A universal digital control concept for end-to-end manufacturing

Abstract

With the implementation of advanced manufacturing concepts (e.g., continuous operation, integration of different process routes and modular manufacturing), control of the process and the related product quality becomes more complex. Unlike traditional in-process controls and end-product testing, modern high-speed manufacturing does not allow for delays due to measurements between unit operations and requires real-time information about the material state and associated control actions. This study presents a universally applicable control concept, which offers the highest level of control together with flexibility for processing, equipment and automation. The concept can deal with a combination of batch and continuously operated process steps, manual and automated operation, different scales of equipment, and digital and manual data acquisition. It was demonstrated using a compact end-to-end manufacturing line for tablets, which included the active pharmaceutical ingredient (API) synthesis, crystallization, filtration and washing, as well as the formulation part. A suspension of API was directly fed into hot melt extrusion (HME), which was combined with direct compaction (DC). The control strategy was based on data not only from classical process analytical technology (PAT) sensors, but also from the equipment and soft sensors (e.g., the content was monitored and controlled using the soft sensors and the feeder data). The process and quality information were fed into a digital twin of the entire process, which executed the model-based control strategy.

Introduction

In the pharmaceutical industry the increasing demand for flexibility, efficiency and quality requires advanced and more innovative process routes and technologies. Traditional manufacturing typically involves sequential steps (e.g., mixing, granulation, tablet compaction, capsule filling and packaging), which are often performed in batch mode. To ensure product quality, safety and efficacy, well-defined protocols and regulatory guidelines need to be fulfilled (Kapoor et al., 2021, Lee et al., 2015, ÓConnor et al., 2017). However, limitations in terms of efficiency, process robustness, flexibility and time to market exist under the traditional approach. These can be addressed by novel concepts, e.g., modular, continuous, high-speed or end-to-end manufacturing (Mascia et al., 2013, Ge and Yuan, 2021, Roggo et al., 2020, ÓConnor et al., 2017).

Modular manufacturing uses interchangeable unit operations or equipment that can easily be integrated and reconfigured to accommodate different process routes (Ge and Yuan, 2021). In continuous manufacturing, the process runs without interruption, continuously producing the target product (Roggo et al., 2020). Such high-speed manufacturing has the potential for rapid production, which leads to increased productivity. End-to-end manufacturing covers the entire production chain, from synthesis of the active pharmaceutical ingredient (API) to the final drug product at one single site, whereby batch and continuous unit operations can be involved (Mascia et al., 2013).

While many continuous secondary production lines have been reported (e.g., via direct compaction (DC), dry or wet granulation) (Singh et al., 2014; Van Snick et al., 2017, Simonaho et al., 2016, Chavez et al., 2022, Conway et al., 2023), studies on continuous end-to-end manufacturing lines are scarce. Domokos et al. (2020) integrated primary manufacturing consisting of continuous flow synthesis, mixed suspension, mixed product removal (MSMPR) crystallization, continuous filtration and drying combined via a continuous filtration carousel (CFC) with secondary manufacturing. The solids process involved blending the API and excipients via a continuous twin-screw multipurpose extruder and compressing the blend into tablets via an eccentric tablet press. In-line monitoring was limited to the blend uniformity control via a near-infrared (NIR) probe directly after the twin-screw extruder. The entire manufacturing line was realized on the lab-scale without a sophisticated control strategy. Mesbah et al. (2017) developed a fully continuous line on the pilot plant scale, including chemical synthesis, purification, formulation and the final compression into tablets. In this manufacturing line, several synthesis steps were accomplished, with the purification of the target intermediates carried out after each step. A control strategy was established based on model-predictive control (MPC) to maintain the critical quality attributes (CQAs) within their defined limits despite possible uncertainties and disturbances and to manage the complex plant-wide dynamics between the integrated process units.

While classical continuous manufacturing entails a rigid process order without room for flexible sequencing of individual unit operations, modular end-to-end manufacturing offers higher flexibility and interchangeability (Mascia et al., 2013, Ge and Yuan, 2021). It allows customizing the production process to the current market situation. This is especially important during global crises and pandemics, when rapid production of certain medications is essential.

A flexible manufacturing line requires a control strategy capable of addressing all potential issues. Therefore, data on the process and (intermediate) product state is required in real-time, which can be accomplished via process analytical technology (PAT) (Fonteyne et al., 2015). Deviations of the CQAs from their specified limits can be detected, and control actions can be triggered. The simplest option to control such quality events is to divert out-of-specification (OOS) material from the process. A deeper process understanding based on the relationship between the input variables, such as the process parameters (PPs) and the material attributes (MAs), and the CQAs as output variables allows more sophisticated control concepts. Predictive process models can be employed to establish a control concept that allows automated adjustments to the PPs (i.e., manipulated variables) to control the target CQAs (i.e., controlled variables) via feed-back loops. A timely reaction to any deviations from the set points can prevent the generation of OOS material, leading to better material and energy efficiency and cost-saving compared to traditional control strategies (Domokos et al., 2020, Vanhoorne and Vervaet, 2020, Malevez and Copot, 2021, Destro and Barolo, 2022, Scott et al., 2023, Boehm et al., 2024, Neugebauer et al., 2024).

This study focused on establishing a universal control concept for an end-to-end manufacturing line, which can handle different operation modes, data acquisition strategies, automation levels and equipment types. The developed control concept was based on real-time data acquisition and process models for all unit operations. It was fully embedded in a digital infrastructure allowing automation of the complete line. In addition, a novel process route was implemented, i.e., substitution of the drying step between the primary and secondary manufacturing processes. The drying step, which typically follows the purification of the synthesized API, is an energy-demanding and time-consuming unit operation, in which the API characteristics and quality can be influenced by attrition, agglomeration or heat (Domokos et al., 2020, Sarkis et al., 2021, Bano et al., 2020, Ouranidis et al., 2021, Conder et al., 2017). Here, the filtered API was re-suspended and directly fed into a hot melt extruder (HME). The extrudates were pelletized and pressed into tablets in a DC line.