Acoustic behaviour of GFRP-PUR web-core composite sandwich panels

Construction and Building Materials 438 (2024) 137195 Available online 1 July 2024 0950-0618/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Acoustic behaviour of GFRP-PUR web-core composite sandwich panels Miguel Proença a , 1 , Pedro Santos b , Luís Godinho b , Albano Neves e Sousa a , Jo˜ ao R. Correia a , * , M´ ario Garrido a , Jos´ e Sena-Cruz c a CERIS, Instituto Superior T´ ecnico, Universidade de Lisboa, Portugal b University of Coimbra, ISISE, ARISE, Department of Civil Engineering, Coimbra, Portugal c University of Minho, ISISE, ARISE, Campus de Azur´ em, Guimar˜ aes, Portugal ARTICLE INFO Keywords: Sandwich panels Glass fibre reinforced polymer Polyurethane foam core Airborne sound reduction Impact sound pressure level ABSTRACT This paper presents an experimental and analytical study about the sound insulation of innovative sandwich panels - named EasyFloor – for the rehabilitation of degraded timber floors in old buildings. Two types of panels were developed: (i) an all-composite sandwich panel, made of glass fibre reinforced polymer (GFRP) face sheets and webs, and polyurethane (PUR) foam; (ii) and a hybrid sandwich panel, where the top face sheet comprises a thin layer of concrete. The study included the comparative analysis of the airborne sound reduction and impact sound pressure level of both types of panels based on experimental tests on full-scale specimens, including a typical floor covering. The acoustic insulation properties determined experimentally were compared with predictions from different analytical models typically applied to sandwich panels, and also with performance requirements defined in building regulation. Among the different analytical models that were assessed, Sharp’s model provided the most accurate predictions when considering the natural and dilatation frequencies as the critical frequency. The comparison of the normalized airborne and impact sound indices determined experimentally with requirements set in building regulation shows that the use of GFRP-PUR panels in the rehabilitation of lightweight timber floors and in new construction, where building comfort requirements are stricter, require mitigation measures to improve sound insulation. 1. Introduction Sandwich panels are composite-section building components that typically comprise thin face sheets bonded to a low-density core [1–3]. Such type of panels is widely used in civil engineering applications mainly as non-structural or secondary components subjected to low load levels, such as cladding or roof sheeting. However, their application as primary structural members has also been investigated, namely in bridge [4–6] and pier decks [7], and in building floors [8], for both new construction and rehabilitation. In contrast to the more typical lightweight configuration, for some heavier applications, sandwich solutions can comprise relatively thick steel sheets and high-density elastomeric cores [9]. Regarding the constituent materials of sandwich panels, alternatives to typical metal sheets include fibre-reinforced polymer (FRP) composites [10–13], such as glass fibre-reinforced polymer (GFRP), which are often combined with lightweight cores, such as polyurethane (PUR) foam [14,15]. Advantages offered by FRP composites when compared to metallic-based solutions include their lower self-weight, high strength-to-weight ratio, improved durability and reduced maintenance requirements [16,17]. The prefabrication possibilities (e.g., vacuum infusion or pultrusion) and low self-weight of FRP composites renders such composite sandwich panels an interesting solution for old building floors, namely in rehabilitation scenarios, where the existing structural members (often stone-rubble masonry walls) are usually not able to withstand significant weight increase; rehabilitation solutions with FRP sandwich panels involve lower overall vertical loads than traditional solutions (e.g., reinforced concrete slabs), which avoids affecting the original seismic performance of such buildings [14]. Despite such advantages, and as a result of their low weight, a weak point in such type of elements is the poor acoustic performance for both airborne and impact sound transmission [18] compared to conventional solutions, which means that additional sound transmission mitigation measures (such as false ceilings and/or floorings supported on a resilient * Corresponding author. E-mail address: [email protected] (J.R. Correia). 1 † deceased Contents lists available at ScienceDirect Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat https://doi.org/10.1016/j.conbuildmat.2024.137195 Received 11 January 2024; Received in revised form 20 June 2024; Accepted 21 June 2024 Construction and Building Materials 438 (2024) 137195 2 layer) should be needed to fulfill user comfort requirements for housing and office buildings. The recent work by Proença et al. [14] on the acoustic performance (for airborne and impact sounds) of homogeneous-core GFRP-PUR sandwich panels for structural applications highlights such need for additional detailing of corrective measures to comply with building code requirements. In the follow-up of the development and study of homogeneous core GFRP-PUR sandwich panels [14], a novel web-core composite sandwich panel design was proposed as a modular system, named EasyFloor. Two variants of this EasyFloor sandwich system were developed for building floors: (i) an all-composite sandwich panel, made of GFRP face sheets and webs, and PUR infill foam (Fig. 1a), for spans up to 4 m; (ii) and a hybrid sandwich panel, where the top face sheet is made of a thin layer of concrete (Fig. 1b), for spans up to 5 m. Adjacent panels can be connected vertically through a snap-fit or horizontally by means of adhesive bonding (without snap-fit). The present paper aims at assessing the acoustic performance of such web-core sandwich panel systems, through the accomplishment of the following goals: analysis of their (i) airborne sound reduction and (ii) impact sound pressure level based on experimental tests on full-scale specimens; (iii) experimental assessment of the impact sound reduction provided by a typical floor covering; (iv) comparison of the acoustic insulation properties determined experimentally with corresponding analytical predictions; and (v) verification of the compliance with building code requirements for the acoustic insulation properties of the sandwich panels. The remainder of the paper is organized as follows: (i) a literature review is presented first, summarizing the main findings from previous studies more directly related with this study; next, (ii) the materials and test methods are described, and (iii) the experimental results are presented and (iv) compared with analytical predictions; (v) the compliance with the design requirements and typical building regulations is assessed; and, finally, (vi) the main conclusions drawn from this study are presented. 2. Literature review In general, wave motion in a sandwich panel subjected to an acoustical excitation may be separated into four basic motions: (i) symmetric motion (often related to longitudinal waves); (ii) antisymmetric motion (relative to flexural/bending waves); (iii) dilatational motion (propagation of a wave through the thickness); and (iv) shear waves. D’Alessandro et al. [18] presented an extensive review of theoretical and numerical models for the acoustic analysis of sandwich panels, including methods such as the Finite Element Method (FEM), the Boundary Element Method (BEM) or Statistical Energy Analysis (SEA). Besides these advanced models, simpler analytical/empirical models for sandwich-type panels were proposed by different authors, such as Krakers [19] and Ballagh [20]. Few studies are available in the literature about the acoustic behaviour of sandwich panels similar to the one developed herein for building applications. Patinha et al. [21] performed tests to characterize the sound insulation properties of hybrid composite sandwich specimens made of jute and glass fibre hybrid face sheets with three different fibre architectures and extruded PUR cores. However, the test setup adopted was quite unique, precluding the comparison of results with experimental data for other sandwich solutions, and pointing out the need for test results obtained via standardized testing procedures. Garay and Pino [22] performed laboratory measurements for the determination of the airborne sound reduction index (R w ) of a sandwich panel made of expanded polystyrene (EPS) core and oriented-strand board (OSB) face sheets, and obtained a R w value of 39 dB. Petrone et al. [23] studied the acoustic power radiated from sandwich panels made of aluminum foam via experimental testing and numerical modelling, and concluded that the models adopted were able to predict the experimental results when assuming a constant damping factor. Wawrzynowicz et al. [24] performed experimental tests and analytical and numerical (including finite element) simulations to determine the airborne sound insulation of a composite sandwich panel made of magnesium-cement face sheets and EPS core. FE models compared well with test results, while the analytical model provided accurate predictions until 1600 Hz. Santos et al. [25] performed airborne and impact sound transmission tests on a sandwich panel made of cross-ply wood layers and a PUR foam core on a reduced-size chamber based on test protocols from ISO 10140–2 [26] and ISO 10140–3 [27]. They found out that analytical models developed to describe the airborne sound insulation of typical sandwich panels (i. e., with thin and rigid face sheets and thick soft core) were able to simulate the experimental behaviour of a metal-sheet sandwich panel, but failed to describe the behaviour of the panels they had developed, which comprised thicker face sheets. From a series of tests on different analytical models, they found out that the Sharp model for homogeneous isotropic elements [28] with an adaptation (the replacement of the typical critical frequency due to bending by the dilatational frequency) provided the best fit to test results. Proença et al. [14] performed standard airborne and impact sound transmission tests on full-scale homogeneous-core composite sandwich panels made of GFRP face sheets and PUR core, designed for residential building floors; the study included also numerical and analytical simulations for the assessment of airborne and impact sound insulations, respectively. The numerical and analytical simulations presented good agreement with the experimental data. Furthermore, results revealed an overall poor acoustic performance of those panels, pointing out the need for additional non-structural elements to fulfill the acoustic requirements set in building codes, for both airborne and impact sounds. Such need was also reported in the experimental study by Nurzynski and Nowotny [15], which assessed different strategies (plasterboard suspended ceilings, different types of floor coverings and floating screeds) to improve the acoustic performance of composite panels for both types of sounds. However, the study focused on non-structural composite panels (total thickness of 35 mm and 60 mm, with very low surface mass, 10–12 kg/m 2 ); the experiments presented in this study were not complemented with any analytical or numerical simulations. The EasyFloor composite sandwich panels present some similarities with previous solutions, namely with the homogeneous-core panel studied by Proença et al. [14]. However, the EasyFloor sandwich panels also comprise relevant differences, namely: (i) its web-core shape of the cross-section; and (ii) the use of a concrete top face sheet in the hybrid variant. These intrinsic differences and the envisaged application justify Fig. 1. EasyFloor sandwich panel variants comprising the snap-fit connection: a) all-composite and b) hybrid. M. Proença et al. Construction and Building Materials 438 (2024) 137195 3 the need to investigate the acoustic behaviour of this novel type of panels, by means of experimental tests, as well as to assess the validity of analytical models available in the literature to describe such behaviour. 3. Materials and test methods 3.1. Materials and specimen’s preparation For both all-composite and hybrid sandwich panels, two different types of geometries were considered for the test specimens: (i) panels with plan dimensions of 1.508 ×1.250 m 2 for the airborne sound reduction tests (Fig. 2); and (ii) panels with plan dimensions of 1.508 × 1.600 m 2 for the impact sound pressure level tests (Fig. 3). The specimens were composed of five segments of the EasyFloor panel, each with 0.30 m of width and length dependent on the type of test (1.25 m for the airborne sound tests and 1.60 m for the impact sound tests), adhesively bonded together along the side edges (this connection had an average adhesive thickness of 2 mm). The reference all-GFRP sandwich panel was composed of a PUR core with density of 70 kg/m 3 and GFRP face sheets and webs with thicknesses of 5.5 mm and 7.7 mm, respectively (average tensile modulus in the longitudinal direction of 31.4 GPa and 26.0 GPa, respectively). The GFRP-concrete hybrid sandwich panel had an additional 20 mm thick layer of C30/ 37 concrete placed over the top GFRP face sheet (average cylinder compressive strength of 42.3 MPa, average flexural tensile strength of 5.5 MPa). The all-GFRP and the hybrid sandwich panels mass per unit area was 41.7 kg/m 2 and 89.7 kg/m 2 , respectively. A two-component thixotropic epoxy adhesive (Sikadur®–31 EF) produced by Sika was adopted to perform the panel-to-panel connection. The epoxy adhesive was applied to the all-GFRP specimens and cured in laboratory conditions (temperature and relative humidity). For this purpose, the side surfaces (lateral edges) of the panels were firstly sanded with sandpaper, to increase the roughness of the surface, and then cleaned with acetone to remove any oils or debris, prior to the application of the adhesive. A base plate made of plywood was prepared with a plastic film on top of it (Fig. 4a), to prevent the adhesive from adhering to the base plate. Then, the two-component adhesive was mixed and applied to the side surfaces of the panel segments (Fig. 4b), and these were placed side-by-side on top of the base plate (Fig. 4c). After placing and carefully aligning all five segments of composite sandwich panels, two straps were placed transversely to the segments, along the web of an I-shaped GFRP profile, and firmly tightened (Fig. 4d). Finally, the excess of squeezed adhesive was removed. The hybrid panel specimens were produced following the same adhesive application methodology. After the 5-day cure of the adhesive, the specimens were prepared for concreting. Similarly to the bonded lateral faces, the top surface of the panels was firstly sanded with sandpaper, to increase the roughness of the surface, and then cleaned with acetone to remove any oils or debris. Then, a formwork was prepared with wooden boards around each panel, fixed by two threaded rods on each side (Fig. 5a). The Sikadur®–32 EF adhesive, produced by Sika, was applied on the (GFRP) top face sheet of the panel (Fig. 5b) to improve the bond between the GFRP and fresh (wet) concrete. Afterwards, the concrete was poured onto the mould (Fig. 5c). Lastly, the concrete was screeded with a flat board to achieve a smooth and flat finish (Fig. 5d). 3.2. Airborne sound insulation tests The airborne sound insulation tests were performed according to ISO 10140–2 [26]. The standard requires the use of two large-scale adjacent reverberation rooms, with a test specimen of about 10 m 2 plan area Fig. 2. Airborne sound reduction test specimen geometry – homogeneous-core sandwich panel (in mm). Fig. 3. Impact sound pressure level test specimen geometry – homogeneouscore sandwich panel (in mm). M. Proença et al. Construction and Building Materials 438 (2024) 137195 4 placed in between the two rooms. The chambers where the tests were performed consist of two contiguous rooms, the source and receiving chambers, with internal volumes of 111 m 3 and 122 m 3 , respectively; these rooms are separated by a high sound insulation wall with an opening where the test specimen was placed. As for the size of the test specimen, in this work, a smaller sized panel (1.5 m x 1.3 m) has been used, due to production and equipment limitations. Indeed, the standard also accounts for this possibility, although some care should be taken in the interpretation of the results in such cases, particularly when the wavelength of the bending wave is higher than half of the minimum size of the panel (0.625 m in the present case). The bending wavelength of an orthotropic sandwich panel (λb) depends on the wave direction and equals that of the isotropic equivalent plate for the case of a flexural wave propagating in the principal directions. For other directions, it is affected by the shear constant [29]. Fig. 6 illustrates the bending wavelength as a function of the direction and frequency, for the all-GFRP (Fig. 6a) and hybrid panels (Fig. 6b). It can be seen that below 1900 Hz in the all-GFRP panel, and 2500 Hz in the hybrid panel, the smaller dimensions of the panel may influence the sound reduction results. The modal behaviour of the tested specimens has also been analysed, and the lower order natural frequencies are shown in Table 1, considering an equivalent plate approach and simply supported edges. Table 1 confirms that only few vibration modes are estimated to occur in this frequency region in the all-GFRP and hybrid sandwich panels, considering an equivalent plate approach, as given in Eq. (1), R ω nxny= π 2  m √ n4 x b4 x Bx+n2 x b2 x•n2 y b2 y Bxy +n4 y b4 y By √(1) where ω nxny are the natural frequencies obtained, for modes (nxny), for a simply supported orthotropic plate with an equivalent thickness h; and Bx, By and Bxy correspond, respectively, to the flexural bending stiffness along the principal directions and the transverse shear stiffness. It can be seen that only few vibration modes are estimated to occur below 2000 Hz, both in the all-GFRP and hybrid sandwich panels. These results indicate that the experimental measurements may be affected by the modal behaviour exhibited by the panels at lower frequencies. The following equipment was used for the tests: (i) a data acquisition device, Symphonie 01 dB; (ii) two ½” microphones, GRAS model 46 A; and (iii) a sound source Brüel & Kjær, model OMNIPOWER 4292. The acquisition and processing of the signal was performed using dBBATI32 software that converts the sound pressure level directly into the frequency domain. The positions of the microphones (labelled 1–5) and the sound sources (labelled F1 and F2) in the acoustic source chamber are shown in Fig. 7 and the five adopted positions of the microphones for the receiving chamber (labelled 1–5) are presented in Fig. 8. These microphones were located at least 0.8 m apart from the wall to mitigate closeto-wall effects, following normative references of ISO 10140–4 [30]. Measurements were performed considering two repetitions for each combination of: (i) source position; (ii) microphone position in the source room, and (iii) microphone position in the receiving room. This involved a total of 100 measurements for each airborne sound pressure test. Fig. 9 shows the receiving and source rooms with the all-GFRP and hybrid sandwich panels placed on the test opening. Regarding the hybrid sandwich panel, the specimen was placed with the concrete layer facing the source room. After placing the specimens inside the opening, compressed mineral wool was introduced around the specimens to mitigate any gaps. Furthermore, a thin line of silicone was applied to both sides of the contour of the specimens, and left to cure for 24 hours. The panels can be considered to be simply supported along all edges, as small displacements and rotations were possible within the compressed Fig. 4. Preparation of the acoustic test specimens: a) base plate; b) application of the adhesive on the edge of one panel segment; c) placement of the panel segments side-by-side; d) tightening of the straps around the five panel segments. M. Proença et al. Construction and Building Materials 438 (2024) 137195 5 a) Panels with lateral formwork b) Application of the GFRP-concrete interface adhesive c) Pouring of the concrete d) Screeding of the concrete Fig. 5. Preparation of the hybrid sandwich panel. Fig. 6. Bending wavelength spectra for different angles of the wave with the longitudinal direction: a) all-GFRP sandwich panel; b) hybrid sandwich panel. Table 1 Estimated natural frequencies for the all-GFRP and hybrid sandwich panels. Panel Natural frequencies (Hz) All-GFRP f 1,1 254.1 f 1,2 438.9 f 1,3 821.5 f 2,1 907.7 f 2,2 1016.6 f 2,3 1285.1 f 1,4 1383.8 f 2,4 1755.5 f 3,1 2010.4 f 3,2 2094.6 Hybrid f 1,1 284.8 f 1,2 541.1 f 2,1 985.4 f 1,3 1052.4 f 2,2 1139.2 f 2,3 1518.1 f 1,4 1794.0 f 2,4 2164.4 f 3,1 2173.6 f 3,2 2289.9 M. Proença et al. Construction and Building Materials 438 (2024) 137195 6 mineral wool with silicone seal around the panels. According to ISO 10140–2 [26], the sound reduction indices in 1/3 octave bands (R) are calculated according to Eq. (2), R=L1−L2+10•log10 S A(2) where L1 is the energy average sound pressure level in the source room, in dB; L2 is the energy average sound pressure level in the receiving chamber, in dB; S is the area of the test specimen, in m 2 ; and A is the equivalent sound absorption area in the receiving chamber, in m 2 . The energy average sound pressure levels were calculated considering Eq. (3), as prescribed by ISO 140–4 [31], L=10 •log10(p2 1+p2 2…+p2 n n•p2 0)(3) where p1,p2, …, pn are the root-mean-square pressures at n different microphone positions in the room and p0 is the reference air pressure, equal to 2 ×10−5 Pa. According to EN 12354–1 [32], the standardized level difference (DnT) and the normalized level difference (Dn) can be calculated according to Eqs. (4) and (5), respectively, DnT =L1−L2+10•log10 T T0 (4) Dn=L1−L2+10 •log10 A A0 (5) where T0 is the reference reverberation time for dwellings of 0.5 s), and A0 is the reference absorption area, taken as 10 m 2 . The determination of the reverberation time was performed according to ISO 354 [33] by placing the sound source in the receiving chamber in two different positions and considering five microphone positions in the receiving chamber (those depicted in Fig. 8). For each combination of sound source and microphone positions, two repetitions were considered. The reverberation times were calculated based on the average of all these measurements for each 1/3 octave frequency band ranging from 50 to 5000 Hz – the results obtained are presented in Table 2. According to ISO 10140–5 [34], the reverberation time (T r ) of the chamber should be higher than 1 s and lower than 2 •(V/50) 2/3 =3.62 s (V is the volume of the chamber, 122 m 3 ). Only for the 50 Hz and 63 Hz frequency band the reverberation time exceeds the normative limit. However, these frequency bands (plus the 80 Hz band) were not included in the airborne analysis of sound reduction indices. According to Sabine’s equation, Eq. (6), the reverberation times can be related with the equivalent absorption area of the chamber as long as the sound absorption coefficient ( α ) is lower than 0.1: Tr=0.161 •V A(6) The equivalent absorption areas for each 1/3 octave frequency band are presented in Table 3. The background noise was also registered in both chambers, but no correction to the results was needed, as the difference between the background noise pressure level and the pressure level at each chamber was far greater than 15 dB (maximum difference value for which ISO 10140–4 [30] requires a background noise correction). 3.3. Impact sound insulation tests The impact sound pressure level tests performed according to ISO 10140–5 [34] require large acoustic chambers with standardized dimensions. Despite that, similar tests have been performed on a reduced test chamber (with an internal volume of 2.73 m 3 ) and results obtained from both standardized tests and in-situ measurements (two contiguous rooms of consecutive levels) for other types of specimens were very similar [25],[35]; thus giving confidence on the reliability of such test setup, whose details can be found in the referred works. In the tests, the specimens were placed on top of the small-size test chamber, and were supported along the 10 cm thick chamber walls. A Fig. 7. Position of the microphones (1–5) and sound sources (F1 and F2) in the source chamber (average height of 3350 mm) (dimensions in mm). Fig. 8. Position of the microphones (1–5) in the receiving chamber (average height of 3150 mm) (dimensions in mm). M. Proença et al. Construction and Building Materials 438 (2024) 137195 7 resilient cork layer was placed between the panel and the chamber walls to reduce flanking transmissions. After placing the panel on top of the chamber, a thin silicone line was applied to the chamber-panel interface in order to guarantee the sealing between the two elements, thus further reducing flanking transmissions. The silicone was left to cure for a period of 24 hours. The following equipment was used for the impact sound pressure level tests (Fig. 10): (i) a 4-channel data acquisition device, from National Instruments, model NI USB 4431; (ii) three ½’’ microphones, from GRAS model 46 A; and (iii) a Brüel & Kjær type 3207 tapping machine. The acquisition and processing of the signals during each test was performed using MATLAB software [36]. Inside the chamber, and following the provisions of ISO 10140–4 [30], the number of microphone positions was equal to the number of tapping machine positions. Five positions, labelled from A to E, are presented in Fig. 11 and were chosen with different distances to each orthogonal face of the chamber to minimize close-to-wall effects. However, the minimum distance of 0.7 m to any of the chamber boundary elements (defined in the standard) was not fulfilled, as the chamber dimensions were too small. Also, three different orientations of the tapping machine were considered for each position: (i) 0◦(aligned with the webs of the panels); (ii) 45◦(forming a 45◦angle with the webs Fig. 9. Airborne sound transmission tests of the EasyFloor panels: a) receiving room view with the all-GFRP panel; b) source room view with the all-GFRP panel; c) receiving room view with the hybrid panel; d) source room view with the hybrid panel. Table 2 Reverberation times of the receiving chamber per 1/3 octave frequency band. f [Hz] 50 63 80 100 125 160 200 250 315 400 500 T [s] 4.65 4.04 3.31 3.26 2.63 2.81 2.64 2.54 2.54 2.53 2.41 f [Hz] 630 800 1000 1250 1600 2000 2500 3150 4000 5000 T [s] 2.44 2.49 2.51 2.61 2.62 2.55 2.36 2.19 1.91 1.67 Table 3 Equivalent absorption area of the receiving chamber per 1/3 octave frequency band. f [Hz] 50 63 80 100 125 160 200 250 315 400 500 A [m 2 ] 4.20 4.84 5.89 5.99 7.41 6.95 7.41 7.68 7.69 7.72 8.09 f [Hz] 630 800 1000 1250 1600 2000 2500 3150 4000 5000 A [m 2 ] 8.01 7.85 7.76 7.49 7.46 7.65 8.28 8.93 10.24 11.72 M. Proença et al. Construction and Building Materials 438 (2024) 137195 8 of the panels), and (iii) 90◦(perpendicular to the webs of the panels). For each tapping machine position, measurements were performed for the five different microphone positions. In total, 75 measurements were performed for each test (3 tapping machine orientations times 5 tapping machine positions times 5 microphone positions). Each measurement was taken for a period of 10 s, longer than the minimum of 6 s stated in ISO 10140–4 [30]. The pressure (p) was used to calculate the impact sound pressure level (Li) using Eq. (7): Li=20 •log10( p p0)(7) After determining the energy average sound pressure level (L), according to Eq. (3), the normalized impact sound pressure level was then calculated using Eq. (8), Ln=Li+log10 A A0 (8) where Li is the impact sound pressure level in dB and A0 is the reference equivalent absorption area in m 2 . Corrections were introduced to reduce the contamination of the measured sound pressure level inside the chamber by airborne transmission following the procedure described in [25]. Indeed, it is known that in impact tests airborne transmission through weak elements of the test chamber may influence the final results; in the present case, these elements include the lateral door and also the tested floor specimen. For such correction, two microphones were placed outside the chamber to measure the outside pressure level: (i) one in front of the chamber door, at about 1 m of distance and aligned with the center of the door, and (ii) another one between the top of the slab and the ceiling (about 0.5 m from the top of the slab). The microphones that were placed outside the chamber and the tapping machine are presented in Fig. 10a and Fig. 10b, respectively. The sound pressure level due to the airborne sound generated by the impact source that crosses the floor and door elements (Li_airborne) was calculated from the energy average sound pressure level measured in front of the door (Lout_door) and the floor (Lout_floor), according to Eq. (9), Li_airborne =10 •log10⎛ ⎝ 1 2⎛ ⎝10Lout_door −Ddoor 10 +10Lout_floor −Dfloor 10 ⎞ ⎠⎞ ⎠(9) where Ddoor and Dfloor are the uncorrected airborne sound insulation of the door and floor, respectively. If Li−Liairborne <10 dB, then Li is replaced by Li_impact that corresponds to the sound pressure level due to impact sound, calculated according to Eq. (10). Li_impact =10 •log10⎛ ⎝10Li 10 −10Li_airborne 10 ⎞ ⎠(10) The correction due to background noise is performed in the same way as described for the airborne sound tests. The effects of a typical floor covering on the impact sound pressure level were also analysed for both types of sandwich panels – all-GFRP and hybrid. For this analysis, a wooden floor covering composed of wood planks with thickness of 15 mm and mass of 10.5 kg/m 2 , supported on a resilient cork layer, was applied to the top face of the specimens. The resilient cork layer had thickness of 5 mm, mass of 172 kg/m 3 and dynamic stiffness of 190 MN/m 3 [35]. Then, the impact sound pressure level test procedures were repeated for the specimen Fig. 10. Impact sound pressure level test setup: a) microphones outside the acoustic chamber; b) tapping machine. Fig. 11. Position of the microphones (A to E) inside the acoustic chamber (height =1500 mm) (dimensions in mm). M. Proença et al. Construction and Building Materials 438 (2024) 137195 9 configuration with resilient floor covering. These tests allowed to calculate the reduction of impact sound pressure level (ΔL) according to EN ISO 12354–2 [37], as per Eq. (11), ΔL=Ln0−Ln(11) where Ln0 and Ln are the normalized impact sound pressure level (in dB) in the absence and presence of floor covering, respectively. Firstly, the all-GFRP sandwich panel was submitted to an impact sound pressure test without any floor covering (Fig. 12a). Then, the resilient floor covering was applied to the top face of the panel (Fig. 12b) and the test was repeated. 4. Results and discussion 4.1. Airborne sound reduction Fig. 13 presents the one-third octave frequency spectrum of the experimentally determined airborne sound reduction of both types of EasyFloor web-core sandwich panels. Also, the results of a similar type of composite sandwich panel, but with homogeneous core (RehabGFRP), are presented for comparative purposes. The RehabGFRP panel had an overall thickness of 134 mm, comprising two identical GFRP face sheets with 7 mm of thickness, and a rigid PUR foam core with thickness of 120 mm and density of ρ c =87 kg/m 3 , presenting a weight of 37.2 kg/ m 2 . It is worth referring that this latter panel was tested in a normalized full-scale chamber, with 10 m 2 of plan area, and therefore experimental results are less affected by the floor modal behaviour than those obtained for the EasyFloor. Further details about the characterization of the acoustic behaviour of the RehabGFRP panel can be found in [38]. The overall shape of the airborne sound reduction curves of the EasyFloor panels are quite similar, only showing differences for frequencies lower than 630 Hz. For the lowest frequency bands, the hybrid panel presents higher airborne sound reductions of about 5–6 dB when compared to the all-GFRP panel. It can be assumed that the composite sandwich panel controls the shape of the sound reduction curve, with the concrete top layer only increasing the noise reduction for frequency bands up to 630 Hz. It is also relevant to analyse the dips in the sound reduction curves, as they are usually related with resonant behaviour phenomena. The all-GFRP panel shows a first dip from 250 to 500 Hz, while the hybrid panel seems to present a double dip along a wider frequency range, from 250 to 800 Hz. Data given in Table 1 shows that the lowest values of modal density are obtained for the one-third octave frequency bands of 250 and 400 Hz for the all-GFRP panel, whereas for the hybrid panel the lowest values occur at 315, 500 and 1000 Hz, which explains the observed experimental behaviour. The all-GFRP panel also presents the initiation of a dip in the 3150 Hz frequency band. Regarding the RehabGFRP panel, a first dip occurs at the 160 Hz frequency band, with a second and third dips occurring at 1000–1250 and 2500 Hz, respectively. The differences on the overall shape of the airborne sound reduction spectra of the EasyFloor and RehabGFRP panels might be related with the test methods and size of the samples, together with the presence of GFRP webs on the EasyFloor panels, as these vertical (shear) strengthening elements of the core entail a significantly diverse structural system, which responds differently to acoustic waves. The application of the rating method described in EN 717–1 [39] yielded a weighted sound reduction index R w (C, C tr ,C 100–5000 , C tr100–5000 ) =38 (-1, −5, −1, −5) dB for the all-GFRP sandwich panel. For the hybrid sandwich panel, the same rating method yielded a R w (C, C tr ,C 100–5000 , C tr100–5000 ) =41 (-1, −3, 0, −3) dB, which is in agreement with the expected improvement after doubling the floor mass. The EasyFloor panel’s weighted sound reduction indexes compare to a much lower value of the RehabGFRP of R w =32 dB, which has a slightly lower mass per unit area than the all-GFRP EasyFloor panel. 4.2. Impact sound insulation tests Fig. 14 presents the experimentally measured one-third octave spectrum of the impact sound pressure level for the all-GFRP sandwich panel with and without a resilient floor covering. Both specimens present a similar trend throughout the frequency spectrum, which shows how the panel governs the noise transmission from the tapping machine to the interior of the chamber, as expected. The addition of the resilient Fig. 12. Impact sound pressure test: a) all-GFRP sandwich panel (no covering) with tapping machine; b) all-GFRP sandwich panel with resilient floor covering; c) hybrid sandwich bare panel. Fig. 13. Airborne sound reduction of the EasyFloor sandwich panels and the RehabGFRP panel. M. Proença et al.