Webinar
on
Multi-dimensional
particle characterisation from in-situ sensors
1st June 2023 at 3pm (UK time)
In-line
PAT are useful for measurement of particle size distributions (PSDs) in a
non-destructive manner and with high time resolution inaccessible off-line.
These measurements can be useful for process monitoring and control, and
population balance modelling. Different in-line sensors have different
preferred operating ranges in terms of particle size. For example, imaging is
limited by resolution when sizing small particles; FBRM can suffer from chord
splitting.
This
work aims to quantify operating ranges to inform choice of sensor and analysis.
PSDs via in-line image and chord length distribution (CLD) analysis are
evaluated against off-line measurements (imaging and laser diffraction) and
manufacturer specification (for standard particles) using representative
metrics. We employ two image analyses: a non-machine learning (non-ML) method,
and Mask-RCNN (ML). We also employ two CLD analyses: a non-ML method and a
ResNet (ML) based approach.
We
obtain error maps which give estimates of error values for a given combination
of nominal particle size and concentration. These maps can be used for choosing
between different analyses or different sensors. We found image analysis to be
affected mostly by particle size for both ML and non-ML methods. Deep learning
image analysis is seen to be insensitive to concentration (within the studied
range). Classical image analysis slightly preferred higher concentrations owing
to few particle detections at low concentrations. Chord length distribution
results show a slight trend down in error with larger particles, but don’t have
any clear trends in error with concentration
This
webinar will be presented by Christopher Boyle, University of Strathclyde.
The
webinar will last no longer than one hour.
To
register please contact Christine.stevenson@strath.ac.uk