Redesigning Turbines to Eliminate Cavitation


When the new turbine runners at the 2,830-MW G.M. Shrum powerhouse suffered cavitation damage, BC Hydro collaborated with the equipment manufacturer to arrive at an acceptable solution for the future 80-year life of the units.

In 2009, Canadian provincial utility BC Hydro and turbine manufacturer Voith Hydro entered into a contract to modernize the Unit 1 through 5 turbine runners at the 2,830-MW G.M. Shrum powerhouse on the Peace River in Hudson’s Hope, British Columbia. The original runners had a long history of cavitation and cracking, and they had reached the end of their working life.1

As part of this work, the Unit 4 runner was replaced first, going into service in February 2013. After about 3,000 hours of operation, BC Hydro personnel took the unit offline for the first dewatered inspection. During this inspection, entrance edge cavitation damage up to 0.7 mm in depth was found on the suction side of each of the 13 blades near the runner band. At the time this was discovered, manufacturing of all five units was complete. The second runner was mid-way through installation, the third runner was in storage on site, and the last two runners were in offsite storage.

During the development stage of this runner replacement project, separate model tests were performed at Voith Hydro’s laboratory and at the LMH (Laboratory for Hydraulic Machines) independent laboratory in Lausanne, Switzerland. Neither model test revealed any entrance edge cavitation inside the operating range. However, cavitation damage was observed on the prototype. The hydraulic profile of the prototype blades was three-dimensionally (3D) scanned and found to be within the allowable tolerances when compared to the model design geometry. In addition, the unit was operated within the defined operating range.

Immediate corrective action

Promptly upon discovery of the cavitation damage, BC Hydro and Voith Hydro began a collaborative process to develop a permanent design solution to address the root cause of the cavitation.2 A runner leading edge profile modification was developed using unsteady computational fluid dynamic (CFD) analysis. BC Hydro and Voith Hydro held weekly design reviews to exchange ideas and concepts. The team had a wide range of experience, including site supervisors, engineers and project management. Once the new design modification was mutually agreed on, it was released for reduced-scale manufacture and verification by model testing.

A second model test was conducted in 2014 on the non-modified and modified runners at the independent hydraulic laboratory in Lausanne, Switzerland, to determine the benefits of the modified design. During the first model test, in 2010, the team could not fully visualize the runner entrance edge at the band. This was a contributing factor in the cavitation seen on the original design prototype.

Enhanced observation techniques were developed to overcome the limitations of the first model test. A model wicket gate was modified to include a window and an internal borescope, to provide a direct view of the entrance edge. Borescopes were also inserted into the flow, and an excellent view of the entrance edge was achieved.

The enhanced observation techniques revealed cavitation in the normal operating range on the unmodified model, as suspected. Some residual cavitation was also seen on the new modified model runner. However, it was in a seldom-used high head-low power zone of the operating range. It was mutually agreed to implement the model tested modification (Modification 1) on all of the Unit 4 blades.

The modified design was implemented on Unit 4 in July 2014 and the unit was operated inside the normal operating range for about 6,000 hours, then inspected in August 2015. Almost half of the operating hours were above rated head. During this inspection, entrance edge cavitation was again found on all of the runner blades despite limited operation in the high head-low power zone identified as potentially problematic during the 2014 model tests.

Root cause analysis

BC Hydro and Voith Hydro re-initiated the collaborative process to resolve the cavitation observed during the Unit 4 inspection in 2015. Upon thorough review of the hydraulic pressure fluctuations in the vaneless space between the runner and wicket gates, it was determined that the difference in cavitation behavior between the model and prototype was due to amplification of the dynamic effects of the interaction of the runner blades and wicket gates at the prototype scale. The inspection of the first modification on Unit 4 indicated that the rate of cavitation damage was about half that of the cavitation experienced with the original blade shape. Thus, the initial modification was not sufficient to eliminate the cavitation.

One interesting result of the August 2015 inspection was that the blade with the least cavitation damage was one that had residual buildup of protective epoxy material on the leading edge from earlier strain gauge testing. Although rough, this epoxy profile modification resulted in a larger inlet diameter, reducing the inlet angle of attack on the runner blade (see Figure 1). The locally larger runner inlet diameter increased the peripheral tangential runner velocity at the inlet edge, resulting in a lower angle of attack for the same absolute water velocity vector coming from the wicket gates. The added thickness of the epoxy also locally modified the suction-side blade profile and moved the core of the cavitation further from the blade surface.

The ultimate goal was still to define and implement a solution that would eliminate the cavitation damage. Many conceptual ideas were suggested during the collaborative weekly meetings, including modifying the surface finish, using coatings, trying compressed air injection and further modifying the profile (following the concept of the epoxy-altered profile).

CFD assessment of runner inlet edge cavitation

Both steady state and unsteady results of the CFD simulations were evaluated to establish what safety margin against cavitation was achieved for different modifications to the runner leading edge profile at the band. Figure 2 shows the steady state Thoma (sigma) number versus the relative blade length for the original design. The sigma value for discrete points along the blade is plotted relative to the plant sigma line. When the sigma values on the blade exceed the plant sigma values, cavitation is expected to occur. The streamlines at the band were the focus because this was the region where the cavitation was occurring.

This numerical simulation shows that the model has 17.0 m of “numerical steady state” margin from incipient cavitation (observed at the spike) at the runner inlet edge, but this was not enough to overcome the dynamic effects observed in the prototype.

Upon this discovery, the team recognized that the lower the sigma value on the leading edge (i.e. larger incipient cavitation protection margin), the less chance of the prototype runner amplifications exceeding the threshold for cavitation (sigma incipient). Figure 3 shows the original design compared to Modification 1 that was field tested and the proposed Modification 2. From the original 17.0 m of numeric steady state safety margin from cavitation, the first modification gained 9 m of safety margin, producing 26.0 m of cavitation safety margin. The second modification gained 17.8 m of safety from the original, producing a total of 34.8 m of protection from sigma incipient (based on steady state analysis).

In addition, unsteady CFD was analyzed to see if it could be used to correctly model the phenomenon observed on the prototype. Analyzing the unsteady CFD required some interpretation because the clock position of the runner blade behind the wicket gate resulted in different levels of the sigma incipient values on the inlet edges of the blades. At any given time, some blade inlets showed cavitation at sigma plant and others did not. Figure 4 shows a bar placed in the region where the first and last blade (of all 13 blades) would start to visually show cavitation on the blade inlets in the numerical unsteady simulation. This bar also helped guide the design to lower steady state values of sigma incipient.

Cavitation margin was observed in the steady state because the boundary conditions circumferentially average the flow coming from the wicket gates. When the unsteady CFD was calculated, it captured the full variation in flow coming from the wicket gates as it entered the runner domain — thus better capturing the cavitation behavior but still not at the correct sigma incipient values. Based on the unsteady CFD, Modification 1 should have resulted in acceptable prototype cavitation behavior because it produced 4.3 m of safety (see Figure 2). Because damage was observed on the prototype after Modification 1, it was clear there were additional dynamic effects on the prototype scale that were not fully captured by the unsteady simulation.

Runner cavitation mitigation alternatives

After evaluating all the concepts, Voith Hydro and BC Hydro mutually agreed the best testing method was to implement several cavitation mitigation alternatives on one prototype runner. This prototype test runner provided a method to quantitatively evaluate the alternatives under the same set of operating conditions and determine the one with the best cavitation performance. The team mutually decided to implement:

• A second runner inlet profile modification, Modification 2 (larger than Modification 1);

• A cavitation-resistant alloy weld repair using Cavitec, an electrode based on an alloy originally developed by Hydro-Quebec specifically for runner repair and now sold and marketed by Eutectic-Castolin;

• A local surface roughening, in an effort to prevent separation of the boundary layer; and

• A standard 309 weld repair as a benchmark.

Blade profile measurements for quality

To ensure proper implementation of the changes to the runner for Modification 2, a 3D photogrammetry scan was used from a third party using the HandySCAN 700 from Creaform. The report analyzed the data and showed section cuts parallel to the distributor centerline at different elevations to ensure that the modified profile was located at the correct diameter and that the profile blending to the existing blade was within tolerance. Figure 5 shows the profile results. The runner was returned to service after the 3D scan revealed the profile was within the tolerance.

Final outcome

In January 2016, the prototype test modifications were implemented on Unit 2. All traces of cavitation were removed and the profile was corrected back to the original shape and contour. Four conceptual ideas were implemented: three blades with Modification 2, three with a Cavitec repair, three with a roughened surface, and four with a 309 repair. The various blade modifications/repairs were evenly distributed around the runner to avoid creating an imbalance. Because cavitation damage depth is never completely uniform on all runner blades, the maximum cavitation depth values recorded during the latest inspection of Unit 2 were used to determine how the different repair concepts were distributed among the blades.

The blades were grouped based on the measured cavitation depth. The blades with the worst damage were grouped “bad” and the blades with the least damage were grouped “good.” At least one “bad” blade and one “good” blade per repair concept were selected to fairly distribute blades based on the expected damage potential.

In November 2016, Unit 2 was inspected after about 4,000 hours of operation, and it was observed that the second profile modification eliminated the cavitation. The roughened blades had the worst cavitation damage. The Cavitec repair had some cavitation damage but was clearly better than the 309 repair.

It was mutually decided to proceed with Modification 2 on all blades of all five units. Although repair with Cavitec is less invasive, the electrode is costly and challenging to use. In addition, repeated cavitation repairs can change the hydraulic profile, leading to a loss of efficiency, and can induce unwanted welding residual stress over time.

Over the anticipated 80-year life of these five units, cavitation repair would have been a significant cost and easily justified the implementation costs and developments for the successful profile modification.

The G. M. Shrum Units 1 through 5 operate on a three-year maintenance cycle, with each isolated every three years for routine maintenance. Cavitation repair work was historically included in these maintenance outages to avoid incurring lost opportunity costs. Based on the level of cavitation damage for the original design, it was estimated that a cavitation repair using Cavitec on one unit would cost about C$110,000 (US$85,000) every three to six years for the lifetime of the unit. Over the anticipated 80-year life of these five units, cavitation repair would have been a significant cost and easily justified the implementation costs and developments for the successful profile modification.

These photos of the November 2016 inspection of Unit 2 show a typical E309L blade with cavitation damage (left) and a typical profile modification (right).
These photos of the November 2016 inspection of Unit 2 show a typical E309L blade with cavitation damage (left) and a typical profile modification (right).

These photos of the November 2016 inspection of Unit 2 show a typical E309L blade with cavitation damage (left) and a typical profile modification (right).

Lessons learned

The G. M. Shrum runner inlet cavitation issue is considered unique. However, there are many lessons learned that can be applied to other hydro projects.

During the Units 1 through 5 turbine upgrade, Voith Hydro worked within the constraints of the existing embedded parts, which is inherently more challenging. Dynamic measurements of the original prototype unit may have been useful to identify any existing site-specific conditions that could result in a deviation from model expected behavior.

The original turbine was model tested by BC Hydro to establish the baseline performance characteristics. The stay ring, spiral case and draft tube from this test were re-used because they were unchanged by this project. As a result, the borescope positioning was not optimized for visualization of the new runner design.

The improved model visualization techniques should be implemented on model tests to improve the view of the entrance edge when all runner clock positions (relative to distributor) are not visible using conventional means.

One of the keys to identifying the successful cavitation solution was the collaborative approach that allowed members of BC Hydro and Voith Hydro teams to actively contribute. The participating team members had a wide range of experience. Having this diversity of individuals committed to the end goal fostered the creativity and motivation needed to succeed. This collaboration approach should be used for all future turbine projects.


1Finnegan, P., “Modernizing 40 Year Old Turbines – Increasing Efficiency and Adding Power,” Proceedings of HydroVision International 2013, PennWell Corp., Tulsa, Okla., U.S., 2013.

2Taylor, K., “Enhancing Cavitation Performance of the GMS Units 1-5 Turbines,” Proceedings of Hydro 2015, Aqua-Media International, Surrey, UK, 2015.


Armstrong, Marie, Zuobin Donald Tang and Indra Kusuma, “Expanding into the Future: GM Shrum Station Service Rehab,” Hydro Review, October 2016,

Katie Taylor is mechanical engineer and Danny Burggraeve is principal engineer with BC Hydro. Brandon Harmer, P.E., is hydraulic design engineer and Stuart Coulson is vice president – modernization technology with Voith Hydro.

Original Source