Behrooz Jalalahmadi and Stephen Steen, Sentient Science Corp
DigitalClone accurately predicted an OEM’s bearing fatigue data that had compiled over a long period.
Behrooz Jalalahmadi
A recent prognostics-based alternative to conventional condition monitoring methods provide cloud-based modeling and clever software to accurately simulate real-world operating conditions for specific machine components. The system uses a digital model to represent a serialized asset such as a single wind turbine, and how it reacts under different operating conditions. This approach lets the system accurately predict how components will perform down to their material microstructure in different applications. The new system also lets users determine how these components will function and when they will fail. Further, the system provides a risk assessment of the asset’s Remaining Useful Life (RUL).
Another big plus is that the new system can simulate what-if scenarios. This assists users with changing operating conditions, such as a change in lubrication, so they can keep performance optimized. The task helps users establish parameters that extend the operating life of a component.
Stephen Steen
A little history
The system was developed for the Defense Advanced Research Projects Agency which requested a new process to more efficiently and effectively evaluate the reliability of equipment while reducing the need for physical testing. The highlight of the program was Sentient Science’s validation of 30 years of gear testing done by NASA.
In a nutshell, prognostic features let crews perform maintenance when it is needed and reduce the amount of data collected and analyzed. The goal is to predictably extend the remaining useful life of the turbine.
The project’s goal was to better predict the remaining useful life of tail rotor gearbox bearings in a UH-60 M Black Hawk Helicopter. The main route to failure is caused by spallation damage in tapered roller bearings in the helicopter’s tail. Testing included correlating vibrations to the actual condition of the bearing.
A physics-based model predicts the rate at which spallation damage initiates and progresses in a bearing under specific operating conditions. This multi-physics software model accurately predicted the rate of spall propagation and the damage pattern observed in experiments.
A second example highlights a comparison between empirical test data, supplied by a bearing OEM, and model data developed by Sentient. DigitalClone, part of the system, developed a model that accurately predicted and validated the historic fatigue failure data from the OEM.
Use in wind turbines
The graph suggests how performance of a wind turbine progresses from its best at launch to functional failure.
The simulation software can be applied to boost the reliability of wind turbine gearboxes, a primary system of failure. Its prognostic features can let users know future component failure rates and modes, which allow taking action to minimize them.
The system follows a five step process:
- Connect to a SCADA and condition-monitoring systems for its historical and current operations data.
- Build a prognostics model using the software for how the wind turbine or wind farm operates.
- Predict the state of the wind turbine or wind power plant which includes determining RUL and failure modes, establishing alarms, and preparing reports that predict when cracks in components will initiate and propagate into operational failure.
- Establish goals for the ROI of the wind turbine or wind farm based on taking different servicing options.
- Prepare control settings to optimize performance of the wind turbine or wind farm.
A rotating system model shows how the software works. Vibration signals from the components propagate to its housing. The model simulates these signals, finds fault levels and locations, and predicts premature failure.
Such prognostics can let wind-farm operators devote little or no headcount to vibration or other CBM systems. The service almost runs itself because it is focused on predicting failure far ahead of time and what that failure looks like in the predictive data instead of looking for indications of failure. The approach is to manage by exception, by only focusing on changes to predictions or sensor data rather than manually monitoring the data for indications of failure. Entire wind turbines, major subsystems, and individual components can be managed this way.
The monitoring system is a parameterized model, meaning a parameter under study can be changed to see how it affects life and reliability. The result: wind-farm operators can make better informed recommendations on how to improve operations through better design analysis, better understanding of suppliers, and determine whether components need remanufacturing for optimization.
The graphs track the failure rates of three bearing suppliers used in a commercial gearbox design plus a fleet wide failure average (blue). The X axis suggests when in 2011 to 2013 turbines from several OEMs entered service. As they work and wear, the Y-axis or failure rate shows it increasing with time. Eventually, in the left graphs, all turbines fail before 2031. The right graphs suggest longer working years with changes to lubrication systems and de-rating strategies. Green, red, and orange turbines may work well past 2031 with such changes.
The system capability spans all mechanical equipment in the wind turbine including the lubricant type, gearbox, pitch and yaw systems, surface finish, and geometry. This facilitates the best future use of the wind turbine and lets the operator plan for future use on a per asset basis.
Beyond routine maintenance
The prognostics approach provides benefits to wind-farm operators from the standpoints of planning preventative maintenance, focusing on evaluating specific parameters, and extending RUL. Conventional approaches to monitoring make it difficult to plan preventative maintenance. The new system is useful as the wind turbine enters the final phase of its operating life.
Over 90% of its operating life, a wind turbine will develop micro cracks that eventually lead to it spending the remaining 10% of its life in a faulted state and then ultimately operational failure. Prognostics can predict and see what failure looks like in components such as the gearbox. This lets DigitalClone predict the future state of the wind turbine under its current set of operating conditions and what can be done to extend RUL.
What-if scenarios
One maintenance objective is to extend the time from when failure can be predicted to when it eventually occurs. An example of how prognostics can be used is to focus on vibrational signals from the wind turbine’s nacelle as in Fault descriptions and sensor locations.
Another benefit: Wind farm operators need not collect extensive amounts of historical data from a monitoring system. Once the turbine is modeled by DigitalClone, an analysis can determine the onset of premature failure and recommend changes in operating parameters to extend and optimize life. A complete gearbox can be modeled in DigitalClone in 60 days to provide fleet wide performance and failure predictions. Serialized or per asset predictions require another 30 days.
DigitalClone can also quantify the impact on life using what-if scenarios that let the wind-farm operators assess their equipment and find the best ways to improve performance and ROI. For example, de-rating a component is a way of extending its remaining useful life. In de-rating a 1.5 MW turbine, its power generation capability can be scaled back so that it produces less electricity, say to 1.3 MW, which reduces loads and torques on the components, thereby extending their operating life. Knowing a de-rate level provides a best balance between ROI and operational failure is new to the market.
Prognostics will be able to assist the wind farm operator with determining what set of conditions to use in de-rating to extend the turbine’s operations to the maximum extent over its remaining operating life. This includes how much the turbine should be de-rated.
Surface treatments to the bearings can also be evaluated by prognostics to determine a best fit for a specific system. The application of a surface finish can lead to improved performance over a longer operating period. The predictive software can determine if options such as a ground finish or a super finish are appropriate for extending the remaining useful life of a specific bearing. Lubricant selection can also be assessed.
Even after a component breaks, the system can help determine a best possible option for replacing or remanufacturing that component for increase system life. One option, for instance, could be to buy all new replacement parts including internal components. A second option: replace specific internal components and reuse as many components as possible.
In addition, the software can accurately predict at what stage of failure remanufacturing should be done to get the most operating life out of the existing turbine. Also, recommendations for type of surface treatment to use can be made during remanufacturing.
The use of prognostics in de-rating, surface treatments, alternative lubricants, the right type of remanufacturing, and the best OEM parts flattens the operating life curves, which means that the wind turbine can conceivably provide optimal performance for time frames in excess of 2031.
An added benefit is that systems analysis can be tailored to wind turbines manufactured by specific OEMs. RULs for each of these OEMs will move in a different fashion due to differences in how each producer manufactures its wind turbines. WPE
Filed Under: News
