Chelan County PUD owns and operates the Rocky Reach hydroelectric project located just outside of Wenatchee, Washington, on the Columbia River in the US. The project consists of 11 generating units, the first of which was installed in the early 1960s. Units C1 to C7 generators were originally commissioned beginning in 1961, and were designed and supplied by Westinghouse. Except for rewinds performed in the 1980s, the 117MVA machines ran fairly reliably for nearly 40 years.

In the late 1990s they began showing numerous signs of ageing, when three major failure events occurred in a two-year period including two winding failures and a rotating strike. Post event investigations found signs of fatigue cracking of the rotor rim and spider, and evidence that the rotor rim floated at rated speed. Machine condition monitoring equipment indicated that the stator shape was not round, creating a wide variation in air gap.

Feasibility studies were launched in 2001 to define the problems. Based on the earlier investigations, it was concluded that the rotors were at the end of their useful life, and the stator windings were at, or near, theirs. Different implementation plans were identified and analysed for the total benefits and impact on future operations. Each plan had different capital costs, outage duration and long term benefits, including impact on O&M cost. Based on the analysis of the different options, a final decision was reached to replace the generators completely. This solution provided for the best balance of addressing the expected root causes, minimising the outage duration, and providing the maximum life extension (anticipated 30-50 years). In order to combine the first unit outage with another planned outage, a fast-tracked implementation was desired, and the project specifications were quickly prepared and the project was awarded to alstom Power in 2001.

New generator design and installation

As a result of the previous rotor-stator contact and subsequent investigations, maintaining a generator air gap that was circular and uniform was important to the PUD. Therefore, the project specifications included technical requirements meant to achieve this goal. In particular, the stator frame movement was to be controlled such that at no time would the air gap have a difference of more than 20%, from the smallest to the largest reading. The total eccentricity of the air gap at any time also could not exceed 10%. These criteria would be difficult to achieve with a large stator (diameter 10.6m, core height 1.8m), but due to the previous rotor/stator contact event and lessons learned from it, the PUD wanted to prevent this type of event from occurring again.

Based on these requirements, along with contractually required studies, and Alstom’s design standards, the new generators were designed to obtain an overall mechanical stiffness higher than the magnetic stiffness. The resulting design has a rotor rim that remains shrunk fit to the rotor spider at speeds up 150% of nominal. The stator frame is designed to be sufficiently stiff, but still without risk of buckling the core. As a result, thermal expansion capabilities of the frame are not hampered. The interface between the stator frame soleplate and its foundation is designed to utilise the existing embedded plates, which helped to keep the installation outage duration short. The soleplates allow only radial expansion while also controlling the eccentricity of the frame. With these design elements, it was believed that a very stable air gap could be obtained.

Alstom erected the new stators and rotors entirely in the erection bay of the powerhouse prior to the outage. When the scheduled outage time arrived, the old generator components were removed, and the new ones were installed in their place. This reduced outage time to between 29 and 38 days (later units were quicker). During installation, particular attention was paid to the alignment of the stator and rotor in order to achieve the air gap variation tolerances required by the specifications.

To provide a means to verify that the specification requirements were met and provide ongoing machine condition assessment, air gap monitoring equipment manufactured by vibrosystm was installed on each generator. The monitoring equipment consisted of sensors mounted to the face of the stator core that were capable of monitoring the air gap between the rotor and stator during operation. Additional sensors to monitor stator frame movement relative to the stator foundation were also installed. The primary items of concern to the PUD were the overall maximum to minimum variation of the air gap, and free movement of the stator frame during thermal cycling of the generator.

Overall, the project was implemented very effectively and was finished on schedule and under budget. However, after installation of the third unit, it was confirmed that the new generators were experiencing air gap variations that were well outside of the specified and expected value. The findings caused the project team to re-evaluate what was thought to be the root cause of previous problems, and brought to light interactions between the civil works and the generators that had not been documented prior to the rehabilitation.

Air gap variation issues

The first unit of the replacement project (at the south end of powerhouse) was installed in March 2003. Operation of this is rarely stopped as it is a fish attraction unit and supply for several critical station service loads. The air gap variation experienced here was slightly larger than specified but not excessive and was primarily chalked up to it being the first unit.

However a fairly dramatic air gap variation increase was noted immediately after commissioning of the second unit in November 2003. With the first unit performing just outside of specified tolerances and the second well outside of them, it was unclear if either unit was an anomaly.

When the third unit returned to service in November 2004 it quickly erased any notion that the air gap variation issue was not real. This unit experienced a variation in stator air gap that approached 35% by mid March 2005 but then decreased to a minimum of about 16% by mid September (almost back at installed condition). However it then began to climb again, following a predictable, repeatable pattern. Data gathered for the first two units confirmed a similar pattern of variation, but to a lesser extent.

When trended, it was apparent that the air gap variation was cyclical in nature and appeared to be tied to the seasonal variation of the temperature of the Columbia River. The water temperature varies between about 3-20ºC, with the minimum and maximum being reached, on average, around 20 February and 20 August respectively. The variation in air gap values peak and bottom out about three weeks after these dates.

It is worth noting that the air gap variation was occurring only to the stator; the rotor maintained its original shape very well throughout the variations. The eccentricity also does not vary much for the stator frame. This is testament to the good performance of the rotor and overall centering of the unit. Polar plots indicated that the shape of the stator frames tended towards an elliptical shape, with the major axis aligned north/south, which is the axis of the powerhouse.

Potential causes

With excessive air gap variation confirmed on the third unit, project team members from the PUD, the PUD’s consultant MWH, and Alstom met to discuss possible causes. Among the items brought up for discussion were:

• Stator frame stiffness.

• Alkali aggregate reaction (AAR) of the concrete.

• Free expansion/contraction of the stator frame.

• Non-uniform stator cooling.

• Stator soleplate design.

• Powerhouse design.

The first three items were deemed to be unlikely, as the stiffness was more rigid than the majority of units designed by Alstom; the powerhouse structure had not experienced any symptoms related to AAR; and inspections did not reveal any external components restricting free expansion/contraction of the frame. The remaining items were thought to be likely contributors to the excessive air gap variation.

During installation of the next two units, attempts were made to reduce the air gap variation to within specified limits by balancing the stator cooling and making modifications to the stator soleplate design. While marginal improvement was achieved, the air gap variation remained outside of specified tolerances. It was also learned that the time of year of stator installation (centering) dictated when the maximum air gap variation would occur. A unit installed in the fall experienced its maximum air gap variation the next March; a unit installed in the spring experienced its maximum in September, which was in phase with the river temperature variation curve. This would lead to an important revelation later.

Faced with only two units remaining to install, the theory of powerhouse concrete expansion and contraction returned to the forefront. Although in great physical condition, the concrete does expand and contract seasonally, with a coefficient of expansion that is similar to steel. The Rocky Reach units are low head units, and utilize large Kaplan runners. With the low head, the water passages are very large, and the affect of the water temperature variations may be more pronounced than on other types of units with smaller passages. There is a very large surface area for the concrete to “soak up” the temperature from the water. These large water passages also tend to make the concrete monolith less rigid and prone to distortion.

Rocky Reach consists of multiple concrete monoliths that make up the powerhouse structure. The expansion of each unit monolith is restricted on the sides by another unit monolith that is similarly trying to expand (except the end units, which are unrestricted on one side). Expansion is less restricted on the upstream and downstream sides. The shape of the concrete where the stator soleplates are attached tends to expand more in the upstream-downstream axis of the unit. This difference in expansion leads to the greatest elliptical shape of the stator soleplates along this axis three weeks after the water temperature reaches its maximum (approx 20 August); it then returns to its most circular shape in mid-March, three weeks after the water temperature reaches a minimum (approx 20 February).

The stator frame is fixed to the concrete via its soleplates at the time of centering the stator during installation. The soleplates consist of a sliding surface and radial dowels to allow for thermal expansion and contraction of the stator frame. Since the concrete is expanding and contracting in this location, and taking a non-circular shape, the radial dowels can become misaligned in the radial direction, causing the sliding feature to be reduced or eliminated altogether. The stator frame will continue to expand with any temperature rise; it just expands where the resistance is least (or least misaligned), resulting in non-uniform movement. In the case of Rocky Reach, this is along the upstream/downstream and north/south axes.

An examination of the shape differences for units installed at different times of the year also bears this out. The fourth unit was installed in the spring, and its shape variation is directly opposite to that of units installed in the fall. In all cases, the best shape, ie smallest air gap variation, occurs near the time when each unit was centered and the soleplates were radially aligned. When installed in the spring, a unit develops its maximum elliptical shape in the fall and vice versa for a unit installed in the fall (also with a 90º axes shift). This reversal of the phenomenon fits a seasonal distortion of the generator foundation.

Based on the foregoing, it was apparent that a different approach was required. It was thought that if the radial dowels in the soleplates could simply be removed from all but the primary axes (US/DS, N/S), the stator frames would likely remain round and experience very little air gap variation. However, due to the particular design of the generators, an electrical fault could result in serious damage, including possible damage to the concrete structure. Instead, increasing the clearance of the radial dowels by 0.5mm was proposed, in the hope that this would prevent the dowels from becoming misaligned or locked.

This was carried out on the sixth unit. The results were similar to previous attempts (some improvement), but the air gap variation still exceeded the specified maximum. With six of the seven units now installed and none of them meeting the specification requirements, another idea that was previously set aside had to be considered.

Presetting the stator frame shape

The notion of presetting the stator frame in such a way as to take advantage of the predictable nature of the air gap variation had initially been rejected, since doing so would not fix the main cause of the variation. However, since a complete redesign of the stator soleplate system was not feasible without major cost and outage time, reducing the overall magnitude of variation became a viable option.

To test the presetting method, the seventh unit was installed in a preset elliptical shape to compensate for the expected movement of the concrete. The goal was to set the stator so it would equally distribute the resulting deformation plus and minus of the perfect round shape. If the presetting is precisely correct, then it should reduce the air gap variation by 50%. The risk of this exercise appeared to be low; if the shape was incorrect and the deformation moved the wrong way, then the unit could be shut down for a short outage to correct the shape with adjustable keys.

A review of the air gap data from the earlier units provided the basis for deciding how elliptical to preset the seventh unit. The air gap sensors located on the primary axes of the generators (the 90s) were seeing approximately 1.5mm of air gap variation, while the ones on the secondary axes (the 45s) were seeing very little variation. In this way, the secondary axes act as inflection points on the ellipse, and are nearly stationary.

In mid October 2006 work on the seventh unit was carried out. Slightly less than half of 1.5mm was used as the value for presetting the frame. Initially, the stator was centered and rounded out to be near perfect (<5% roundness). Then jacks were used at the north and south locations of the stator frame to push the stator in about 0.75mm. An outward deflection of 0.75mm was obtained at the upstream and downstream locations, with very little movement noted at the 45s (as predicted). Rotational air gap measurements confirmed the desired preset shape had been obtained, which was slightly elliptical, with the US/DS axis as the major axis. The stator frame was then fixed to the embedded soleplates per the procedure used for the other units.

When the unit returned to service in late October 2006, it was obvious that a change had been achieved. Instead of the air gap variation immediately beginning to increase dramatically, it decreased. It continued to decrease into December and then the variation started to increase. The maximum variation occurred in mid March 2007 as predicted, but at a value that averaged approximately 27%, which was well below maximums experienced on the other units. Another minimum occurred in late June that matched the December value. The variation then increased into September and then started the cycle again.

After the seventh unit was returned to service, the third unit, which had confirmed the existence of an excessive air gap variation issue, also had its stator frame preset. After it was returned to service, the unit experienced a very similar variation pattern to the seventh unit. The air gap variation was reduced from over 35% to approximately 27%.

Over a 12 month period it became clear that a slightly more aggressive presetting could have been used which would have lowered the maximum variations further. However it was decided best to err on the side of caution on the first attempts until a positive result was achieved. Based on the results, a larger displacement may be considered for presetting other stator frames as required over the life of the new generators.

Confrontational remedies

The C1-C7 generator rehabilitation project specifications were written based on addressing observations (assumptions) about the root cause of recent problems with the units. Though a thorough analysis and new design were provided by Alstom per the specifications, the problems with an excessive air gap variation persisted. It was only after trying several remedies, and carefully monitoring the units for three years, that a primary root cause became clear.

As time progressed, and no modifications solved the problem, there were opportunities for both the PUD and Alstom to choose more confrontational remedies. However, a good working relationship was the key to keeping everyone focused on the problem, and ultimately finding a solution. This relationship was developed early on in the project, and it cannot be stressed enough how important this was to arriving at a satisfactory conclusion to this issue.

Once the concrete movement theory was satisfactorily proven, presetting the stator frame shape in order to achieve a reasonable air gap variation was the logical choice. Redesigning and rebuilding the stator frames and/or soleplates was not a feasible alternative. The results achieved have been positive. One other stator frame has been preset by the PUD (October 2008), and another is scheduled.

Previously, most generator suppliers considered the foundation to which the stator frame is attached as a stationary point. The experience on this project highlights that normal seasonal concrete expansion and contraction should be expected and assessed. Depending on the design of the powerhouse; number and type of units; temperature swings experienced locally; and the physical size and arrangement of the units, seasonal concrete movement will likely have some impact on stator shape and alignment. Being aware of this in the design stage can lead to a generator design that accommodates this movement, and eliminates the excessive variation altogether.

The authors are Carey Schenck, Douglas County PUD; Steve Sembritzky, Chelan County PUD; Don Erpenbeck, MWH Global; and Daniel Chenard, alstom-hydro Canada