To determine the condition of the generators at the 212-MW Powerhouse One at its Rock Island project, Chelan County Public Utility District installed an air gap monitoring system. After discovering generator alignment problems, the utility installed instrumentation to measure stationary displacement and temperature. Analysis of the data collected let Chelan County PUD improve specifications for unit rehabilitation.
By Matthew T. Davis, Paul J. Wolff, and Patrick A. March
Chelan County Public Utility District (PUD) installed a monitoring system on the ten generators at the 212-MW Powerhouse One at its 624-MW Rock Island Project in 2002. The purpose of this system was to explore the condition of the generators before the utility tendered a rehabilitation contract. Chelan County PUD also wanted to use the monitoring system to produce alarm thresholds for detecting abnormal online operating states.
The initial generator monitoring system at Rock Island consisted of four capacitive sensors used to measure the rotating stator-to-rotor air gap. After the data produced using this system indicated generator alignment problems, Chelan County PUD installed additional instruments to measure displacements of non-rotating components, to measure temperatures for evaluating temperature transients, and to collect unit operating information such as power and flow. The data generated was stored in three separate databases.
This separation of the data made it difficult to understand the root cause of the generator misalignment. Analyzing the data was further complicated by the significant transients that occurred as a result of unit start up and shutdown. Determining the root cause of the generator misalignment required several data analysis techniques, including:
1) Synchronizing the disparate data sets;
2) Filtering the data to reduce thermal and start-up transients; and
3) Correlating patterns in the data to pinpoint the source of the misalignment.
A paradox with many projects is that, although automated data acquisition and archival systems are essential for measuring key information to diagnose equipment problems, these systems produce so much data that it becomes extremely difficult to extract useful information. Uncovering useful information often requires specialized analyses that go well beyond simple trending of the data and specialized tools to automate data processing. In this case history, once multiple measurements and data sets were synchronized and filtered, this information was used to diagnose the generator alignment problem.
Establishing a generator monitoring program
Generator failures at Chelan County PUD ’s 1,287-MW Rocky Reach were leading to lengthy forced outages. The utility suspected conditions at Rock Island Powerhouse One might be similar, meaning that facility was in danger of experiencing similar expensive outages. To help avoid this situation, in 2002 Chelan County PUD established a comprehensive generator monitoring program at its Rock Island Project.
Rock Island ’s 50-year-old generators, manufactured by Allis Chalmers, operate at a power output of 25 megavolt amperes (MVa) and a rotational speed of 100 revolutions per minute (rpm). The air-to-water cooled generators are driven by vertical shaft Kaplan turbines from Allis Chalmers. The generators ’ design rotor-to-stator air gap is 0.47 inch.
Figure 1: The air gap data gathered on the generators at the 624-MW Rock Island Project showed significant scatter that was attributed to dimensional changes occurring during start up.
A key component of the generator reliability program at Rock Island involved ensuring that the air gap operated within recommended tolerances. A non-uniform air gap can create large magnetic and inertial forces and place significant load on the generator frame and bearings. In the worst case, contact between the stator and rotor causes a catastrophic failure of the equipment, leading to a lengthy forced outage.
Initially, air gap monitoring systems supplied by VibroSystM of Longueuil, Quebec, Canada, were installed in 2002 on four of the ten units at Rock Island. The measurements recorded indicated that the air gaps on these units were well below industry standards and that the air gaps gradually drifted over time.
Figure 2: After filtering the air gap data in Figure 3 to remove the start-up transients, the charts show gradual drifting of the air gaps at all sensors. The drifting is most apparent at 60 and 330 degrees.
To understand the root of the problem, Chelan County PUD installed a second instrumentation system consisting of proximity probes and thermocouples to measure the displacements and temperatures of stationary components. In addition, unit operating information consisting of power output, generator operating parameters, water flow, blade angle, wicket gate position, headwater, tailwater, temperatures, and time was available in an existing OSIsoftâ„¢ Plant Information (PI) database. Chelan County PUD ran the project with this equipment installed for about one year to collect sufficient data.
The air gap sensors measure the online air gap between the stator and rotor. The sensors on each unit consist of four capacitive probes mounted around the stator circumference. With 0 degrees upstream and using a counter-clockwise rotation, Sensor 1 is at 60 degrees, Sensor 2 at 150 degrees, Sensor 3 at 240 degrees, and Sensor 4 at 330 degrees.
The system samples the air gaps at 6,000 Hertz (Hz) for one sensor and 750 Hz for the other three sensors. With a generator rotational speed of 100 rpm, each sensor acquires a minimum of 450 samples per revolution. The air gap data used for the analysis are the minimum measurements acquired from each of the sensors over five-minute intervals.
Proximity probes were installed to measure the relative displacements of primary generator components. The components that were instrumented include the stator frame, stator core, generator guide bearing, and generator lower bracket (which consists of the thrust/guide bearing housing and two H-frame brackets, with one supporting the other). These displacement measurements were made using probes installed at 42 locations on each unit, and the data included both displacements relative to ground and displacements relative to other generator components.
The data analyzed from these instruments were acquired by using an Omega Engineering MultiScan 1200 data logger and associated computer. Although data were sampled at various frequencies over the course of several months, Chelan County PUD determined that a 20-minute sampling interval was sufficient to characterize the displacements for the stationary components.
Temperature measurements were made using thermocouples at 12 points on the generator lower bracket of each unit. Temperature is an important measurement because thermal transients produce a significant amount of deformation. As a reference, Chelan County PUD personnel measured and recorded ambient temperatures in the turbine wheel pit. These data were logged using the same MultiScan data logger and at the same frequency as the stationary displacement data.
Finally, Rock Island unit operating data was obtained from Chelan County PUD ’s PI archival database over the same time period as the VibroSystM data, at a five-minute sampling interval. Operating data included time, power, flow, wicket gate, blade angle, generator operating parameters, and unit temperature measurements. These data were acquired by the supervisory control and data acquisition (SCADA) system.
Analyzing the data collected
Because the systems installed at Rock Island were in their initial demonstration phases, Chelan County PUD had not fully integrated the separate data sets into one centralized database. As a result, the data collected from these various systems was contained in about 60 different files, in three different formats, and in three different databases. Manually processing this large quantity of data would have required a prohibitive amount of time. Chelan County PUD used DataWolffâ„¢, an Excel-based data analysis program, to automate this processing. The general data analysis engine is configured for specific analyses using scripts. For example, the data import template specifies which data to import and filters out bad readings. The calculation template provides the instruction set that specifies which calculations to perform. The chart template, in conjunction with the chart library, creates trends of all data in Excel charts. This lets the user zoom in, zoom out, and scroll in time through the data.
As part of the automated data processing, the analysis script configures the analysis engine to process the data in a batch mode, with no additional manual operations required. When the analysis script is loaded, it prompts the user for files to analyze. The analyses then proceed automatically, and the software produces an Excel workbook containing the desired calculations and charts needed to evaluate the results.
Results of the analysis
Determining the probable root cause for the generator alignment problems required several iterations of viewing the data in different groupings, filtering the data to remove transients, and identifying important patterns in the trends. The air gap data were analyzed first because alarms and generator outages are based on these data. Once the air gap data were analyzed, the long-term patterns observed in the trend plots helped to identify the generator component responsible for the misalignment.
Chelan County PUD used the system to collect air gap data over a period of about one year. Figure 1 on page 50 shows significant scatter in the data. Upon closer review, this scatter resulted from dimensional changes occurring during start up. Because the air gap dimensions and the differences between the air gaps at the various circumferential locations were not within recommended tolerances, Chelan County PUD performed a realignment in August 2003, placing both the turbine and generator in their best centers. This realignment produced the sudden shift in air gap dimensions shown in Figure 1.
Figure 4: Measurements of the mid-span displacements of the upper portion of the lower bracket in the direction of water flow show that transverse displacements of the bracket change as the ambient temperature changes.
Figure 2 on page 50 presents the ambient temperature data and rotating air gap data after removing the start-up transients. All sensors show gradual drifting of the air gaps, but the drift is most apparent for the sensors at 60 and 330 degrees. For example, the air gap at 60 degrees gradually increases from mid-February 2003 until mid-July 2003. A step change in the air gap then occurs, corresponding to when the unit was realigned. The air gap then gradually decreases from the end of August 2003 until mid-December 2003, when the data ends.
Trends in the air gap data, especially for the sensors mounted at 60 and 330 degrees, correspond closely to the trends in the temperature data. As the ambient temperature increases, the air gap at 60 degrees increases. Decreasing ambient temperature corresponds to a decreasing air gap. The air gap at 330 degrees displays an opposite trend. As the ambient temperature increases, the air gap decreases; as the ambient temperature decreases, the air gap increases. These trends provided a clear indication that the ambient temperature has a primary influence on generator alignment.
Stationary displacement and temperature data
Both the displacement data for the stationary components and the temperature data contained large transients that obscured other trends. The transients decayed to steady-state values after several hours when the unit was not operating. Because it was apparent that heat generated during unit operation caused these transients, a thermal-function filter was applied to the stationary displacement data to segregate all transients and include only those steady-state data obtained when the unit had not operated for several hours.
The project team then reviewed the stationary displacement data to find long-term trends that cycled with the ambient temperature and that were similar to the trends observed in the air gap displacement data. Of all the displacement data, the eight displacement probes measuring the movement of the upper portion of the generator ’s lower bracket displayed trends similar to the ones observed in the air gap data.
Figure 3 on page 52 presents measurements from four displacement probes (referred to as side displacement probes in Figure 5). These displacements measure the movement of the upper portion of the lower bracket perpendicular to the direction of water flowing through the water passage. Long-term patterns in these displacements are very similar to the ambient temperature cycle.
Figure 4 on page 53 presents the mid-span displacements of the upper portion of the lower bracket in the direction of water flow (referred to as upstream/ downstream displacement probes in Figure 5). The transducers ULB-UTP and ULB-UBT are on the upstream side of the bearing assembly and measure displacements perpendicular to transducers ULB-NS1 through ULB-NS4. Transducers ULB-DTP and ULB-DBT are on the downstream side of the bearing assembly, opposite ULB-UTP and ULB-UBT.
Figure 4 demonstrates that transverse displacements of the bracket change as ambient temperature changes. The displacements measured by the 60-degree and 150-degree probes increase and decrease as the ambient temperature increases and decreases, while the displacements measured by the 240-degree and 330-degree probes move in opposition to the temperature changes. This is especially apparent for a few months beginning in August 2003, when the average ambient temperature decreases from its peak value.
Application of the results
These data analyses indicate that the stator-to-rotor air gaps and the unit alignment are strongly dependent on the ambient temperature. Displacement of the upper portion of the lower bracket also changes with ambient temperature. The movement of this component is particularly relevant to generator alignment because it houses the generator guide bearing and thrust bearing.
Figure 5: Final analysis of data gathered for the generators at the 624-MW Rock Island Project indicated the upper portion of the generator ’s lower bracket was deforming in an upstream direction.
Figure 5 shows a plan view of a Rock Island unit. It also shows the location of the displacement probes and stabilizing bolts, and an exaggerated deformation of the lower bracket inferred from the displacement measurements. Displacement of the upper portion of the lower bracket, measured by the side displacement probes, is in phase with the ambient temperature cycles (i.e., the displacement increases with increasing ambient temperature and decreases with decreasing temperature). The displacement is measured with proximity probes attached to steel mounting brackets, which are rigidly attached to the concrete in the turbine wheel pit. A probable cause for this displacement is the force produced by the stabilizing bolts that anchor the lower bracket to the turbine wheel pit along the length of the lower portion of the bracket. Thermal expansion of the bolts or the turbine wheel pit concrete could cause the lower portion of the lower bracket to displace transversely, which also would cause a displacement of the upper portion of the lower bracket.
The information gained from these analyses helped Chelan County PUD in several ways:
— The utility has awarded a comprehensive unit rehabilitation contract for at least five of the ten units at Rock Island. The information from these analyses helped to improve the unit specifications and performance requirements.
— Based on the Rock Island data analyses and other phenomenon measured at Rocky Reach, Chelan County PUD has concluded that the primary influence on concrete and unit ambient temperature variations is the seasonal change in river water temperature, which can fluctuate as much as 30 degrees Fahrenheit. The concrete expansion and contraction arising from these variations must be taken into account. Provisions for accommodating these effects need to be incorporated in future planning.
— Unit outages for generator alignment have been avoided because the utility now understands that the misalignment is linked to ambient temperature cycles.
— Monitoring instrumentation on other units has been streamlined because Chelan County PUD now has better insight into the mechanisms causing generator misalignment.
Mr. Davis may be reached at Public Utility District No. 1 of Chelan County, P.O. Box 1231, Wenatchee, WA 98807; (1) 509-661-4052; E-mail: matthew. firstname.lastname@example.org. Dr. Wolff may be reached at WolffWare Ltd., P.O. Box 554, Norris, TN 37828; (1) 865-494-9653; E-mail: email@example.com. Mr. March may be reached at Hydro Performance Processes Inc., 614 Grandview Drive, Maryville, TN 37803; (1) 865-603- 0175; E-mail: firstname.lastname@example.org.
Matt Davis, project electrical engineer at Chelan County Public Utility District ’s 624-MW Rock Island Project, was project manager and lead engineer for the online monitoring work at Powerhouse One. Paul Wolff, PhD, president of WolffWare Ltd., performed the specialized data analyses to determine the cause of the generator misalignment. Pat March, principal consultant with Hydro Performance Processes Inc., provided consultation and technical guidance.