Traceable Torque Measurement under Rotation in Nacelle Test Benches : A Good Practice Guide

Traceable Torque Measurement under Rotation in Nacelle Test Benches – A Good Practice Guide – Paula Weidinger Gisa Foyer Physikalisch-Technische Bundesanstalt (PTB), 2019 2 DOI: https://doi.org/10.7795/530.20190111 This guide has been produced within the EURAMET project entitled Torque Measurement in the MN m 1 range. More information about this collaborative research project can be found on the project’s website https://www.ptb.de/emrp/ind14-home.html. The aim of this guide is to provide practical information and advice about torque measurement and especially about torque calibration in nacelle test benches for nacelle test bench operators. Disclaimer Any mention of commercial products within this guide is for information only; it does not imply recommendation or endorsement by the partners in this project. The views expressed in this guide are those of the authors and of the EMPIR 14IND14 project team. Acknowledgement of funding The production of this guide was funded by the European Metrology Programme for Innovation and Research (EMPIR). The EMPIR initiative is co-funded by the European Union’s Horizon 2020 research and innovation programme and the EMPIR Participating States. Authorship Preparation of this guide was led by Paula Weidinger and Gisa Foyer of the Physikalisch-Technische Bundesanstalt (PTB), Braunschweig (Germany) with extensive input from all members of the EMPIR 14IND14 project team. The discussion and input of all the partners in the project and their colleagues are greatly appreciated. Suggestion for the quotation of the references Weidinger, Paula and Foyer, Gisa, 2019. Traceable Torque Measurement under Rotation in Nacelle Test Benches: A Good Practice Guide. Physikalisch-Technische Bundesanstalt (PTB). DOI: https://doi.org/10.7795/530.20190111 This document and all parts contained therein are protected by copyright and are subject to the Creative Commons user license CC by 4.0 (https://creativecommons.org/licenses/by/4.0/). Project logo All copyright and related or neighbouring rights to this logo are waived using the CC0 Public Domain Dedication (https://creativecommons.org/publicdomain/zero/1.0/). DOI: https://doi.org/10.7795/920.20190111 1 The unit for torque can also be written as MN·m. 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 3 Contents 1 Introduction 4 1.1 Glossary 4 1.2 Symbols and their meaning 4 2 Calibration set-up 6 2.1 Place to measure the reference torque 6 2.2 Torque transfer standard 6 2.3 Data acquisition 7 2.3.1 Amplifier and filter settings 7 2.3.2 Timewise synchronisation of different data sets 7 2.4 Example of a set-up 8 3 Performing a torque calibration 11 3.1 Preparatory operations and tests 11 3.2 Ambient and boundary conditions 12 3.3 Signal evaluation 13 3.4 Determination of the zero signal 13 3.4.1 Static zero-point determination 14 3.4.2 Rotational zero-point determination 14 3.5 Quasi-static torque calibration 15 3.6 Characterisation maps 15 3.7 Crosstalk effects 17 3.8 Calibration interval 18 4 Evaluation of the calibration result including a measurement uncertainty 19 4.1 Evaluation of the transducer to be calibrated 19 4.1.1 Relative indication deviation 19 4.1.2 Relative repeatability 20 4.1.3 Relative resolution 20 4.1.4 Relative reversibility 21 4.2 Measurement uncertainty 21 4.2.1 Measurement conditions 21 4.2.2 Uncertainty contribution of the resolution 22 4.2.3 Uncertainty contribution of the repeatability 22 4.2.4 Uncertainty contribution of the transfer standard 23 4.2.5 Expected value for the indication deviation 23 4.2.6 Example of an uncertainty calculation 23 5 Calibration output and benefit for the test bench operator 24 Glossary 25 6 References 26 4 DOI: https://doi.org/10.7795/530.20190111 1 Introduction Prior to the market launch of wind turbines (also called nacelles), large gearboxes and generators, extensive tests, which can be performed on a nacelle test bench, are of great importance. One such test is the efficiency determination of the device under test. For direct efficiency determination, the mechanical input directly at the flange of the device under test and the electrical or mechanical output of the device under test are to be measured. While the output can already be measured with a sufficient uncertainty, measuring the input consisting of torque and rotational speed poses a problem. A broad variety of measuring instruments is available for measuring rotational speeds. The torque that appears in nacelle test benches, however, ranges up to several MN m and cannot easily be determined due to the lack of traceability possibilities. This good practice guide introduces a method for tracing large torque measurement up to 5 MN m in nacelle test benches using a torque transfer standard. This good practice guide comprises the following sections. A list of the general requirements on the performance of a torque calibration in nacelle test benches including the calibration set-up is given in section 2. A description of the preparatory work, tests and instructions on a quasi-static torque calibration and a torque calibration using so-called characterisation maps to cover the entire operation range of a nacelle test bench as part of the performance of a torque calibration are outlined in section 3. Guidelines for the evaluation of the calibration results including the determination of a measurement uncertainty are introduced in section 4. Section 5 again states the benefits of traceable torque measurement in nacelle test benches for the operators of the test benches and the feasibility of the calibration results. Further information about understanding measurement and measurement uncertainty can be found in the Guide to the expression of uncertainty in measurement (GUM) [7] and the International Vocabulary of Metrology (VIM) [5]. General information about operating procedures in calibration laboratories is given in DIN 17025 [2], while information about static torque calibration can be taken from EURAMET cg-14 [6], one of the European calibration guidelines. 1.1 Glossary We have provided a glossary of technical terms that arise in the following description of a torque calibration in nacelle test benches at the end of this document. 1.2 Symbols and their meaning Important symbols and their meaning regarding torque calibration in nacelle test benches are listed in Table 1. Table 1 Symbols and their meaning. Symbol Unit Meaning a % Relative resolution of the torque measuring instrument in the test bench aF % Relative resolution of the torque measuring instrument in the test bench under load aZ % Relative resolution of the torque measuring instrument in the test bench after load release b % Relative repeatability of the torque measuring instrument in the test bench fsample Hz Sampling frequency of the DAQ HNTB/TTS % rH Relative humidity measured close to the torque measurement instrument in the test bench (NTB) and close to the torque transfer standard (TTS) k - Amplification factor to calculate the expanded uncertainty based on the combined uncertainty l - Integer number of revolutions that are averaged over M kN m Increasing torque load indicated by the torque transfer standard M' kN m Decreasing torque load indicated by the torque transfer standard Mi kN m Increasing torque load indicated by the torque measuring instrument in the test bench 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 5 Mi ' kN m Decreasing torque load indicated by the torque load in the test bench Mi    ,M  kN m Arithmetic mean of several measurements of Mi and M at the same load step Mi,0 kN m Residue indication of the torque measurement instrument in the test bench after load release ML kN m Torque load step nmax min-1 Maximum rotational speed of the test bench considering the device under test nNTB min-1 Rotational speed measured in the test bench n NTB min-1 Arithmetic mean of the measurement of rotational speed q  % Relative indication deviation of the torque measuring instrument in the test bench averaged over several repetitions q j % Relative indication deviation of the torque measuring instrument in the test bench per load cycle qmax % Maximum value of q at each load step qmin % Minimum value of q at each load step r kN m Resolution of the torque measuring instrument in the test bench rrot ° Envisioned signal resolution per rotation S mV/V Measurement signal Szero.rot mV/V Rotational torque zero signal Szero,stat mV/V Static torque zero signal tdwell s Dwell time before measurement theat s Time to heat up all components until stable conditions are reached tmeas s Measurement time depending on the minimum rotational speed and the number of revolutions that are averaged over tramp s Time to ramp up or down torque or rotational speed tstep s Total measurement time per torque step uc % Combined absolute measurement uncertainty ui % Uncertainty component urep % Uncertainty component due to the repeatability ures % Uncertainty component due to the resolution ustd % Uncertainty component due to the deployed torque transfer standard U % Expanded absolute measurement uncertainty v % Relative reversibility of the torque transducer in the test bench ϑNTB/TTS °C Temperature measured close to the torque measurement instrument in the test bench (NTB) and close to the torque transfer standard (TTS) 6 DOI: https://doi.org/10.7795/530.20190111 2 Calibration set-up Due to several constraints on nacelle test benches, the operators use different methods in different places to measure the input torque to the device under test. An example of a nacelle test bench including the possible positions for measuring torque is depicted in Figure 1. In nacelles, the main axis is the x-axis. Consequently, the main torque measured in nacelle test benches is Mx. Although the main axis of torque measurement in metrology is the z-axis, in the following description and instructions, the main axis is the x-axis and the main torque to be measured is Mx since this good practice guide is directed at test bench operators. Figure 1 Examples of torque measurement options in test benches and the coordinate system of a nacelle and, therefore, the test bench [9]. 2.1 Place to measure the reference torque To determine the efficiency of a device under test using the direct measurement method by determining the input and output of the device under test, the torque should ideally be measured directly at the connection flange between the test bench and the device under test. In most test benches, specially designed adapters are required to install this additional torque transducer to measure the reference torque. A calibration of the test bench is not possible without a device under test, because only then is a torque generation possible. For a calibration of the typical testing range (torque and rotational speed) of the test bench, a suitable test object representing frequently tested objects is to be installed on the test bench. Ideally, this test object provides full access to its control system to allow the initiation of several control scenarios dependent on the controlled change of electrical braking power asserted by the generator. All torque transducers, the reference torque transducer as well as the torque transducer in the test bench, must be aligned very precisely to minimise parasitic permanent loads such as longitudinal and lateral forces (Fx, Fy and Fz) and bending moments (My and Mz) on the torque transducer. Possible misalignments are caused by a lack of planarity on the adapters, a lack of concentricity of the entire drive train, and a wrong distance between the load application system and the device under test. 2.2 Torque transfer standard For measuring the reference torque, to which the torque measurement in a test bench is then compared, a so-called torque transfer standard is needed. A transfer standard is a measurement gauge that bears a defined relation between a physical quantity, which in this case is torque, and a unit of measurement, here kN m. This torque transfer standard is to be calibrated according to EURAMET cg-14 [6], a European calibration guide or DIN 51309 [3], the German torque standard. A torque transfer standard has to meet several requirements. First, the transfer standard must have a sufficient measurement range to cover the operational torque range of the test bench, and it is to be calibrated over this measurement range. The best physical principle working in the MN m torque range is the detection of strains by means of alternating electrical resistance using strain gauges. The strain gauges are glued to a deformation body made of steel. To gain the required stiffness of the transfer Prime mover Load application system Main gearbox Generator (load machine) Device under test Torque based on electrical power and rotational speed Torque based on electrical power and rotational speed Torque on low speed shaft Torque on low speed shaft with additional loads Torque on high speed shaft z y x 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 7 standard’s deformation body and a sufficient sensitivity at the same time, a hollow shaft-type deformation body is best (project result taken from [10]). Moreover, the transfer standard must withstand the occasionally occurring additional mechanical loads created by the load application system as it is installed directly at the hub flange of the device under test and is, therefore, exposed to all simulated wind loads caused by the load application system. Even when no explicit additional loads are applied to the drive train, the load application system deploys loads to stabilise the drive train, leading to power losses and additional friction torque along the drive train during rotation. For information about these additional loads and for surveillance reasons, the transfer standard should be equipped with additional measuring bridges to sense bending moments, and longitudinal and lateral forces. Additional knowledge about the behaviour of the transfer standard under alternating temperature (and humidity) is advantageous, since the ambient conditions in a test bench do not comply with the laboratory conditions during the calibration of the transfer standard. In general, the transfer standard has to meet special requirements regarding its dimensions and weight. These requirements are not only limited by the test benches, but also by the available torque calibration machine that is used to characterise the torque transducer and lead to it being a transfer standard. In order to induce the torque load up to several MN m correctly and to ensure the easy mounting of the transfer standard, a flange with bore holes for bolts is an appropriate type of connection. As mentioned above, this might call for specially designed adapters to install the transfer standard in the test bench. 2.3 Data acquisition A timewise synchronised data recording of the torque measured by the transducer of the test bench and by the transfer standard can be realised either by one shared or by two synchronised data acquisition systems. Regardless of the data acquisition system used, all recorded data should be timestamped, and both the start and end time of the measurement are to be documented. Other than for static calibrations, the signals are to be recorded continuously to enable an investigation of the angular acceleration and its impact on the torque signals. The quantities to be measured during a calibration are listed in Table 2. Table 2 Quantities to be measured for a torque calibration in test benches. Quantity to be measured Symbol Reading Unit Torque measured by the transducer in the test bench Mi mV/V kN m Rotational speed measured by instruments in the test bench or the TTS nNTB V min-1 Torque measured by the transfer standard M mV/V kN m Temperature measurement close to the test bench transducer and the transfer standard (temp. meas. inside the transfer standard) ϑNTB, ϑTTS V °C Humidity measurement close to the test bench transducer and the transfer standard (humid. meas. inside the transfer standard) HNTB, HTTS V % rH Temperature ϑ and humidity H are to be measured as close to both torque transducers as possible. Ideally, temperature and humidity are measured inside the transfer standard, where the strain gauges are located, and not only certain temperature points but also the temperature gradients over the transducers and the adjacent components are to be recorded. All transducer measurement signals should be corrected for changes in the environmental conditions. 2.3.1 Amplifier and filter settings All signals should be recorded using an adequate sampling frequency taking the rotational speed nNTB and the envisioned resolution rrot of both torque signals and the rotational speed into account. The resolution aimed at is rrot = 1° for the maximum rotational speed nmax . The lowest limit of the sample frequency fsample can be calculated accordingly: fsample ≥ nmax /rrot ⋅ 360° , (1) where the sampling frequency fsample is in Hz and the rotational speed nmax is in s-1. Furthermore, a low pass filter should be implemented to avoid aliasing effects. 2.3.2 Timewise synchronisation of different data sets To synchronise all components of one data acquisition system or two different data acquisition systems including all their components, either a collective time server, where all associated components use the same network time protocol or a simple synchronisation signal, which is recorded and timestamped by the different components and data acquisition systems, can be deployed. Concerning the accuracy, both methods are sufficient. 8 DOI: https://doi.org/10.7795/530.20190111 When assigning the square-wave method to synchronise the data sets, the square-wave signal is used to shift the data in post-processing and to erase the temporal shift between the two data sets. The square-wave can have, e.g., an amplitude of u sync=±5 V and a frequency of fsquare = 0.2 Hz. While a rough alignment of the data sets is achieved by the logged start time and by distinctive signal changes as shown in Figure 2a), the fine synchronisation is based on the square-wave signal, which is depicted in Figure 2b). An advantage of the timewise synchronisation using a square-wave signal is its easy implementation in different test benches with all kinds of data acquisition systems. Figure 2 Timewise synchronisation of the data sets of the torque transfer standard and the nacelle test bench using an ideal square-wave signal, which is recorded by both data acquisition systems, for a precise alignment of the different data sets. 2.4 Example of a set-up In the example of a calibration set-up in Figure 3, the torque transfer standard that measures the reference torque is placed directly at the hub flange of the device under test as required. Moreover, an overview of the requirements on the transfer standard based on a survey of test bench operators within the EMPIR 14IND14 project is listed in Table 3, while the general requirements regarding the dimensions and the weight of the transfer standard are listed in Table 4. Figure 3 The reference torque transducer to calibrate the torque measurement in a nacelle test bench is installed directly at the hub flange of the device under test, where the mechanical input to the device under test is to be determined (4 MW nacelle test bench of the Center for Wind Power Drives at RWTH Aachen). Torque transducer Reference torque transducer Motor Load application system Nacelle 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 9 Table 3 Requirements of the RWTH Aachen test bench on torque transfer standards for calibrating nacelle test benches regarding the carrying capacity taken from [8]. Torque Nominal torque kN m ± 6000 Maximal torque kN m ± 7000 Torque measurement accuracy % 0.1 Forces and moments Parasitic loads Operating loads Forces (Fx, Fy, Fz) kN ± 100, ± 100, ± 100 ± 1450, ± 2000, ± 2000 Maximum forces (Fx, Fy, Fz) kN ± 100, ± 100, ± 100 ± 2000, ± 3000, ± 3000 Moments (My, Mz) kN m ± 100, ± 100 ± 7500, ± 7500 Maximum moments (My, Mz) kN m ± 100, ± 100 ± 15000, ± 15000 Forces/moments must be transmitted Requirement Measurement of forces/moments Optional Lifecycles Static stress assessment (cycles @ max. torque) 5000 Fatigue stress assessment (cycles @ max. torque) Failsafe 1012 Kinematic Nominal rotational speed min-1 12 Maximum rotational speed min-1 25 Measurement during rotation Requirement Measurement in both rotational directions Requirement Table 4 Dimensional requirements of PTB’s 1.1 MN m torque standard machine on a torque transducer to be calibrated and, afterwards, to be deployed to calibrate the torque measurement instrument in the nacelle test benches of RWTH Aachen. Dimensions Maximum flange diameter for calibration machine m 1.25 Maximum flange diameter for test bench m 4.5 Minimum length for calibration machine m 0.6 Maximum length for calibration machine m 2.2 Maximum weight for calibration machine t 2 Flanges: through-hole Requirement • 30 x M36 on bolt circle Ø 900 mm Requirement • For test benches, an adapter is required Optional Additionally, there is a list of metrological requirements on the behaviour of the transfer standard in Table 5. Table 5 Metrological requirements on a torque transfer standard to become an adequate torque transfer standard. Metrological parameters Non-linearities ≤ ± 0.05 % Hysteresis ≤ 2.9·10-5 mV/V Interpolation deviation ≤ |0.019| % Drift over time ≤ 8·10-3 % Creep (short term creep) ≤ |3·10-3| % An example of a torque transfer standard for nacelle test bench calibration is shown in Figure 4. This transfer standard is owned by PTB, the national metrology institute of Germany, and was produced by Hottinger Baldwin Messtechnik GmbH. It has a measurement range of 5 MN m and is statically calibrated up to 1.1 MN m with an expanded relative uncertainty of 8.8·10-4 (k = 2) according to DIN 51309 (which is comparable to EURAMET cg-14) using the world’s largest torque standard machine located at PTB. Due to the very small non-linearities of the torque transfer standard, a linear regression curve was ascertained for clockwise torque load applied in increasing and decreasing steps: M = 3851.1 kN m/(mV/V) ⋅ S . (2) Above 1.1 MN m, a calibration is not possible. To predict the relation between the output signal and the applied torque, including a predicted measurement uncertainty taking the prediction itself into account, an extrapolation method was developed. 16 DOI: https://doi.org/10.7795/530.20190111 So-called characterisation maps provide a simple solution for rectifying this problem. A similar procedure to this has been suggested for rotatory power measurement [1]. Most test benches have a defined main direction of rotation. For these test benches, only one direction of torque load is to be calibrated. However, if downwind nacelles are tested on the test bench, both torque application directions are to be calibrated. Moreover, since negative torque can appear during braking scenarios, the effect of applying negative torque on the hysteresis behaviour and the zero return should be analysed for all nacelle test benches. To define the measurement points of the characterisation maps, the typical operating range of the test bench, depending on the commonly tested devices, is to be listed. The typical operating range is depicted in Figure 9 in the form of a blue rectangle. It should be pointed out that the range of the characterisation map is limited by the installed device under test during the calibration measurements. For that reason, the device under test is to be picked accordingly. Figure 9 Schematic depiction of a characterisation map based on the typical testing range of a nacelle test bench with a uniform distribution of the measurement points. When there are no particular points within the operating range which need to be calibrated explicitly, such as typical maximum torque loads for certain devices under test, the load steps should be spread evenly as shown in Figure 9 in the form of the Xs. In general, the calibration should cover the entire operating range of the test bench by repeating the standard calibration with different combinations of torque load and rotational speed (Figure 10). To investigate the influence of rotational speed on the torque measurement, there are two different categories of characterisation maps: for CM1 (CM1a and CM1b in Figure 10), the torque is periodically fixed while the rotational speed is altered stepwise; and for CM2 (CM2a and CM2b in Figure 10), the rotational speed is kept constant periodically while the torque load is increased and decreased. All forms of the characterisation map have the same operating range. To determine the hysteresis of the test bench torque transducer, each calibration point should be met twice during one load cycle: first when raising the increment and second when attenuating it again. Attention is to be paid to omitting rotational speed steps close to eigenfrequencies of the system, which would lead to undesired dynamic effects. Here again, both the torque and rotational speed can be increased or accelerated respectively stepwise as a ramp, a step or a sine, whereby the ramp is the most common one. Moreover, if this type of combined calibration is not sufficient, single calibrations for each typical load, so-called partial range calibrations, can be performed. For a statistical evaluation of the calibration measurements, each characterisation map is to be repeated twice or each load step is to be passed at least four times. Should a calibration like this not be possible at all due to a lack of access to the control system of the device under test, the calibration has to be performed in the median of the typical testing range (see Figure 9c)) with a suitable standard device under test. Because of the set-up and the permanently acting control systems in a test bench, the determination of the transducer creep is not possible, which should not pose a problem. At least one map per category (CM1 and CM2) is to be performed in order to investigate the influence of the rotational speed on the torque measurement. If a closed loop control is not possible, then performing only CM2 is sufficient, since the reproducibility of CM1 would not be good without a closed loop control anyway. Torque Rotational speed typical testing range measurement point Torque Rotational speed Torque Rotational speed 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 17 Figure 10 Example of different load step sequences for a calibration with torque being increased (and decreased) and rotational speed being accelerated and decelerated at each torque step and vice versa. 3.7 Crosstalk effects The assembly process and the integration of the nacelle into the test bench are challenging, because of the large masses and dimensions of nacelles. For example, a 5 MW nacelle has the following dimensions: 6.5 m height, 6.5 m width and 17 m length and a total mass of 290 t. Even a very accurate assembly leads to an axis misalignment between the main shaft of the test bench and the hub of the nacelle. This axis misalignment may be in the range of up to 0.5 mm and, thereby, it may generate parasitic loads in the form of forces (Fx, Fy, and Fz) and bending moments (My and Mz) under rotation. The control system of the load application system tries to keep these parasitic loads to zero. However, due to the inaccuracy of the control system, the perfect limitation of the parasitic loads to zero is not possible in most cases. Consequently, an application of pure torque during the calibration is not possible. Figure 11 shows examples of the additional loads on the torque transfer standard within the averaging sequence over an integer number of distinct rotations (here six rotations). Additional influences on the torque transducer are not desired as they not only have an influence on the intended measuring bridges but also on the other measuring bridges, which causes an unintentional signal alteration. Figure 11 Parasitic loads during the calibration at the torque step of 1000 kN m and rotational speed of 6.5 min-1. 18 DOI: https://doi.org/10.7795/530.20190111 The close surveillance of additional loads is, therefore, necessary to avoid unwanted influences on the torque measurement. However, most periodic influences of additional loads can be neglected due to the averaging of the torque signal. To purposely study crosstalk influences on the torque signal, characterisation maps similar to Figure 12 are suggested. Figure 12 Schematic principle for investigating the crosstalk influence on the torque signal caused by additional loads. 3.8 Calibration interval For test benches involved in development tests of test benches in general or in the development of the devices under test, the calibration interval should not exceed 12 months. The calibration interval of the torque measurement instrument in the test bench highly depends on the set-up and the usage of the test bench and the quality procedure in the test bench. For test benches, which are mainly used to determine the efficiency of the device under test, the recommendation for the calibration interval is rather long at up to 24 months. After every change of the underlying regression curve or any maintenance or alignment concerning the torque measurement instrument, a recalibration is required. Ideally, a calibration at the beginning and at the end of a device under test being assembled and tested would be preferable. In reality, however, the utilisation and activity of the test bench need to be considered as well. 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 19 4 Evaluation of the calibration result including a measurement uncertainty The set-up for calibrating a torque transducer in a test bench by using a torque transfer standard is comparable to a reference torque calibration machine. In a reference torque calibration machine, a calibrated and very well-known torque transducer is deployed to calibrate another torque transducer. Due to the similarity of the set-up, the effects during the calibration are analogous. Other than the load cycles, which are based on EURAMET cg-14, the evaluation of the calibration result is based on ISO 7500-1 [4]. ISO 7500-1 describes the calibration of a force measurement system in material testing machines. Here, the deployed method does not necessarily demand that the exact load steps are met. Instead, an indication deviation between the transducer to be calibrated and the transfer standard is calculated. For the torque calibration in test benches, the calibration result is adapted to the calibration result in ISO 75100-1 and consists of: • the relative indication deviation q, • the measurement uncertainty uc for the indication deviation, and • the reversibility (in case increasing and decreasing torque load was applied). The calibration result is not a classification of the transducer calibrated as is common for calibrations. The calibration result can be used in the form of: • a look-up table (to correct the result by consulting the table), • a regression curve (to correct the result in post-processing), or • a correction of the underlying regression curve of the test bench transducer (requires a second calibration of the test bench transducer using the corrected regression curve to determine the measurement uncertainty). The evaluation of an example of a torque calibration under constant rotation can be found in the following publications: P. Weidinger, G. Foyer, S. Kock, J. Gnauert, R. Kumme 2018 Procedure for torque calibration under constant rotation investigated on a nacelle test bench in Proc. of Sensoren & Messsysteme 2018, Nuremberg, Germany P. Weidinger, G. Foyer, S. Kock, J. Gnauert, R. Kumme Calibration of torque measurement under constant rotation in a wind turbine test bench submitted to JSSS. 4.1 Evaluation of the transducer to be calibrated The performance of the torque transducer to be calibrated can be evaluated by the relative indication deviation q, the relative repeatability b, the relative resolution a, and the determination of the relative reversibility v when a calibration of decreasing torque is desired. All the following subsections are adapted from ISO 7500-1. 4.1.1 Relative indication deviation The relative indication deviation qj(ML) is the deviation of the torque measurement instrument in the test bench from the torque transfer standard relative to the real torque value measured by the torque transfer standard per load step ML. It is to be calculated for each of the at least two repeated load cycles j (1 and 2) or characterisation maps: qj(ML) = Mi,j(ML) - Mj(ML) Mj(ML) ⋅ 100% , (14) with j = 1,…, n , where Mi is the increasing torque load indicated by the torque measuring instrument in the test bench, and M is the increasing torque load indicated by the torque transfer standard. The calibration result is represented as a relative indication deviation q (ML), which is the arithmetic mean of the relative indication deviation per repetition qj(ML): q (ML) = 1 n∑qj(ML) n j = 1 . (15) Figure 13 is an example of a depiction of the relative indication deviation for a quasi-static load cycle under constant rotation. To assess the calibration result and the interrelation between the applied torque 20 DOI: https://doi.org/10.7795/530.20190111 and the rotational speed, not only is the relative indication deviation to be calculated, but also the standard deviation for this value. Depending on the standard deviation for the repeated measurements, a distinction between the torque deviations contingent on the different applied rotational speeds can be made. Figure 13 Example of indication deviation for four repetitions and averaged indication deviation. 4.1.2 Relative repeatability The quality of repeating the torque measurements in the test bench is described by the relative repeatability b. This relative repeatability is calculated for each load step separately and is the difference between the maximum indication deviation qmax and the minimum indication deviation qmin per load step: b(ML) = max j(qj(ML)) - min j(qj(ML)) . (16) The relative repeatability is an important contribution to the measurement uncertainty and an example of this is illustrated in Figure 13. 4.1.3 Relative resolution Most test benches are state of the art and display the applied load digitally. Therefore, the focus regarding the resolution r lies on digital torque displays. In general, the resolution depends on the resolution of the amplifier’s A/D converter and the data saving format. However, if the indication fluctuation exceeds this resolution, the resolution is to be summed up by half the span of the indication fluctuation. The indication fluctuation is to be determined with the prime mover and the device under test switched on, but with no load applied. To determine the indication fluctuation, the signal is to be averaged per revolution over the averaging sequence as depicted in Figure 14, then the minimum and maximum of the averaged signals per revolution are to be found and the difference is calculated. Subsequently, the relative resolution a is to be determined for each load step depending on the applied load: a(ML) = r(ML) M(ML) ⋅ 100% , (17) with r (ML) = max lrl(ML) - min lrl(ML) 2 , (18) where M is the torque load for the observed load step (averaged values of all repetitions). The resolution is to be represented in the same unit as the measured torque. repeatability reversibility 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 21 Figure 14 Schematic figure on how to determine the indication deviation per load step. 4.1.4 Relative reversibility For the special case that a regression curve for decreasing torque load is demanded, the relative reversibility v for the decreasing load can be determined. However, it is to be emphasised that the reversibility highly depends on the maximum torque load applied beforehand. To determine the relative reversibility, the deployed torque transfer standard must be calibrated for decreasing torque load. Moreover, for the reversibility measurements, only one decreasing load cycle is to be performed. Based on the deviation between the torque measurement values for increasing and decreasing load steps (Figure 13), the relative reversibility is calculated: v = M - M' M  ⋅ 100% . (19) 4.2 Measurement uncertainty As mentioned above, the prevailing calibration set-up is comparable to a reference calibration machine, where the transducer to be calibrated is compared to a national standard. Thus, in general, for an NTB calibration, the influences on the transducer to be calibrated are the same as those in a reference torque standard machine but even larger. The measurement uncertainty itself is based on the aforementioned factors and consists of the following uncertainty contributions: • the resolution, • the repeatability, and • the torque transfer standard. All this results in the best expected value for the indication deviation and the corresponding expanded measurement uncertainty. The following evaluations are based on ISO 7500-1 and are adapted for a torque calibration under rotation. 4.2.1 Measurement conditions In general, the influences on the torque transducer to be calibrated are the same as those in a reference torque standard machine but even larger. The Ishikawa diagram in Figure 15 lists examples of possibly occurring influences affecting the calibrated torque measurement in a nacelle test bench. Most influences can be quantified and categorised by the aforementioned preparatory operations and tests. Additionally, all set-up-specific effects, i.e., misalignments and friction in different components are directly part of the calibration result and have already been considered in the measurement uncertainty by determining the repeatability. Emergency braking, which may occur and is to be tested under normal testing conditions, should be avoided during the calibration process. After emergency braking, the test bench is to be operated with maximum torque load in the driving direction for at least 3 min and a new zero signal is to be determined. In doing so, the hysteresis effect due to the opposite torque load is minimised. 1 2 3 4 5 6 22 DOI: https://doi.org/10.7795/530.20190111 In the case that different amplifiers were used for the calibration of the torque transfer standard than for the calibration in the nacelle test bench, the difference between the diverse amplifiers is to be accounted for in the measurement uncertainty or even the indicated torque value. Because of the signal averaging over a certain number of revolutions, the dead-weight influence on the torque signal can be neglected. Figure 15 Ishikawa diagram listing the influences during the calibration of the torque measurement in nacelle test benches. 4.2.2 Uncertainty contribution of the resolution The uncertainty contribution of the relative resolution of the torque measurement instrument in the test bench ures for each examined load step is the square root of the sum of the following components squared: • the uncertainty component due to the relative resolution of the torque measurement instrument in the test bench under load, which is denoted as aM divided by 2√3 because of the assumption of a rectangular distribution, • the uncertainty component due to the relative resolution of the torque measurement instrument in the test bench after load release, which is denoted as aZ divided by 2√3 (same assumption of a rectangular distribution). The relative resolution of the torque measurement after load release is part of the indication deviation of every load step because of the aforementioned taring of the torque values for every load step. The uncertainty contribution of the relative resolution can be calculated for every load step as follows: ures(ML) = √(aM(ML) 2√3)2 + (aZ 2√3)2 . (20) As the relative resolution a is expressed in the unit %, the absolute uncertainty contribution of the relative resolution ures is expressed in % as well. 4.2.3 Uncertainty contribution of the repeatability As mentioned before, the relative repeatability is considered as an uncertainty contribution to the total uncertainty for the indication deviation. The uncertainty contribution of the repeatability urep is the standard deviation of the best expected value qi in % and the relative mean indication deviation q  in %, where n is the number of repetitions per load step: urep(ML) = √1 n (n - 1) ∑ (q j (ML) - q (ML))2 n j = 1 . (21) Sampling frequency Dead-weight Misalignment Humidity Temperature/ Temp. gradients Pressure Drift Measurement uncertainty Environmental conditions Creep Torque transfer standard Calibrated torque in the wind turbine test bench Stability Resolution EMC Measurement conditions Misalignments Parasitic loads LAS Torque inversion Friction in diff. comp. Amplifiers Torque transducer Zero-point deviation Repeatability Reversibility Creep Filter Negligible General effects Set-up specific effects Rotational speed 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 23 4.2.4 Uncertainty contribution of the transfer standard The uncertainty contribution of the torque transfer standard ustd is given by the relative calibration uncertainty of the transfer standard ucal, which is stated in its calibration certificate, and additional possible uncertainty contributions (A, B, and C) due to deviating temperature and humidity during the calibration measurements, drift over time, and deviations because of the deployed combined linear regression curve for increasing and decreasing torque load. The uncertainty contribution of the torque transfer standard is calculated according to the following: ustd(ML) = √(ucal(ML))2 + A2 + B2 + C2 . (22) As the uncertainty contribution of the torque transfer standard is shown as a relative uncertainty, it is expressed as a percentage (%). 4.2.5 Expected value for the indication deviation The result of the torque calibration is the best expected value for the mean relative indication deviation q (ML) of the torque measurement instrument in the test bench. Moreover, an expanded measurement uncertainty U(ML) belongs to this relative indication deviation, which is the product of the coverage factor k and the combined uncertainty uc(ML): UU(ML) = k ⋅ uc(ML) = k ⋅√∑(uj(ML))2 n j=1 = k ⋅√(ures(ML))2 +(urep(ML))2+(ustd(ML))2 . (23) Additional uncertainty contributions may be influences of the deployed amplifiers of the torque transfer standard or significant influences of the friction torque. It is recommended to use a coverage factor of k = 2. In certain cases, k can also be calculated based on the number of effective degrees of freedom as stated in the GUM. Annotation: the measurement uncertainty is an absolute measurement uncertainty of the relative indication deviation, which is represented in the unit %. The absolute measurement uncertainty has the same unit as its reference value. Consequently, the unit of the absolute measurement uncertainty is %. 4.2.6 Example of an uncertainty calculation An example of a calculation of the measurement uncertainty for torque calibration under constant rotation is given in P. Weidinger, G. Foyer, S. Kock, J. Gnauert, R. Kumme Calibration of torque measurement under constant rotation in a wind turbine test bench submitted to JSSS. This paper covers a complete consideration of a torque calibration under constant rotation in a nacelle test bench including the measurement uncertainty budget and the temperature difference between the calibration laboratory, where the torque transfer standard was calibrated, and the conditions during the nacelle test bench calibration. Moreover, a detailed calculation of the relative indication deviation for four measurement repetitions is given. The expanded (k = 2) absolute measurement uncertainty U(ML) for the best expected value for the relative indication deviation is calculated including the uncertainty contribution of the torque transfer standard, which comprises the drift of the transducer, the temperature influence and the uncertainty of the amplifier calibration (special case). 24 DOI: https://doi.org/10.7795/530.20190111 5 Calibration output and benefit for the test bench operator The result of a torque calibration in nacelle test benches can be expressed in different forms: • a look-up table (post-processing), • a regression curve (post-processing), or • a correction of the torque transducers underlying the transfer curve (additional calibration required afterwards). In general, the possibly occurring indication deviations in the torque measurement of a nacelle test bench can be corrected by all the different forms of the calibration result. Moreover, the calibration gives a measurement uncertainty for the calibrated range, which is again a contribution to the measurement uncertainty of the efficiency determination. Test bench operators can benefit from knowing the deviation between their internal torque measuring instrument and a traced national torque standard and correcting this deviation to obtain a more accurate torque measurement result. Furthermore, after a calibration, the precision including the repeatability of the torque measurement instrument in the nacelle test bench is determined. Based on this, not only can the efficiency of devices under test be determined more reliably, but also the data to validate simulations and the repeatable component testing gain credibility. With the standardised calibration of torque measurements in nacelle test benches, efficiencies of devices tested on different test benches can be compared to each other reliably. 14IND14 – Torque Measurement in the MN·m range Good Practice Guide DOI: https://doi.org/10.7795/530.20190111 25 Glossary • Calibration: operation that, under specified conditions, in a first step, establishes a relation between the quantity values with measurement uncertainties provided by measurement standards and corresponding indications with associated measurement uncertainties and, in a second step, uses this information to establish a relation for obtaining a measurement result from an indication (VIM 2.39). • Characterisation map: load cycle covering the typical operation range of a nacelle test bench. The characterisation map is limited by the device under test installed during the calibration. • Crosstalk effect: influence of additional multi-axial loads not only on the respectively intended measuring bridges but also on the other measuring bridges, which causes an unintentional signal alteration. • Load cycle: range of applied load on a measuring instrument during calibration. • Measurement uncertainty: non-negative parameter characterising the dispersion of the quality values being attributed to a measurand, based on the information used (VIM 2.26). • Torque transfer standard: calibrated and very well-known torque transducer, which is used to trace torque measurement to the national standard. • Zero signal: offset value of a transducer in non-loaded condition. The zero signal is used to tare all measurement signals of the following load cycle.