,550 (GBP) |
Organisation | UCB Pharma |
Sector | Private |
Country | United Kingdom |
Start | 09/2018 |
End | 09/2021 |
Title | CSD 1903343: Experimental Crystal Structure Determination |
Description | Related Article: Eric Janusson, Noël de Kler, Laia Vilà-Nadal, De-Liang Long, Leroy Cronin|2019|Chem.Commun.|55|5797|doi:10.1039/C9CC02361E |
Type Of Material | Database/Collection of data |
Year Produced | 2019 |
Provided To Others? | Yes |
URL | http://www.ccdc.cam.ac.uk/services/structure_request?id=doi:10.25505/fiz.icsd.cc21wl5q&sid=DataCite |
Title | CSD 1903344: Experimental Crystal Structure Determination |
Description | Related Article: Eric Janusson, Noël de Kler, Laia Vilà-Nadal, De-Liang Long, Leroy Cronin|2019|Chem.Commun.|55|5797|doi:10.1039/C9CC02361E |
Type Of Material | Database/Collection of data |
Year Produced | 2019 |
Provided To Others? | Yes |
URL | http://www.ccdc.cam.ac.uk/services/structure_request?id=doi:10.25505/fiz.icsd.cc21wl6r&sid=DataCite |
Description | 3rd International Conference on ML/AI in (bio)Chemical Engineering |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | An online research conference. |
Year(s) Of Engagement Activity | 2020 |
Description | DaM Predictive Scalability |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | A conference organised by this project and another EPSRC-funded project under the auspice of Dial a Molecule Grand Challenge Network. The conference was held at GlaxoSmithKline (Stevenage) and involved international speakers. |
Year(s) Of Engagement Activity | 2018 |
Description | ML and AI in (bio)chemical engineering |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | Conference organised by this research project. Conference links with practitioners interested in application and development of ML/AI tools specifically for chemistry and chemical (bio) engineering. |
Year(s) Of Engagement Activity | 2019 |
URL | https://www.ceb.cam.ac.uk/news/events/machine-learning-chemical-engineering |
Description | Open Research Meeting |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | Machine Learning in (Bio)Chemical Engineering - a 1 day open scientific workshop, including lecture delivered via Web conference from Singapore. |
Year(s) Of Engagement Activity | 2018 |