“Smarter” manufacturing control systems are needed to cope with the increased use of outsourced production and multi-product facilities, according to new research. The study, by Sagnik Mitra, a doctoral researcher, and Ganti S. Murthy, PhD, professor, at the Indian Institute of Technology in Indore, examined whether established process control technologies are a good fit for modern biopharmaceutical manufacturing strategies.
And the findings were not positive, according to Murthy, who says “the inadequacies of classical and even some modern control systems and techniques are becoming apparent.”
The challenge, Murthy says, is that traditional systems are designed to monitor and control a fixed number of unit operations in a fixed facility with little modification, which is at odds with the more flexible approaches being used to make medicines.
“The move to contract manufacturing and a need for rapid scaleup and flexible manufacturing are big drivers. It is in such situations that you need more advanced scale-free control systems that can be deployed at larger scales,” he explains. “A few trends that are driving this are: increased use of single-use technologies, the need for reliable supply chains, scalable and just-in-time manufacturing, and increased needs for pharmaceutical products especially biologicals which are more difficult to produce.”
AI versus ML
Fortunately for the industry, the technologies needed to develop more effective process control technologies are already available, notes Murthy.
“With the availability of genome-scale models and advances in metabolic engineering technologies together with the increased computation power, the process control systems can be built to be ever more sophisticated and include not only the real-time process variable control objectives but also advanced strategies that will include the economic and the environmental objectives into the process control,” he tells GEN.
Murthy predicts that artificial intelligence technologies (AI) will play a role in developing expert systems to make some of the real-time choices. Machine learning (ML) systems may also have a role in the development of smarter control systems; however, their use is likely to be more limited, according to Murthy.
“ML technologies may be limited to the development of state estimators—also known as software sensors—rather than the complete plant and process models. The primary reason being that in the pharmaceutical industry we would like to have models that are completely known inside out rather than black-box models,” he says.
“Until we can have explainable ML technologies more widely available, I see limited use of ML technologies in process control. However, AI technologies can play a role in developing expert systems to make some of the real-time choices.”
Biopharma will need to work with the technology sector on the development of smarter control systems, points out Murthy, adding that “More interdisciplinary collaboration between the drug developers, bioprocess engineers, and control system engineers is needed to bring some of the advanced control methods to market.”
With the technologies in place the next hurdle to the wider adoption and acceptance of smart process control will be overcoming industry resistance to change.
According to Murthy, although there would be regulatory challenges associated with the introduction of newer control technologies “the regulatory hurdles are secondary to the risk aversion from the industry as they may not want to move away from the more established practices.”