In March, APCI CEO and Co-Founder Dr. Joe Pekny traveled to the Bioprocess International Conference in San Diego to deliver a keynote presentation discussing how biopharma manufacturers can optimize their production processes and create synergy with artificial intelligence with a factory digital twin.
Dr. Pekny began his talk by explaining how factory digital twins are built and the ways biopharma facilities can use them to create more accurate production plans and schedules. He then explored how digital twins are likely to evolve over the next several years as use of artificial intelligence continues to grow. Dr. Pekny discussed the possibilities of pairing AI with digital twin capabilities to create enhanced benefits and streamlined data input processes for biopharma plants. “What we’re learning is that if you take MES data or ERP data and use AI to coordinate it with a library of hundreds of existing models, the digital twin can get a draft digital model,” said Dr. Pekny. “It’s a pretty decent model, but the human still needs to audit it, run it, and see what it predicts in terms of throughput.”
Layering AI Capabilities with Human Knowledge to Enhance Factory Digital Twins
As manufacturing organizations grapple with how to implement AI most effectively to enhance their productivity and profitability, Dr. Pekny highlighted the importance of retaining human “natural intelligence” and creativity when using AI tools. In his presentation, Dr. Pekny noted that it’s common for digital twins to produce up to 15% improvement in revenue enhancement and cost savings. However, digital twins have often been a more realistic option for big companies who have the time and manpower needed to create a detailed plant model.
In the future, Dr. Pekny sees AI becoming a tool that can take on the time-consuming, monotonous tasks involved with creating biopharma process models, which could help make digital twins a more accessible application for smaller companies with a limited labor force. Human experts can instead use their time to identify further improvements to the model based on their unique knowledge. “The computer can look through all the data, which would take months of time for a human, and does it in a matter of minutes,” said Dr. Pekny “Then the human can look and say, ‘This is wrong, go back and give me a better model in that area.’ It’s a perfect application of pair programming to get a draft digital model.”
Dr. Pekny also sees biopharma sites using AI to quickly read and analyze years’ worth of comments on production schedules to identify patterns and indicate problematic areas. Humans can then use the AI’s pattern recognition to direct their efforts toward engineering solutions for the processes that frequently create challenges. “When you ask AI questions with considerable creativity implicit in the answer, or certain ambiguity in the answer, AI falls apart,” said Dr. Pekny. “If you ask it to do something with a well-known pattern, AI does great. As a human being, I’m going to use it in areas where I can let it do the drudgery, and we do the creativity.”
