Capturing diagnostics, detecting faults and optimizing operations is difficult to begin with, but it’s even harder with fewer and less-experienced people. To make data meaningful for users and companies unable to do inhouse diagnostics programming, Fluke Reliability’s Azima DLI subsidiary developed its Watchman software and services for standardized, automated and transparent data capture, and advanced diagnostics.
Once these analytics are complete, Fluke has two other programs that perform follow-up tasks. After a process fault is identified, eMaint software executes workflow, handles computerized maintenance management systems (CMMS), and verifies and documents reporting. Next, Pruftechnik’s hardware device can be utilized to perform solutions on the identified faults, and validate with Azima DLI.
Get your subscription to Control's tri-weekly newsletter.
“Previously, we mostly got monthly or infrequent snapshots of collected data, but now users need more details daily or even continuously, so Azima’s software scrubs and prioritizes input to tightly monitor automation accuracy, find faults, and determine actions that need to be taken,” says Michael DeMaria, product management director at Azima. “We monitor and analyze over 560,000 machine tests per year. This lets users move from responding to individual events to implementing a continuous, automated analytics stream that can quickly react to the complex, analytical datasets.
“Building data analytics used to be very after-the-fact. Users would gather information, and go back and look at what to do. AI requires far more information to produce useful results, but many applications and organizations can’t generate enough to make AI workable. Users must also choose between onsite or cloud-based data storage, how much processing power is sufficient, and whether to add AI tools and when. Azima’s advantage is that we’ve tagged and stored 100 trillion data tags since the 1990s, and can apply that knowledge to help users immediately make the most suitable choices for them.”
Lining up the data ducks
For instance, Azima DLI has a centralized data lake, which uses AI to quickly sort through large volumes of data, find useful details, and develop representations that can be compared to actual operations and equipment performance.
“These details include internal component details, such gear-mesh frequencies, vane counts on pumps, rotor bars on motors, but all these complex information sources need AI to look through their databases and find correlations,” explains DeMaria. “We previously had to wait for faults to occur, and learn what we could, or we maybe had to remain blind. Representations with AI mean we can detect more problems sooner, with greater accuracy, and confidently take recommended actions earlier and plan further ahead.”
DeMaria adds that Azima’s software can take vibration data from wireless sensors on most machine types, beyond just balance-of-plant equipment, portable devices for complex or standard routes, and online systems for hard-to-access and critical applications.
“If a user bought all these components and capabilities separately, it would be hard to integrate them, and even more difficult to get them to uniformly deliver output results,” adds DeMaria. “Azima can coordinate these data types and sources, and use it AI resources to develop techniques it can add to its diagnostics software. Because Azima is built on data-agnostic principles, input comes in, and runs through its AI engine, which recognizes patterns, such as how initial vibrations fit into the context of a machine’s overall profile.”
Because capturing wireless vibration data daily or more often produces too much material for humans to review manually, DeMaria reports that 92% of the input Azima receives is handled automatically by its continuous monitoring system, so only 8% is usually examined by people.
“Not only does automation identify emerging and priority faults with plain-language actions to mitigate them, it also evaluates data quality; reviews fault stability to determine if a fault is stable or not before issuing a workorder; and uses persistence of events to automate recurring events through,” adds DeMaria. “Everyone would like to get down to fully automated analytics that require zero review by humans, but customers ultimately want a human setup of eyes confirming these findings before making that multi-million-dollar repair decision. Azima handles this volume at scale with highly accurate automation and human analysts confirmation.”
