• By Megan Buntain
  • Digital Transformation
  • IIoT

With the right hybrid architecture, manufacturers can analyze data in its native locations and formats, avoiding the complexities of data transfer or replication.

By 2028, cloud computing and the Internet of Things (IoT) in manufacturing will be poised to achieve the “plateau of productivity,” or the phase when they drive transformational impact on business outcomes, according to business analyst firm Gartner. At this point in their digital transformation journeys, many manufacturers have completed their Industrial Internet of Things (IIoT) pilot projects and are approaching mid- to late-stage adoption in operations.

While the term “IIoT” was coined just a few years ago, the large volumes of data associated with it are familiar to the process control and automation industries. For decades, manufacturers have generated and collected more data than they know what to do with via sensors, legacy digital networks, and various host systems.

But a great deal of data was stranded in process historians and other databases, collecting dust. Today, manufacturers can fully benefit from this data and information in the cloud by using hybrid data architectures coupled with advanced analytics applications.

Agile production and the IIoT

Transitioning to agile production requires optimizing the entire supply chain, from improving overall equipment effectiveness and asset reliability to reducing inventory. IIoT implementations can help organizations clear common optimization hurdles, because they empower staff to access, collect, and analyze more data in near real time. This enables process experts and operators to make timely and productive decisions to enhance product quality, optimize operations, and reduce waste.

With Internet connectivity, IIoT implementations can directly access the vast computing power and scalability of the cloud. Each year, the variability, speed, and volume of process data grows exponentially, rendering IIoT architectures as the only suitable options for compute-intensive Industry 4.0 projects. 

Some of the leading cloud applications and components include digital twins, machine learning (ML) tools, autonomous robot artificial intelligence (AI) repositories, and augmented reality simulators. Each of these use cases requires high CPU processing power, which can be difficult for on-premise servers to provide because information technology (IT) teams cannot scale up the required computing resources on demand.

Cloud computing for manufacturing operations

According to Gartner, when it comes to cloud computing for manufacturing operations, the industry is currently in a “trough of disillusionment,” or a state of lowered expectations. This mindset is largely a result of the unproven idea that IIoT and related databases must feed a central data lake, which is intended to serve as the single source of truth and common access point for all users worldwide.

If this were true, cloud-based data lakes would need to replace all existing process historians—along with other host systems such as those used for asset management, laboratory information, or inventory tracking—to provide the data required for analysis. In reality, this is not the best approach because many legacy on-premise servers, such as those hosting process historians, collect and store highly valuable operational technology (OT) data. The context housed in these rich data archives is required to ensure Industry 4.0 initiatives, such as predictive maintenance via ML, succeed. Attempts to move or copy this OT data to the cloud are often time consuming and costly.

To properly aggregate and analyze the data produced by legacy sensors and infrastructure alongside new “born-in-the-cloud” IIoT sensor data, a bridge is required.

Hybrid approach to IIoT implementation

To address this issue and provide combined access to OT, IIoT, and other data, process manufacturers use a hybrid data architecture approach to:

  • effectively leverage on-premise data, regardless of whether it is connected to the Internet
  • take advantage of ML and advanced analytics opportunities by streaming select data to the cloud
  • use the process domain expertise and skills of their existing workforce
  • reduce in-house data acquisition, storage, access, and maintenance costs
  • increase data availability
  • integrate AI/ML into industrial processes.

This is not a rip-and-replace approach but is instead a bridge connecting traditional manufacturing data infrastructure with cloud-native data to leverage the best data from both sides by creating a continuum of access. Process automation systems can continue to use on-premise or edge data for real-time decision making where low latency is required. Simultaneously, the hybrid model empowers organizations to apply global reporting and compute-intensive tasks, like ML, to cloud-native IIoT data (figure 1).

This approach requires a data abstraction layer to facilitate traffic flow among various data sources (figure 2).

Figure 1
Figure 1. Hybrid data architectures empower manufacturing organizations to leverage IIoT in the cloud for compute-intensive processes, while executing real-time process control using on-premise data.
Figure 2
Figure 2. Data abstraction layers facilitate data access and transfer among multiple data sources, including on premise and cloud databases.
Data abstraction indexes and facilitates access to data in its native locations, a key differentiating point from data-lake functionality. Because data is not copied or moved, its management is significantly simplified. Once data abstraction is implemented, organizations can add advanced analytics applications to simultaneously query and make use of information from multiple, and often previously disparate, data sources. This improves awareness and predictive maintenance capabilities across the organization.

For example, when training and executing ML models, organizations must access maintenance records and historical process data. Staff must then access results to proactively identify issues and adjust the operational model. Abstraction makes it easy for personnel and software applications to access multiple datasets through a single source.

Hybrid data architecture for asset monitoring

Asset monitoring is a critical task for many process manufacturers. For common assets—including pumps, valves, heat exchangers, and others—manufacturers deploy a variety of maintenance methods to maximize productivity over the asset’s life. At the two extremes, these methods include “run to fail” in the most basic case, and condition monitoring for predictive maintenance in more advanced situations.

By monitoring asset performance to detect anomalies in near real time, manufacturers can identify potential issues before failure, reducing unplanned downtime and maintenance costs. When these anomalies are detected, advanced analytics software can generate alerts to inform personnel, so they can schedule inspections and maintenance of affected assets.

These monitoring applications can be scaled to hundreds of assets across multiple sites. Therefore, it is critical to normalize data before generating alerts and to streamline notification paths so the right personnel are informed.

By working together, OT and IT teams can use a hybrid data architecture to achieve these asset monitoring goals. First, OT teams must deploy suitable sensors, in addition to data acquisition and storage technologies, to populate asset hierarchies with data for grouping equipment and devices of a common process or location. These asset hierarchies include sets of metadata collected for each asset of a common taxonomy. Once the hierarchies are in place, assets can be analyzed within process groups, rather than individually, or solely as unrelated assets of the same type.

Next, OT works with IT personnel to ensure the former group can access this data securely by implementing cloud data storage, advanced analytics, and workflow automation tools. IT and data science teams collaborate with OT subject matter experts to configure ML models that create insights and effectively predict asset failure, generating intelligent alerts to improve issue remediation and decrease downtime.

Consider hybrid data architecture

When evaluating hybrid data infrastructure, organizations should consider these questions before implementation:

  • Does the software solution provide access to both OT and IIoT data sources simultaneously, for both querying and analysis?
  • Can users generate dashboards and reports in near real time?
  • Does the infrastructure provide on-premise OT data to the digital transformation team for use in Industry 4.0 applications, such as ML and digital twins?
  • Is there significant cost and effort associated with implementing and maintaining the solution?
  • Can the infrastructure support the organization’s current and future data architecture as it inevitably changes?

Hybrid data architectures empower process manufacturers to more quickly realize the business benefits from their cloud and IIoT investments. By using IIoT data and pipelines, on-premise process data, abstraction, and advanced analytics, organizations can quickly pass through the trough of disillusionment and reach the digitalization plateau of productivity.

All figures courtesy of Seeq

Reader Feedback

We want to hear from you! Please send us your comments and questions about this topic to InTechmagazine@isa.org.

Like This Article?

Subscribe Now!

About The Authors

Megan Buntain is the director of cloud partnerships at Seeq Corporation, a company building advanced analytics applications for engineers and analysts that accelerate insights into industrial process data. She was formerly a consultant with analytics, IoT, and blockchain software and services companies, and prior to that worked at Microsoft for 15 years.