Index Test Box Setup and Diagnostic Tools For Kaplan Turbines 


Winnebago, Illinois

January 14, 2016



Kaplan gates and blades must be positioned accurately and robustly in order to maximize efficiency and service life. The first phase of any index test is a thorough checkout of the unit’s control system.

The Index Test Box (ITB) was developed at Woodward Governor Company in 1984 as a dual-purpose test instrument; the primary function is analysis of Kaplan type hydroelectric unit efficiency in order to optimize (or tune-up) the 3-D cam surface profile. The secondary purpose is diagnostic analysis of the turbine control system behavior. 

A third function for Condition Monitoring has recently been added to enhance the ITB’s utility value. 

The ITB provided remote diagnositc analysis and index-testing capabilities using off-the-shelf IBM Pc type computers and data streamed from the existing powerplant SCADA system over normal office LAN or Ethernet communication.

This allows the engineers back in the office to observe and diagnose turbine control systems behavior in the field, in real-time over the Internet.


Setup and Instrumentation

Modern powerplant SCADA systems typically have the necessary instrumentation for an index test, except in some cases the Winter Kenney taps are non-existent or blocked with sediment. Every unit in every powerplant will be different. 

To setup a unit for these tests the governor is modified to provide signals to the SCADA system for gate setpoint, blade setpoint, blade position and governor inputs are added for manual gate and blade positioning. Many turbine governors don’t bring out these signals to the SCADA system. In every case, the usual powerplant’s engineers and technicians have been able to do this setup work. Next the SCADA system is configured to bundle the required signals and route them to the existing powerplant data logger.


         1.        Forebay

         2.        Tailwater

         3.        Gate Setpoint

         4.        Gate stroke

         5.        Blade Setpoint

         6.        Blade angle

         7.        Flow (Winter Kennedy or other flowmeter signal)

         8.        Power (Watt transducer output).

         9.        Grid frequency

      10.      Grid voltage


The powerplant data logger is reprogrammed to scan the above listed signal channels, add a scan-count and date/time stamp and record scans twice/second. Data is recorded continuously while the machine is operated normally.

Whenever an interesting event occurs, data files are sent by sneakernet to ATECo for analysis.

The first concern is to determine if the govenor is robust, accurate and precise enough for index-testing "on the governor."

The ITB software is not needed for most of these checks - simply looking at the data in an Excel spreadsheet and knowing what to look for will show up many of these problems.


Kaplan Off-Cam Index Testing Procedures

There are 3 ways to test a Kaplan turbine off-cam. 

The most common is Fixed Blade, swept gates and power

The second is Fixed Gate, swept blades and power.

The third is the Constant Power, moving gates and blades 

The Fixed Blade method is the most commonly used because on most existing governor cabinets the gates are easier to control using the Gate Limit than the blades are to control by climbing up inside the governor cabinet to make adjustmenst to the thumbscrews on the blade hydraulic-amplifier's floating lever.

method is vastly superior for testing a single run of the river turbine with small forebay and tailwater.


ITB Prototype Test Demonstration (1985-1986)

ITB Connected to Kaplan Governor Cabinet at Clarence Cannon Dam

Woodward’s first ITB was field tested in conjunction with the acceptance test of a 33MW vertical Kaplan at USACE’s Clarence Cannon Dam. USACE test engineer Don Sachs was a proponent of the Constant Power index testing method that the ITB embodies, but HDC insisted that the classical Fixed-blade test method be used for his index test. This author was recruited into his index-testing crew as the “governor man.” A12-man USACE test crew performed the index test according to the procedures laid out in HDC’s “Red-Book.”


Because Lee’s tutorial on off-cam Kaplan index testing was the recipe used to design the original ITB, both sets of test data were forwarded to Lee at BPA for evaluation. Lee prepared a comparative analysis report of these two data sets, prompting BPA to purchase the first ITB from Woodward 6-months later.


Demonstration of Woodward’s ITB at Bull Run Dam (1987 to 1988)

When the original ITB was developed at Woodward Governor Company a U.S. Patent was acquired to protect the new instrument. The first test was at Portland General Electric’s (PGE) Bull Run Dam (PGE-PHP-2) located about 30-miles East of Portland, Oregon. The field-test at Bull Run went exactly to plan, according to field-test reports by Lee Sheldon, Gary Hackett (senior staff engineer from PGE) and Terry Bauman (The engineer that Woodward hired to take over the ITB project after I returned to Woodward’s aircraft development lab). 

Because he didn't have a dog in the fight - Gary’s conclusion was the most compelling of the three:

An article in Hydro Review pitched the new ITB from Woodward.

After this successful test BPA offered to procure ITBs for all of USACE’s large Kaplan turbines in a project estimated to exceed $30-million to acquire the new governors, 3-D Cams and ITBs, but instead of accepting this FREE proven equipment that BPA would purchase from a world-class supplier (Woodward) using DOE's funds, HDC engineers attempted to duplicate Woodward’s ITB with their own design at DOE expense. A BPA report on this project was acquired years later; despite the report’s encouraging words the project had failed to achieve its objective and was quietly abandoned. Woodward shelved and abandoned the ITB project after the government didn’t buy them.

This author quit Woodward in 1992 to startup the Actuation Test Equipment Company (ATECo) to resurrect the ITB as a new enterprise. When a prototype was ready Woodward was approached to add a proven automatic self-optimizing capability to their Kaplan governor. Threats of lawsuits kept the ITB out of the market. Woodward didn't want it, but they didn't want anyone else to  benefit from it either. This and other blunders by Woodward’s hydro managers led the division to becoming so unprofitable that corporate managers sold the entire division to GE for $1 to sweeten the deal when negotiations were underway to sell Woodward’s Engines and Turbines division to GE. GE got the Patent with the hydro division, and when they were approached with the ITB a month later they were also not interested. So the ITB sat idle until the patent expired 17 years later.

Experience with ITB at USACE HDC (2003 to 2008)

In 2003 Lee was working at HDC as a rehired annuitant tasked with index testing all 113 of USACE’s large Kaplan turbines on the Columbia River under the asupices of the fish-mortality lawsuit in Federal court. The patent had expired 2-years early because GE neglected to pay the 3rd renewal fee. Rod Wittinger said he had been watching the Patent and when he saw it had expired, directed Lee to contact this author to acquire the detritus of Woodward’s ITB project as the start of a government project to resurrect the ITB - but ATECo had already beaten them to it. HDC purchased a 2nd generation ITB from ATECo on a government contract after 2 sole-sourced solicitations (#1  #2). A U.S. Copyright was acquired to protect the software source code intellectual property. This version of the ITB consisted of an IBM PC with a National Instruments I/O board and Software Toolbox TopServer OPC communication program. The ITB was connected directly to the GDACS SCADA system via a Cat-5 LAN cable and TopServer OPC program.



Figure 1 ATECo Pressure Transducer Cart with Electric Flushing and Integral Air Compressor

A pressure transducer cart was constructed by ATECo to facilitate moving the ITB from machine-to-machine in a powerhouse. The ITB was deployed to McNary and Ice Harbor Dams. At McNary, 2 ATECo personnel attended the test conducted by government personnel. At Ice Harbor, government personnel ran the ITB without anyone from ATECo present. The desired configuration had the ITB residing in the control room and the transducer cart moved from machine to machine.


ITB Test at McNary Dam (December 2005)

Rod Wittinger’s report on the test at McNary explains how the turbine was manually exercised to the required gate and blade positions while the ITB monitored the SCADA system’s data, and when Steady-state operation was detected at each test point, the ITB automatically started recording data scans. Many steady-state scans are collected at each test point and evaluated later by a PostProcessor step to pick out the best test points. The ATECo test report explains further.

Text Box:  
Figure 3 Final Graph of Test Data

Figure 2 ITB Test Data from McNary Dam

ITB Test at Ice Harbor Dam (February 2006)

For this test the ITB was only used to monitor and record data in parallel with a classical Fixed Blade manual index test conducted by HDC’s test engineers using their customary methods and tools. When the test was completed, the “canned” data from the ITB was emailed to ATECo for analysis. When the reduced result was returned, HDC’s engineer charted the comparison of both results in an Excel® spreadsheet.


Figure 3 Index Test Results for COE and ATECo Instrumentation

His internal HDC memo reported on the test. The graph above has a slight offset in the Winter Kennedy calibrations to show the similarity of the results. If the calibrations had been equivalent, the efficiency curves would have laid right on top of each other. In the subsequent test report to Bonneville Power Administration (BPA) Hydro Optimization Team (HOT) USACE personnel said that the ITB results were “Virtually identical to those obtained using COE data acq system,” and that the ITB was “Ready for unattended automated data collection.”  Lee Sheldon* commented that the ITB data had less scatter than the conventional data, falling closer if not right on top of the efficiency curves for individual fixed-blade sweeps. After this successful test, instead of buying the proven ITB from ATECo, government engineers initiated a project to reverse-engineer the ITB in an internal government project. This all happened 10 years ago and by every indication (Periodic FOIA requests are keeping tabs on them), they’ve been un successful just like when they attempted to duplicate Woodward’s original ITB in 1990 with their “automatic index testing device,” submitted to BPA as a potential surrogate to Woodward’s ITB.

(*Please feel free to call Lee to verify any of this, we’re currently working together on the project to index test the 5MW turbine discussed herein: 503-356-8302)


Experience at Clergue Powerplant in Sault Ste Marie (2009 to 2013)

The best use of the ITB is when it can be connected to the SCADA system so that sample rates of a KHz (more or less, depending on the computer and I/O board selected) allow higher sample rates. A cavitation monitor and alarm in the ITB would be a valuable feature, but due to powerplant politics and policies a direct connection is forbidden. Most powerplants prohibit connecting strange, new equipment (like the ITB) from unknown suppliers (such as ATECo) to their SCADA system without first vetting the supplier and scanning for viruses and malware - and in many cases there’s no money in the budget for it. This was the case at Clergue in 2009. The ITB was demonstrated to work with North American Hydro’s (NAH’s) new governors planned for the 3 turbines there. Dave Kornegay, NAH’s project engineer had worked alongside me in 1984 when I developed the first ITB at Woodward. Dave knew that the ITB would work for his governor designs at NAH and wanted to add a “self-optimizing” feature to his Kaplan governor. Unfortunately, Brookfield Renewable Power (BRP) had a policy of not allowing any computing equipment or internet connections to their SCADA system and the money for the new ITB hardware had not been approved. Dave really wanted this feature on his governors, so along with Andrew Punkari (Andy) BRP’s project engineer, Dave setup the governor software and prepared specifications for the governor, powerplant instrumentation and SCADA system software changes; Andy programmed the powerplant data logger to capture scans at a 2-Hz sample rate. All of this was accomplished by the powerplant personnel using the pre-existing powerplant equipment and sensors; most of the necessary signals were already in the SCADA system so setting up the recorder was simple and straightforward. Data was collected for several months and sent to ATECo periodically.


Another prohibition at Clergue was because their contract with the power marketing agency stated that if the machine was run in a test-mode (i.e. off-cam) the price paid for the power was halved. BRP managers balked at the reduction in their revenue stream so off-cam testing was disallowed. To get around this prohibition, a blade-offset was added to the governor and HMI so that slightly off-cam data could be acquired. Blade offsets were changed every few days, keeping within a range of +3.0%. When the ITB test results were compared to the Hatch Acres 2006 index test results good correlation was found. Unfortunately we were unable to finish the testing because Andy was dispatched to make repairs at another powerplant 200 miles distant and was gone for over a year, so the ITB testing waited. Before he returned, Alstom bought NAH and Dave jumped-ship to move to Innovative Automation - who did not have a spot for the ITB in their product line.


Necessity is the Mother of Invention

By modifying the existing governors, instrumentation, SCADA system and data logger to collect the required data, and having the operators position the gates and blades manually while the powerplant data logger stored 2 scan per second, the data for a Kaplan off-cam index test can be collected in 4-hours. Low cost CDs, thumb-drives and high-speed broadband Internet transfer of data files provide good alternatives to a direct connection to the SCADA system. The most beneficial result of this is the logistics of sending expert test personnel to the dam are no longer necessary.


This short video was prepared to publicize our method:

Steady State Scan

Figure 4 ITB SteadyState Analysis Display and Link to Clergue Video

Click here: Http://


Current Field Test (December 2015 to present)

Lee Sheldon got the job to index test a 5MW vertical Kaplan turbine and engaged with ATECo to utilize the ITB’s Hybrid Index Testing capability for the job. The facility is a flood control dam that was electrified after 63 years. An index test was sought for the new 5MW vertical Kaplan to develop an optimized 3-D Cam surface for this machine. An off-cam index test was needed to correct the myriad issues that affect the optimum shape of a Kaplan turbine’s 3-D Cam head and gate to blade surface.


Pre-Index Test Control System Checkout

A proper index test includes a dewatered inspection of the unit so that the surface condition of water passageways and turbine working surfaces can be checked and the full range of gate and blade motion can be verified from squeeze to full open while the normal powerplant instrumentation and index testing equipment are adjusted and calibrated. In lieu of such a comprehensive inspection, with the above described modifications to the system software a modest review of the data characteristics and time-response behavior can determine if gate and blade movements conform to ASME and IEEE performance standards. Below are a few examples of how this remote data observation capability discovered problems with the Kaplan control system.


This first check found a problem with the synchronizer. Getting a unit in sync with the grid before closing the breakers on startup is a tedious and precise requirement of the turbine control system. Most governor systems use an external accessory such as Basler Electric’s BE1-25A Automatic Synchronizer for this. This data from the 5MW Vertical Kaplan shows that rotor bounces against the rotating magnetic field when the breaker is closed on startup, which indicates the synchronization may not be adequate. This was seen in typical operation data recorded during a normal startup.


Figure 5 ITB Cartesian coordinate display during normal startup

The green lines on the left side are efficiency spikes that indicate the rotor was not spinning at the same exact speed as the stator’s magnetic field when the main breaker was closed. The unit should be at “speed-no-load” with flow but no power generation i.e. a slightly negative efficiency value. What is happening here is when the breakers close the rotor bounces against the rotating magnetic field until it is pulled into synchronization. This may or may not damage the unit, depending on how far out of sync the unit is when the breaker closes. In any event, the control system should synchronize the unit to the grid better than this for reliable long-term operation.


Another problem was found in the stripchart traces from viewing the datalogger files using Microsoft Excel®.

Figure 6 Excel® Stripchart Display Showing Blade Instability

At first glance the bottom trace in the chart above was noticeably thicker than the others, so it was magnified at each blade value used for the index test to investigate further. The Y Axis on these trace are blade angle in percent and the X Axis is number of scans where each scan is 0.51 seconds


Figure 7 Excel® Stripchart Display of a Magnified View of the Blade Instability

This blade trace display is magnified to show the amplitude and frequency of the instability. The blades are always moving about 0.6% to 2.0% p-p with about an 8-second period. This is most likely caused by a positive lapped pilot valve on the blade servo.


Another problem found with the PLC-based Kaplan governor with a 3-D Cam and blade controller and an external AGC in the SCADA system. This next graph shows the normal startup of this machine. When generation is ramped up, the gate/blade indicator moved from left to right. When it got to the 40-ft head line, the blades tracked up the 41ft blade cam line instead of 46 ft, then the gates stop moving at 85% and the blades continue going straight up, causing further increases of flow and power after the gates stopped. At this point the AGC sees that power is too high and the gates are closed a bit, and then the blades come back down a short time later. This chart shows how the gates and blades continue to chase each other before finally settling at the 41 ft on-cam line at the far right.


Figure 8 ITB Cartesian Coordinate Display showing 3-D Cam Output Problem

This chart shows conflicts between the governor, sluggish blade to gate relationship and Automatic Generation Control (AGC). The two red lines immediately above and below the heavy blue "On-Cam" line are the ASME prescribed 1.0% deadband for proper blade positioning.


A few more examples of problems detected at Clergue are included below:

XYGraph Unit1

Figure 9 ITB Display of Gate/Blade motion of Kaplan Bulb Unit with NAH Governor

The chart above shows the actual gate/blade relationship when the new governors were installed. The blades and gates mostly track the on-cam line at the high end, but something screwy is happening at the low end. Below 30% gate the blades change direction and go up to about 45% blade angle instead of tracking down along the on-cam line. When this was showed to the governor engineer he consulted with the programmer and this got fixed.


Clergue at 7.35m only

Figure 10 ITB Display of Gate/Blade motion of Kaplan Bulb Unit with NAH Governor 3-D Cam Data Table Bug

A second problem was seen in the on-cam line when head was above 6.8m. The cause was an error in the lookup table at the highest head. Again, as soon as the problem was seen, the programmer slapped his forehead and fixed it in a jiffy. Without this Cartesian coordinate display he wouldn’t have caught this problem until it was in the field, and then he’d be fixing this under the customer’s nose.



These real-world problems are not presented herein to be critical of anyone’s workmanship; instead they are here just to show the utility value of observing the gate/blade behavior of a Kaplan unit with an X-Y Cartesian coordinate display in the field. If the problems can be observed and repaired during installation and setup of the governor, everything will go smoother. In many cases, the decision makers are business types or civil engineers; unfamiliar with governors and their dynamic requirements.  


Sincerely yours,

Douglas Albright

Actuation Test Equipment Company

(815) 335-1143