Prescriptive maintenance (RxM) couples highly sophisticated analytics with Big Data to prescribe corrective actions. New and emerging prescriptive analytics solutions are making RxM a viable goal. Their automated intelligence and ability to reconcile exponentially growing data from the internet of things (IoT), far-flung existing data silos, and untapped workforce knowledge has these tools drawing increasing interest.
AI augments the analytics
The use of artificial intelligence (AI), machine learning, and digital twins has been aimed at improving the predictive capability of preventive maintenance solutions and the development of prescriptive methodologies, says Louis Halversen, CTO of Northwest Analytics (NWA). NWA offers a real-time analytics solution for monitoring process equipment health in addition to other applications.
“Once you add ‘assignable cause’ and ‘corrective action’ information in response to the events detected, the information base for prescriptive maintenance is being created based on actual events at the local plant,” explains Halversen. “These can then be shared with other facilities with similar processes and equipment.” He suggests that the next phase for high-value targets can be to develop multivariate models based on fairly straightforward techniques such as principal components analysis (PCA), which can be analyzed with a statistical process control chart.
Pegasystems (Pega) views prescriptive maintenance as preventive maintenance technology with built-in intelligence. Pega’s AI-powered digital transformation platform supports “continuous monitoring of sensor events, aggregation of sensor data insights, and orchestration of people, systems, and things in end-to-end, automated value streams,” says Steve Rudolph, vice president and business line leader at Pegasystems.
With Pega’s Decision Hub and Dynamic Case Management tools enabling what it calls digital prescriptive maintenance (DPM), organizations can proactively optimize the cost and duty cycle of devices and systems. “With today’s increasingly intelligent and responsive IoT technology, maintenance is evolving from minimizing machine downtime to predicting potential failures before they occur,” Rudolph observes.
AI is also leveraged in the Falkonry LRS machine learning system, which detects hidden patterns in existing operational data, enabling resident operational experts to easily visualize and provide context to the data-driven analysis. With the added context, Falkonry can discover behaviors. The resulting predictive and prescriptive insights help users to avoid unplanned downtime and improve quality, safety, and yield.
“Plant operators are able to discover patterns and recognize conditions in the multivariate time-series data generated in their industrial operations,” says Nikunj Mehta, founder and CEO of Falkonry. “These early warning signs empower them to apply their domain knowledge and intervene to prevent undesired operating events.”
Knowledge consolidation extends the value
Information from smarter and more-connected assets, coupled with existing knowledge retained from the aging workforce, create the opportunity for analytics that are prescriptive and powered by machine learning, remarks Sandra DiMatteo, a global director at Bentley Systems. Prescriptive analytical tools “can not only predict what is likely to occur, but ... can offer ‘what-if’ analysis of alternatives to provide a scenario that can help the outcome,” she says.
Bentley’s AssetWise CONNECT Edition software, powered by the Microsoft Azure Machine Learning platform, uses situational intelligence to provide early warning of asset and operational issues, augment and expedite decision-making, and guide actions.
Emerson’s emerging asset platforms, such as Plantweb Optics, are data aggregators that eliminate data silos within plants and across the enterprise. Mani Janardhanan, Plantweb IIoT product director at Emerson, explains that most plants today have data silos, many of which are functional. They successfully provide alerts to asset faults that could lead to outright failures. “Prescriptive maintenance asks us to take the next step,” he says.
“Now we are working with customers to move beyond predicting failure to prescribing a resolution that is business- and safety-efficient,” says Janardhanan. This is achieved by connecting “asset data and history from many plants with analytics to create insight-driven, prescribed responses.”
Greg Perry, senior consultant for Fluke Accelix, believes RxM represents the dawn of autonomous reliability. “With machine learning and improved data analysis, powerful insights into autonomous corrective, preventive, or predictive actions can be precisely aligned,” he says. “As technology improves and organizations adopt these advanced condition-based methods, downtime and breakdowns will result in smaller margins of random failures.”