Shuo-Huan Hsu

Modeling and Control of Roller Compaction for Pharmaceutical Manufacturing. Part I: Process Dynamics and Control Framework

We derive a dynamic model for roller compaction process based on Johanson’s rolling theory, which is used to predict the stress and density profiles during the compaction and the material balance equation which describes the roll gap change. The proposed model considers the relationship between the input parameters (roll pressure, roll speed, and feed speed) and output parameters (ribbon density and thickness), so it becomes possible to design, optimize, and control the process using the model-based approach. Currently, the operating conditions are mostly found by trial and error. The simulation case studies show the model can predict the ribbon density and gap width while varying roll pressure, feed speed, and roll speed. The roll pressure influences the ribbon density much more than roll speed and feed speed, and the roll gap is affected by all three input parameters. Both output variables are very insensitive to the fluctuation of inlet bulk density. If the ratio of feed speed to roll speed is kept constant, neither ribbon density nor gap width change, but the production rate changes proportionally with feed speed. Based on observations from simulations, a control scheme is proposed. Furthermore, Quality by Design of the roller compactor can be achieved by combining this model and optimization procedure.

High fidelity mathematical model building with experimental data: A Bayesian approach

Abstract Mathematical models of physicochemical systems are usually built in an iterative fashion during the course of an experimental investigation. In this paper, a novel Bayesian approach to model building is presented. This approach is now feasible because of breakthroughs in Monte Carlo sampling procedures and high performance computing, that make it possible to deal directly with the nonlinear mathematical models themselves instead of their linear approximations. By including an error model for experimental data, it is further possible to use nonlinear statistical concepts to test a given model for adequacy against experimental data and prior knowledge, and to place realistic confidence limits on the resulting model parameters. In this paper a model building work flow that takes advantage of these recent advances to enable high fidelity mathematical modeling is proposed. A set of models and their parameters are needed to initiate the process. Probability distributions for the models and their parameters based on available quantitative and subjective information must also be supplied. Finally, an error model describing the heteroscedasticity in the data along with probability distributions for the error model parameters must be generated from exploratory data. Then experiments are designed and data collected. Using Bayes’ theorem, Monte Carlo (MC) or Markov Chain Monte Carlo (MCMC) methods are used to generate a sequence of samples of parameter values for each postulated model. These sets of samples are then used to discriminate among the models using the criteria introduced in this paper. Once discrimination is achieved, a global lack of fit test is introduced to determine model adequacy. After a single adequate model is selected, highest probability density (HPD) intervals are determined for the individual parameters and HPD density regions are constructed for all model parameter pairs. Experiments are then designed to reduce the uncertainty in the joint posterior probability HPD regions. Finally, a sampling procedure is described to properly represent uncertainties in predictions made from the model. The proposed approach is demonstrated by an illustrative problem where three simple models are discriminated and the parameters in the most suitable ones are estimated rigorously.

Modeling and Control of Roller Compaction for Pharmaceutical Manufacturing

Roller compaction is the major process of dry granulation which is attractive to heat or moisture-sensitive pharmaceutical products. Currently, the product quality of roller compaction is analyzed off-line in the quality control lab. In this work, we demonstrate how online process control can be applied on roller compaction using the simulator built in Part I of this paper. Different control strategies are discussed: multi-loop proportional–integral–derivative, linear model predictive control (MPC), and nonlinear MPC. The MPC strategy provides a systematic approach to design the multivariable control system. The simulation results show that the linear MPC can serve as a high-performance control strategy for roller compaction with the trade-off between the control performance and computational complexity. Such enhanced process control facilitates the FDA’s process analysis technology initiative.

A domain-specific compiler theory based framework for automated reaction network generation

Catalytic chemical reaction networks are often very complicated because of the numerous species and reactions involved. Hence, automating the network generation process is necessary as it is quite labor intensive and error prone to write down all the reactions manually. We present an automated integrated framework for reaction network generation based on domain-specific compiler theory using a knowledge base of chemistry rules. The chemistry rules represent basic reaction mechanisms that the reactants can undergo. The systems domain-specific compiler takes the rules and initial reactants as inputs, parses the rule text, generates the intermediate representation, and finally produces the reaction network by interpreting the intermediate representation. We chose the Abstract Syntax Tree (AST) as the intermediate representation because of its transparency and ease of search. The system executes the AST using the initial reactants, and generates the reaction network. The Reaction Description Language (RDL) has been extended to describe the chemistry rules for catalytic systems, and the molecules are represented by Simplified Molecular Input Line Entry System (SMILES). This framework separates the molecules and the behavior of catalysts, represented by the chemistry rules. This approach accelerates the speed of generating hypotheses for building the kinetic models for catalytic systems.

Onto MODEL: Ontological mathematical modeling knowledge management

Abstract In this paper we describe OntoMODEL, an ontological mathematical model management tool that facilitates systematic, standardizable methods for model storage, use and solving. While the declarative knowledge in mathematical models has been captured using ontologies, the procedural knowledge required for solving these models has been handled by commercially available scientific computing software such as Mathematica and an execution engine written in Java. The interactions involved are well established and the approach is intuitive, therefore not requiring model user familiarity with any particular programming language or modeling software. Apart from this key benefit, the fact that OntoMODEL lends itself to more advanced applications such as model based fault diagnosis, model predictive control, process optimization, knowledge based decision making and process flowsheet simulation makes it an in dispensable tool in the intelligent automation of process operations. This paper also discusses the shortcomings of existing approaches that OntoMODEL addresses and also details its framework and use.

Excipient interaction prediction: application of the Purdue Ontology for Pharmaceutical Engineering (POPE)

Abstract A drug product consists of a drug substance and one or more excipients that play specific roles in rendering desired properties to that product, from improvement of flow to control of the release of the drug substance. Inter-excipient and drug substance-excipient chemical reactions are to be avoided and formulators often use heuristics and past experience to avoid potential interactions during drug product development. Multiple tools are present to mechanistically predict chemical reactions: however their utility is limited due to the complexity of the domain and the need for explicit information. In this work, the Purdue Ontology for Pharmaceutical Engineering (POPE) was used to develop an excipient reaction prediction application that made use of structural, material and environmental information to predict reactions

A systematic approach for automated reaction network generation

Abstract In this work, we propose a systematic approach to gather reaction mechanism knowledge and automatically generate reaction networks based on this knowledge. Ontologies are created to model all related information and knowledge. Ontologies for molecule patterns, as well as elementary reaction operations have been created. The reasoning capability provided for an ontology is used to classify molecules or fragments to predefined molecule patterns. A reaction mechanism is modeled as a set of individuals of the defined ontology. The semantic consistency between the elementary steps in reaction mechanism is also validated using the reasoning capability. An execution engine has been developed to automatically generate reaction network, given the reaction mechanisms and molecules. The reaction mechanism knowledge stored in the system can also be easily reused to create new reaction mechnisms.