According to analyst firm Gartner, within the next five years two manufacturing technologies will achieve the “plateau of productivity,” or the stage where they drive transformational impact on business outcomes: the internet of things for manufacturing operations, and cloud computing in manufacturing operations.1
This means that manufacturers have largely completed their industrial internet of things (IIoT) pilots and are reaching mid- to late-stage adoption, with many having implemented IIoT and cloud projects in support of operations.
In the process industry, while the term IIoT may have emerged through IT-led initiatives over the last decade, the concept is certainly not new. Operational leaders have been managing real-time process control networks for decades and generating growing amounts of data.
Subscribe to PII and receive:
Our long-established bi-monthly digital magazine *Process Industry Informer*, featuring exclusive articles, news and updates across the Process Industries, plus commentary from our three industry experts:
Sean Moran
Gavin Smith
Dave Green
PLUS:
- Process Industry Update — our weekly e-newsletter
- PII Podcast
- Special messages from our advertising partners
Despite their history with the IIoT, many process industry organisations are still facing data architecture challenges that prohibit them from realising the full benefit of their IIoT and cloud investments. Fortunately, a solution is at hand in the form of hybrid data architectures coupled with advanced analytics applications.
IIoT and the cloud
While the earliest wave of IIoT was for use in smart products and machines, IIoT has many applications to address common industrial operational challenges. These include optimising supply chain management and reducing inventory, improving overall equipment effectiveness and asset reliability, transitioning to more agile modes of production, and other initiatives.
IIoT implementations ensure more real-time data can be accessed, collected, and analysed quickly. This enables process engineers and operators to make just in time—or better yet proactive—remediation decisions to improve product quality, reduce waste, and optimise operations.
Via the internet, IIoT implementations can also directly access the near limitless computing power and scalability of the cloud. The volume, variability, and speed of process data is growing by an order of magnitude every few years.
Therefore, only IIoT architectures are suited for compute-intensive Industry 4.0 innovations, such as machine learning, digital twins, augmented reality simulations, and autonomous robots. On-premises server capacity cannot be scaled up on-demand, making it very difficult for these types of systems to keep up with the CPU loads required by the aforementioned Industry 4.0 use cases.
Cloud computing for manufacturing operations
In the arena of cloud computing for manufacturing operations, Gartner indicates the industry is currently in the “trough of disillusionment.” These lowered expectations are largely based on the unproven idea that IIoT and other databases need to be fed into an industrial datalake as the single source of truth, with ubiquitous access provided to all users worldwide.
Following this path, the bridging of operational technology (OT) and information technology (IT) means replacing existing process historians with a cloud-based data lake. However, on-premises OT servers, such as these historians, often hold the most valuable operational data.
These are the rich data archives, with context, needed to ensure initiatives, such as predictive maintenance via machine learning, are successful. Experiments to move or replicate this OT data in the cloud have proven to be expensive and time-consuming for most organisations.
The primary issue to solve, then, is blending the legacy technology of on-premises brownfield sensors and data infrastructure with the “born in the cloud” IIoT greenfield model, where sensor data is streamed directly to the cloud for storage.
Hybrid data architectures for IIoT implementation
To address this issue, process manufacturers should take a hybrid approach with their data architectures to harness the benefits of combined access to OT, IIoT and other data. This will enable them to:
- Make use of on-premises data, whether it is connected or disconnected from the internet.
- Stream selected data to the cloud to take advantage of advanced analytics and machine learning opportunities.
- Integrate artificial intelligence/machine learning into industrial processes.
- Leverage the skills of employees with process domain expertise.
- Use the cloud to lower data acquisition, storage, and access costs.
This stage between traditional manufacturing data infrastructure and cloud-native data is known as hybrid. A hybrid data architecture is not a rip-and-replace approach, but instead leverages the best of the cloud and the on-premises (or edge) data to provide a continuum of access.
In the hybrid model, on-premises or edge data is used for real-time decision-making where low latency is required. Cloud-native IIoT data is added for global reporting or compute-intensive tasks, like machine learning (Figure 1).
This hybrid approach is only possible when there is a data abstraction layer in place to access these various data sources (Figure 2).
Figure 2: Data abstraction solutions, such as Seeq Cortex, provide on-demand data access to one or many heterogenous data sources, whether on-premises and in the cloud, and whether located at remote sites or in centralised servers.
A significant advantage of data abstraction versus a data lake is that all data is indexed and accessed where it is today. The data is not copied or moved, making it less costly and complex to manage. With data abstraction in place, the next step is to add analytics applications to query and make use of the data for various departments and employees. Use cases such as predictive maintenance benefit significantly from this approach.
For example, historical process data and maintenance records often must be accessed to train and execute predictive machine learning models. The downstream results of the model then need to be shared in field service applications so maintenance teams can be dispatched to proactively address issues.
To get started, teams should begin with one or two high-impact operational challenges where historical data has already been collected and stored in a historian, and where new machine sensor data is available in a cloud database. The teams can then review the options for data abstraction solutions that don’t require the historian data to be moved to the cloud.
Evaluating the hybrid data architecture
When determining whether hybrid is the right approach to bring OT and IIoT data sources online, consider the following questions:
- Can both OT and IIoT data sources be accessed, queried, analysed at the same time, with reports and dashboards generated in near-real time?
- Is the on-premises OT data available to the Industry 4.0 team for use in machine learning, digital twin, or other applications?
- What is the cost and effort involved to implement and maintain the infrastructure?
- Does the infrastructure support the company’s current and future data architecture, which will inevitably change?
The goal of hybrid data architectures is to enable process manufacturers to receive more business benefit from their IIoT and cloud investments sooner. This can be accomplished by using IIoT, data abstraction, and advanced analytics to quickly traverse through the trough of disillusionment to eventually reach the plateau of productivity.
