FEATURE Manufacturing Data & Acquisition
The PAT knowledge manager is connected to all different elements of the PAT system
found in some processes. These are Design of Experiments (DoE) and multivariate analysis (MVA). The fi rst tool, DoE, allows manufacturers embarking on a PAT implementation journey to design an experimental plan with the objective of collecting the data and samples necessary to build multivariate calibration models. In practice, this strategy defi nes the most eff ective way to collect key data – gathering the maximum amount of information while using the minimum amount of time and resources. When preparing to design and execute a DoE, it is crucial to clearly defi ne a suitable analytical target profi le (ATP). This identifi es what should be measured and why, in addition to when and how the measurements should take place. By having a clear answer to these points, it is possible to select the most eff ective analytical instruments, sensors and analysers. Once relevant, high-quality measurement
data are collected, as defi ned by the DoE set-up, a knowledge management software can ease the building of data sets and analytical results. These can then be exported to the selected MVA package, to build the instrument calibration models. It is important to note that MVA is data- driven, i.e., it does not rely on a priori system knowledge, therefore it generates information even in the absence of an existing fundamental or mechanistic understanding of the process. Even more, it minimises biases from operators and chemometricians. To create a suitable MVA and predictive model, it is possible to use diff erent methods, such as principal component analysis (PCA), partial least squares (PLS), discriminant analysis (DA), factor analysis (FA), multivariate curve resolution (MCR), maximum autocorrelation factors (MAF) or a combination of these. Each alternative can provide a diff erent insight into the
automationmagazine.co.uk
The PAT knowledge manager dashboard, with overview of CPPs, CQAs and real-time reporting of the manufacturing process
relations between the data collected and CQAs. For example, PCA provides a linear combination of variables, while FA off ers a measurement model of a latent variable. Therefore, it is crucial to select a suitable MVA method that can provide the most relevant insight into a specifi c process. Once a suitable MVA model has been
fi nalised and validated, it can be used as a component of a digital twin of the physical process. Provided with suitable data sets, this allows manufacturers to avoid running real experiments to test and refi ne PAT methods without the need to run a physical process and incur raw material usage and staffi ng costs. In eff ect, businesses can use the digital twin to simulate the process itself and vary its CPPs without using any physical resources. Furthermore, as more operational data are collected, the MVA and digital twin can be improved, delivering ever more accurate actionable insight and supporting continuous improvement.
Knowledge management The MVA and digital twin are also key components to establishing a multivariate statistical process control (MSPC)-based closed-loop, quality-centric control system within a PAT knowledge manager, such as Optimal’s synTQ. The PAT knowledge manager needs to connect to physical systems such as instruments and unit operations, interface to third party control and software systems, run complex PAT methods (orchestrations) and interface with all users, in order to make this quality- driven control approach a reality. In this way, it is possible to collect measurements and make quality predictions in real time, while the manufacturing process is taking place, and to check if the data is within specifi ed parameters to deliver high-quality results. If not, then it is possible to manually or automatically adjust CPPs to maintain optimal conditions and obtain products of
consistent, high quality. Setting up a control system for
continuous processing is more complex than for batch processing. In a batch process, it is possible to break down the control functions into diff erent, independent and confi ned unit operations. The control system for continuous processes, however, needs to make the various stages work together simultaneously and seamlessly. If something is not in sync, it can aff ect the entire manufacturing line. Therefore, similarly to batch processes, it is important to make sure that the operator interface can provide a comprehensive view of the process, from incoming raw materials to end products, in order to easily detect any unwanted deviation – and provide information on the reason for the deviation. This platform, also known as a “PAT knowledge manager”, is connected to all diff erent elements of a PAT system and should off er an easy to use dashboard, where all users, from operators to management, have an overview of CPPs, CQAs and real-time reporting of the manufacturing process. In addition, it should clearly fl ag anomalies and parameters that require attention, so that any issue can be dealt with before it irremediably aff ects end-product quality. By combining all the diff erent elements of PAT together, PAT knowledge management solutions can support complex PAT methods, making the diff erent stages work in complete synchronisation. As a result, manufacturers can ensure that they have a holistic monitoring and control system for their continuous processes.
CONTACT:
Optimal Industrial Technologies
www.optimal-tech.co.uk
Automation | April 2021
15
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46
