Smart Industry: What is your approach to predictive maintenance?
Burt: Our approach is to offer decision support to site, regional and global industrial-enterprise managers. Site-level maintenance teams performing mechanical break-fix tasks receive accurate and timely predictions to plan repairs and eliminate unplanned events; and site, regional and enterprise managers tasked with managing maintenance budgets and capital expenditures receive benchmarking reports that correlate maintenance activities with avoided costs and overall machine health to ensure the highest-possible production efficiencies and asset-life expectancy. Azima DLI reaches a number of industrial-enterprise audiences by capturing data at massive scale from a number of condition-monitoring technologies, aggregating those data from across the plant (and across plants throughout the enterprise) and using that data both locally and regionally to standardize demonstrated best practices across the enterprise.
Smart Industry: What advantages does this approach offer?
Burt: The Azima DLI approach to PdM varies from traditional approaches in two important ways:
First, we are the only PdM service company to have built a fully integrated, end-to-end solution that incorporates data collectors, a cloud-based support network, an expert diagnostic system and the world’s largest and most experienced team of machine-health analysts. That means we can support customers who want to run programs themselves, but want the comfort of secure data back-up and access to diagnostic experts if required. We can also support customers seeking a fully outsourced, turnkey PdM solution, as well as customers who mix and match in-house with outsourced programs depending on the site.
Second, Azima DLI hosts machine-test data in perpetuity, which means the company retains an unmatched store of machine and site-performance history that is the basis for valuable industry-performance benchmarking information. Benchmarking reveals opportunities to draw on the experience and practices of comparable sites to achieve the best possible yield on maintenance spending and capital expenditures. So while PdM has always been invaluable as a site-level tool, it now has the potential to impact performance across the enterprise.
Smart Industry: How is digital transformation affecting the world of predictive maintenance?
Burt: Digital transformation is the catalyst for unleashing all the benefits of visibility that are already commonplace to other markets, like retail and banking and media. Today’s informed consumer would never buy without comparing options, performance and pricing online. Industry is proving much slower to adopt, mainly because the full implications of digital transformation at the plant are poorly understood.
Despite the hype around machine learning and IoT, value propositions have been elusive, the one big exception being predictive maintenance. PdM has always been effective in the hands of sophisticated users, but very difficult to scale because sophisticated users were rare and their influence constrained to within the four plant walls. Cloud-based solutions change everything by making the successes of sophisticated users visible and replicable.
Our business is increasingly concentrated among sophisticated global/industrial production operations that have recognized the potential for leveraging best practices exposed by the digital transformation.
Smart Industry: How important is it for an enterprise to benchmark itself against peers/competitors regarding maintenance? How do these challenges spur growth?
Burt: Benchmarking drives competition in every other market—why should it not fuel the competitiveness of industry? Automotive companies compete head-on for ratings from well-known independent-ratings agencies, as do television shows for viewers (also tracked by ratings agencies), and restaurants for guidebook star ratings. Even the fitness industry has embraced personal benchmarking with step- and calorie-counting devices and apps. Benchmarks are a ubiquitous method of measuring relative performance, and with digital transformation, benchmarking will become important to industry as well, in at least three important ways:
- Self-benchmarking: The impact of unplanned events that lead to down time, unanticipated costs and lost production is easily measured. If the event can be predicted and the costs avoided, what organization would say “No”? Organizations that commit to rigorous internal benchmarking discover surprising opportunities to standardize on the practices of their most-skilled managers and operators.
- Competitor benchmarking: Nobody likes to be compared to a competitor and found wanting, but if the alternative is staying uninformed about of how competitors are winning in the marketplace, companies that play to win would rather know. Digital transformation will bring competitive benchmarking to industry, and companies committed to leadership will embrace it. PdM is rife with opportunities to move from average to top-quartile performance—from data-collection strategies to response times, from maintenance-planning strategies to tactics for avoiding costs. The potential for improving yield on production assets is material and the gains achieved relatively easily.
- OEM benchmarking: Benchmarking applications are built on compiling data that originates at the individual asset level. If enough data exists to self-benchmark and benchmark against competitors, then enough data exists to compare the individual performance of like assets. OEM benchmarking presents opportunities to improve the performance of installed assets, and to be smarter about investing in new assets.
Smart Industry: What role does big data play in the smart factory?
Burt: The problem with big data is that it’s BIG. Factories are only smart if they can see the data that’s meaningful in the sea of data they’re drowning in. The machine-learning systems that IoT promoters crow about are still learning, and like the driverless car, may be many years from reality. That is not to say that big data will be useless forever, just that the smart in smart factory still refers to people—domain experts—who know which data matters and how to make sense of it.
Much of the knowledge that domain experts apply in day-to-day operations can be coded into expert systems to lighten the load of mundane functions, but the judgment required to address exceptional circumstances can be hard to automate because systems can’t weigh and act on the impact of extraneous variables and events. A good analogy might be today’s commercial-airline autopilots, some of which can taxi from the gate, take-off, land at the intended destination, and taxi to the arrival gate. But I wonder if any autopilot could have landed safely on the Hudson River saving all souls aboard, the way Captain Chesley “Sully” Sullenberg did?
Factory automation will increasingly account for the mundane, repeatable tasks that humans used to perform, but when circumstances diverge from normal operating conditions, decision-making will revert to experts—the so-called “expert assist” model, where humans can override a system that loses its way in unfamiliar circumstances.
Smart Industry: You're a champion of reliability-as-a-service. Why?
Burt: Our best reliability-as-a-service case histories share a common trait: the creation of a solution to a customer’s unmet need in data that we routinely collect to deliver PdM services. Customers have been the genesis for tools to visualize risk, manage compliance, measure avoided cost, and track repair-response times, among other applications. Delivered as a service, reliability data can be repurposed to go beyond typical maintenance functions and remove stubborn obstacles to optimizing production operations.
For instance, a major oil-and-gas company told us, “I know there’s risk in our operations we can’t see. Can you develop a tool that shows me where the risks are?” The result was the Azima DLI Composite Risk Index, or “CRI” report, that rank-orders sites by likelihood of a mechanical failure, and rank-orders the assets in each site by the severity of the risk they present to the production operation.
Another case involved a global industrial gases company trying to address division-level cultural resistance to predictive maintenance. By developing a tool to measure compliance in each operating division, and at each site within each division, the company was able to correlate high compliance with low unplanned maintenance and capital expenditures, and low compliance with abnormally high unplanned spending. Strongly held (but unsubstantiated) beliefs were no match for the economic evidence.
