(243d) Sensor Modeling

Singh, R., Rutgers, The State University of New Jersey
Sam Cherian, C., Rutgers University
Ramachandran, R., Rutgers University

Process model is an important computer-aided tool that has been widely used for different applications such as virtual experimentations, design of sensing and control architecture, process optimization, sensitivity analysis, operating space identification, feasibility and flexibility analysis. The measurable model outputs should ideally match with the corresponding sensors outputs that have been used for real time monitoring of any plant.  In order to achieve this, process model standalone is not sufficient and sensor models are also needed. Sensor noise and specifications such as accuracy, precision, response time, operating range, resolution, sensitivity and drift can have significant effects on process monitoring and thereby on model performance [1]. Extensive work has been done in process modeling field [2]. However, no attempt till date has been made for sensor modeling. Therefore, a sensor modeling framework is highly desired to develop a sensor model to be integrated with process model. Process model integrated with sensor models should truly represents the plant and thereby the integrated model can be used more effectively for control system design, optimization and any other applications.

A process is well understood when all critical sources of variability are identified and explained and where the product quality attributes can be accurately and reliably predicted. Continuous pharmaceutical manufacturing processes often consist of a series of unit operations. To ensure acceptable and reproducible results, consideration must be given to the quality attributes of incoming materials and their processing ability for each unit operation. This can be tracked with the help of an inline/online monitoring tool. Process-sensor integrated model can be ideally used for virtual process monitoring if all source or critical process variability have been identified and included in modeling.

In this work, a systematic framework including the methods and tools has been developed for sensor modeling. The framework includes a generic methodology, a model library and an ontological knowledge-based system.  The generic methodology provides step-wise guidelines to develop a sensor model. The model library consists of process model that needs to be integrated with sensor model. Examples of process models included in library are unit operations involved in continuous and batch tablet manufacturing process [2], API manufacturing process and fermentation process. Ontological knowledge base consist of commercially available sensors and corresponding performance specifications such as accuracy, precision, response time, operating range, response time, resolution, sensitivity, drift [1]. The process model provides the inputs to a sensor model while the knowledge base system provides the necessary sensor performance specifications. The applications of the framework have been demonstrated through modeling of sensors involved in continuous pharmaceutical manufacturing process.

The inline monitoring tool under consideration here is the Near-Infrared Spectrometer (NIR) [3]. NIR is the most commonly used sensor for monitoring of different variables of the continuous tablet manufacturing process such as blend uniformity, drug concentration and powder bulk density. Therefore, NIR has been considered here to demonstrate the sensor modeling approach. The temperature sensor has been also considered for modeling since it has been widely used in many processes including pharmaceutical. The NIR sensor model is proposed, developed and integrated with the existing unit operation models in the continuous tablet manufacturing process. The main motive of this paper is to simulate and match the process output results with the actual sensor outputs. Efforts have been made to take into account the sensing delay and the noise variation in the sensor models. A detailed analysis of simple sensors like thermocouples, thermistors and resistance thermometer (RTDs) are first taken into consideration. Variables like accuracy, precision, response time, operating range, resolution, sensitivity and drift are considered for the thermocouples, thermistors and the RTDs. This information is then utilized to model the more complex NIR sensor with the same variables. The performance of integrated process-sensor model for control system design has been compared with the one based on without sensor model and found to be better.

The objective of this presentation is two-fold. First to highlight the systematic framework for sensor modeling, and second to demonstrate its applications through continuous pharmaceutical manufacturing case study.


  1. Singh, R., Gernaey, K. V., Gani, R. (2010). An ontological knowledge based system for selection of process monitoring and analysis tools. Computers & Chemical Engineering Journal, 34(7), 1137-1154.
  2. Singh, R., Ierapetritou, M., Ramachandran, R. (2013). System-wide hybrid model predictive control of a continuous pharmaceutical tablet manufacturing process via direct compaction. European Journal of Pharmaceutics and Biopharmaceutics, 85(3), Part B, 1164-1182.
  3. Singh, R., Sahay, A., Karry, K. M., Muzzio, F., Ierapetritou, M., Ramachandran, R. (2014). Implementation of a hybrid MPC-PID control strategy using PAT tools into a direct compaction continuous pharmaceutical tablet manufacturing pilot-plant, International Journal of Pharmaceutics, 473, 38–54.