“If It Ain’t Broke, Prevent It From Breaking:” IoT and IoE in Prescriptive Maintenance for Turbines

How IoT and IoE are Impacting Power Generation via Prescriptive Maintenance of Turbines

From a mechanical standpoint, turbines are engineering marvels. These massive machines have thousands of delicate moving parts, each with precision tolerances necessary to achieve maximum efficiency and complex maintenance routines. Even a minute disturbance in the turbine’s system can trigger a catastrophic breakdown. Yet, because of the cost of maintenance, unplanned downtime, and concerns involved with repairs (which could just as likely be performed incorrectly and result in another breakdown), companies generally service turbines only when necessary.

Despite planned maintenance strategies, 82% of all assets still fail randomly. While some assets can be run to failure, that’s not an option when dealing with turbines. Due to the operational cost of unplanned downtime for turbines and the importance of these assets, it is crucial they operate optimally. Traditional maintenance strategies have simply proven to be of limited help. However, incorporating Internet of Things (IoT), Internet of Energy (IoE), and artificial intelligence (AI) technologies can boost the top-line production abilities of turbines, adding an estimated $1.3M in value annually, as well as aiding the bottom-line savings.

Before exploring these effects on production and costs, it’s worth exploring the shortfalls of traditional maintenance strategies, which rely heavily upon availability of Subject Matter Experts (SMEs). Typically, SMEs develop static, first-principle models for failure detection—which can take months to develop, are difficult to maintain, and aren’t reflective of every scenario a turbine will encounter. In fact, the most valuable data from turbines can come from transient events like startups and coast-downs, as many critical issues first materialize during these periods. Traditional models are not mathematically equipped to work with transient events, as they occur over an indeterminate length of time. On the other hand, AI technology is well-suited to the challenge, as algorithms are able to normalize events over time, providing the operational intelligence needed to implement prescriptive maintenance.

Another issue with the traditional modeling approach is the human resource intensity needed to maintain and update these first-principle models. Given the difficulty of finding SMEs in an industry with a shrinking talent pool, this can be particularly arduous. AI eliminates these organizational constraints by analyzing volumes of data (at speeds that humans can’t) and then providing actionable insights to improve energy production. In turn, AI then frees time for SMEs to perform higher-level tasks, and codifies SME knowledge for the future. An ROI research study from Tuck School of Business at Dartmouth recently estimated operations and maintenance savings of $400,000 annually for a combined cycle gas turbine power plant.

To implement an AI strategy, a turbine fleet must be equipped with sensors to provide operators with an accurate, real-time view of the system. The sensors can monitor data points such as vibration, temperature, or stress on parts, pull additional information that would affect machinery, like weather conditions, and organize it into an Asset Performance Management System. The same Tuck study found this increased insight into system performance in three main ways: reducing downtime (by 50%), decreasing response time to failure (by 25%), and reducing catastrophic incidents (by 35%). The study estimated the increase in production would amount to $1.3M annually for a combined cycle gas turbine power plant. Below is a more specific comparison demonstrating the benefit of predictive methods using AI versus traditional models.



AI is currently providing immediate and significant value to utilities, and the future of the Internet of Energy holds even more exciting innovations. Operators will be able to approach a machine using augmented reality, get an immediate list of alarms and readouts, diagnose issues with appropriate manuals and work order history integrations, dialogue with remote or virtual SMEs, and schedule necessary repairs. For equipment like offshore wind turbines, the ability to virtually interact with the AI system reduces difficult treks to the site and makes workers more productive, ensuring the process is cost-effective and safer.

Adopting a prescriptive maintenance strategy significantly improves production via decreased downtime, incidents, and failure response time. In addition, SMEs are provided with needed information in a timely manner, decreasing the burden of repetitive tasks and allowing focus on larger production issues. As the Internet of Energy continues to grow, the quick wins will add to the system’s overall knowledge, cascading to make artificial intelligence into an indispensable resource.

About the Author

Stuart Gillen is Director of Business Development at SparkCognition, a global provider of cognitive computing analytics. His areas of specialty include IoT architectures, platforms, and technologies. Stuart focuses on how advanced data analytics can be applied to Asset Performance Management to optimize efficiency, maintenance and safety on critical infrastructure.