We present a framework for the development of design and control spaces that simultaneously considers the raw material property space (Z), the critical to quality process variable space (X), and the critical quality attribute space (Y). The importance of jointly defining all of these spaces and simultaneously considering the eventual process feedforward-feedback control system is illustrated. It is shown that changes in any one of these spaces or in the control system will greatly affect the other spaces. Justification is provided for the use of multivariate principal component analysis and projection to latent structures methods to define more meaningful raw material design spaces and the use of statistical process control concepts to redefine control spaces.
In the process industries Big Data has been around since the introduction of computer control systems, advanced sensors, and databases. Although process data may not really be BIG in comparison to other areas such as communications, they are often complex in structure, and the information that we wish to extract from them is often subtle.Multivariate latent variable regression models offer many unique properties that make them well suited for the analysis of historical industrial data. These properties and use of these models are illustrated with applications to the analysis, monitoring. optimization and control of batch processes, and to the extraction of information from on-line multi-spectral images.
This paper investigates an approach to modeling and optimizing an industrial tablet manufacturing line for different API and excipient formulations. Multi-block partial lease square (PLS) models are built from historical data on a given class of drug products. The data blocks consisted of data on the mass fractions of API and 11 excipients used in the different formulations, the roller compaction process variables, the tablet press settings and the measured final product quality attributes (tablet weight, hardness, and disintegration time). More than 400 runs are used in the modeling. The multi-block PLS models are first used to show which process blocks and which variables in each of the process blocks are most influential on the product quality variables. An optimization is then performed in the latent variable space of the PLS model to find the optimal combination of settings to use for the critical to quality roller compaction and tablet press variables in order to achieve the desired final tablet properties for a specified drug formulation. This optimization can be used to set up the tableting line prior to running a new formulation or can be used in an on-line mode for making small corrections to the operation of the tablet presses in response to small variations in formulations, raw material properties, and roller compaction operation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.