It’s a challenge to transition to the cloud under most circumstances, but the unique nature of manufacturing poses even more difficulties.
A recent guest on Unleash IT is Arila Barnes, Chief Cloud Architect at Standard Industries, a company with a 140-year history in modern industrialism and a rich understanding of what it takes to succeed in the world of cloud.Among many other things, Arila and host André Christ discussed how manufacturing companies can transition to the cloud without stopping their production lines and while managing terabytes of data per week.
They also spoke about:
Here's a summary of their conversation.
“Unlike an on-prem situation, where there needs to be a lot of planning and a lot of investment up front of the computer resources, cloud gives the advantage of being able to scale that, as needed, based on demand.” — Arila Barnes, Chief Cloud Architect, Standard Industries
Cloud offers manufacturing groups the potential to enhance visibility across entire fleets of plants. Of particular benefit to Arila and her team, cloud adoption also helps drive standardization via key performance indicators and by synchronizing and providing data for new types of analytics.
For Standard Industries and its numerous operating companies, such data extends to a spectrum of traditional and emerging manufacturing use cases, processes, and supply chains. In the last couple of years, for example, the business has made a significant investment in bringing industrial IoT platforms and capabilities to its plant operations.
That being said, many things have naturally changed in their attempt to visualize and manage variables across its heterogenous production sites — chief among them, cloud and visualization algorithms. Utilizing the cloud, machine learning, and IoT platforms to improve visibility in this process is nonetheless a huge undertaking.
“Most visualization algorithms out there actually involve a lot of coordination," explained Arila. "a lot of trial and error, a lot of partnerships with universities, and cross-organizational partnerships with subject matter experts.”
These challenges are outweighed by the benefits of having readily available data and scalable computational power. “Unlike an [on-premises] situation requiring a lot of planning and a lot of investment upfront of computer resources, cloud gives the advantage of being able to scale that, as needed, based on demand,” Arila said.
Complete data centers in the past had to be sized for temporary projects and then be consistently updated. Today, however, cloud computing allows for a stable and adjustable approach to migrating plant components — all of which without fear of oversizing systems and on a company's own schedule. The cloud also provides data centers with initial setup advantages and ongoing maintenance that keeps obsolescence at bay.
But beyond using cloud for dynamic storage options, Arila shared that Standard Industries is investing in infrastructure-as-code to become cloud agnostic in the future. They are also building a data platform with Google's managed services to store and format data (raw and transformed) to enable various machine leaning reporting capabilities.
A major challenge for manufacturing is the nature of the business (i.e., production lines must run 24/7).
“Planning when to introduce changes can be challenging to coordinate around plant downtime and to be able to distill the use cases that can justify this step of taking the risk,” Arila said. Manufacturers are understandably risk averse to anything that can potentially disrupt day-to-day operations.
To protect the production line, Arila advises taking the path of least resistance. For example, if there are backup processes already in place, she recommends leveraging these services at the same time one determines which incremental changes are needed. As data gets configured and validated, pipelines can then be expanded with other datasets.
Other strategies include having parallel streams that are independent of production systems that can then be integrated in a coordinated fashion.
“We are not solving all of those pieces ourselves,” Arila said. “We rely on partners that have deep expertise in the industrial IoT platform to make sure that those concerns are addressed.”
From an industrial perspective, the importance of having good network connections is on full display when it comes to establishing performance analytics on certain production segments with high sensor volumes. Arila emphasized that it’s difficult but far from impossible to coordinate the local load of analytics when leveraging cloud to improve performance analytics of wind farms, solar farms, and hydro plants.
Other challenges of cloud adoption in manufacturing include but aren't limited to:
For analytics purposes, though storing the volume of sensor data at Standard Industries is a monumental task, Arila explained that she and her team focus considerably more time on how to actually qualify data. Upon grading data (e.g., by date, compliance measures, etc.) it then becomes much easier to rationalize cost-intensive cloud storage options, identify common infrastructure models to use as foundations for digital spaces, and set clearer schedules for data engineering teams.
“Data engineering is the people that wrangle all those data because it's not perfect. That's the challenge of industrial data. It's not perfect.” — Arila Barnes
These architectural challenges are solved thanks to numerous teams at Standard Industries.
Data engineers, Arila explained, can efficiently understand the sources of data, their destination, and the minutiae of transformations that occur during their handling. To preserve both raw data and various other levels of data depending on transformation purposes (i.e., for privacy concerns), they are particularly adept at applying Extract, Transform, and Load (ETL) patterns.
The size of the data engineering team depends, of course, on the size of the use case and a company's stage of transformation. Team sizes can range from, say, three people if you’re starting on a proof of concept to a team of hundreds responsible for managing data pipelines in the cloud. These are the people who wrangle the data in whatever state it arrives. It can easily take 300 engineers to convert raw data into meaningful analytics.
“That's the challenge of industrial data. It's not perfect. It's dirty.” — Arila Barnes
DevOps teams are also an essential component of Standard Industrial's cloud journey. This team's priority is configuring the infrastructure itself in regard to compute, storage, and security controls. By having all this in code, it makes it easy to bring about repeatable patterns, provide audits, and layer security concerns — not to mention avoid lock-in to cloud vendors.
Containerization is an essential technology for delivering consistent, manageable, and portable analytics via automated processes. To support this technology, IoT field engineers at Standard Industries help define all plant- and device-specific setups (and constraints...) to better align manufacturing domains to what's actually in the cloud.
In the next three to five years, Arila predicts the biggest changes in cloud technologies in manufacturing will be in the maturation of industrial IoT platforms and their specialization for specific industrial use cases or industrial verticals.
Another trend expected is more investments in analytics given the surfeit of connected data. Depending on a manufacturer's needs, this could mean more complex analytics, vision analytics, or even virtual reality.
And at the plant level, Arila predicts more autonomous equipment similar to self-driving cars. “Coordinating all of that and making sure all of that flows in a system is a very interesting challenge, and that's the kind of puzzle I like to solve,” she said.
To hear this interview and many more like it, subscribe to the Unleash IT Podcast on Apple Podcasts, Spotify, or our website.