Digitalization and artificial intelligence in pharmaceutical development and manufacturing

Increased digitalization and artificial intelligence have made inroads in pharmaceutical development and manufacturing for the last decade. The promise of advanced algorithms to streamline and optimize operations, using the vast data pharmaceutical companies produce daily, could usher in a new age of safe and reliable drugs.

 Yet, regulatory and organizational uncertainty may hinder companies from realizing AI's full potential. Model-based operator or human-in-the-loop concepts aim to ease that transition by integrating advanced analytics with human operators' implicit knowledge and experience. It creates opportunities to detect and correct deviations standard control concepts cannot mitigate, e.g., human errors like filling feeders with faulty materials.

 Within CAPRI, RCPE has developed a set of operator concepts for the ConSigmaTM-25 wet granulation line. The proprietary algorithm uses data from a cognitive sensor for granule composition to detect and classify faults in the granulation unit. Additional API- and liquid-content measurements support root-cause-analysis (e.g., equipment failure or material disturbance) of the detected deviation. The algorithm runs in real-time and parallel with the process. The algorithm also uses the data to compare the predicted dissolution of the currently manufactured product with the desired target dissolution. If a deviation exceeds a threshold, the algorithm recommends an appropriate action for the process operator, such as discarding the out-of-spec product or adjusting process settings. The responsible process operator receives notifications as pop-up messages in the Cognitive Automation Platform (CAP). That allows human operators to be more decisive and commence corrective measures earlier. Ultimately these operator concepts will contribute towards more sustainable production practices and significantly shorten production times.

Figure 1: Based on the information in the Cognitive Automation Platform (left), the ConsiGmaTM-25 plant operator monitors the process and eliminates the active disturbances.