Experimental Test Plan—1:6 ScaleReference Model 2 Cross-Flow Turbine

P. Bachant1

M. Wosnik1

B. Gunawan2

V. Neary2

1Center for Ocean Renewable EnergyUniversity of New Hampshire

Durham, NH2Water Power Technologies

Sandia National LaboratoriesAlbuquerque, NM

December 15, 2014

Prepared by:Center for Ocean Renewable Energy

University of New HampshireDurham, NH

Prepared for:Wind and Water Power Technologies Program

Office of Energy Efficiency and Renewable EnergyU.S. Department of Energy

Washington, D.C.

Contents

1. Introduction 31.1. Study goals and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2. Experimental setup and methods 62.1. Facility and instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1.1. Calibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1.2. Synchronization of instrumentation subsystems . . . . . . . . . . . . . . 92.1.3. Tare drag and torque compensation . . . . . . . . . . . . . . . . . . . . 9

2.2. Turbine model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.3. Test parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.4. Determining tank settling time . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.5. Determining Reynolds number independence . . . . . . . . . . . . . . . . . . . 112.6. Measuring the effects of strut drag . . . . . . . . . . . . . . . . . . . . . . . . . 122.7. Near-wake characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.8. Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.8.1. Uncertainty analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3. Research deliverables 173.1. Experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.1.1. Directory structure and naming conventions . . . . . . . . . . . . . . . . 173.1.2. Metadata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.2. Management and archiving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.3. Licensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4. Summary 204.1. Milestones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5. Acknowledgements 21

A. Sample test plan matrices 24

B. Data acquisition software, file content and organization 25

2

1. IntroductionSandia National Laboratory (SNL) developed the cross-flow turbine, Reference Model 2 (RM2),to provide an open-source point design for MHK research and development, including its useas a test object for scaled model testing to generate power performance data that can be used tovalidate open-source design tools. More details on the reference modeling effort are provided inNeary et al. [1].

Dimensional analysis provides scaling laws that are used to upscale model test data into per-formance and design information for a full-scale prototype turbine. For hydrokinetic turbines,hydrodynamic similitude is achieved when the Reynolds number, Froude number and tip-speed-ratio of the model and the full scale device match. The Reynolds number is defined as

ReL =ULν

, (1.1)

where U and L are characteristic length scales, and ν is the fluid kinematic viscosity. The Froudenumber is defined as

Fr =U√gL

, (1.2)

where g is the gravitational acceleration, and the tip speed ratio is defined as

λ =ωRU∞

, (1.3)

where ω and R are the rotor angular velocity and radius. For hydrokinetic turbine testing, Re, notFr, is the dominant scaling parameter. However, the Froude number (based on blade tip submer-gence) should remain small (significantly below one). It is rare to achieve perfect similitude forRe, but a threshold value should be exceeded, such that scale model results can be extrapolated.

As a cross-flow turbine, the RM2 blades will experience large variations in angles of attack asthey rotate about their axis (“cross-flow turbine” means the axis of rotation is perpendicular toflow direction—either vertically or horizontally). These variations become larger as the tip speedratio decreases [2]. Typically, the maximum angles of attack are sufficiently large that the bladesoperate under dynamic stall, which is a complex unsteady process and deviates significantly fromstatic foil behavior, during part of the rotation. Furthermore, the higher the solidity of a cross-flow turbine, the lower the optimal tip-speed ratio at which it operates. Marine Hydrokinetic(MHK) cross-flow turbines typically have higher solidity than cross-flow wind turbines (VAWT),and hence operate at lower tip-speed ratios. Since MHK turbines operate in a higher density fluid,unsteady dynamic effects related to the blades’ pitching motion and flow curvature also becomemore important when compared to wind turbines.

3

The performance of cross-flow MHK turbines thus depends on both Reynolds number andsolidity (note that these issues are related, since an average blade chord Reynolds number,Rec,avg ≈ λU∞c/ν , can be expressed in terms of tip speed ratio, the optimal value for whichis a function of solidity). If numerical models are validated with physical model data that wasobtained at insufficiently high Reynolds numbers, it cannot be determined whether problemswith model predictions are caused by Reynolds number effects, issues related to higher solidity,or both. It is uncertain whether numerical models validated with physical model data obtained atlow Reynolds number should be considered validated at all, since the scale at which the modelwill be applied is orders of magnitude larger. One way to overcome this uncertainty is to showthat the scaled physical model test has become Reynolds number independent, and thereforevalidation efforts should be relevant at full-scale.

For example, the effect of Reynolds number on average power output was shown to be signif-icant on the 2 m Sandia Research Darrieus turbine in wind tunnel testing [3]: Maximum powercoefficient, CP,max, increased with Reynolds number, Rec, along with a shift of the location ofCP,max toward lower tip speed ratios due to delayed blade stall. The effects of Reynolds numberwere quite dramatic over a relatively small range of Rec ≈ 1.1× 105–2.9× 105. More recently,Bachant and Wosnik [4] showed that turbine performance and near-wake characteristics becomeReynolds number independent at Rec ≈ 2×105.

The need for experimental data that is relevant to full-scale behavior stems from the need tovalidate numerical models—most importantly mid-fidelity models—desirable for MHK devel-opers to predict the performance of their cross-flow turbine designs, since physical modeling atappropriate scales can be prohibitively expensive in the early stages of engineering. Further-more, Navier–Stokes-based computational fluid dynamics (CFD) simulations require modelingin 3 dimensions, which generally necessitates high performance computing—a resource that isnot commonly available, and more expensive.

To date, attempts to validate SNL’s mid-fidelity CACTUS vortex line model [5] have reliedon measurements from the Saint Anthony Falls Laboratory (SAFL) [6] and the University ofNew Hampshire (UNH) [7, 8]. For the SAFL experiments (RM2), the chord Reynolds number,Rec ∼ 104, was below the threshold needed to properly simulate lift and stall characteristics. Forthe UNH experiment [9], the chord Reynolds number was sufficiently high at Rec ≈ 2.7× 105,but the chord/radius ratio and solidity (13.4%) created instability in the free-wake evolution inthe model, which caused significant overestimation of power coefficient.

The present task is to acquire a new dataset for the lower solidity RM2 turbine, but at higherReynolds numbers than those achieved in the experiments at SAFL. It is also suspected that thestrut drag model in CACTUS could be the source of some discrepancy. The strut drag will bedeliberately modified in the physical model to provide data to help answer that question. Thisdataset will be publicly available for both validation of CACTUS and other numerical models.This report details the experimental test plan for acquiring, processing, and archiving this data,which includes development of a scaled physical model, and descriptions of the experimen-tal setup and procedure to be performed in a towing tank at the University of New Hampshire(UNH).

4

1.1. Study goals and objectivesThe overarching goal of this project is to collect a Reynolds number independent performancedataset for a 1:6 scale RM2 turbine, which can be used to investigate the accuracy of CACTUSand other numerical models. This project will provide insight on the physics of hydrokineticcross-flow turbines, including the importance of blade strut drag on turbine power output, whichwill inform investigations on the present limitations of the CACTUS model and potential mod-ifications to improve its accuracy. The strut drag investigation will also give turbine designerssome perspective on the need to mitigate parasitic strut drag effects to improve rotor efficiency.The experimental dataset will be shared publicly and can be used for validating other numericalmodels/codes used by developers and DOE partners. The study goals can be distilled into thefollowing main objectives:

• Design and fabricate a 1:6 scale model of the RM2 turbine.• Perform measurements with the RM2 1:6 scale model in the UNH-CORE tow tank turbine

test bed to:

– Document the dependence of the turbine power coefficient on Reynolds number, andthe conditions under which it becomes independent of Reynolds number.

– Observe the influence of strut drag on turbine performance.– Characterize the near-wake at maximum power coefficient at the Reynolds number

independent state.

• Write a technical report to document the study objectives, methods, results, and conclu-sions.

• Analyze and archive measurements as an open model validation dataset.

5

2. Experimental setup and methodsTurbine performance is characterized by the nondimensional power coefficient

CP =P

12ρAU3

∞

, (2.1)

where P is the mechanical power output (the product of the shaft torque T and angular velocityω), ρ is the fluid density, A is the turbine’s frontal area, and U∞ is the free stream velocity. Alsoof interest is the overall drag (a.k.a. thrust) coefficient the device imparts on the flow in whichits placed, defined as

CD =FD

12ρAU2

∞

, (2.2)

where FD is the streamwise component of the force on the rotor.A performance curve, which is a plot of CP versus nondimensional rotation rate—tip speed

ratio λ—describes the device’s behavior over its range of operation. An example of is shownin Figure 2.1. In this experiment we will create similar curves by measuring turbine powercoefficient over a range of prescribed λ , where the tow speed, or turbine diameter Reynoldsnumber ReD is held constant. By creating curves for various ReD, we seek a condition where thepower coefficient at the optimal tip speed ratio begins to converge asymptotically to some limit,similar to the approach used in [4], the data from which is plotted in Figure 2.2.

2.1. Facility and instrumentationExperiments will be performed in the UNH tow/wave tank, a 36 m long facility with a 3.66 mwide by 2.44 m deep cross-section, capable of tow speeds up to 3 m/s1, pictured in Figure 2.3.The turbine will be mounted in a frame built from NACA 0020 struts, attached to the tow carriageby four linear bearings, which transfer all streamwise force to a pair of S-beam load cells. Theturbine shaft RPM will be controlled by a servo motor system, which allows prescription of theturbine tip speed ratio. The load torque will be measured by an inline rotary torque transducerand a load cell mounted at a fixed distance from the servo motor, providing a redundant mea-surement. Turbine shaft angle will be measured using the servo drive’s emulated encoder output,set to 105 counts per turbine shaft revolution. Carriage speed, and therefore inflow velocity willbe measured using a linear encoder with 10 µm resolution. All of these performance-related

1Note that though 3 m/s is the technical limit, the practical limit for achieving substantial tow durations is approx-imately 2 m/s.

6

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

λ

−0.05

0.00

0.05

0.10

0.15

0.20

0.25

0.30

CP

Figure 2.1.: Example of a turbine performance curve.

quantities will be sampled as 2 kHz, while the tow tank’s motion controller will provide redun-dant measurements of the carriage speed and turbine angular velocity sampled at 1 kHz. Turbinewake measurements at 1 turbine diameter downstream will be measured with a Nortek Vectrino+acoustic Doppler velocimeter, sampling at 200 Hz. A list of the sensors to be used in the ex-periment is shown in Table 2.1, instrumentation in Table 2.2, and a drawing of the experimentalsetup is shown in Figure 2.4.

Measured quantity Device type Mfg. & model Nominal accuracyCarriage position Linear encoder Renishaw LM15 10 µm/pulse [10]

Turbine angle Servo encoder output Kollmorgen AKD 105 pulse/rev [11]Turbine torque Rotary transducer Interface T8-200 ±0.5 Nm [12]

Turbine torque (2) Load cell (& arm) Sentran ZB3-200 ±0.2 Nm [13]Drag force, left Load cell Sentran ZB3-500 ±0.6 N [13]

Drag force, right Load cell Sentran ZB3-500 ±0.6 N [13]Fluid velocity ADV Nortek Vectrino+ ±0.5% ±1 mm/s [14]

Table 2.1.: Details of the sensors to be used for the experiment. Note that “(2)” denotes a sec-ondary redundant measurement. “Turbine torque (2)” nominal accuracy estimatedby combining load cell accuracy and arm machining tolerances (±1× 10−4 m) asroot-sum-square.

7

0.2 0.4 0.6 0.8 1.0 1.2 1.4ReD ×106

0.65

0.70

0.75

0.80

0.85

0.90

0.95

1.00

1.05

CP/C

P0

×105

0.5 1.1 1.6 2.1 2.7 3.2 3.7Rec,ave

Figure 2.2.: Normalized turbine power coefficient plotted versus Reynolds number, adapted from[4].

Figure 2.3.: Photos of the UNH towing tank and turbine test bed.

2.1.1. CalibrationsBefore collecting data, traceable calibration certificates will be obtained for the Interface T8-200torque transducer and NI 9405 and NI 9237 modules. The torque transducer calibration will beused directly in data processing, while the drag measurement load cells will be calibrated usingan additional Sentran ZB S-beam load cell and indicator, a package which will also have its owncalibration certificate. This load cell will be mounted to a fixture that allows varying load on eachdrag load cell—while the linear bearings are installed—by a lead screw. The redundant torquemeasurement load cell/arm system will be calibrated in a similar fashion using the Sentran loadcell and indicator, by attaching it to a fixture with a distance from the axis of rotation known to

8

Measured quantity Device type Mfg. & modelCarriage position Differential counter NI 9411

Carriage velocity (2) Motion controller ACS NTMTurbine angle Differential counter NI 9411

Turbine RPM (2) Motion controller ACS NTMTurbine torque Analog voltage input NI 9405

Turbine torque (2) Analog bridge input NI 9237Drag force, left Analog bridge input NI 9237

Drag force, right Analog bridge input NI 9237

Table 2.2.: Details of the instrumentation to be used for the experiment. Note that “(2)” denotesa secondary redundant measurement.

within approximately 0.005 inches.

2.1.2. Synchronization of instrumentation subsystemsThe three data acquisition instrumentation subsystems—motion controller, NI DAQ (perfor-mance measurements), and Vectrino+ (wake velocity measurements)—will begin sampling atprecisely the same time each run, after being triggered by a TTL pulse created by the motioncontroller. This strategy retains synchronization for all performance signal samples (tow speed,torque, drag, angular velocity), ensuring precise calculation of, e.g., power coefficient. Sincethere is also synchronization of the initial sample from each three subsystems, correlation ofevents in the performance and wake signals is also possible.

2.1.3. Tare drag and torque compensationTare torque and drag runs will also be performed to measure the shaft bearing friction torque andturbine mounting frame drag, respectively. These data will be similar to the turbine performancedata, omitting torque measurements for the tare drag runs and vice versa. Tare drag runs willbe performed for each tow speed in the experiment, for which the mean value is used in dataprocessing. Tare torque runs will be performed by rotating the turbine shaft (without blades)in air at constant angular velocity for a specified duration, over the range of angular velocitiesused throughout the experiment. Tare torque will then be fit with a linear regression versus shaftangular velocity, and added to the measured turbine torque in post-processing.

2.2. Turbine modelThe turbine is to be a 1:6 scale model of the RM2 rotor. Turbine geometry is to be scaled from theRM2 “rev 0” design report [15], with the exception of the shaft diameter, which will be a scaledversion of the SAFL RM2 shaft [6]. The hub design is also similar to the SAFL model, which

9

A

A

3.66 m

1.24 m

2.44 m

Vectrino probe positioning system

Turbine

Torque transducer

2X load cells

Servo motor

Hydrofoil frame

Guy wires

Tow carriage

Figure 2.4.: Illustration of the experimental setup.

may aid in comparison of the results, though this is not a top priority. Geometric parametersare shown in Table 2.3 and a drawing of the turbine design is shown in Figure 2.5. The turbinemodel components—blades, struts, shaft, and center hub sections—will be fabricated from 6061-T6 aluminum, which will be hardcoat anodized per MIL-8625-A, type III, class 2 specifications.

2.3. Test parametersData collection runs will take place during individual tows, for which all independent variables—tow speed, tip speed ratio, velocity probe position—are held constant. These runs can then begrouped into logical test matrix “sections,” in which typically a single independent variable isvaried. For example, a test matrix section will be dedicated to collecting turbine performancecurves for a single tow speed, or turbine diameter Reynolds number, where only tip speed ratio isvaried. Note that the upper limit of tip speed ratio will be determined in “shakedown” runs, andchosen to be just above the value at which turbine power coefficient reaches zero. Additionalperformance curve test matrix sections will be created/executed for different tow speeds. For

10

Full-scale Model (1:6)Diameter (m) 6.450 1.075Height (m) 4.840 0.8067Blade root chord (m) 0.4000 0.06667Blade tip chord (m) 0.2400 0.04000Blade profile NACA 0021 NACA 0021Blade mount 1/2 chord 1/2 chordBlade pitch (deg.) 0.0 0.0Strut profile NACA 0021 NACA 0021Strut chord (m) 0.3600 0.06000Shaft diameter (m) 0.2540 [16] or 0.4160 [6] 0.06350

Table 2.3.: RM2 turbine geometric parameters.

wake characterization, test matrix sections will vary the cross-stream position of the velocityprobe, providing a single profile transverse, and multiple profiles will be taken at varying verticallocations. Examples of test matrices for a performance curve and a wake profile are shown inTable A.1.

Parameters for each test matrix section will be created in the Config/Test plan folder insidethe experiment working directory, saved in comma separated value (CSV) format as input to theTurbineDAQ2 experiment automation software, which was developed specifically for vertical-axis turbine measurements in the UNH tow tank (see Figure B.1 for a screenshot). These CSVsalso provide information about each run to be used in post-processing. Note that the softwareis designed to read turbine properties (from Config/turbine_properties.json), e.g. radius andheight, so inputs for turbine rotational speed (tip speed ratio) and Vectrino probe location are innondimensional form.

2.4. Determining tank settling timeSample tows will be done to determine the amount of time taken between runs such that the tankhas settled adequately, i.e., background turbulence and any large scale mean flows have beendissipated. This will be assessed by towing the turbine, then allowing the Vectrino to continuerecording velocity data, monitoring the mean and standard deviation of the signals. The settlingtimes will be stored in the experiment configuration—one value for each tow speed.

2.5. Determining Reynolds number independenceIn order to establish that we have reached a Reynolds number independent regime for turbineperformance, we will acquire multiple performance curves, starting at a relatively slow tow speed

2https://github.com/petebachant/TurbineDAQ

11

0.807 m

1.075 m

Figure 2.5.: Illustration of the UNH RM2 scaled physical model.

(on the order of 0.2 m/s), incrementing up to the maximum practical tow speed. For the scaleof the RM2 model, we expect to see convergence in the measured power coefficient, similarto Figure 2.2, around a tow speed of 1 m/s, which corresponds to a turbine diameter Reynoldsnumber ReD = 1.1× 106 and an approximate blade chord Reynolds number Rec = 2.0× 105

at λ = 3. This will be determined by post-processing the data concurrently. See Table 2.4 forReynolds numbers corresponding to specific tow speeds.

2.6. Measuring the effects of strut dragIn order to measure the effects of strut drag, we will acquire an additional performance curveat the Reynolds number independent tow speed U0 for the device with cylindrical strut coversslid over the struts to increase their drag by several orders of magnitude. We will also measurethe rotor torque with the blades removed and strut covers installed, while rotating the turbineover the range of angular velocities seen in the performance curve, providing a comparison forthe drag coefficient of the cylinders in a rotating flow. An illustraction of the turbine with strutcovers installed—both with and without blades—is shown in Figure 2.6.

12

Tow speed (m/s) ReD Rec at λ = 30.2 2.2×105 4.0×104

0.4 4.3×105 8.0×104

0.6 6.5×105 1.2×105

0.8 8.6×105 1.6×105

1.0 1.1×106 2.0×105

1.2 1.3×106 2.4×105

1.4 1.5×106 2.8×105

Table 2.4.: Turbine diameter and approximate blade chord Reynolds numbers corresponding totow speeds.

EndcapStrut cover

Figure 2.6.: Illustration of strut cover design with (left) and without (right) blades installed.

2.7. Near-wake characterizationAfter Reynolds number independence has been found, we will perform a characterization ofthe near-wake at one turbine diameter downstream, acquiring a series of cross-stream profilesat varying height, mapping out the upper half of the wake, measured from the turbine center.The Vectrino will be in a fixed location (relative to the turbine center) for each run, necessitatingmultiple runs at varying cross-stream and vertical coordinates with the approximate ranges y/R=±2.8 and z/H = 0–0.75, respectively—chosen based on previous experiments with similarlysized turbines, e.g. [9], to visualize the complex three-dimensional flow field present in the near-

13

wake. The probe will be mounted to a NACA 0020 strut, attached to an automated positioningsystem, shown in Figure 2.7, which is driven by stepper motors with resolutions of approximately3×10−5 meters per step, and controlled by the tow tank’s main motion controller.

Figure 2.7.: Photograph of the Vectrino probe positioning system.

2.8. Data processingTurbine shaft angular velocity and carriage speed must be computed by taking the derivativesof the measured position time series. This will be done using a second order central differencescheme, and may be subsequently filtered with a moving average filter to help reduce noiseintroduced by the differentiation. To check that excess high frequency energy was not removedduring filtering, the velocity signals can be compared to their redundant measurements recordedfrom the motion controller.

After the velocity time series are calculated, they are used to calculate time series for instan-taneous λ , CP, and CD, from which mean values are computed.

Data collected for each run will include a transient period, where the tow carriage accelerates,after which turbine torque and wake velocities settle to become stationary or periodic. It is this“steady-state” duration that we are interested in. The duration will be identified for several runsat each tow speed by manual inspection. The tail end will then be truncated such that the data tobe processed includes an integer number of blade passages, to minimize bias from periodicity.The time interval will be recorded as part of the reduced data for each run, along with the numberof blade passages and revolutions.

Wake velocity data will be post-processed using methods outlined in Gunawan et al. [17].Corrections will include, for example, eliminating and/or replacing spikes [18], Doppler noise[19], and filtering [20].

14

A CSV table of derived data will be created for each section, with one row per run contain-ing the mean and standard deviation for the run’s measured tow speed, tip speed ratio, powercoefficient, drag coefficient, u, v, and w components of wake velocity, along with u′v′, u′w′, andv′w′ Reynolds stresses, turbulence kinetic energy k, and 95% confidence expanded uncertain-ties for λ , CP, and CD (see 2.8.1 for the uncertainty analysis procedure). These CSV files willbe stored in the Data/Processed directory, which will be included in the experiment’s versioncontrol repository, to track changes to derived data from, e.g., modifications to the processingcode.

2.8.1. Uncertainty analysisAs mentioned previously, one of the primary uses for the data collected is comparison withresults from numerical modeling, and this comparison requires an estimate for what range ofvalues from the experimental result includes the true value. This range—or uncertainty—resultsfrom a combination of systematic and random errors. The random error can be inferred from thesample standard deviation and the systematic from the sensor calibrations or datasheets. Com-bining both sources of error, along with their propagation into quantities derived from multiplemeasurements, will follow the procedures outlined in Coleman and Steele (2009) [21], describedbelow.

An expanded uncertainty interval with 95% confidence can be computed for λ , CP, and CD

U95 = t95uc, (2.3)

where t95 is the value from the Student-t distribution for a 95% confidence interval and uc isthe combined standard uncertainty. Combined standard uncertainty for a given quantity X iscalculated by

u2X = s2

X +b2X , (2.4)

where sX is the sample standard deviation, calculated as the standard deviation of the mean perturbine revolution, and bX is the systematic uncertainty, computed by

b2X =

J

∑i=1

(∂X∂xi

)2

b2xi, (2.5)

where xi is a primitive quantity used to calculate X (e.g. T , ω , and U∞ for calculating CP), and bxi

is the primitive quantity’s systematic uncertainty, estimated as half the value listed in Table 2.1.Selecting t95 requires an estimate for degrees of freedom νX , which can be obtained using the

Welch–Satterthwaite formula

νX =

(s2

X +∑Mk=1 b2

k

)2

s4X/νsX +∑

Mk=1 b4

k/νbk

, (2.6)

where νsX is the number of degrees of freedom associated with sX and νbk is the number ofdegrees of freedom associated with bk. νsX is assumed to be (N−1), where N is the number of

15

independent samples (or turbine revolutions). νbk will be estimated as

νbk =12

(∆bk

bk

)−2

, (2.7)

where the quantity in parentheses is the relative uncertainty of bk.From previous measurements of a similarly-sized vertical-axis turbine with the same experi-

mental setup, we expect expanded uncertainties for λ , CP, and CD to be approximately 0.008,0.01, and 0.02, respectively, at a tow speed U∞ = 1.0 m/s. These values are well within the dis-crepancies between previous measurements and predictions by CACTUS [8], making the datafrom the experiments described here suitable for model validation.

16

3. Research deliverablesThe research deliverables for this project include the following:

• A 1:6 scale RM2 vertical-axis turbine rotor model, 3-D CAD models of its geometry, and2-D manufacturing drawings detailing how it was fabricated and assembled.

• Raw turbine performance data for power coefficient curves acquired at multiple Reynoldsnumbers. The data are described in detail in Section 3.1 below.

• Raw turbine performance data showing the parasitic effects of high-drag support struts.• Three components of raw wake velocity data characterizing the upper half of the wake a

one turbine diameter downstream.• Tabulated processed data from each run and the software used to compute them.• A final report detailing the goals of the experiment, how it was conducted, descriptions of

the results, and conclusions reached.

3.1. Experimental dataEach data collection run (one tow) will produce raw data for the turbine torque, drag force onthe submerged equipment, turbine shaft angle, carriage position, and Vectrino velocity measure-ments, along with the relative times for the samples. Raw data will be stored in HDF5 for-mat, additional Vectrino raw data in Nortek’s binary format (*.vno), run metadata in text-basedJavaScript Object Notation (JSON), processed data as CSV, and processing code in Python.These formats were chosen for their platform- and language-independence, which will allowbroadest usage of the data by other researchers. A summary of data file types, names, and con-tents is presented in Table B.1.

3.1.1. Directory structure and naming conventionsA sample directory structure is shown in Figure B.2. In the Config/Test plan folder are theCSV files containing the tabulated test parameters for each test matrix section, whose names areinferred from the CSV file names. Raw data files are stored in the Data/Raw/{section}/{run}

subfolders, while processed data are stored in the Data/Processed subfolder—one CSV file pertest matrix section.

17

3.1.2. MetadataEach run’s metadata file (metadata.json) will include:

• Nominal tow speed• Prescribed tip speed ratio• Test matrix section and run number, e.g., “Perf-0.8 run 4”• Time created• Turbine name, radius, and height• TurbineDAQ software version• Vectrino location• Vectrino metadata (Note: Additional Vectrino metadata will be included in the *.vno files.):

– Coordinate system– Sample rate– Velocity range index

• NI DAQ device channel metadata:

– Sample rate– Scale names, slopes, y-intercepts, and units

3.2. Management and archivingThis experiment will likely produce on the order of 10 GB of data. The two guiding principlesfor the management of this data are openness and usefulness, i.e., we want potential users toknow exactly how the data was created, and be able to reuse the data as conveniently as possible.We will make available all raw and processed data, along with any software written for theprocessing, analysis, and visualization of this data. It is also important that we provide the dataand code such that they are maintainable, both by ourselves and by other users.

Since the dataset is of moderate size, we will split the data into two parts—one containingthe processed data, and one the raw data. This will allow users to first obtain the processeddata, processing code, and documentation—most relevant for generating figures or comparing tonumerical modeling results—without having to download an excessively large archive. However,if users want to reanalyze the raw data, the processing code will be written such that raw datafiles are downloaded as needed, rather than one bulk download, though it will still be possibleto download the entire raw dataset, if desired. This strategy is convenient, for example, for aresearcher looking to compute wake spectra at a single location, since he or she would only needto download ∼ 10 MB of data, rather than the entire ∼ 10 GB.

Raw data files will be uploaded to figshare1, which will then give the dataset a Digital Object1http://figshare.com

18

Identifier (DOI) and therefore a persistent URL. Processed data, processing code, and documen-tation will be stored in a Git2 version control repository, hosted on GitHub3 to help facilitateupdates to the analysis by both the investigators and third parties. Tagged versions or “releases”of the Git repository will be uploaded to figshare such that they will have their own DOIs, andcan be cited appropriately in publications.

3.3. LicensingThe research products described above—documentation, data, code, CAD models, etc.—willbe freely available under a Creative Commons Attribution (CC-BY) license, which allows freeaccess, further sharing, and adaptation so long as the original source is credited.

2http://git-scm.com3https://github.com/UNH-CORE/RM2-tow-tank

19

4. SummaryIn this study we will set out to acquire a Reynolds number independent dataset for the RM2 cross-flow turbine, producing an extensive collection of raw data, along with open repository of pro-cessed data, code, and documentation to support numerical model validation efforts. This objec-tive requires that we design and build a turbine model at a scale that is expected to reach Reynoldsnumber independence within the UNH tow tank’s tow speed range. We will then acquire fulldevice performance curves (power and drag coefficients) at multiple tow speeds—or turbine di-ameter Reynolds numbers—to attempt to find convergence. Once convergence is found, we willacquire measurements to characterize the three-dimensional nature of the near-wake (one turbinediameter downstream) at a threshold Reynolds number Re0 using acoustic Doppler velocimetry.Lastly, we will observe the effects of strut drag on turbine model performance—measuring theparasitic torque from support struts by rotating the turbine in still water, both with and withoutcylindrical tubes slid over the struts to significantly increase their drag coefficients. We willthen acquire a performance curve at Re0 with the high-drag struts to provide a benchmark casefor assessing the importance of accurately capturing the effects of strut drag inside numericalmodels.

4.1. MilestonesThe progress of this project will be marked by the completion of the following:

1. November 7, 2014 – Experimental test plan draft and turbine model design reviewed.2. January 16, 2015 – Turbine model assembled.3. March 30, 2015 – Experimental data collected, processed, and basic report written.

20

5. AcknowledgementsThis study was supported by the Department of Energy (DOE), Office of Energy Efficiency andRenewable Energy (EERE), Wind and Water Power Technologies Office (WWPTO). Sandia Na-tional Laboratories is a multi-program laboratory managed and operated by Sandia Corporation,a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’sNational Nuclear Security Administration under contract DE-AC04-94AL85000.

21

References[1] V. S. Neary, M. Previsic, R. A. Jepsen, M. J. Lawson, Y.-H. Yu, A. E. Copping, A. A.

Fontaine, K. C. Hallett, and D. K. Murray, “Methodology for design and economic analysisof marine energy conversion (MEC) technologies,” Tech. Rep. SAND2014-9040, SandiaNational Laboratories, March 2014.

[2] I. Paraschivoiu, Wind Turbine Design with Emphasis on Darrieus Concept. Montreal, Que-bec, Canada: Polytechnic International, 1st ed., 2002.

[3] B. Blackwell, R. Sheldahl, and L. Feltz, “Wind tunnel performance data for the Darrieuswind turbine with NACA 0012 blades,” Report SAND76-0130, Sandia National Laborato-ries, Albuquerque, NM, May 1976.

[4] P. Bachant and M. Wosnik, “Reynolds number dependence of cross-flow turbine perfor-mance and near-wake characteristics,” in Proceedings of the 2nd Marine Energy Technol-ogy Symposium METS2014, (Seattle, WA), April 2014.

[5] J. Murray and M. Barone, “The development of cactus, a wind and marine turbine per-formance simulation code,” in Proceedings of the 49th AIAA Aerospace Sciences Meetingincluding the New Horizons Forum and Aerospace Exposition, 2011.

[6] C. Hill, V. Neary, B. Gunawan, M. Guala, and F. Sotiropoulos, “U.S. Department of En-ergy Reference Model Program RM2: Experimental results,” tech. rep., St. Anthony FallsLaboratory for Wind and Water Technologies Program, Office of Energy Efficiency andRenewable Energy, U.S. Department of Energy, June 2014.

[7] V. Neary, A. Fontaine, P. Bachant, M. Wosnik, C. Michelen, R. Meyer, B. Gunawan, andW. Straka, “US Department of Energy (DOE) national lab activities in marine hydrokinet-ics: Scaled model testing of DOE reference turbines,” in Proceedings of European Waveand Tidal Energy Conference EWTEC, 2013.

[8] C. Michelen, V. S. Neary, J. C. Murray, and M. Barone, “CACTUS open source code forhydrokinetic turbine design and analysis: model performance evaluation and public dissem-ination as open source design tool,” in Proceedings of the 2nd Marine Energy TechnologySymposium, 2014.

[9] P. Bachant and M. Wosnik, “Performance and near-wake measurements for a vertical axisturbine at moderate Reynolds number,” in Proceedings of the ASME Fluids EngineeringDivision Summer Meeting, no. FEDSM2013-16575, (Incline Village, NV), July 2013.

22

[10] Renishaw, “LM15 linear encoder datasheet.” http://resources.renishaw.com/download.

aspx?lang=en&data=49238&btn=1, Accessed December 2014.

[11] Kolmorgen, “AKD user guide.” http://www.kollmorgen.com/en-us/products/drives/

servo/akd/_manuals/kollmorgen-akd-user-guide-en-rev-m/, Accessed December 2014.

[12] Interface Inc., “T8 general purpose rotary torque transducer specifications.” http:

//www.interfaceforce.com/index.php?T8-General-Purpose-Rotary-Torque-Transducer&

mod=product&show=65, Accessed December 2014.

[13] Sentran, “ZB S beam load cell datasheet.” http://www.sentranllc.com/pdfs/zb1.pdf, Ac-cessed December 2014.

[14] Nortek-AS, “Vectrino 3D water velocity sensor lab probe datasheet.” http://www.

nortekusa.com/lib/data-sheets/datasheet-vectrino-lab, Accessed December 2014.

[15] M. Barone, T. Griffith, and J. Berg, “Reference model 2: “rev 0” rotor design,” Tech. Rep.SAND2011-9306, Sandia National Laboratories, November 2011.

[16] M. J. Beam, B. R. Elbing, T. K. Fetterolf, B. L. Kline, D. F. Kerstetter, J. A. Mickey, A. A.Fontaine, and W. A. Straka, “Power take-off system design SNL reference turbine 2,” tech.rep., Penn State Applied Research Lab, 2011.

[17] B. Gunawan, V. S. Neary, and J. R. McNutt, “ORNL ADV post-processing guide andMATLAB algorithms for MHK site flow and turbulence analysis,” Tech. Rep. ORNL/TM-2011/338, Oak Ridge National Laboratory, September 2011.

[18] D. G. Goring and V. I. Nikora, “Despiking acoustic Doppler velocimeter data,” Journal ofHydraulic Engineering, 2002.

[19] G. Voulgaris and J. H. Trowbridge, “Evaluation of the acoustic Doppler velocimeter (ADV)for turbulence measurements,” Journal of Atmospheric and Oceanic Technology, vol. 15,1998.

[20] C. Garcia, M. I. Cantero, Y. Nino, and M. H. Garcia, “Turbulence measurements withacoustic Doppler velocimeters,” Journal of Hydraulic Engineering, vol. 131, no. 12,pp. 1062–1073, 2005.

[21] H. W. Coleman and W. G. Steele, Experimentation, Validation, and Uncertainty Analysisfor Engineers. John Wiley & Sons„ third ed., 2009.

23

A. Sample test plan matrices

Run U∞ λ y/R z/H0 1.0 0.2 0.0 0.01 1.0 0.4 0.0 0.02 1.0 0.6 0.0 0.03 1.0 0.8 0.0 0.04 1.0 1.0 0.0 0.05 1.0 1.2 0.0 0.06 1.0 1.4 0.0 0.07 1.0 1.6 0.0 0.08 1.0 1.8 0.0 0.09 1.0 2.0 0.0 0.0

10 1.0 2.2 0.0 0.011 1.0 2.4 0.0 0.012 1.0 2.6 0.0 0.013 1.0 2.8 0.0 0.014 1.0 3.0 0.0 0.015 1.0 3.2 0.0 0.016 1.0 3.4 0.0 0.017 1.0 3.6 0.0 0.018 1.0 3.8 0.0 0.019 1.0 4.0 0.0 0.0

Run U∞ λ y/R z/H0 U0 λ0 -2.75 01 U0 λ0 -2.5 02 U0 λ0 -2.25 03 U0 λ0 -2 04 U0 λ0 -1.8 05 U0 λ0 -1.6 06 U0 λ0 -1.5 07 U0 λ0 -1.4 08 U0 λ0 -1.3 09 U0 λ0 -1.2 010 U0 λ0 -1.1 011 U0 λ0 -1 012 U0 λ0 -0.9 013 U0 λ0 -0.8 014 U0 λ0 -0.7 015 U0 λ0 -0.6 016 U0 λ0 -0.5 017 U0 λ0 -0.4 018 U0 λ0 -0.3 010 U0 λ0 -0.2 020 U0 λ0 -0.1 021 U0 λ0 0 022 U0 λ0 0.1 023 U0 λ0 0.2 024 U0 λ0 0.3 025 U0 λ0 0.4 026 U0 λ0 0.5 027 U0 λ0 0.6 028 U0 λ0 0.7 029 U0 λ0 0.8 030 U0 λ0 0.9 031 U0 λ0 1 032 U0 λ0 1.1 033 U0 λ0 1.2 034 U0 λ0 1.3 035 U0 λ0 1.4 036 U0 λ0 1.5 037 U0 λ0 1.6 038 U0 λ0 1.8 039 U0 λ0 2 040 U0 λ0 2.25 041 U0 λ0 2.5 042 U0 λ0 2.75 0

Table A.1.: Sample test matrices for acquiring a performance curve (left) and a cross-streamwake profile (right). U0 is a constant a tow speed at which performance has reachedReynolds number independence, and λ0 is the tip speed ratio at which turbine powercoefficient is maximum.

24

B. Data acquisition software, file contentand organization

Figure B.1.: Screenshot from the TurbineDAQ experiment automation software, developed forvertical-axis turbine measurements in the UNH tow tank.

25

Data Sample rate File type File nametime

carriage_pos

turbine_angle

torque_trans

torque_arm

drag_left

drag_right

2 kHz HDF5 nidata.h5

time

carriage_vel

turbine_rpm

1 kHz HDF5 acsdata.h5

time

u

v

w

corr_u

corr_v

corr_w

snr_u

snr_v

snr_w

200 Hz HDF5 vecdata.h5

vecdata.dat

vecdata.hdr

vecdata.pck

200 Hz Vectrino binary vecdata.vno

Table B.1.: Raw data file description.

26

RM2-tow-tank/

Config/

Test plan/

perf_0.8.csv

wake_0.25.csv

tare_drag.csv

tare_torque.csv

settling.csv

perf_strut_drag.csv

stat_strut_torque.csv

turbine_properties.json

settling_times.csv

raw_data_urls.json

Data/

Processed/

perf_0.8.csv

Raw/

perf_0.8/

0/

metadata.json

acsdata.h5

nidata.h5

vecdata.h5

vecdata.vno

1/

metadata.json

acsdata.h5

nidata.h5

vecdata.h5

vecdata.vno

Documents/

Test plan/

Final report/

Scripts/

README.md

Figure B.2.: Sample directory structure for experiment configuration, data, and documentation.

27