A Novel Continuous Pharmaceutical Manufacturing Pilot-Plant

Advanced model predictive control

Ravendra Singh,  C-SOPS, Department of Chemical and Biochemical Engineering Rutgers, The State University of New Jersey

Currently, pharmaceutical industries are going under paradigm shift from traditional batch to novel continuous manufacturing. Such a continuous plant has been built at C-SOPS which is being adapted by several pharmaceutical companies. Real time process control is highly desired for efficient Quality by Design (QbD)-based continuous pharmaceutical manufacturing. A control system ensures the predefined end product quality, satisfies the high regulatory constraints, facilitates real time release of the product, and optimises the resources. In this work, an advanced Model Predictive Control (MPC) system has been developed and implemented into direct compaction continuous pharmaceutical tablet manufacturing pilotplant. The closed-loop process performance has been practically demonstrated through experiments. It has been found that, MPC performs better than PID.

Continuous Manufacturing (CM) is evolving into a preferred platform for pharmaceutical products involving solid dosage forms. The US Food and Drug Administration (FDA) has recently approved some pharmaceutical products to be manufactured in continuous line, with several others on the way1. Therefore, the pharmaceutical industry is going through a paradigm shift from conventional batch manufacturing to advanced CM2. One advantage of CM is that the product quality can be controlled in real time thereby opening up the possibility of achieving Quality by Control (QbC) and Real Time Release (RTR) paradigm. Indeed, CM has a strong impact on drug quality1.

The objective of this work is to demonstrate the real time advanced model predictive control of a novel continuous direct compaction pharmaceutical tablet manufacturing pilot-plant.

1. Process and Pilot-plant description

The snapshot of the pilot-plant developed at Rutgers along with the control system overview is shown in Figure 1 (whole plant is not shown).The pilot plant is built in three levels at different heights to take advantage of gravitational material flow. The top level is used for feeder placement and powder storage, the middle level is used for delumping and blending, and the bottom level is used for compaction. Each level consists of 10x10 square feet working area. There are three gravimetric feeders (K-Tron)-with the capability of adding more- that feed the various formulation components (API, excipient, lubricant etc.). A co-mill (Glatt) is also integrated after the feeder hopper primarily for de-lumping the powders and creating contact between components. The lubricant feeder is added after the co-mill to prevent over lubrication of the formulation in the co-mill. These feed streams are then connected to a continuous blender (Glatt) within which a homogeneous powder mixture of all the ingredients is generated. The chute is placed in between blender and tablet press. The chute has interface to integrate the sensors. Finally, the outlet from the blender is fed to the tablet press via feed frame. The pilot-plant consist of some inbuilt local level control system in feeder (to control mass flow rate) and tablet press (to control main compression force) unit operation. In this article, the implementation of supervisory control system consisting of feedback and feed forward loops is described.

2. Advanced hybrid model predictive control architecture

2.1. Development of control architecture

A control architecture for direct compaction continuous tablet manufacturing process has been developed as shown in Figure 2. A combination of MPC and PID has been used to utilize the advantages of both control algorithms. As shown in the figure, the drug concentration has been measured using NIR sensor. The measured drug concentration is the input for the master controller, which generates the feeder ratio set point. Based on this ratio set point and the total powder flow rate, the individual flow rate set points for API, excipients and lubricant feeders are calculated and then controlled by manipulating the respective feeder RPMs using built-in feeder controllers. The powder level in chute has been controlled by manipulating the turret speed. In the tablet press, the tablet weight is controlled through a cascade control arrangement using one master loop and one slave loop. Master loop is used to control the tablet weight that provides the set point of slave controller which has been designed to control the Main Compression Force (MCF). MCF has been controlled by manipulating the fill depth. A feedforward control loop has been also added. The real time measured powder bulk density is the input to feedforward controller (FFC) that’s manipulating fill depth. FFC has been added to take proactive actions to mitigation the effects of variations in powder bulk density.
The tablet hardness has been controlled by manipulating the punch displacement.

2.2. Implementation of control architecture

Two types of sensors are used for real time monitoring and control of continuous pharmaceutical tablet manufacturing process: spectroscopic and non-spectroscopic. In terms of control system implementation, the basic difference between these type of the sensor is that the spectroscopic sensors need addition tool for real time prediction while the signal from a non-spectroscopic sensor can be directly send to the control platform. The implementation procedure of a control system is shown in Figure 3. The implementation of the control loop in case of spectroscopic sensor has been previously reported and therefore has not been repeated here2. A non-spectroscopic sensor can be directly integrated with the control platform as shown in Figure 3. A sensor that generates 4-20 mA signal has been considered to demonstrate the concept of control system implementation. As shown in the Figure, the sensor is integrated with the control panel at relay through serial ports. From relay, the signal transmitted to charm and from charm to controller. From controller block (placed in control panel), the signal transmitted to control platform where the control loop is implemented. In the control platform, the mA signal is converted into a relevant variable to be monitored and control. From control platform, the signal goes to the plant using the standard communication system (OPC, serial ports, profibus).

3. Results and Discussions

The advanced model predictive control system has been successfully used to control the continuous direct compaction pharmaceutical tablet manufacturing pilot-plant. The closed-loop response in case of non-spectroscopic sensor has been evaluated here. The closed-loop performance of MPC for spectroscopic sensor has been previously reported2. The powder level control has been considered here as a demonstrative example. The powder level is measured in real time using Triflex (Fluidwell Ins.) sensor. The closed-loop response is shown in Figure 4. As shown in the figure, the powder level was controlled at 50 per cent. Then a step change in powder level set point has been made from 50 per cent to 60 per cent. The result shows that, the MPC brings the powder level signal back to the new set point. At steady state, the difference between set point and actual value is very less meaning that the perfect control has been achieved. No overshoot has been overserved and settling time is very less. There is some process delay and therefore powder level took some time to response. But, MPC has taken this dead time into account very efficiently. When the signal become stable at 60 per cent powder level then a bigger step change from 60 per cent to 40per cent has been made. The MPC again bring the signal back to new set point efficiently meaning that the developed model predictive controller is valid for a bigger range of powder level. Finally, the powder level set point has been changed again to original set point (50 per cent). As expected, the MPC was able to bring the signal back to 50 per cent. The developed MPC is robust as it can be seen in actuator response. The results demonstrate that; the powder level can be controlled at desired set point (50 per cent) very effectively through MPC. The controller can track the set point and can handle the fluctuations in flow rate. The performance of the MPC has been compared with PID and found to be better.

4. Conclusions

An efficient control system is required to operate the continuous tablet manufacturing pilot-plant safely and to achieve the predefined end product quality consistently. A systematic procedure to implement the control system in case of both spectroscopic and non-spectroscopic sensors have been developed and applied to the continuous pharmaceutical manufacturing process. A novel method for powder level control has been developed. It includes, an advanced model predictive controller and an electric filed based sensor. The powder level exhibits an integrating response meaning that it is a non-self-regulating process and must be controlled. The proposed systematic control framework supports the paradigm shift of pharmaceutical tablet manufacturing from conventional QbT-based batch-wise, open-loop production to QbD-based continuous, closed-loop production.


This work is supported by the National Science Foundation Engineering Research Center on Structured Organic Particulate Systems, Rutgers Research Council and US Food and Drug Administration (FDA).


1. Yu, L. (2016). Continuous Manufacturing Has a Strong Impact on Drug Quality. U.S. Food and Drug Administration (FDA). https://blogs.fda.gov/fdavoice/index.php/2016/04/continuous-manufacturing-has-astrong-impact-on-drug-quality/

2. Singh, R., Sahay, A., Karry, K. M., Muzzio, F., Ierapetritou, M., Ramachandran, R. 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. 2014b; 473, 38–54.

--Issue 28--

Author Bio

Ravendra Singh

Ravendra Singh is Research Assistant Professor at C-SOPS, Department of Chemical and Biochemical Engineering, Rutgers University, NJ, USA, working in Pharmaceutical System Engineering research field. He is also serving as a manager and key researcher of “multi million dollars projects funded by NSF, FDA and pharmaceutical companies. He is the recipient of prestigious EFCE Excellence Award given in Recognition of an Outstanding PhD Thesis, from European Federation of Chemical Engineering. He has published more than 44 research papers, written 7 book chapters, presented at over 89 international conferences and currently editing one pharmaceutical book from Elsevier. He is actively serving as a conference session chair, Journal reviewer and guest editor.