Panel Discussions

Chair: Stephen Bysouth

Panel: Chris Lampard / AkzoNobel, Ian Riley / Labman, Ian Tovey / Sygenta, John Carroll / CPI, Jim Cawse / CawseAndEffect, Mark Baker / Unilever

Errors in HTE

Commonly HTE is considered to remove all errors, obviously this is not the case and in reality errors need to be considered from the start of any HTE campaign.  Correct consideration of errors will result in considerable benefits from the reliability of HTE systems.  Good practice is to carry out regular calibration and assessment of errors in weighing / dispensing etc..  The challenge in HTE is ensure that the system is engineered to pick up "stupid" errors, i.e. when raw materials run out, this system should behave appropriately.  Equally once a campaign is complete any data which contains systematic errors (caused by equipment failing, wrong raw materials having been used etc.) must be marked as such and failings in meta-data also recorded, otherwise the quality of your dataset will be downgraded.  Best practice is to add meta-data to mark the quality of the data and its meta-data, i.e. bronze, silver, gold data - allowing selective use.

Good statistical practise should be used whenever possible, however in order to achieve efficient robot operation it may be necessary to break from best practise, i.e. the campaign may run samples in a not entirely random due to the inability / inefficiency of making all materials available for the whole campaign, such decisions must be considered when analysing the data generated.

Implementation of Industry 4.0 practises such as self-reporting of system components would increase reliability, i.e. balance saying it has gone out of specification, would ensure the HTE process is always in control.

While HTE is all about new technology it is very important to consider cultural / organisational challenges.  HTE's successful implementation also relies on soft skills, as such similar considerations are needed as for any other change management process.  The adoption of HTE not only drastically changes the working practises, but also the way / even type of data which is presented.  So important to that laboratory workers are on-board with the implementation, good practice is to involve them in the design process, equally higher levels need to properly understand how HTE implemention changes what is possible, i.e. asking for large number of replicates may be easy, changing an ingredient from a liquid to a solid may have big implications on throughput.  While jobs change experience is that no jobs are lost, but throughput increases, allowing more thinking time, i.e. 5 x rheology for Unilever.

Barriers to Effective Introduction of HTFS

Introduction of HTE will result in a different skill balance being needed across the organisation, so mechatronic engineers, data scientists are required to be part of the team.  However, formulation also requires a flexiblity which is not the case with drug discovery or routine medical screening, so the appropriate industrial formulation skills are still required to define the HTE experiments effectively and to achieve scale-up etc..

A key tip is to work with some of the least enthusiastic worker as early adopters can rapidly achieve widespread acceptance.

How Does HTE Change The Way You Work?

HTE 1.0 was all about doing what we were already doing, but just doing it faster, there is some evidence that HTE 2.0 is changing the way in which we are doing formulation.  Differences seen in two ways, people are not just optimising one formualtion anymore, but trying to understand the formulation space, this is producing robust formulations where we now know what the effect of changes in materials or quantities of materials in a formulation will be, so avoiding potential in service issues and we are now mining the space around our formulations looking for new products, a bit like post-its being a failed adhesive.  More prosacically but equally valuably large data wharehouse are now available for data mining, i.e. experimental campaigns for free, data re-use / re-purposing.  Data quality is even more important for HTE 2.0.

Are The Necessary Workers Skilled In HTE Available?

Few universities are producing the required hybrid graduates who have an appreciation of the science, data sciences, mechanical engineering, mechatronics, software engineering and statistics certainly at the undergraduate level, but even at the postgraduate level few are available.

The implication of "Internet of Things" could see products being designed based on direct customer design and / or data being recieved directly from for example a washing machine.  This sort of futurology could be as successful as nuclear 50's vision, to be overtaken by a new disruptive technology, i.e. the internet was not envisaged in 1950's.

Closing Thoughts

Jim Cawse, The Elizabeth Colbourne Memorial Lecturer - As my PhD Advisor told me when I started in experimental statistic many years said "I don't want lots of data, I want good data".

Simon Gibbon, Chair Formulation Science and Technology Group left us to ponder that while HTE 2.0 is going to help close the UK's productivity gap it can't do it alone, so perhaps Future Formulation in May will provide some more pieces of the jigsaw.