How manufacturers can create a normalised view across production lines

Imagine a manufacturing site with three production lines making the same product, following the same processes, and using the same raw materials.  In theory, these lines should perform comparably, but ask a production manager how their facility is running and the response probably sounds something like,

“Production Line 1 has been off to a bit of a slow start this week, but that is normal for Line 1 and we should still be able to meet our target.  Line 2 is always a problem child and we’re having some quality issues again this week.  The causes never seem to be the same, and we have to watch that line closely.  Line 3 always meets its targets and is running great. It is our most dependable line.”

Even though this site has a talented team with decades of experience, they struggle to identify how to increase the throughput of Line 1 to match that of Line 3 or what should be done to reduce the frequency of quality issues on Line 2.

In many cases, the data needed to answer questions like these already exists but is spread across multiple data sources.  Each source contains a piece of the story.  Pulling together the data needed to address the throughput issue on Line 1 can be extremely resource-intensive and those investments are rarely helpful in addressing the quality issue on Line 2.  To optimise the performance across production lines, we need to break down the data silos and create a normalised view across our lines.

---

 

*Image courtesy of Depositphotos

---

Achieving a normalised view requires addressing a series of data challenges that deliver accessible, trusted data to develop insights across our production lines:

1. Breaking down data silos

This is a well-documented challenge for manufacturers related to integrating OT and IT data, extracting data across generations of equipment, and handling both structured and unstructured data types.  While the majority of manufacturers report that integrating data across generations of equipment is one of their most pressing challenges, increasing adoption of data lakes and growing integration solutions are attempting to support manufacturers’ ability to overcome this challenge.

2. Contextualising and understanding data

One level of complexity deeper, data integrated into data lakes needs to be contextualised so that data consumers can understand which data is linked to which production line.  Naming conventions for ERP, MES, and historical data commonly vary across each system.  Contextualising the data allows a user to quickly find all data related to a specific process on any production line without having to understand every contributing data source and its structure.  While many are attempting to manually address this challenge, automating this process will be critical to understanding industrial data at scale across every production line.

3. Creating a dynamic data model

In short, the goal is not to create a universal data model that can apply to every production line use case, but instead to focus on high impact use cases and deliver dynamic models that will support these use cases. Although data modelling is a well-documented topic (e.g. ISO and ISA-95 standards), implementing standards often proves to be challenging. They are often too simple to meet all the needs of the organisation or too complex to implement in practice.  When thinking about how to model production line data, we must balance the need to be comprehensive with the need to accelerate time to value when delivering use cases.  The data model for digitalisation must have composable and reusable components at the core, but also be customisable to fit the unique needs of a consumer or use case.

While manufacturers have been successfully managing the variance across production lines for decades, those who are able to normalise their production line data will have distinct advantages in increasing throughput, improving quality, and reducing utility consumption across their enterprise.