554
R. W. Kessler and W. Kessler
23.1 Process Analytical Technology (PAT): A Systems
Approach
23.1.1 Road Map for PAT
In PAT, complex data (e.g., spectra) are obtained in real time, then processed and
evaluated using the modern techniques of chemometrics, multivariate data analysis or data mining (see previous chapters in this book). This requires a strongly
interdisciplinary and transdisciplinary approach, because process managers, chemometric data analysts, electronics engineers, process specialists, and chemists with
domain knowledge, e.g., of spectroscopy have to be brought together in a business
environment.
There have been dramatic changes in the manufacturing and processing industry
during the past years. Increasingly sophisticated and holistic quality management
systems like quality by design (QbD) have been developed over the years and strongly
promoted by the US Food and Drug Agency (FDA). However, the present-day
strategy focuses strongly on process optimization and the evaluation of safety risks
in the chemical industry. Much less emphasis is laid on the possibility of producing a
product which matches exactly the customer profile and expectations of personalized
products. This process- and product-functionality design allows a precisely defined
manufacture of the widest range of products with minimized costs. When computerdriven, the inverse model enables the manufacturer to develop new products and
techniques within the defined process limits. Multi-objective optimization can be
quickly, reliably, and cheaply achieved, because the essential steps are known from
the model algorithm and are known as “knowledge-based process management.”
As described in the next chapter, concepts like Industrie 4.0 (i40) or the Industrial Internet of Things (IIoT) are a strategy which describes in detail the holistic
approaches needed to achieve these goals [1–3].
In-process quality and process optimization will only be successful when it is
based on appropriate process understanding, i.e., the analysis of the connection
(cause and effect) between process parameters and the quality characteristics of
the final product with its specifications. PAT within this framework means therefore
understanding the causal relation between measurement and response.
Very often in chemometrics and modeling, only descriptive or statistical knowledge is produced which fits the specific dataset but cannot be used as a general or
global model based on molecular information. Figure 23.1 visualizes the different
levels of knowledge for process understanding [1, 4].
Complex products are usually produced in lines involving several steps.
Depending on the application of the product, each step can be adjusted specifically
for the quality of the incoming material and for the quality of the end product [4].
Furthermore, personalized products and materials are needed to fulfill the customers
demand for smart functionalities. This is the concept of the future production.
R. W. Kessler and W. Kessler
23.1 Process Analytical Technology (PAT): A Systems
Approach
23.1.1 Road Map for PAT
In PAT, complex data (e.g., spectra) are obtained in real time, then processed and
evaluated using the modern techniques of chemometrics, multivariate data analysis or data mining (see previous chapters in this book). This requires a strongly
interdisciplinary and transdisciplinary approach, because process managers, chemometric data analysts, electronics engineers, process specialists, and chemists with
domain knowledge, e.g., of spectroscopy have to be brought together in a business
environment.
There have been dramatic changes in the manufacturing and processing industry
during the past years. Increasingly sophisticated and holistic quality management
systems like quality by design (QbD) have been developed over the years and strongly
promoted by the US Food and Drug Agency (FDA). However, the present-day
strategy focuses strongly on process optimization and the evaluation of safety risks
in the chemical industry. Much less emphasis is laid on the possibility of producing a
product which matches exactly the customer profile and expectations of personalized
products. This process- and product-functionality design allows a precisely defined
manufacture of the widest range of products with minimized costs. When computerdriven, the inverse model enables the manufacturer to develop new products and
techniques within the defined process limits. Multi-objective optimization can be
quickly, reliably, and cheaply achieved, because the essential steps are known from
the model algorithm and are known as “knowledge-based process management.”
As described in the next chapter, concepts like Industrie 4.0 (i40) or the Industrial Internet of Things (IIoT) are a strategy which describes in detail the holistic
approaches needed to achieve these goals [1–3].
In-process quality and process optimization will only be successful when it is
based on appropriate process understanding, i.e., the analysis of the connection
(cause and effect) between process parameters and the quality characteristics of
the final product with its specifications. PAT within this framework means therefore
understanding the causal relation between measurement and response.
Very often in chemometrics and modeling, only descriptive or statistical knowledge is produced which fits the specific dataset but cannot be used as a general or
global model based on molecular information. Figure 23.1 visualizes the different
levels of knowledge for process understanding [1, 4].
Complex products are usually produced in lines involving several steps.
Depending on the application of the product, each step can be adjusted specifically
for the quality of the incoming material and for the quality of the end product [4].
Furthermore, personalized products and materials are needed to fulfill the customers
demand for smart functionalities. This is the concept of the future production.