Summary
ECG partners with a major power utility to validate the new Predict-It Pathway module. Using advanced statistical techniques, Pathway can detect equipment problems during steam turbine start-ups and shutdowns, providing functionality that other predictive analytics solutions cannot. With Pathway, the customer’s data showed ample early warning of a developing steam turbine bearing failure, which would have allowed for planned maintenance instead of a more costly forced outage.
Background
Our energy partner provides the energy that powers millions of homes and businesses across eight Western and Midwestern states, serving more than 3.7 million electric customers and 2.1 million natural gas customers. A progressive leader in the industry, our partner was an early adopter of predictive analytics to help improve operations and equipment health.
Challenge
Our customer experienced a forced outage on their Rankine Cycle Plant due to a steam turbine bearing failure on 9/24/21. Although the company was using real-time competing Advanced Pattern Recognition (APR) predictive analytics software, APR did not provide early warning of the problem. The customer believed they needed to get a better understanding of what went wrong, and suspected the problem started developing early in start-up or shutdown, areas not visible to its current solutions. The customer sought a tool to supplement its APR steady-state solution by identifying and analyzing start-up/shutdown problems in steam turbines.
Solution
Predict-It’s Pathway module was under development and ECG sought to validate the value of the new addition to their predictive analytics software. The Pathway module can be used to identify operating anomalies early in start-ups, shutdowns, and other batch production processes with defined start and end parameters. The solution could be used across all types of processes and equipment, including steam turbines.
ECG suggested that our utility partner be an early tester of Pathway via a post-event analysis of the Unit 1 steam turbine bearing failure. The customer was confident in ECG’s decades of predictive analytics and diagnostic experience, power industry expertise, as well as ECG’s commitment to partner-clients success and agreed to test Pathway through a post-event analysis of their steam turbine failure.
ECG imported The customer’s data for Unit 1 from 48 coastdowns, each with 71 variables, spanning the time frame of Q4 2020 through 2021. Pathway was used to study normal operation for each variable (e.g., temperatures, vibrations) individually and to condense all the variables’ normal data into one Multivariate Data Analysis (MVDA)
model. From there, Pathway could identify individual variables’ deviations from normal, as well as the overall impact each had on the process from beginning to end.
The customer also provided ECG with similar data for its Unit 2 steam turbine to see if Pathway could identify any past anomalies in its data.
Results
Unit 1
The Pathway software alerted to a Unit 1 bearing issue on 12/11/20 — more than 9 months prior to the 9/24/21 bearing failure. Bearings 4 and 5 experienced abnormal metal temperature spikes late in the 12/11/20 coastdown, beginning at ~550-200 RPM. Pathway also continued to alert to bearing high metal temperature deviations during
coastdowns on 6 additional dates’ shutdown events. Showing these results well in advance of the bearing failure validates the effectiveness and value of the Pathway module for finding and alarming equipment anomalies in start-up and shut-down events, giving companies time to take corrective actions.
Unit 2
The Pathway software alerted to a similar developing bearing issue in Unit 2. Its #5 bearing experienced an abnormal metal temperature spike late in the coastdown, at ~550 RPM, on 1/26/21. This was a very early warning of a bearing issue that was developing and was able to be addressed by the company through a planned outage as a result of ECG’s analysis. Unit 2 had not yet experienced a failure and the customer was able to take corrective action and avoided unplanned downtime.
“Since equipment changes and resultant indications were occurring during the startup/
Randy ~ M&D Center Manager
shutdown periods, these changes were not detectable in the APR, until very late into the
failure process. We wanted to go back and look at startup/shutdown data but didn’t have
a good solution to do that. Testing Pathway provided us a clear analysis of our steam
turbine problem, which proved to be effective in aiding our maintenance planning and
decision making. The potential to automatically analyze these batch events will be a key
criteria for us when evaluating a predictive analytics solution.”