Replacing the gut feeling-driven APM with a prognostics approach
- Jun 26, 2020 1:33 pm GMT
In the next five years, 50% of professionals in the energy field will retire.
They’re taking with them valuable expertise, including the quasi-mystical ability to know when equipment is going to fail, or what they call the “gut feeling.” While it’s hard to lose that experience, new asset performance management (APM) techniques have evolved to take over for intuition.
Oftentimes, predictive diagnostic solutions are used to detect anomalies before they turn into an actual malfunction, but they can’t supply the valuable “what if?” An advanced APM approach called prognostics is condition-based and can anticipate how an asset will fail and, more importantly, what will happen because of it. Leveraging smart maintenance tools and improved data techniques, power companies can use prognostics to incorporate future equipment conditions into maintenance decisions.
By moving to a prognostics approach, energy utilities can arm their operators with objective foresight, leading to more insight about what to do before assets fail and how to adjust given those failures.
Steps to Prepare for a Prognostics Approach
To fully reap the benefits of prognostics efforts, power organizations need to take several preparatory steps. This will help them gain new understanding of equipment and, likely, identify fresh opportunities to technically optimize condition monitoring.
#1 Target Prognosis
Determine which equipment to target to gain prognostic information. To gain the most value, select equipment that is truly critical for continuous plant operations and involves considerable costs and effort in the event of malfunction or failure. The best targeted equipment also generates condition and process data that is recorded and available for further quantitative processing.
#2 Specify Malfunctions
Malfunctions don’t always mean a complete failure or outage. To identify suboptimal states, specify the top ten malfunctions of the equipment type selected. It’s important to examine these malfunctions in depth to determine the true root cause and how they are typically detected in the data.
#3 Review Data
Having prioritized the equipment and its related malfunctions, identity all related data sources. Ensure the data is reliable and accurate. How much data history is exported will vary depending on the prognostic horizon demanded and the type of malfunction. Generally, three to five years should be sufficient.
Review the data by visualizing it over time and identify data patterns that explain malfunctions. Apply advanced stochastic methods when needed and verify observations with experienced plant engineers.
#4 Make Formulations
After distinguishing between direct and indirect causal relationships, specify correlations of data to particular malfunctions. Formulate useful condition parameters that reflect diagnostic rules of thumb involving arithmetic or logical functions. Consider the period of data observation and observe contingencies and dependencies on other rules. Determine a set of value ranges for formulated parameters.
#5 Compute RUL
To determine Remaining Useful Life (RUL), you must feed condition and process data, as well as the complementary malfunction and parameter specifications, into a stochastic model. The computations will:
- Project equipment condition over an explicit time horizon
- Apply diagnostic rules at different future time stages
- Infer malfunction likelihoods based on prognostic and diagnostic results
- Obtain a set of malfunction-specific RUL distributions
- Consolidate data for a total equipment RUL distribution
- Convert distribution to a meaningful illustration for maintenance planning
#6 Validate Results
Conduct a retrospective analysis validating the results with empirical observations from the past. Continuously monitor prognostic results and ascertain their plausibility.
APM Prognostics in Action
Putting this into practice, a major Swiss power generation and distribution utility had legacy assets that were nearing end-of-life. Operators deployed APM at a 40+ year old hydropower station last refurbished in the late 1990s. They used all available historical asset data to configure and train the stochastic model for computing asset RUL. Working with on-site plant engineers, the APM team reviewed asset condition data histories and correlated condition parameters with critical asset malfunction modes to determine risk profiles. With information on asset RUL, the scope of the scheduled maintenance could be limited to the truly required services, reducing maintenance costs. In addition, operators gained insight on how to optimally operate these assets to maximize utilization over RUL.
Now, the prognostics-based APM approach successfully computes future malfunction risk and estimates RUL when presented with current condition and process data. The results are visually summarized on a prognostics dashboard, which enables operators to quickly spot potential upcoming failures and identify exact parts that need replacement. Based on future simulation results, operators can also safely operate high risk assets at an appropriately adjusted load, wait for a better maintenance window or availability of parts and crew, or extend the RUL of the entire plant. Now, outages can be scheduled to have the least financial impact.
Taking the critical preparatory steps to ready assets and data for prognostics APM is important to effectively analyze how current reliability management processes can be leveraged using the insights from prognostics. With newly transparent information about future risk profiles, operators can improve operations and maintenance as well as increase maintenance scheduling efficiency. Advanced prognostics tools provide foresight about the future state of assets. In doing so, APM can provide a basis of essential information for advanced asset management decision-making – demystifying uncertainty around how key assets will fail and the impacts of those failures – which in turn will reduce costs, better determine risk, optimize maintenance strategies and increase uptime.