Contextualize and Integrate Multidimensional, Segregated Data in Explaining Unplanned Downtime at Multipurpose Lines

The fourth industrial revolution, or Industry 4.0, is widely known for increased productivity and equipment uptime. However, it is minimally embraced by the pharmaceutical industry, which is denoted as the Pharma 4.0 gap. Loes Bekkers master thesis; ‘Contextualize and Integrate Multidimensional, Segregated Data in Explaining Unplanned Downtime at Multipurpose Lines: A Case Study at Bayer’s Product Supply Center in Grenzach’ focuses on the first two stages of the trajectory towards Industry 4.0: visibility and transparency.

Where it started

One important aspect that comes up with the Pharma 4.0 gap is unplanned downtime. Machines break down, need unexpected maintenance or don’t work at full capacity. This is often due to a lack of preparation or ad hoc expertise. However, previous research also linked unplanned downtime to the use of multipurpose lines. Whereas other industries renounce multipurpose lines, pharma is highly dependent on them because of the wide variety of products: 800 products on 10 lines is no exception. Simultaneously, manufacturing performance is particularly lower, while profits are significantly higher, compared to other (manufacturing) industries. By coupling pharma’s inherent characteristics with Industry 4.0 technologies, the focus of this research was on unplanned downtime reduction on multipurpose production lines by the deployment of multiple databases.


As a first step, vertical integration of databases was applied to establish a single overview of manufacturing, in which unplanned downtimes could be compared. This is also essential for future feature selection, as individual databases perform worse than combined. With feature selection, the diverse product characteristics of a batch were related to unplanned downtime during production to infer underlying relations. This selection was done with a combination of four different methods: spearman correlation, VK correlation, mutual information and random forest. This selection ensures that univariate and multivariate statistics are combined with feature’s importance, strength and direction. The impact on unplanned downtime was smaller than anticipated. Nonetheless, it was found that specialty products resulted in more unplanned downtime than generic products, and variable effects were found between product lines.
Further, an impact analysis was carried out on six different scenarios. The scenarios were analyzed both qualitatively and quantitively through interviews, scenario- and sensitivity analysis with regards to (1) unplanned downtime; (2) capacity; and (3) profit. The manufacturing changes are separated in two types: (i) product changes and (ii) line changes, visualized on a heat map by likelihood and impact. If was found that:
– Product changes had higher financial impact but a low likelihood.
– Line changes were assessed with a high likelihood but small financial impact.
– Manufacturing changes related to product characteristics (with their respective influence on unplanned downtime) have substantial effects, despite minor effect sizes in feature selection.

Further advice

By using Industry 4.0 methods, the pharmaceutical industry can reduce unplanned downtime and thereby increase productivity and profit. The three-stage methodology proved relevant to analyze multipurpose lines. By using extra databases or new features, further manufacturing possibilities can be explored. With the rising demand for pharma’s products, capacity increases become a necessity. However, the endless flexibility of multipurpose lines can become a burden through unplanned downtime and simplicity is the key to success.