Big Data 631128639b8d8

Drowning in Raw Data? Lean Principles Can Help

Sept. 1, 2022
The average factory generates 1 terabyte of data each day, says IBM.

A manufacturing plant produces more than just goods. It’s also a data factory. Every asset, product, process and system on the plant floor creates and contributes to a staggering volume of data. One programmable logic controller alone can produce tens of thousands of data tags. According to IBM, the average factory generates 1TB of production data every day.

While plant floor data has always been used for process control, it’s increasingly being leveraged by teams across the enterprise to improve product quality, reduce waste, predict maintenance, prevent downtime and deliver new services and products to customers. Manufacturers who have adopted the Industry 4.0 mindset—or who are at least on their way—are beginning to “manufacture” information from this raw data.

But we all know raw, unstructured data has its challenges. Because it was never intended to be used beyond process control, the data is not correlated for use by enterprise systems in the cloud. It lacks context and standardization. It must be modeled in motion. It may be unstructured or structured in a variety of ways (e.g., machine, transactional and time series). The volume, velocity, and variety of raw industrial data are unparalleled and ever-increasing. In short, it’s a difficult raw material to work with.

Fortunately, manufacturers already have the framework they need to optimize data production and preparation. The same lean manufacturing concepts that have transformed manufacturing over the past three decades also apply to data management.

Consider the three key steps of lean, as defined by James Womack and Dan Jones, two of the founding fathers of the lean movement. According to the Lean Enterprise Institute (LEI), they include purpose, process, and people. Let’s take a look at the three steps and how they can be applied to lean data management.

1. Purpose. A lean initiative should target a customer value, such as price, quality or product availability. Information production is no different. Manufacturers should think about who and what their data is serving. Different customers within the organization need access to this information to solve a variety of issues. Some of these internal stakeholders might include quality assurance, maintenance, finance, supply chain management or order management.

2. Process. After defining the purpose, manufacturers can focus on how they will deliver on their objective. In lean manufacturing, we refer to this as value-stream mapping, which typically involves product and process development, fulfillment and product/customer support. According to LEI, each process should demonstrate value to customers, be capable of producing positive results, be available when needed, enable continuous flow or operation and have the flexibility to move multiple types of products without batching or delays. Here’s a look at how manufacturers can apply similar principles to data preparation:

  • Map the data flow: In this analogy, manufacturers are looking for ways to reduce architectural complexity commonly found in industrial environments where multiple systems are delivering data across the organization. The process may also reveal security gaps and technical debt, which occur when manufacturers have deployed software using the fastest, easiest code available without consideration for future technology needs. This creates inefficiencies when trying to connect new systems and maintain them over the life of the factory.
  • Create data flow: Here, manufacturers are looking at ways to prevent interruptions to data by establishing standardized data models of machines, processes and products before moving that information to consuming applications, such as an ERP or business intelligence system.The organization must contextualize thousands of industrial data points by merging them with information from other systems, adding meta data, standardizing data attribute names and lists and normalizing units of measure.
  • Establish a pull system: This is a type of signal-based system, often using color-coded cards or lights, commonly used for material replenishment. When this concept is applied to data, lean tells us the operations team doesn’t need to push all data to IT. Instead, the organization targets a specific use case and defines how it plans to receive that information based on a set frequency or event. This ensures that manufacturers only store and process the data they need to accomplish that use case in the cloud before moving on to the next use case. Manufacturers can establish this pull system with a combination of IIoT-connected devices and modern data-modeling software that standardizes the data formats and contextualizes the information for consuming applications, such as ERP systems and cloud data lakes.

3. People. To keep lean projects on track, Womack and Jones suggest “frequent improvement cycles for each process” and the creation of a dedicated value stream manager. The same concept holds true for data preparation, though in this case we’re homing in on data governance and cross-functional communication. Manufacturers will need to determine who will oversee the data project and get it past the pilot stage. New jobs titles like data engineer, solution architect and digital transformation director are becoming more popular in manufacturing to meet this cross-functional skills demand.

Enabling Lean Data

Similar to lean manufacturing, lean data management requires the orchestration of people, processes and technology. Consider the case of an auto supplier we recently worked with at HighByte. This particular customer wanted to provide real-time data about defects to its quality team.

It had data streams coming from various operational systems, including OPC unified architecture servers, a quality test stand and a SQL database. Without a standardized way to view the data coming from these systems, the auto supplier needed days, even weeks at times, to collect and curate the data.

It was a resource-intensive process. Lacking a streamlined data flow with the context the quality team needed to initiate continuous improvement activities, the auto supplier implemented a data-modeling solution that allowed it to collect data in real time directly from different sources, standardize the data and then feed that information to Power BI dashboards in Azure.

The data framework freed workers from manual data collection, allowing them to address quality issues faster and improve first-pass yield.

By applying lean principles to data management, manufacturers can reduce data waste—irrelevant information that leads to data overload and hides the real problem they’re trying to solve. Manufacturers can start their lean data journey by first understanding the problem they’re trying to address and then map how that data flows from a specific operation to its final destination.The addition of technologies that can help manufacturers merge data flows from multiple sources, send that information at specified frequencies and create standard data models are key enabling tools for successful data continuous-improvement initiatives.

If data is the raw material of business, the plant floor is ripe with this resource. Let’s use the tools that have transformed manufacturing over the past three decades to transform our industrial data into information and open the doors to the lean data factory.  

Sponsored Recommendations

Voice your opinion!

To join the conversation, and become an exclusive member of IndustryWeek, create an account today!