Artificial Intelligence Takes Manufacturing Efficiency to the Next Level

Jennifer Markarian

Jennifer Markarian is manufacturing editor of Pharmaceutical Technology.

Equipment and Processing Report

Equipment and Processing Report, Equipment and Processing Report-11-21-2018, Volume 11, Issue 12

Big data collection, advanced analytic tools, and artificial intelligence enable ongoing improvements to overall equipment effectiveness.

In the pharmaceutical industry, increasing price pressures are driving the need for significant and sustainable improvements in manufacturing efficiency. One of the many areas that can be targeted for efficiency gains is overall equipment effectiveness (OEE). Fortunately, Industry 4.0 tools-including sensors that can connect production equipment to data collection systems via the industrial Internet of Things (IIoT), cloud storage for large data sets, advanced analytics to make sense of the data, and software to make the data understandable and visible to those who can use it-are commercially available. 

OEE at Novo Nordisk

Traditional analytical tools use statistical models, and a platform from Bigfinite also provides artificial intelligence (AI) tools to analyze manufacturing data. The platform collects all types of data and stores it as unstructured data, in an easily accessible manner that can be used to train and use AI models. AI models provide visibility to patterns within the data and predictive insights that traditional statistics cannot.

At the ISPE Annual Meeting & Expo (Nov. 4–7, 2018), Novo Nordisk and Bigfinite presented a recent project that aimed to obtain the promised productivity gains of Industry 4.0 in pharmaceutical finish and packaging. “OEE had ‘flat-lined,’ and Novo Nordisk wanted more and sustainable improvement,” noted Gilad Langer, vice-president of Business Development at Bigfinite, in the presentation (1). For its initial pilot project, Novo Nordisk chose the assembly and packaging of injection pens. 

The first phase of the packaging pilot project was to gather data from the equipment about line speed (i.e., performance), unplanned stops (i.e., availability), and scrap (i.e., quality). In the second phase of the project, advanced analytics was applied to these data to build predictive models for OEE that could predict quality, performance, and availability. Data from the line and different shifts were used to identify anomalies and what was normal to then train the model. Data from production, including operator comments, were included to try to help identify the causes of problems based on pattern recognition. 

Once implemented, the system provides real-time failure diagnosis, explained Langer. “The tools show behavior of the equipment in real-time so the technical support team can identify if the equipment performance is under control or not. The model can predict that a downtime period is likely to occur-in the next 30 minutes, for example-and the operators can then take action to fix it and prevent the downtime.” Using AI, the software “learns” to recognize unusual events, and the model is continually improved. Some of the lines where the system was implemented have shown an OEE increase of 10%, reports Langer.

Novo Nordisk’s goal with this initial project was to develop a system that could be replicated to other manufacturing areas “‘Agile development’ means you make small changes, learn from those, then iterate,” said Langer. “This mindset is different from the typical pharmaceutical industry way of doing things. You need to find a place to start and take the risk.”

Engineering productivity at Eli Lilly and Company

Eli Lilly is employing Industry 4.0 technologies in a variety of manufacturing areas to obtain greater efficiencies, as detailed in a recent Lilly presentation at ISPE Annual Meeting (2). One of these applications is improving engineering productivity (EP) using an information management (IM) system. This EP/IM tool, being developed in-house, will use advanced analytics and will accommodate additional “plug-and-play” tools to enable collection and analysis of large amounts of information, including both company and vendor data (numeric, graphical, and textual). Future releases of the tool are expected to incorporate AI.

The tool is expected to deliver a 5% boost in engineering productivity and have a payback period of less than a year (2). Among the tool’s benefits are the ability to minimize downtime and provide early warning on asset performance issues using predictive and condition-based maintenance. Connecting data sources using IIoT to remotely gather asset operational data is expected to save 75% of the time currently required for maintenance technicians to field-collect measurements that can be used for predictive maintenance.


1. G. Langer and K.M. Larsen, "Applying Quality and Predictive Control," presentation at the ISPE Annual Meeting & Expo (Philadelphia, PA, 2018). 

2. R. Schad and J.Tunell, “Applications of Digital Plant in Pharma Manufacturing, presentation at the ISPE Annual Meeting & Expo (Philadelphia, PA, 2018).