Experiences with source characterization methods within and beyond the scope of EN 15657

Fabian Schöpfer 1 , Tobias Kruse, Johanna Weinzierl, Andreas Mayr, Ulrich Schanda Rosenheim Technical University of Applied Sciences Hochschulstr. 1, 83024 Rosenheim, Germany

ABSTRACT Prediction of machinery noise in buildings requires knowledge about the airborne sound radiation and the structure-borne sound excitation from the source of interest. Laboratory methods to determine the airborne sound power of sources are well established since many years. However, standard methods for the characterization of structure-borne sound from building service equipment were first described in EN 15657-1:2009 which was then superseded by EN 15657:2017. In recent years the test methods described in this standard were applied in several research and development projects in the laboratory for sound measurement (LaSM) at Rosenheim Technical University of Applied Sciences. For this purpose the experimental equipment and the facilities of the laboratory were expanded which involves a variable reception plate test rig that was built in 2017. This paper describes the experience with test methods within and beyond the scope of the EN 15657:2017 using the available laboratory resources.

1. INTRODUCTION

For the prediction of machinery noise according the installed power in the building is required. This requires the knowledge of the equivalent blocked force or the free velocity of the source and the equivalent source mobility. EN 15657 describes laboratory methods to determine these quantities. One possibility is an indirect measurement using the so-called reception plate method. The source is therefore connected to an isolated plate and driven in the desired stage. This plate is considered as an SEA subsystem that allows to set up the power equilibrium as follows

W in = W d = E ωη = m ⟨ ˜ v 2 ⟩ s ωη (1)

Where W in is the unknown input power of the source, and W d the plate power that can be determined from the vibrational energy (i.e. the mass m and the spatial average root-mean-square velocity ⟨ ˜ v 2 ⟩ s ) and the total loss factor η for the frequencies ω . In this paper the limitations of the reception plate test rig according to EN 15657 are discussed based on project experience with building service equipment at the LaSM from recent years. A new variable test-rig that can be adapted to the requirements of di ff erent sources is presented. On this test-rig a 3D scanning Laser Doppler Vibrometer (LDV) was used to measure the response at

1 fabian.schoepfer@th-rosenheim.de

a slaty. inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS O ¥, ? GLASGOW

a huge number of positions across the plate for a known power input. This data is used to assess practical sampling strategies with drastically reduced set of measurement points. In addition the determination of the total loss factor is discussed and the application of the power injection method (PIM) to determine the loss factor is presented and comapred to the results with the decay method.

2. TEST-RIG

In 2015 a reception plate test-rig based on the suggestions in the Annex of EN 15657:2009 was built in the facilities of the LaSM. However research and development projects mainly with manufacturers from the buildings service equipment sector showed, that this test rig has limiation for a couple of sources (Figure 1). For example there is the problem with roller shutters that are mounted on top of a window frame before it is installed. This source comprises the roller-shutter including the motor, the housing, the shield and the rails but also the window frame which is finally the element that is screwed into the building structure. Without a window opening in the reception plate such sources can not be characterised. An then there is also the question whether this source predominately excites bending waves [1]. Another example is a sanitary pre-wall installation as described in [2] for example. These installations cover a wide area of the surface of the reception plate which is problematic in terms of the positioning of accelerometers across the plate. However the use of three isolated plates gives the possibility to determine the power input into each coupled wall or floor separately. Other sources that can not be fully characterised with the described test rig are waste water pipes. Although the pipe clamps can be fixed to the reception plate and the power as well as the blocked force can be determined, it is not possible to characterize the pipe feed-through in floors with the typical mock-up. And because the horizontal plate is typically placed on the laboratory floor it is not possible to characterise the power input of horizontal pipe sections underneath a floor with fixing to the concrete ceiling.

Figure 1: Limitations of a typical reception plate test-rig.

So the main aspects to improve were: (a) a plate with an opening for windows or similar, (b) a bigger plate that allows enough space even for big service equipment (n.b. that a bigger plate also improves the modal density at low frequencies), (c) plates with opening for a pipe feed-through and

Figure 2: Left: Variable reception plate test rig with crane and rack to store plates for di ff erent purposes. Right: Current mock-up with four plates and a lifted floor plate to characterise sanitary system including pipe feed-through and clamps fixings underneath the floor.

(d) a horizontal plate that is lifted from the ground to attach source such as horizontal pipe sections from underneath. And there may be other requirements for other sources that were not considered yet. Hence it is required to develop a variable reception plate test rig to adjust the rig to the requirements of the source. In 2016 the founding of a new technology centre at the Rosenheim Technical University of Applied Sciences gave the possibility to realise these aspects in a new variable reception plate test rig. The so called roteg (Rosenheimer Technologiezentrum für Energie und Gebäude - Rosenheim Technology Centre for energy and building systems) comprisis laboratories from di ff erent discplines in the field of technical equipment of buildings. This gives the chance to use synergies and share knowledge and facilities to run an test service equipment on the various test-rigs which involves vibro-acoustic testing. In the roteg laboratory the reception plate test-rig also comprises three individual plates. To mount the two vertical plates, a corner from 20 cm reinforced concrete was cast in-situ. This allows to walk behind the test-rig and hide supply pipes and water reservoirs that are necessary to run various systems. In addition one wall has a rectangle opening that can be ussed to reach the back of the actual reception plate which can be useful to mount accelerometers when there is not enough space on the front. The opening can also be used in combination with a reception plate with an opening to characterise sources like roller shutters that require such openings as described above. A variable test-rig rig requires a facility that allows the handling of heavy prefabricated plates. The laboratory hall is equipped with a 3 . 2 t bridge crane and a rack to store plates for di ff erent purposes. The latest mock-up uses a fourth reception plate that rests on top of the l-shaped concrete walls and a support at the free edge. This plate allows the characterisation of sanitary systems with both feed- throughs (floor and ceiling) as well as the connection to the vertical wall.

3. SAMPLING STRATEGIES

The spatial average velocity is the main quantity to measure when determining the reception plate power. It is therefore required to find a strategy to sample measurement points in a way to get the best estimate for the energy carried by bending waves in the plate. A precise result can be achieved by sampling the whole plate in a fine grid of points. However this is not practicably applicable. It is therefore required to find the best compromise between e ff ort and precision by reducing the measurement points with given guidelines.

2 . 5

2

dimension in m

1 . 5

1

0 . 5

0 0 . 5 1 1 . 5 2 2 . 5 3 3 . 5 0

dimension in m

Figure 3: Horizotal reception plate (2 . 9 m × 3 . 7 m). Excitation (diamond) and measurement points (corner: circle, edge: square and center zone: triangle).

3.1. EN 15657 EN 15657 [3] refers to ISO 10848-1 [4] when considering the sampling position for the spatial average velocity. At least six positions should be chosen. As in ISO 10848-1 the minimum distance between the positions should be 0 . 5 m and the minimum distance to the source (exciation point) should be 0 . 1 m. In EN 15657 no information is given on relevant distances to the boundaries of the plate. However in ISO 10848-1 the minimum distance to the boundary should be 0 . 25 m. Späh and Gibbs [5] showed that ist is required to use positions near the edges. EN 15657 however mentions, that the number and the position of the chosen sampling points should be validated using the power substitution method.

3.2. Strategy applied in the LaSM Based on the guidelines given in EN 15657 a sampling strategy was developed and applied for the test-rig in the LaSM. With a known power input from a point excitation, the velocity on the plate was measured with 20 accelerometers. Four of these were placed exactly in the corners ( < 5 cm). One accelerometer was placed randomly on each edge with a boundary distance of ( < 5 cm). The remaining 12 were randomly distributed across a centre region of the plate with a minimum boundary distance of 0 . 25 m, a minimum distance of 0 . 5 m between the points and at least 0 . 3 m between the excitation point and the nearest accelerometer (see Figure 3). From these measured velocities it was aimed to find a sampling strategy with a set of 12 accelerometers. This suits best to the available equipment in the laboratory but can also be considered as a feasible number in terms of e ff ort for a laboratory charaterisation method. In post processing in M atlab two loops were used to randomly draw positions and find the combinations that give best agreement with the input power. The spatial average velocity was the only parameter. The mass of the plate and the total loss factor were not changed in this process. The applied loss factor was determined from the measured structural reverberation time. In the inner loop a random set of 12 points were drawn out of these 20 positions. No weighting or priority was given to corner, edge or centre positions. All 20 points were treated the same way. This gives a total number of  20 12  = 125 970 possible combinations. For each set of 12 drawn measurement points the reception plate power was calculated and compared against the known input power. The arithmetic mean of the absolute di ff erence for the one-third-octave bands from 50 Hz to 3150 Hz (i.e. the frequency range required in EN 15657) was then calculated and saved before the next pick.

20

ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 11, co: 1, ed: 0 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 11, co: 1, ed: 0 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 10, co: 1, ed: 1 ce: 9, co: 1, ed: 2 ce: 9, co: 1, ed: 2 ce: 11, co: 1, ed: 0

10

L W in − L W d

0

− 10

63 125 250 500 1k 2k − 20

f in Hz

Figure 4: Di ff erence between the measured input power and the plate power for the best random sets of 12 out of 20 points. The legend indicates the number of centre (ce), corner (co) and edge (ed) position for each set.

After 1000 cycles the sampling points for the smallest average di ff erence were saved. The outer loop repeated this procedure 100 times. In each cycle it was checked if the the chosen sampling that gave the best agreement with the input power was already drawn in an earlier cycle. If not the sampling points were saved. Figure 4 shows the unique sets of good sampling points after 100 cycles. The results show that are are less than 100 unique sets. The remaining sets show that for 12 out of the 20 points (4 corner, 4 edge, 12 centre), 9-11 should be in the centre zone, 0-2 on edges and at least 1 point in a corner. From this results the the following procedure was derived for the application in the LaSM with a set of 12 accelerometers: One position in a random corner ( < 5 cm distance to boundary), one position on a random edge, randomly distributed between the corner with a minimum distance of 0 . 25 m and ten positions in the centre zone with a boundary distance of > 0 . 25 m 0 . 5 m between the points and a distance of at least 0 . 3 m to the excitation point.

3.3. Sampling strategy suggested by Reinhold and Hopkins Reinhold and Hopkins [6] suggested a sampling procedure derived by analysing of the vibration field of a validated finite element model of a horizontal reception plate. A validated model gives the possibility to quickly determine the response at a huge number of points for a range of excitation positions. It is suggested to use an area weighted average from corner, edge and centre positions. Corner zones are defined as squares 10 cm × 10 cm, edge strips are defined as rectangles with a width of 10 cm between the corner zones and the central zone is the rest of the plate. The strategy requires 20 positions: one in each corner zone, two on each edge strip and eight in the central zone. On the edge strip and in the central zone, the minimum distance between points should be 0 . 8 m. The area weighted velocity, L v , w [6] is then given as

L v , w = 10 lg S Corners 10 L v , Corners / 10 + S EdgeStrips 10 L v , EdgeStrips / 10 + S CentralZone 10 L v , CentralZone / 10

! (2)

S Total

3.4. Application and comparison of sampling strategies A recent study in the LaSM used a 3D scanning Laser Doppler Vibrometer (LDV) to investigate the response of one of the vertical plates of test rig described in section 2 on a huge number of

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

Figure 5: Left column: Available measurement point from the LDV scan. Top: edge excitation, middle: corner excitation, bottom: random excitation point in the centre zone. The excitation point is idicated by the black cricle. Nearfield points within a radius of 0 . 3 m are excluded. Shadows of the shaker stand or metal fixing plates of the vertical plate are the reason for the gaps in the grid. Right column: Corresponding di ff erences between the input power and the reception plate power.

measurement positions. This study was also used to investigate the relative importance of in-plane waves for out-of-plane and in-plane excitation [7]. However, in this context the out-of-plane response for out-of-plane point excitation is used to assess the sampling strategies described above. Three excitation positions were considered: one in a corner, one on an edge and one randomly positioned in the centre zone of the plate. The excitation signal was white noise. The power input was measured with an impedance head an averaged across all individual measurements for each point. Depending on the shadow from shaker stand the number of total points varied for the di ff erent excitation position but was above 400 points for each. In figure 5 the three excitation positions and corresponding measurement positions can be seen. It is assumed, that the spatial average velocity from all measured points is a good estimate for the vibrational energy in the plate. The mass of the plate was measured with a crane scale, however the uncertainty of the mass can be considered negligible compared to the uncertainty of the spatial average velocity. In the following the results from the sampling strategies are therefore compared against the result from the full set of point from the LDV scan. Using the guidelines given in sections 2 and 3 a random set of position was drawn from the full set of available measurement positions. This was again performed in a loop of 100 cycles. Hence 100 random sets were considered.

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

Figure 6: Sampling points and results with the LaSM sampling strategy. Left: example sets of sampling point and indication of the minimum boundary distance. Right: Di ff erences between the input power and the reception plate power.

For the comparisons a measured total loss factor was used in equation Equation 3. This loss factor was determined using the decay method and reverse filtering of the impulse response. For one-third- octave bands at and below 160 Hz the octave band loss factor was used. The applied total loss factor is shown later in Figure 8. Figure 5 also shows the di ff erence between the input power, measured with the impedance head and the reception plate power using the full set of sampling points. It can be seen that for all three excitation positions the reception plate power underestimates the power at frequencies < 80 Hz and in the 100 Hz one-third-octave band. From 125 Hz the agreement is very good with an average absolute di ff erence of < 0 . 5 dB in the range from 125 Hz to 1250 Hz. Frequencies above 1250 Hz were not considered due to due a low signal-to-noise ratio. Figure 6 shows the di ff erence between the input power and 100 random draws following the LaSM sampling strategy described in section 2. The Figure shows the average di ff erence and the standard deviation of the di ff erence for the 100 random draws. It can be seen that the frequency characteristic is very similar to the result using all positions from the LDV scan. At low frequencies there is also an underestimation of the reception plate power. Towards high frequencies the reception plate power tends to overestimate the power slightly with an arithmetic deviation of < 1 dB in the range from 125 Hz to 1250 Hz. The standard deviation for the 100 random draws is < 1 dB on average across the considered frequency range.

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

2 . 9

10

dimension in m

L W in − L W d

2

5

1

0

31 . 5 63 125 250 500 1k − 5

0 1 2 3 4 0

f in Hz

dimension in m

Figure 7: Sampling points and results with the sampling strategy by Reinhold and Hopkins. Left: example sets of sampling point and indication of the corner, edge and central zones. Right: Di ff erences between the input power and the reception plate power.

Figure 7 shows the results of this comparison using the strategy suggested by Reinhold an Hopkins (section 3). Again the frequency characteristic is very similar and the deviations towards high frequencies tend to be smaller compared to the LaSM sampling strategy. The standard deviation for the 100 random draws is < 1 dB on average across the considered frequency range.

4. TOTAL LOSS FACTORS

As mentioned in the section before the energy determined from the full set of sampling points from the LDV scan can be considered as very good estimate. As the mass was measured with a crane scale, the remaining uncertainty is mainly justified by the measured total loss factor. Due to the low modal density at low frequencies the decay method tends to have high uncertainties. As the LDV scan gives a good estimate of the plate power, the power injection method can be used to determine the total loss factor of the considered system.

η = W in m ⟨ ˜ v 2 ⟩ s ω (3)

Figure 8 shows the total loss factor from the power injection method compared to the result from the decay method. As expected the loss factors are very similar towards high frequencies, were the agreement shown in section 3 between reception plate power and input power was good. Towards

120

η determined from T η determined usin PIM

115

10 lg( η/ 10 − 12 )

110

105

31 . 5 63 125 250 500 1k 100

f in Hz

Figure 8: Total loss factor from determined from the measured structural reverberation time compared with the total loss factor determined using the power injection method.

low frequencies there are major discrepancies between the loss factors determined with these two methods. It seem that the decay method severely underestimates the total loss factor. In the following the reception plate is calculated fro the two sampling strategies (LaSM and Reinhold and Hopkins) using the the PIM total loss factor. The di ff erences between the input power and the reception plate power is shown in Figure 9. Except the 31 . 5 Hz one-third-octave band the agreement is now very good with average absiloute di ff erences of < 0 . 5 dB for the Reinhold-Hopkins strategy and < 1 dB for the LaSM strategy.

5. SUMMARY

Recent project experience showed that for the characterisation of service equipment it is useful to have a variable reception plate test-rig than can be adapted to the requirements of the source mounting. Such a test-rig was developed and realised in the Laboratory for Sound Measurement at Rosenheim Technical University of Applied Sciences. For the measurement of the spatial average velocity to determine the reception plate power various sampling strategies exist. this paper used measurement data from a 3D scanning LDV to assess the strategy developed and applied in the LaSM and a recently published suggestion by Reinhold and Hopkins. Both showed good agreement with the input power. However the method by Reinhold and Hopkins tends to have the best better agreement but it requires more measurement positions. Besides the spatial average velocity, the loss factor is important to determine the plate power. Typically this is determined from measured structural reverberation time, sometimes in combination with the half power bandwidth at low frequencies. As the full over sampled set of measurements point from the LDV scan reduces the uncertainty from the spatial average velocity to a minimum, the power injection method was applied in case study to determine the total loss factor. With this total loss factor the agreement between the plate power and reception plate power is very good for both sampling strategies. Further work is required to asses the application plates with other dimensions. It is also worth considering to use the sampling strategies to determine loss factor using PIM as the strategies, in particular the one suggested by Reinhold and Hopkins give very good agreement with the input power and is almost identical to the result from a full LDV scan.

10

Reinhold and Hopkins LaSM

L W in − L W d

5

0

31 . 5 63 125 250 500 1k − 5

f in Hz

(a) Edge excitation

10

Reinhold and Hopkins LaSM

L W in − L W d

5

0

3 1 . 5 63 125 250 500 1k − 5

f in Hz

(b) Corner Excitation

10

Reinhold and Hopkins LaSM

L W in − L W d

5

0

31 . 5 63 125 250 500 1k − 5

f in Hz

(c) Excitation in the centre zone

Figure 9: Di ff erence between input power and plate power for both sampling strategies but using the loss factor from PIM.

ACKNOWLEDGEMENTS

The authors are greatful to Müller-BBM Vibroacoustic Systems GmbH for their support with measurement equipment.

REFERENCES

[1] Ulrich Schanda, Michael Hoßfeld, Fabian Schöpfer, and Andreas R. Mayr. In-plane excitation of reception plates according to en 15657:2017. In Proceedings of ICA 2019 , 2019. [2] Ste ffi Reinhold, Jochen Scheck, Heinz-Martin Fischer, Andreas Ru ff , and Carl Hopkins. Laboratory characterisation and prediction of structure-borne sound transmission of sanitary installations in heavyweight buildings. In Proceedings of Euronoise 2015 , 2015. [3] EN 15657:2017. Acoustic properties of building elements and of buildings - laboratory measurement of structure-borne sound from building service equipment for all installation conditions. [4] ISO 10848-1:2017. Acoustics – laboratory and field measurement of flanking transmission for airborne, impact and building service equipment sound between adjoining rooms: Part 1: Frame document. [5] Moritz Michael Späh and Barry M. Gibbs. Reception plate method for characterisation of structure-borne sources in buildings: Assumptions and application. Applied acoustics. Acoustique applique. Angewandte Akustik , 70:361–368, 2009. [6] Ste ffi Reinhold and Carl Hopkins. Sampling procedures on reception plates to quantify structure- borne sound power from machinery. Applied acoustics. Acoustique applique. Angewandte Akustik , 172:107649, 2021. [7] Johanna Weinzierl, Tobias Kruse, Andreas R. Mayr, Fabian Schöpfer, and Ulrich Schanda. Leistungsvergleich gebäudeähnlicher strukturen bei in-plane und out-of-plane anregung. In Deutsche Gesellschaft für Akustik e.V., editor, Fortschritte der Akustik - DAGA , 2022.