(537e) Cloud-Based Control of a Robotic Manufacturing Process for Personalized Medicines
AIChE Annual Meeting
Wednesday, November 18, 2020 - 9:00am to 9:15am
Based on these considerations, the availability of a fully automated architecture (i.e. hardware and software) capable of manufacturing personalized pharmaceutical products at scale would be of utmost importance. In this respect, we have developed a modular manufacturing system and a supporting cloud-based software architecture for its control. The modular manufacturing system, referred to as ârobotic clusterâ, consists of i) different manufacturing modules, each of them performing specific fabrication steps, and ii) a transfer system that moves work-in-progress units (WIPUs) from module to module in a distributed and parallelized way. During the manufacturing process, WIPUs are shuttled into a given module, which performs a specific step of the process. Then, the WIPUs are moved into another module, and this process is repeated for all the following operations, including quality control. The sequence of manufacturing operations for a small batch of WIPUs is performed by moving it through the modules until all needed operations are completed. This sequence can be performed in a fully-planned, fixed-scheduled, a flexible-scheduled, or dynamic manner.
Each manufacturing module is connected to a dedicated control system and to other equipment via its control box, enabling the input of commands for the next fabrication steps while transmitting out the data pertaining to the work that has already been performed. All critical process parameters are monitored locally and in the cloud. The strategy pursued here not only ensures the security of the manufacturing system, but also increases its versatility and flexibility. It also makes this architecture suitable for scaling, by deploying many robotic clusters at the same time. Moreover, unlike flexible manufacturing systems described in literature, ours is comprised of a set of swappable units that are readily changed, updated, repaired and maintained. This way, complex operations are divided into simpler steps and for each of them there is a corresponding custom-built module. Therefore, the principal application of this technology is in mass customization, where products can be fully personalized to each order and large quantities of products can be fabricated in parallel to achieve the same scale and throughput of traditional serial manufacturing.
As a proof of concept, we evaluated the suitability of the discussed architecture for the manufacturing of multi-compartment capsules for personalized therapy. The capsules here proposed have a modular structure that consists of multiple inner compartments, which are independent of one another. Each compartment can be filled with different active ingredients and with a range of doses, as well as formulations, for any given drug. Multi-compartment capsular devices are especially advantageous for therapies involving multiple drugs, because they simplify the dosing schedule and have a positive potential impact on overall patient compliance.
Personalized filling of each compartment of the capsule shells was performed thanks to a custom-built robotic cluster, composed of a set of automated filling modules working simultaneously, in parallel. A robotic arm (designed to safely collaborate with human operators) moves the capsular devices form one filling station to the next, and quality control machines check the deposition of the desired amount of active ingredients after each step.
Within this proof of concept, we performed an optimization on the manufacturing steps as a fixed-scheduled sequence. We created a list of variable orders i.e. virtual patient prescriptions of varied precise dosages of multiple drugs, and input them into a schedule optimization algorithm. Said algorithm outputs the sequence of robot arm and module actions which maximizes throughput of the cluster. Generally, this algorithm maximizes the amount of work being done in parallel at once (many batches worked on in separate modules), and schedules the order in which the cluster outputs patient-scale batches to minimize the time taken to complete all batches. To ensure quality control we bound the time between manufacturing step and quality control: in-process controls should be measured more-or-less contemporaneously with the action. No WIPU can wait for longer than 10 min - say, sitting in a buffer or idle in a module - before being processed by the corresponding quality control module. For example, we bound the time between a filling module dosing powdered formulations and a weighing module recording the dosed weight. This scheduling optimization algorithm enables the parallelism of the robotic cluster. It is through this novel planning algorithm that we achieve the same scale and throughput of traditional serial manufacturing.
The automated deposition stage allows the rapid and accurate adjustment of the combination and dosage strength of the active ingredients. Cross-contamination is avoided by enclosing each filling station in a dedicated clean room. The capsule filling systems were validated with preliminarily filling trials, which used powders with different particle size distribution and bulk density. We considered 25 batches of 50 capsule each for each powder formulation. In all cases, the amount of active ingredient deposited by the robots had an average coefficient of weight variation (i.e. the ratio between the standard deviation and the average value of the weight of the deposited powder) below 3%. These results demonstrate the accuracy and the flexibility of the powder deposition systems. Additionally, we use spectroscopic techniques to check the weight uniformity and identity of the pharmaceutical formulations. Every measurement is collected and saved in the cloud based quality control system, allowing the creation, in real time, of a digital master batch record with all the information relevant to each manufacturing step (e.g. time, date, type of machine, weight of drug deposited), which could be accessed only by verified personnel and regulatory entities.