Issue link: https://www.e-digitaleditions.com/i/1483907
www.biopharminternational.com Manufacturing and Facilities 2022 eBook BioPharm International ® 17 OperatiOns feedback for alarm setpoint alterations, and helps keep the customer informed of needed calibrations. These are some of the functions of a data analytics and machine-learning platform that are being lever- aged for use with the end-user. These functions en- able the SME to view all operations of the equipment to provide operational history, process acumen, KPI analysis, troubleshooting knowledge, and data trans- parency. When these types of operational changes are enacted, the end-user benefits financially. Financial benefits from data analytics With these technical analyses, it can be shown that having a machine-learning solution monitored by an OEM's SME can provide significant operational ben- efits, training, and knowledge management, which results in financial benefits to the end-user. These benefits can be capital circumvention, quality as- surance, alarm avoidance, decreased labor cost for troubleshooting, and reliability. The OEM's SME can work very closely with the end-us- er's customer to provide alarm resolution of nuisance alarms and fix operational issues. This decrease in alarms results in the financial benefits of less trouble- shooting time for operators and more uptime for produc- tion along with increased operator capabilities. For example, one end-user experienced nearly 250 notifications in two and a half months. Af ter one field service trip, the OEM's SME was able to decrease the number of notifications to 27 over the next five months. This was immensely beneficial due to the increase in system uptime and reduced work load of the operators in triaging and resolving the noti- fications. Financial benefits were also realized in decreased labor costs and emergency field callouts, with the added benefit of training opportunities for the customer's personnel. Conclusion The experience highlighted in this real-world ap- plication of machine learning, AI, and outsourcing SME support is the beginning of a new age in phar- maceutical water service that will shape the future of service. Water systems can be monitored 24/7 using machine learning models, anomalies can be detected in real-time, and automated alerts can lead to a timely adjustment to prevent downtime. With this type of support, manufacturers can significantly reduce the time and energy devoted to their water systems. It is difficult in today's market environment to find and retain skilled employees who understand how to operate and maintain a pharmaceutical water system. The case described here demonstrates how a small team made up of one SME and on-site technicians can utilize online models to investigate issues and prevent system downtime. As the model matures, machine learning and AI will adapt to the history and perfor- mance of the equipment on site, providing knowledge and expertise that would be similar to an experienced technician. New deviations are logged (allowing them to be predicted in the future), KPI baselines are up- dated, and system optimizations can be personalized to meet a specific organization's needs. When all of these factors are considered, a manu- facturer can maintain optimized operations while fo- cusing on more value-added opportunities knowing that the SME can be the steward of their water needs and provide immense technical and financial bene- fits while promoting knowledge management within their organization. Reference 1. M. Rober, "Stealing Baseball Signs with a Phone (Machine Learning)," https://youtu.be/PmlRbfSavbI (June 30, 2019). ■ October November December January February March April May Month of Year 0 50,000 100,000 150,000 200,000 250,000 Water Production By Month Gallonage (gpm) Analytics Analytics Series 3 / 6 Series 3 / 6 RO\AI-202 RO Product Conductivity AI-202 RO Product Conductivity RO\AI-202 RO Product Conductivity AI-202 RO Product Conductivity Min: 3.94 Mean:9.84 Range: 13.42 Max: 17.35 Median: 9.32 Standard Deviation: 3.28 Upper Quartile: 12.05 Lower Quartile: 7.31 Middle Quartile: 9.32 Min: 5.30 Mean: 11.79 Range: 12.13 Max: 17.42 Median: 12.08 Standard Deviation: 2.77 Upper Quartile: 13.80 Lower Quartile: 10.24 Middle Quartile: 12.08 Histogram Distribution Top Ten Distribution Histogram Distribution Top Ten Distribution 4 2 0 % 5 10 15 Point Full Distribution General Distribution 5 2.5 0 % 5 10 15 Point Full Distribution General Distribution FIGURE 4. Presents statistics on the months of December 2021 and January 2022 along with each month's consumption rate.