Internships

Gabriela Czuprynska

Universidade de Coimbra, under the supervision of Luis Godinho

Sound-absorbing materials enhance acoustic comfort by controlling reverberation and reducing noise levels. Porous materials like open-cell foams excel at absorbing medium to high frequencies but are less effective at lower frequencies, while resonant structures such as perforated panels offer durability and customization. Such panels normally use simple perforation shapes.
This study focuses on improving sound absorption by exploring panels with varied internal geometries such as tapered, hourglass, inverse-hourglass, perforations with rounded bubble-shaped extensions, and perforations with embedded resonant structures. The influence of rounding the edges and filling the air gap with porous materials is also investigated. An analytical model based on the transfer matrix method and a finite element model are implemented to analyse the systems. The analytical results show good agreement with the FEM results; however, an angle limitation in a tapered perforation case is identified, which can be due to the 1D simplification assumed in the TMM. FE models are used for examining the structures with embedded resonators. Parametric studies are performed to identify the influence of varying geometrical parameters on absorption.
A design method based on optimisation of the geometrical characteristics of a tapered panel is proposed. Three optimal samples with tapered perforations and one with an embedded torus-shaped resonator with arbitrarily chosen dimensions were 3D printed. The samples were experimentally tested using the procedure presented in ISO 10534-2. A good agreement between the predictions and measurement results was found.

Johana I. Dominguez

Arbane Groupe, under the supervision of Damien Jacquet

Sound spatialization can be achieved by making use of different panning techniques. NESS allows the user to model virtual sound sources based on a combination of DBAP and WFS, which assigns specific gains and delays to speakers. The software was used to design snapshots for a listening experiment conducted at the École Centrale de Nantes amphitheatre. The experiment design considered three different loudspeaker configurations and five different virtual source positions using three different auditory stimuli. The listeners were asked to sit at six different positions on the amphitheatre’s left or right side.
In this article, we show the results of the performance of two sound localization prediction models: TDM and PEM. When compared to real sound perception data from the listening experiment, we evaluated the predicted (models) or perceived (experiment) direction of the virtual source, width, and confidence.

Nazan Erdogan

GeoEND laboratory (Gustave Eiffel Univ.), under the supervision of Odile Avraham and Pierric Mora

In the domain of concrete structures, the Nonlinear Coda Wave Interferometry (NCWI) technique emerges as a potent tool for crack detection and structural assessment. Departing from conventional methods, NCWI employs ambient noise for passive pumping, aiming to replace active pump sources and streamline experimental setups. This study aims to explore the feasibility of ambient noise as a passive pump source for NCWI, leveraging its sensitivity to heterogeneous medium changes to enhance concrete health monitoring and infrastructure assessment. Through meticulous data analysis, intricate behaviors driven by factors like temperature fluctuations, pump levels, and traffic noise are unveiled. Visual representations offer insights into velocity variation, decorrelation coefficient, and other dynamics, providing valuable information for crack detection and structural health evaluation. The study further dissects data to uncover nuanced patterns within NCWI measurements, distinguishing behaviors between crack and uncracked zones. Notably, correlations emerge between traffic noise peaks and variations in key parameters, warranting rigorous investigations into potential artificial influences.
This research underscores the potential of Nonlinear Coda Wave Interferometry, particularly in the context of passive pumping through ambient noise, as a promising avenue for precise crack detection and comprehensive understanding of concrete structures. This approach paves the way for advancements in nondestructive testing and infrastructure health monitoring.

Jose Karotte Sani

Institute for Hearing Tech. and Acoustics (RWTH Aachen Univ.), under the supervision of Lukas Aspöck and Pierre-Olivier Mattei

Simulation of sound propagation that is both physically reliable and perceptu- ally plausible is a crucial requirement in the fields of room acoustic simulation and (real-time) auralization of Virtual Acoustic Environments. While the basics of sound propagation in indoor and outdoor spaces are widely investigated, simulation models are still not able to reproduce the acoustic perception of complex scenarios such as street noise with numerous static and dynamic sound sources and complex diffraction patterns. The thesis aims to utilise simulation tools developed at the Institute of Hearing Technology and Acoustics, RWTH Aachen University, to achieve a plausible simulation of outdoor sound propagation. Issues related to complex acoustic phenomena, such as scattering, absorption and edge diffraction, and optimal simulation parameters are being investigated.
In this study, a hybrid approach in Geometric Acoustics for room acoustic simulation techniques is adapted for outdoor propagation simulation. A combination of physically based synthesis of sound models and anechoic recording is used for source modelling. Propagation in an urban environment is simulated and auralized alongside a visual counterpart. The results of the auralization analysis have been examined, and suggestions for further enhancements have been proposed.

Michal Kubicki

Research Institute for Integrated Management of Coastal Areas (IGIC, Polytech. Univ. Valencia), under the supervision of Rubén Pico and Javier Redondo

This research delves into the intriguing realm of wave propagation within time-modulated metamaterials, leveraging the FDTD method as the simulation tool of choice. It focuses on 1D simulations of wave propagation through a medium with time varyng bulk modulus. The analysis is undertaken by examining the generation of forward and backward waves, their spectral analysis using FFT, power distribution and insertion gain, nature of periodicity including dispersion and band diagrams using plane wave expansion and space-time diagrams showing the evolution of the waveforms in time.

Marco Ribera

Amplitude Acoustics, under the supervision of Rui Ribeiro and Paulo Amado Mendes

The PU probe technique allows in-situ surface impedance measurements of porous materials with minimal sample preparation and measurement setup. However, this technique comes at the cost of multiple additional factors, which can influence the results if not carefully addressed. In this work, a sensitivity analysis is performed for two different materials, Rockwool and Melamine foam, aiming to provide a set of guidelines on how to perform measurements with the PU probe technique regarding sample size, sound field model and probe location. Below 800 Hz, the influence of these factors proved to be significant. Nonetheless, results showed to be accurate at higher frequencies, yielding errors smaller than those obtained by the impedance tube, in reference to the Johnson-Champoux-Allard (JCA) and the Delany-Bazley-Miki (DBM) equivalent fluid models.
The PU Probe technique was then compared with the current standardized procedures by converting all measured normal incidence sound absorption data into random incidence absorption coefficients of infinite lateral dimensions. This was achieved by a model fitting procedure applied to both Impedance Tube and PU Probe normal incidence measurements. The model fitting procedure enabled the inverse estimation of the main non-acoustical macroscopic parameters of the materials employing the JCA and the DBM models, finding good agreement with the macroscopic parameters determined through direct methodologies. Furthermore, this procedure enabled obtaining broadband normal and diffuse field sound absorption coefficients from the PU Probe measurements. Excellent agreement was found between all measured and reference curves for both materials at normal incidence. In a diffuse field, despite the non-diffuseness found in the measurement chamber, the measured absorption after a size correction was found to be oscillating around the reference curve, from which good agreement between the Impedance Tube and the PU Probe techniques was found.

Rodrigo S. Motta

Laboratory of Mechanics and Acoustics (Aix-Marseille Univ., CNRS, Central Med.), under the supervision of Régis Cottereau and Cédric Bellis

In various branches of engineering, it is essential to rapidly and non-destructively examine whether a structure possesses a specific characteristic, such as a defect, like an inclusion with distinct properties or a crack. For example, ultrasonic waves are commonly used to inspect gas ducts to detect cracks that could cause failure. In more complex scenarios, such as in petroleum engineering, specific structures like salt domes are actively sought after as they indicate the presence of rock formations that may contain oil reservoirs. This project aims to explore the potential of neural networks in assisting with classifying structures and determining whether they possess a given feature or not. One of the focuses is on understanding the optimal approach for inputting data into the neural network, exploring whether time series data should be directly fed without preprocessing or if preliminary steps can enhance the process. To achieve this, a collection of 2D computational simulations was generated to enable investigation, utilizing both defect-free and defective rectangular samples. Excitation and measurements were executed on the sample surfaces, similar to methods found in non-destructive evaluation. Subsequently, an ANN was created, having its performance, learning, and robustness analyzed.
This project took place in the Laboratoire de Mécanique et d’Acoustique in Marseille, France, and it had financial support from the Centre National de la Recherche Scientifique (CNRS). In addition to the building facilities of the laboratory, it was provided access to the Aix-Marseille University Mésocentre, a high-performance computing server, which contributed to more efficient calculations in this project.
In this report, the scientific context will be presented initially, followed by the description of dataset construction and the development of the artificial neural network (ANN). In the subsequent section, an analysis of the learning process and robustness will be conducted.

Dmitry Solodov

ReceNDT GmbH (Linz, Austria), under the supervision of Hubert Norbert

Additive manufacturing (AM) is an advanced fabrication process based on the sequential layering of materials, which is driving a transformative shift in modern manufacturing by creation of the components with enhanced geometrical complexity, reduced material waste, and improved structural integrity. AM of metal components, specifically using Wire Arc Additive Manufacturing (WAAM), involves heating of the metal to its melting point, leading to alterations in the material's chemical structure and creation of structural defects. Consequently, Non-Destructive Testing (NDT) for these materials requires both chemical analysis and physical evaluation through new non- traditional NDT techniques. This master's thesis investigates the suitability and precision of Laser- Induced Breakdown Spectroscopy (LIBS) for monitoring chemical defects in the produced material, while Laser Ultrasound (LUS) is employed to examine structural defects. The research aims at exploring the applicability and accuracy of LIBS and LUS as complementary approaches in the NDE of metallic samples. This thesis comprises four chapters, each addressing different aspects of the experimental methodology for NDE of metal components produced by the WAAM technique.

Nadeen Ayyash

DICAM - Structural Mechanics (Univ. Bologna), under the supervision of Antonio Palermo

The objective of this study is to investigate how installing a metasurface would affect Scholte waves. A Scholte wave is a type of surface wave that propagates along a solid-fluid interface. Hence, it exists due to the interaction between the solid and fluid. It exhibits a sinusoidal displacement pattern, and it is a dispersive wave. Its dispersion behaviour is characterized by a frequency-dependant phase velocity; hence the different frequencies of the wave will travel at varying speeds.
A metasurface, is a structured material consisting of an array of oscillators engineered to control and manipulate surface waves either Scholte waves or Rayleigh waves by altering their propagation properties at subwavelength scales. The research body consisted of two parts, the dispersion analysis, and the harmonic investigation. Each part consisted of the same four scenarios: the Rayleigh waves, the Rayleigh waves interacting with a metasurface, the Scholte waves, and the Scholte waves interacting with a metasurface.
The first part, initially, the dispersion relation [Frequency vs. Wavenumber] for Rayleigh waves on a free solid surface showed linearity. However, the instalment of an oscillator array to the surface, namely, a metasurface, showed a notable difference to the dispersion relation. At lower frequencies, linearity persisted until the resonance frequency, beyond which wavenumber increased while frequency remained fixed.
After that, the dispersion relation for Scholte waves on a free solid surface showed nonlinearity in the lower frequencies, and linearity in the higher frequencies. This is when the base difference between Rayleigh waves and Scholte waves was observed. However, the instalment of a metasurface on the solid-fluid interface, affected the Scholte waves dispersion relation almost as much as it affected the Rayleigh waves dispersion relation. A key difference between the two dispersion relations was how the two waves reacted when they encountered the bandgap.
The second part, later, a frequency domain analysis to investigate the wave phenomena validated the dispersion plots findings, ensuring their coherence. The study additionally provided evidence of wave attenuation after the instalment of a metasurface, reaffirming the metasurface principle and verifying the emergence of a bandgap in Rayleigh wave-metasurface interactions.

This internship has led to the research article
Zeighami Farhad, Quqa Said, De Ponti Jacopo Maria, Ayyash Nadeen, Marzani Alessandro and Palermo Antonio Elastic metasurfaces for Scholte–Stoneley wave control, Phil. Trans. R. Soc. A.38220230365, 2024.

Hossein Heidrai

Silicon Austria Labs GmbH, under the supervision of Javad Abbaszadeh

This study consists of two parts. In the first part, PZT Piezoelectric Micromachined Ultrasonic Transducers (pMUTs) were studied under the application of DC voltages. It was shown that their output sound pressure level was improved by 10.7dB and their mechanical bandwidth increased by 53 per cent under a 20-volt DC bias. The second part of the study focuses on the performance and application of AlScN pMUTs for object detection. An elaborate characterisation process was performed on a set of pMUTs and the most suitable one for air-coupled object detection was selected. The selected pMUT was used with the Synthetic Aperture Focusing Technique (SAFT) for object detection.
This thesis starts by introducing the environment in which the work was developed. Then pMUTs will be introduced and their application for object detection will be discussed. Next, The major design elements of pMUTs will be elaborately discussed. Then, the results of both parts of this study will be presented. In the end, the conclusions and perspectives will be presented with some points for future work.

David P. Ortega

4a Manufacturing GmbH, under the supervision of Patrick Hergan, Paulo A. Mendes (Univ. Coimbra) and Domenico Foglia (Foliumtec)

Diaphragm materials for accurate sound reproduction in electrodynamic speakers have been a central research topic since the loudspeakers were created. Nowadays, with the new improve- ments in designs, manufacturing, and engineering, composite materials are used to outperform traditional materials and improve acoustic radiation. Therefore, assessing their mechanical and acoustic properties is essential to prove their advantages and make good product development and manufacturing decisions.
This work is embedded in the context of composite materials and the main mechanical proper- ties directly responsible for their performance, such as the bending modulus or the loss factor. The main goal is to be able to measure such properties with an experimental setup relying on the impulse excitation technique and the free vibrations of a cantilever beam concept, which is described and compared to other existing setups. The setup is proven to be valid for test- ing specimens with low thicknesses and already provides results with less than a 10% difference compared to the Klippel MPM setup. Furthermore, as dealing with layered composite materials often involves using viscoelastic layers, the mechanical properties of these materials are believed to vary depending on the frequency of excitation. Thus, an additional step is taken to measure the dependency of the aforementioned mechanical properties on the frequency of excitation, as the acoustic application is always targeted. The development of this experimental technique is an ongoing work, only partially validated in this work. However, its problems are pointed out in the discussion section with a proposal for further improvements to the system and other possible techniques to measure these properties.
In the course of the experimental section, and as new testing setups are developed, COMSOL Multiphysics software package is used to prove the concepts before and after the setups are built, having a virtual equivalent model of the physical device which allows further improvement of these setups as well as comparisons and validations.

Maria Perez

Univ. Arizona, under the supervision of Sami Missoum

The sound of a trumpet depends on various parameters such as the mouth pressure and the dynamic properties of the lips of the player. The main objective of this project is to explicitly identify the regions of the parameter space (i.e., the playing conditions) to play in tune for a given range of notes.
In this work, a description of the sound generation mechanism in the trumpet is first pro- vided, where the lips act as a basic mechanical oscillator with a single degree of freedom. This oscillator is non-linearly coupled to the air column of the instrument by an airflow equation. All of these components combine to form a model consisting of three ordinary differential equations (ODEs) that must be solved simultaneously. By solving these equations, it becomes possible to obtain the pressure profile for various sounds generated by modifying some of the governing parameters in the acoustics of the trumpet. With the pressure profile it is possible to obtain information regarding the nature of the oscillation, like its frequency and periodicity.
Afterwards, a space of parameters is defined for study. Specifically, two parameters are deeply analyzed: the pressure in the mouth of the trumpet player and the frequency of their lips. By varying these parameters, a wide range of different notes can be produced, even without changing the configuration of the trumpet (i.e., without any of the pistons being touched).

Julio Rangel

CDM Stravitec, under the supervision of Reinhilde Lanoye and Marina Rodrigues, and Luis Godinho (Univ. Coimbra)

The purpose of this work is to propose and evaluate a methodology for drop impact testing on dry, lightweight floating floor systems in laboratory, based on the state of the art in the building industry, and targeting approximation to the boundary conditions of an application in situ. The procedure was implemented by means of a mockup of the Stravigym XP system, using experimental data from in situ tests as reference, and general simulations to validate the results.
It is studied how the weight’s mass, shape, height of the drop, and type of impact contact (single-point, double-point) influence the behavior of the system. Given the lack of an standard methodology, we experimented with different sources to generate insights on the system’s performance variations; the selection was made to mimic the most commonly dropped weights on gyms and training centers.
It is also addressed how to represent the existing loads of a gym (equipment, furniture, others) in the mockup. We handled by placing sand bags over its surface, which worked normally for the single-contact weights but, caused undesired interactions with the double-contact ones.
The analysis showed that the shape of the body is not relevant but the number of contact points and the stability of the bouncing are, to have results’ consistency when dropping from different heights. Regarding repeatability, the approach of keeping the looseness of the top layers is counterproductive due to the soon displacement of the top layers; such feature is key for the system under test, hence the current observations sum to its study.
The current methodology generated relevant information about the functioning of the floating floor on a laboratory environment, as well as some points of comparison to a real escenario, provided the limitations of using different slabs — making them two different systems. Despite their differences, they do offer insights to reach better correspondence in the future.

Naeem Ullah

CDM Stravitec, under the supervision of Patrick Carels and Reinhilde Lanoye, and Andreia Pereira (Univ. Coimbra)

The installation of floating floors is a widely used practice to mitigate airborne and impact noise in buildings. Floating floors help mitigate impact noise and vibrations to any adjacent room, especially the room below, by isolating the impact source in the source room. Generally, during the design, both the floating floor (FF) and structural floor(SF) are considered rigid, which means that the floating floor is considered as a single mass supported by springs on a rigid support (structural floor). The system is treated as a single degree of freedom (SDOF) system and the transmissibility curve is used to predict the sound and vibration reduction performance of the installed floating floor; in such a system the mitigation starts at frequencies higher than √2 times the resonance frequency. However, in a lot of cases the floating floor and the structural floor could be flexible, with considerable bending modes influencing the performance of the floating floor especially at low frequencies, making the SDOF approach to overestimate the system performance considerably.
This thesis investigates the low frequency performance of a flexible floating floor installed over a flexible structural floor by using FEM by means of COMSOL Multiphysics. The effect of the thickness of structural floor on floating floor is studied and the results show that it can affect the isolation performance of the floating floor due to the contributions of its bending modes.
The stiffness of the bearings(springs) affects the performance of the floating floor as it does for the simplified SDOF system and is addressed in the thesis. Increasing the stiffness shifts the SDOF mode as well as some of the bending modes of the floating floor to higher frequencies and vice versa. Increasing the stiffness negatively affects the performance by decreasing the insertion loss and vice versa, at frequencies above the SDOF mode.
The negative effects of the bending modes, especially of the floating floor on the performance of the floating floor system can be avoided by arranging the bearings in a specific combination. Three such combinations are studied. The coincidence of bending modes from floating floor and structural floor can result in prominent negative effects in the performance of the floating floor and hence should be avoided.
A simplified 3-DOF system is also developed, to predict the behavior of the floating floor with results in close agreements to those from COMSOL. This simplified system can be used to predict the behavior of the system with three important resonance modes, using MATLAB instead of COMSOL, allowing to save computational time.

Teodor Wolter

4a Manufacturing GmbH, under the supervision of Patrick Hergan and Domenico Foglia (Foliumtec)

The loudspeaker, a ubiquitous device integral to modern life consists of various components crucial to the success of its acoustic performance. Dating back to the early 20th century, the refinement of electrodynamic loudspeaker components has evolved significantly, driven by advancements in materials and signal amplification. The loudspeaker is made up of a conglomerate of parts whose collaborative goal is to transform an electric signal to sound. Among these components, the diaphragm plays a pivotal role in translating the voice coil’s movement into sound. This paper delves into the development of a novel aluminum-carbon-composite diaphragm material for high-fidelity (hi-fi) speaker applications. The research involved meticulous experimentation, encompassing material characterization, mechanical property testing, and the exploration of various fiber orientations. Additionally, parameter variation trials of an in-house thermal pressure mold were performed for elementary deformation of basic layups of the material. The study also explores the intricate trade-offs in diaphragm design, balancing parameters like stiffness, density, and damping. Through comparison between theoretical insights with empirical findings, this research contributes to the ongoing pursuit of high-performance loudspeaker diaphragms, working toward bridging the gap between acoustic theory and real-world applications.

Mohamed Chakib Drias

Saint Gobain Research, under the supervision of Matthieu Gallezot

This master thesis aims to improve the Mechanical Impedance Measurement (MIM) procedure used to evaluate the mechanical properties of plasterboards. The method involves using flexural resonances of a beam excited by a shaker to determine the equivalent Young's modulus and loss factor of gypsum boards. These properties can be frequency dependent. Finite element models are developed based on thin and thick plate formulations to enhance the identification of the equivalent Young's modulus. The standard BA13 plasterboard is used as a case study to examine the influence of the identification model on the measured Young's modulus and sound transmission loss simulations. The identification models are validated against measurements following the MIM procedure on the same panel, and the results are compared to small-scale testing data. The analysis is then extended to other plasterboard references, including standard and complex boards. The simulations reveal that certain assumptions in the MIM procedure (thin beam kinematic model and sample-shaker link boundary condition) are only valid under specific circumstances. The choice of kinematic model impacts the stiffness of the model and introduces variations in the eigenfrequencies. The use of a thin plate kinematic model results in overall stiffening of the model, while a thicker plate model shows softening at high frequencies. Simulating the link between the sample and shaker accurately proves challenging, especially for smaller samples. The fixed constraint assumption is no longer valid when using a thick plate model. Further investigations also show that added mass due to impedance head and attachment screw does not affect the impedance peaks used for characterization. The identification of Young's modulus for different plasterboard references confirms the influence of the identification model on the frequency-dependent modulus variation. Globally, using a softer kinematic model reduces the slope of modulus variation. The discrepancies between identification models become more significant with increasing board thickness. Simulations of sound transmission loss demonstrate that using an averaged value for Young's modulus following NF EN 16703 underestimates the critical frequency, particularly for thicker and more complex boards. Consistency between the identification model and the model used for simulating sound reduction index is necessary.

Marwa Al-Esseili

4a Manufacturing GmbH, under the supervision of Patrick Hergan, and Paulo Amado Mendes (Univ. Coimbra)

The development of high-quality loudspeakers is crucial for achieving superior audio ex- periences. Central to this technology are the diaphragms, such as tweeter and compression driver domes, which play a vital role in sound reproduction. Traditional materials, while effective, often face limitations in balancing acoustic performance and mechanical durability. To address these challenges, this thesis focuses on the exploration and optimization of com- posite materials, specifically carbon fiber reinforced polymers and aluminum composites, aiming to enhance both sound quality and material resilience.
Based on a combination of experimental research, analysis of secondary data, direct ob- servation and discussion with experts, this thesis investigates the impact of advanced com- posite materials on loudspeaker performance. Rigorous tests were employed, including the Impulse Excitation Technique to characterize mechanical properties such as Young’s mod- ulus and bending stiffness. Additionally, acoustical tests were conducted to assess breakup frequencies and analyze sound radiation behavior across the frequency range. The study also integrates principles from signal processing and materials science, deepen the under- standing of the relationship between material properties and acoustic behavior.
Key findings from this study highlight that pre-impregnated fiber simplifies the mate- rial preparation process, ensuring consistent quality. The anisotropic nature of carbon fiber plays a significant role in its acoustic performance. However, it’s crucial to acknowledge that multiple material properties and setup conditions can influence the measurements. For future research, isolating each factor will enable a more detailed comparison and under- standing of their individual impacts.

Fawad Ali

Laboratory of Mechanics and Acoustics, under the supervision of Cédric Maury, and Teresa Bravo (CSIC Madrid)

Low-frequency noise mitigation presents substantial challenges due to the inherently long wavelengths, which require bulky acoustic devices for effective solutions. Unaddressed, this noise detrimentally impacts both human and ecological systems, for instance noise emission from a boat engine not only disturbs nearby environment but also disrupts marine bioacoustics. Among strategies like active noise cancellation and dissipation devices, Acoustic Black Hole (ABH) technology has emerged as a novel solution. ABH phenomena, first observed in tapered elastic beams, exploit the reduction in flexural waves propagation speed to near-zero at the beam’s apex, a result of gradual spatial reduction in bending stiffness. This principle can be replicated in acoustic wave propagation by employing metamaterials to progressively reduce compressibility, thus enabling wave trapping with negligible reflection. While ABH offers potential for efficient HVAC and automotive mufflers, it introduces challenges including increased drag and limited performance in low frequency range.
The current thesis addresses these issues by incorporating an optimized microperforated panel (MPP) in the ABH structure to enhance dissipation across 200-800 Hz range, simultaneously reducing drag. Aeroacoustic performance in the presence of mean flow is also considered. The research is segmented into three key objectives: Firstly, drag analysis within introduced design parameters using Computational Fluid Dynamics (CFD) tools like COMSOL Turbulent Flow and ANSYS Fluent. Secondly, enhancement of low-frequency acoustic dissipation using the Finite Element Method. Finally, aeroacoustic evaluations and time-domain acoustic analysis under 30 m/s mean flow conditions are conducted. For drag calculations, the study employs solving conservation equations within a hexa-structured computational domain, applying fully coupled numerical schemes for efficient convergence. To account for circulating and jet flow, the Reynolds-averaged Navier–Stokes equations are resolved using the Realizable k − ϵ model. Similarly, the acoustic evaluations focus on the dissipation curve area, derived via frequency-domain linearized Navier-Stokes equations, accounting for visco-thermal effects. Also, the aeroacoustic impacts and time-domain acoustic analyses are obtained by using broadband noise source model and implicit time-domain CFD, with turbulence handled via Large Eddy Simulations (LES) and the Wall-Adapting Local Eddy (WALE) subgrid scale model, respectively. High-fidelity interpolations, primarily Second-Order Upwind except in high-gradient scenarios where Third-Order MUSCL schemes are applied to ensure numerical stability and accuracy.
Findings indicate that high perforation ratio adversely affects drag performance, while the acoustic dissipation patterns are significantly influenced by Helmholtz resonance and ABH effects. Aeroacoustic analyses reveal MPPs as primary contributors to noise, with distinct tonal peaks at 860 Hz due to resonant cavity interactions. In mean flow scenarios, the MPP-ABH model’s transmission loss peaks shift to higher frequencies with reduced amplitude and flattened curve, demonstrating the flow’s impact on acoustic behavior. An optimal ABH configuration with a 15 cm square channel, fifteen cavities each 9 mm wide, separated by 1 mm spacing, following profile (M5), and 0.5 mm thick MPPs of 0.7 mm diameter and 4.5 mm spacing, was identified, offering manufacturability and enhanced performance.

Viktoriia Boichenko

GEOMAR, under the supervision of Prof. Jens Greinert, Dr. Mario Veloso and Christian Kanarski

This master’s thesis focuses on developing a theoretical model for the acoustic frequency response of multi-bubble compounds, such as air bubbles or methane gas bubble seepage from natural reservoirs. The objective is to simulate and analyze the acoustic behavior of these compounds in water columns and to reconstruct the acoustic response of the received bubble signals. The research begins with a review of existing single- and multi-bubble acoustic response models, integrating these with dynamic bubble movement and multiple scattering effects between bubbles. Initial models are simplified and progressively refined based on empirical data from previous studies of Anderson and Thuraisingham formulas for the backscattering strength of a single bubble. Furthermore, the model is extrapolated for multiple bubbles involving the multiple scattering effect. Experiments is conducted in a water tank with a bubble generator to estimate multi-bubble responses, with data used to try to validate and adjust the theoretical model. The final model can be incorporated into existing simulation software, such as KiRAT, to demonstrate its application as a digital twin for underwater column simulations and for generating SONAR data to train AI systems.

Yu-Lin Lai

Tymphany Acoustic Technology, under the supervision of Jessica Wild, and Prof. Francisco Castells

Nowadays, there is a wide range of electroacoustic transducers on the market, making it challenging for products to stand out to consumers. This work focuses on the design and optimization of an automotive subwoofer which emphasizes features such as a shallow mount, excellent acoustic performance, and competitive cost.
The goal is to develop an 8-inch, 2 Ω subwoofer with a 1.5-inch motor, capable of handling around 150W of power. The target acoustic performance includes an sound pressure level (SPL) of 87 dB, a BL factor of over 7.6 (T·m), and an Fs of approximately 40 Hz. Initially, simulations using FINEMotor software were conducted to optimize the component design for good acoustic performance. Next, more advanced simulations were performed using COMSOL Multiphysics, including detailed sound pressure level (SPL) curves, impedance curves, BL curves, and Kms curves, taking into account the detailed geometry of the subwoofer. Afterward, all components were confirmed with suppliers to ensure manufacturability and feasibility. The Thiele-Small (TS) parameters were recalculated based on theory and cross-checked with the simulations to validate the results.
Once the optimal sample based on the simulation was built, its acoustic properties were measured using Audio Precision and KLIPPEL equipment. The experimental results are then compared to the simulation outcomes to analyze performance and guide the development of next versions.
In conclusion, the experimental results generally aligned with the simulations, achieving the targeted acoustic performance. However, a dip in the SPL curve was observed, which, upon investigation, is due to the basket design and definitely requires improvement. Future work will focus on enhancing the SPL and reducing Total Harmonic Distortion (THD), and progressing to power and reliability testing to ensure a high yield ratio for this automotive subwoofer.

Karan Manoj

IGIC Research Inst. for Integrated Management of Coastal Areas, under the supervision of Dr. Victor Espinosa

The design of an acoustic underwater communication and positioning system for an Autonomous Underwater Vehicle (AUV) is always a challenge, due to the restrictions imposed by sea water as a propagation medium. The attenuation of acoustic signals by sea water is a function of frequency and also the environmental properties of the acoustic channel. In shallow water channels, acoustic signals are reflected by the sea floor and sea surface, and so the same signal may be received multiple times with different arrival times and amplitudes. These aspects must be considered when designing signals to be used for communication and positioning purposes.
In the present work, various types of signals were studied for use in underwater communication and positioning systems. As a novelty, sine waves modulated with Maximal Length Sequence (MLS) envelopes were proposed. They offer the possibility of encoding information in the envelope, and identification of the emitting beacon if each beacon uses a different carrier frequency.
A theoretical comparison of signals was performed, on the basis of their ease of detection in noise, and Signal-to-Noise Ratio (SNR). The influence of signal amplitude and duration on these parameters was evaluated. In addition effect of reflected signals on the received signal strength was also studied.
An analytical model was developed to calculate signal propagation and losses in an underwater positioning system. It uses the method of Received Signal Strength Indicator (RSSI), which estimates the distance of an emitter by its signal strength. To account for losses due to the sensitivity of each device in the system, they were first calibrated separately, and their calibration curves were incorporated into the analytical model. The predicted signal strength of this model was compared to actual measurements of the strength of signals propagated in the sea.
It was found that although chirp signals were the best for communication theoretically, they did not propagate as well as the novel MLS carrier signals did. This is because MLS carrier signals have most of their energy concentrated at their carrier frequency, so they can travel larger distances without significant attenuation. In contrast, chirp signals have their energy spread evenly across a frequency range, so they are prone to greater attenuation with distance. It was also concluded that Time Stretched Pulse (TSP) signals were the most resistant to interference from reflected signals, so they can be used for positioning, which requires a stable and predictable decrease of signal strength with distance.
The model developed in this work could be used for the preliminary design of an underwater communication and positioning system. It allows different signals to be tested, and effect of each device on the system performance can be understood. A new device can easily be introduced in the model by including its calibration curve.

Simon Masson

IGIC Research Inst. for Integrated Management of Coastal Areas, under the supervision of Dr. Victor Espinosa and Noela Sánchez (CONICET)

In this work, we investigate the application of new generation broadband echosounders for the seabed characterization. Single beam and multibeam echosounders are well known and widespread for bathymetric studies and for the sea bottom characterization. However, there are other types of echosounders that have an alternative approach based in their broad frequency content and which application have been only scarcely studied for seabed classification and object detection inside the sediments. These broadband scientific echosounders are more commonly used for fisheries acoustics. In collaboration with the Patagonic studies centre CONICET of Puerto Madryn (Argentina) and the Universitat Politècnica de València in Gandı́a (Spain), we investigate new signal analysis techniques and algorithms for sediment characterization, as well as the detection of buried objects in soft sediments. We consider here different models for the underwater acoustic propagation and the interface with the seabed. We include several losses due to absorption and spreading in the water. Then, we introduce different concepts related to signal processing that apply to our echosounder signals.
In particular, we detail how to deal with broadband signals where frequency-dependent corrections need to be applied. We also develop the deconvolution inverse problem. This lets us compute the impulse response of the seabed, without the transduction from the echosounder nor the propagation in the water column. The information of the seabed lets us obtain a characterization of the sedimentation content of the seabed as well as a detection of buried objects. The different steps of the process are detailed in one example, and then presented for measurements obtained during the study, in the ocean and from a boat. We summarize the results of this study with openings and perspectives for further investigations.

Monisha Muralidharan Menon

GeoAzur, under the supervision of Dr. Anthony Sladen and Dr. Diego Mercerat

Instrumentation of the sea-bottom is complicated and not widespread. The inability of electro-magnetic waves to propagate beyond a certain depth has made continuous monitoring a challenge. The varying pressure and temperature with depth presents difficulties that are not easily handled by instruments. With the use of DAS technology, these problems are partly overcome. Researchers are closer to understanding submarine structures and the composition of the sediments that cover the DAS cable.
Distributed acoustic sensing(DAS) is a technology that makes use of telecom optic fiber cables to record seismic and other dynamic events in its vicinity. The anomalies in the cable caused during cable manufacturing allows back-scattering of optical signals. The strain gradient along the cable due to variation in pressure or temperature, aids the back-scattered signal in recording events and understanding the environment the cable is laid on. This makes DAS a robust technology which can be deployed in harsh environments. DAS is currently being used in monitoring ship traffic, tracking and movement of marine mammals among many applications.
Over the course of the master thesis, the aim was to analyse the seafloor off the coast of Marseille. The study is conducted using an 81.5 km long telecom cable that was deployed by Alcatel Submarine Networks(ASN) as part of the 2Africa project from Marseille towards Africa.
The project aims to bring internet connectivity to parts of Africa by extending optical fibers using a series of optical repeaters. The telecom cable was recording events as part of testing and research for 2 months beginning october 2022 before the commencement of Telecom operations. The analysis of the seafloor is performed using this data. Since the start of recording, there have been earthquakes, UXO explosions and quarry blast events, that have originated at large distances of up to 320 kilometres from the cable. The seismic or hyrdoacoustics waves of these events were captured by the cable based on the intensity of magnitude of the event. 10 such events along with 2 windows of ambient noise recordings were analysed. Considering that the events generate low frequency waves, the recordings are studied in the frequency range of 2 Hz to 20 Hz. The data reveals the behaviour of the seafloor and the cable to different types of excitation and offers information on the source and the medium. The structure of the seafloor composed of sedimentary basins offers insight into hazards that may occur during high magnitude earthquakes such as ground motion amplifications and soil liquefaction. These hazards are noticeable in regions containing thick sedimentary deposits where site effects dominate. Three sedimentary basins were identified along the length of the cable. These structures are imaged and analysed in the time and frequency domain. Auto-correlation functions and power spectrum density are used to understand its characteristics.

Hedi Merhaben

Laboratory of Mechanics and Acoustics, under the supervision of Dr. Zine Fellah, Dr. Erick Ogam, Dr. Cédric Maury, Abdellah Bouchendouka and Christian Meso

This thesis investigates the acoustic characterization of porous materials, focusing on determining high-frequency parameters such as tortuosity and porosity. Conducted at the Laboratory of Mechanics and Acoustics (LMA) in Marseille, the study utilizes ultrasonic transducers to analyze various foams provided by an automotive industrial group. The primary objective is to develop and validate methods for accurately estimating these material parameters, which are crucial for applications in fields such as automotive sound insulation.
The research employs both direct methods and inverse problem approaches to extract the desired parameters from the transmission of ultrasonic waves through the materials. Significant emphasis is placed on the sensitivity of these methods to various experimental conditions, such as the angles of insonification. The results indicate that the tortuosity parameter can be reliably estimated with reasonable accuracy, with deviations between developed methods and inverse problem results ranging from 4% to 11%, depending on the material’s attenuation properties.
However, challenges remain in accurately determining porosity. The results for porosity were inconsistent and often outside the expected range, suggesting that the parameter’s sensitivity or negligible influence in the transmission mode may require further investigation, potentially in reflection mode or using alternative experimental setups.
Overall, this study provides a promising approach to estimating tortuosity in porous materials but highlights the need for further research to overcome the limitations encountered in porosity estimation.

Gustavo Nunes Queiroz

SLAM Acoustique, under the supervision of Philippe Schmid, and Prof. Luis Godinho (Univ. Coimbra)

When constructing buildings near railways infrastructures, assessing the vibration levels within these structures is crucial to ensure both safety and comfort. As the number of such buildings continues to grow, it becomes increasingly important to have a method for evaluating ground-borne vibrations that minimizes computational time. This study begins with a review of the fundamental principles of vibration transmission and the impact of these vibrations on buildings. Various predictive engineering approaches, including the FTA, SBB, BAM, CSTB, and DB models are analyzed highlighting their strengths and limitations in different contexts. The primary objective of this study is to implement a simplified method for calculating ground-borne vibrations in building structures efficiently. It focuses on the development and application of a 2D Finite Element Method (FEM) model that incorporates soil-structure interaction (SSI) as a spring-damped system to simulate the dynamic response of buildings to railway-induced vibrations. A key aspect of this research is the use of real-world vibration data, used as the excitation input in the model, specifically measured vibration data from a site in Saint-Cloud, France, where a building will be constructed 20 meters away from a tram line. To enhance the accuracy of the model, two different approaches, based on the works of Lopez et al. and Auersch for calculating stiffness and damping, are employed allowing for a more nuanced representation of soil-structure interaction. The FEM model is applied to various structural scenarios, ranging from simple two-story buildings to complex multi-story frames. The results of these simulations are compared with those from FTA empirical method.

Valerie Ritter

Müller-BBM Building Solutions GmbH, under the supervision of Dr Eckard Mommertz, and Prof. Paulo Amado Mendes (Univ. Coimbra)

Echoes, which are the audible repetition of sounds resulting from strong delayed reflections, present a significant challenge in the acoustic design of multipurpose halls. These echoes can disrupt speech intelligibility and diminish the overall acoustic experience, necessitating effective prediction and mitigation strategies. This thesis investigates methods to predict and eliminate rear wall echoes in multipurpose halls by employing a combination of room and electroacoustic simulations.
The study begins with an in-depth analysis of echo perception, incorporating psychoacoustic principles and various echo criteria, including those proposed by Dietsch and Kraak [1], as well as advanced loudness-based methods by Frey [2].
Simulations under varying conditions, including different rear wall materials and reverberation times, are used to identify key factors influencing echo generation. These simulations are validated through in-situ measurements in a real multipurpose hall, ensuring the reliability of the findings. The research highlights the critical role of reverberation time and rear wall acoustic treatments, demonstrating how varying levels of absorption and diffusion can effectively mitigate echoes.
Furthermore, the study evaluates the effectiveness of different post-simulation methodologies, comparing results using both the well-known Dietsch and Kraak echo criterion and sophisticated loudness-based methods. The latter offers more accurate echo prediction by taking into account the subjective perception of sound, making it a superior tool for assessing unwanted echo disturbances.
Finally, the findings offer practical strategies for optimizing acoustic design in multipurpose halls, particularly through tailored treatments for rear walls. This work enhances the theoretical understanding of echo mitigation and provides concrete solutions that acousticians and engineers can implement in the design and refurbishment of multipurpose spaces using a loudness-based model for realistic echo prediction.

Stelios Tsampourakis

LIS-lab, under the supervision of Prof. Hervé Glotin.

Conventional techniques for monitoring insects such as trapping, sweep netting and others are time consuming and prone to human error. More recently, acoustic techniques for monitoring insect populations have been successfully applied for species detection, population estimations and mappings of their distributions, while additionally providing the advantage of non-invasive monitoring and greater detection distance compared to traditional methods. Although limited to insects that emit sound, these techniques have shown increased popularity, especially due to advances including the rise of autonomous recording units (ARUs) which have led to Passive Acoustic Monitoring (PAM) techniques that allow for the monitoring of ecosystems at large temporal scales. The amount of data generated by PAM have introduced difficulties, which are currently being addressed by the rapidly evolving field of Machine Learning (ML), with Deep Learning (DL) techniques showing the way forward. The recently emerged field of Ecoacoustics includes all of the above under its umbrella, by combining traditional bioacoustic knowledge with advanced methods, forming a multidisciplinary bond between ecologists, acousticians, engineers and data scientists.
In this work, an experimental study was conducted in order to assess the effectiveness of automatic detection algorithms and localisation techniques, by utilising five time-synchronised recording units with GPS/PPS synchronisation. The annotation procedure involved four different insect species of the Orthoptera order (grasshoppers) by utilising the Raven Lite software. The classification of insect mixtures was attempted with up to three concurrent vocalisers of different species (interspecific) by utilising the YOLOv5s Deep Learning classifier model based on the Convolutional Neural Network (CNN) architecture, achieving good detection accuracy. A pipeline was then developed for unmixing the vocalisations of multiple sources through sound source localisation, by utilising two-stage localisation methods and a hyperbolic algorithm, which calculates the time delays of arrival (TDOAs) between pairs of microphones from their cross-correlation functions. This was achieved through the OpenSoundscape (OPSO) framework and a custom approach, showing promis- ing preliminary results.
The developed pipeline offers the advantages of modularity and scalability, allowing for its seamless integration with future experimental setups involving different a number of microphones, sources or even involving more species, and could even possibly generalise to simultaneously involve multiple taxa. Such techniques can help in monitoring animal populations and potentially be used for estimating animal density and abundance, in which occurrences of mixtures of multiple sources is often the norm.

Silvija Zvirblyte

Threshold Acoustics, under the supervision of Scott D Pfeiffer.

Architectural acoustics plays a pivotal role in the design and functionality of concert halls, where the quality of sound significantly impacts the audience experience. The geometry and shape of a concert hall are primary factors in determining its acoustic performance. These elements influence how sound waves travel, interact with surfaces, and ultimately reach the listeners ears. By carefully considering the halls geometries and shapes, architects and acoustic engineers can create environments that enhance sound clarity, richness, and spatial distribution, ensuring that every seat in the house enjoys optimal auditory conditions.
Before diving into acoustical context of concert halls I would like to highlight that music venues are not only about an architecture perspective of the space, but about music architecture from itself. The essence of this statement lies in the understanding that there is no definitive ’right’ or ’wrong’ way to design a concert hall. Rather, its suitability is contingent upon factors such as the musical genre being performed, the arrangement of instruments, stylistic considerations of the era, and numerous other elements. This research endeavors to explore the realm of architectural acoustics, with a particular empha- sis on concert halls. The objective is to gain a comprehensive understanding of the complexity involved in the design and optimization of concert hall environments as well as understand the difference in sound perception depending on the concert hall design.
Firstly, it will conduct a review of existing literature on architectural acoustics, focusing on acoustic fundamentals, anomalies, treatments, and design. Furthermore, the research will delve into the calculation methodologies and simulation techniques employed in the analysis and design of concert hall acoustics. This includes a comprehensive exploration of computational methods and specialized software tools utilized for modeling concert hall geometries and simu- lating acoustic properties. The research will investigate various acoustic treatment options and applications aimed at optimizing the acoustic environment within concert halls.

Marcelo Alejandro Argotti Gómez

Wölfel-Gruppe (IMMI–dBEL), under the supervision of Dipl.-Ing. Janosch Blaul

The accurate prediction of low-frequency noise from open-air music events remains a long-standing challenge in environmental acoustics. Standard engineering methods, such as ISO 9613-2, treat sound sources as incoherent, works in limited octave band frequencies above 63 Hz and consequently underestimate diffraction, ground interaction, and phase-related effects that dominate below 250 Hz. This work addresses these deficiencies through a two-stage research program. First, a Morris screening sensitivity analysis was conducted on an ISO 9613-2 model implemented in the commercial software IMMI. The study, based on an outdoor concert case, revealed ground impedance and source directivity as the dominant parameters, while temperature and humidity exerted only minor influence. Discrepancies exceeding several decibels were observed between model predictions and field measurements across the 25 Hz–250 Hz band, underscoring the need for a more physically rigorous approach.
To that end, a three-dimensional wide-angle parabolic equation (3D WAPE) model was developed and implemented in Python. The solver employs an Alternating Direction Implicit scheme and incorporates detailed source directivity, atmospheric absorption, wind profiles, and ground impedance (Miki model with optional roughness corrections). Validation comprised a detailed confrontation with measurements from the outdoor concert campaign. The model reproduced spectral levels and spatial attenuation within±4 dB over the 25–250 Hz range, capturing forward-beam narrowing and interference patterns that elude conventional methods.
The results demonstrate that wave-based numerical modeling can substantially improve low-frequency noise forecasts for large-scale sound reinforcement systems. By bridging coherent source directivity with atmospheric and ground effects in a computationally tractable framework, this work lays the foundation for integrating advanced propagation tools into environmental noise assessment and regulatory practice.

Tiago Belletti Romero

GAMBA, under the supervision of Tony Lethuillier and Gautier Gillot

This thesis presents the development and evaluation of an alternative methodology for Rayleigh wave analysis within the framework of Multichannel Analysis of Surface Waves (MASW). The approach aims to improve computational efficiency by avoiding the traditional inversion process, which is typically computationally intensive, and instead focuses on direct curve fitting through an optimized procedure to identify and compare the five best-fitting dispersion curves. It also seeks to reduce operator bias, ensuring more consistent and objective subsurface model estimations.
The theoretical foundation of the method is based on Rayleigh wave propagation in layered, elastic media, including the modal dispersion phenomenon and its sensitivity to soil stratification. The method uses post processing tools as adapted moving windows techniques to account for dispersion related wave packet spreading, and bow filter to enable more precise isolation of the target dispersion mode. It follows with a mode-tracking algorithm that constrains results within velocity limits and avoids abrupt jumps. A database of analytically generated main modes is created by varying soil layer properties within realistic constraints, and the extracted mode is compared to this database using mean squared error to identify the best-fitting dispersion curve. This best fit provides soil parameters for subsequent velocity profile generation and validation. The approach is tested on four measurement cases, highlighting the influence of higher modes, possible mode jumping, and the limits of the database when layer inversion occurs, while generally showing strong convergence and consistent high-frequency alignment between measured and simulated curves.
The accuracy of the proposed method is assessed by validating the best-fit soil profiles it produces against measured transfer functions for four different MASW datasets, with results compared to those from the company’s current approach. For each case, the synthetic transfer function derived from the best-fit model is compared to field measurements across both frequency and distance domains. Results show mixed performance: in M1, both methods perform similarly; in M2, the new approach improves mid-frequency agreement at key distances; in M3, the lack of damping in the database limits accuracy, making the previous method superior, especially at higher frequencies; and in M4, the proposed method achieves a generally closer match across multiple frequencies and distances. Overall, the validation highlights the new method’s potential to match or exceed the current approach in certain scenarios, while also revealing specific limitations that could be addressed to improve robustness.
The proposed method offers significant improvements over the traditional inversion-based approach for dispersion curve analysis, notably reducing computation time and minimizing operator intervention. By not using the inversion process, it delivers results in under 90 minutes compared to several hours previously, and the automation of main mode tracking enhances consistency and repeatability. The flexibility of the approach allows operators to choose between five generated solutions or refine one further, combining the strengths of the company’s previous method that allows experienced operators to optimize complex measurement scenarios. Although the method does not directly addresses the non-uniqueness problem, it seeks to minimize its effects by obtaining five best-fitting curves for operator verification. However, this characterizes a limitation of the method, as well as the requirement of a laterally homogeneous soil, since the measurement relies on the reciprocity principle.

Andrea Carbajo Rebollo

Institut ∂’Alembert (Sorbonne University   CNRS), under the supervision of Pablo Abehsera Morell and Brian Katz

Spatiality in room acoustics defines the sense of width, depth, and distance of the listener within a sound field. Among many descriptors, two have proved robust and interpretable: Lateral Fraction (LF) and the Inter-Aural Cross Correlation Coefficient (IACC). This master thesis investigates the physical meaning of these spatiality parameters in diffuse sound fields. Both acoustic parameters are examined, seeing how these metrics interact and relate to spatial perception. Spatial distributions are modified through controlled variations in source-receiver positions, and spread reduction is evaluated using squash-per-axis contributions. Real room measurements are conducted to develop a method for spatial characterization and used for parameter correlation and analysis. A new formulation for IACC is proposed using full 3D room impulse responses, aiming to eliminate the need for binaural recordings or external Head Related Transfer Function (HRTF) convolutions for channelsreconstruction.

Amadou Diallo El Hadji

Laboratory of Mechanics and Acoustics, under the supervision of Zine Fellah and Christian Meso

This work presents theoretical and experimental methodologies for determining key properties and wave propagation characteristics of sound-absorbing materials.
The first contribution introduces a direct airborne ultrasonic characterization method (centered at 100 kHz) for rigid open-cell sound-absorbing media such as melamine and polyurethane foams. The method enables the determination of the viscous and thermal characteristic lengths without imposing a fixed ratio between them. It combines analytical formulations of transmission and reflection coefficients with experimental data acquired at various incidence angles, including measurements of the Brewster angle. The magnitude of the transmission coefficient is obtained from transmission experiments, while reflection measurements identify the Brewster angle, where reflection vanishes. Incorporating this angle into the theoretical reflection coefficient yields a linear relation between the two characteristic lengths. Combined with the transmission equation, this results in a solvable system of two equations with two unknowns, eliminating the classical assumption that the thermal length is three times the viscous one.
The second contribution involves the experimental observation of propagation modes in air occurring above critical incidence angles in expanded and extruded polystyrene composite plates. Broadband airborne transmission measurements (4–60 kHz) were performed using a ribbon loudspeaker and a 1/8-inch microphone. These phenomena cannot be explained by classical elastic theory but are consistent with the micropolar elasticity theory, which accounts for microrotational effects. This study provides new insights into the dynamic behavior of such materials and highlights their potential for advanced noise-control applications.

Giordano Gatti Gomide

Microflown Technologies, under the supervision of ani Fernandez

This dissertation presents the development of a Gaussian Mixture Model (GMM) based detection system for identifying characteristic acoustic signatures in aviation noise data. The work was carried out during an internship within the Erasmus Mundus master WAVES, hosted by Microflown Technologies in partnership with an aviation company, focusing on condition monitoring and fault detection in aerospace applications. The methodology integrates a complete signal processing pipeline: audio data loading and labeling, feature extraction through time–frequency analysis and octave-band transformations, statistical modeling via two-class GMMs, and post-processing to refine detections. Data scarcity and class imbalance, particularly for non-nominal signatures, required tailored training strategies, including k-fold cross-validation and normalization anchored to nominal conditions. The findings indicate that the classification performance is strongly dependent on the adequacy of feature selection and the normalization strategy adopted. Nevertheless, the results confirm the feasibility of using GMM-based approaches for operational condition monitoring, and the proposed architecture offers a scalable foundation for integration with alternative statistical or machine learning models.

Kevin Hendinata Laurentius

CDM Stravitec, under the supervision of Reinhilde Lanoye and Stijn Moons

Flanking sound transmission in mass timber structures, particularly in cross-laminated timber (CLT) buildings, is primarily influenced by junction configurations and the mechanical properties of structural connections. The vibration reduction index ($K{}_{}$), as outlined in EN ISO 12354, serves as a fundamental measure for quantifying energy transmission loss at these junctions. Resilient interlayers are recognized as effective acoustic treatments that can significantly improve the vibration reduction performance. This study uses finite element modeling (FEM) in COMSOL Multiphysics to simulate structural junctions under harmonic excitation, following the evaluation procedures of EN ISO 10848-1 to extract vibration level differences and calculate $K{}_{}$ across a range of frequencies. The study considered X-, T-, and L-junctions with rigid continuity, with the X-junction serving as the baseline and further analyzed in both rigid and elastomerically decoupled forms. Results show that rigid junctions showfrequency-dependent modal interactions, with pronounced resonance dips and anti-resonance peaks shaped by orthotropic natureand wave-mode coupling.The five-layer orthotropic model captures these details more precisely than the ESL approach while requiring greater computational effort.Moving from L-to T-to X-geometries lowers $K{}_{}$ as additional flanking paths redistribute vibrational energy.The elastomeric interlayers significantly increase vibration reduction, providing average $K{}_{}$ improvements ($\Delta {}_{}$) of about 20 dB due to the elastomer’s viscoelastic damping and the mechanical impedance mismatch it introduces. The interlayer acts primarily as a shear-dominated layer, with resonance and antiresonance frequencies closely matching theoretical predictions based on its thickness and material properties, explaining the characteristic resonance dip in transmission.While airborne reradiation remains negligible for fully rigid junctions, minor influences are observed in decoupled systems, especially within the mid-frequency range. These results confirm the importance of junction topology, material orthotropy, and connection compliance in controlling flanking transmission in CLT junctions.

Daniella Torres Hernandez

Fraunhofer IDMT, under the supervision of Christian Rollwage

We study sound source localization for overlapping sources in reverberant rooms using an ACCDOA-based SELDNet with a binary multi-label activity head [stationary, head]. The application focuses on distinguishing between broadcast speech or music, which are typically stationary, and human speech, which is typically moving, while estimating the direction of arrival (DOA) for each class. A custom dataset was generated by convolving audio with simulated room impulse responses. All experiments were performed in simulation. Evaluation used 1-second segments, wherein the SELD score displayed DOA and sound activity performance, and the classifier was assessed using the standard metrics.
Results show that the model tracks the moving source but struggles to detect inactivity, exhibiting leakage in the stationary DOA. The classifier yields reliable predictions for the moving class, whereas the stationary class has low precision. Analyses with clarity indices and the direct-to-early ratio (DER) indicate sensitivity to strong early reflections, which can blur spatial cues and induce coupling between class outputs. In the DNN context, coupling is consistent with both heads sharing the exact representation and lacking an explicit process to keep them separated. We conclude that performance is limited by weak class separability in overlapping, reverberant acoustically related signals. Future work is highly recommended.

Kristina Manachinskaia

Laboratory of Mechanics and Acoustics, under the supervision of Tom Colinot and Nathan Szwarsberg (Buffet Crampon)

This thesis designs and optimizes a child‐friendly pocket saxophone using a fast Transfer Matrix Method (TMM) coupled with least‐squares tuning of musically relevant resonances. The modeling stack is built incrementally—cylinder (lossless/lossy, radiation), cylinder with a side tonehole, truncated cone, then conical bores with one and multiple holes—and benchmarked against OpenWind FEM; across cases, the lossy TMM reproduces the first resonances within 10 cents, which is adequate for intonation‐driven optimization. Experimental validation uses 3D‐printed resin prototypes (one‐ and two‐part bodies, with/without holes); input impedance is measured with an impedance head in a semi‐anechoic room, and repeatability is assessed across re‐assemblies. Measurements reveal a nearly uniform downward shift for open‐hole states, highlighting the role of auxiliary volumes and small leaks. An entrance‐volume correction—implemented as a short series cylinder matched to the small‐end bore—accounts for this bias and aligns the adjusted model with averaged played notes for most fingerings. The optimizer (MATLABl sqnonlin, bound‐constrained) targets the first resonance per fingering over G4–F#5, yielding a compact, manufacturable geometry that meets ergonomic and printing limits while preserving stable low‐order resonances. The work delivers:

1. a validated, modular TMM library for conical multi‐hole instruments;
2. an end‐to‐end pipeline from modeling and FEM benchmarking to measurement and geometry optimization;
3. a functional pocket‐sax prototype.

Limitations (high‐frequency stability, simplified entrance model) and next steps toward a data‐driven, playability‐aware optimizer are outlined.

Kacem Meddeb

Laboratory of Mechanics and Acoustics, under the supervision of Régis Cottereau.

The continuous energy demand globally drives dam construction and specifically around active seismic areas. In such cases, the numerical simulation of dams excited by a seismic wave helps to assess such risks. However, Seismic design of dams can be quite complex due to interactions between soil fluid and the dam and due to nonlinear behaviour observed between the concrete blocks.
The frequency domain numerical simulation is quite limited as it fails to portrait these interactions. The time domain simulation can take these effects into account, but it exhibits a large computational cost due to the size of the mesh and the small size of the time step allowed. Coarse meshes can be established due to the size of a seismic wavelength, yet the dam geometry imposes finer mesh around the structure. This, with the stability condition of the scheme used, limits the maximum value of the time step that can be used globally, and consequently increases its processing cost.
This work presents a recent time scheme using higher-order approximation of the time derivative to increase the maximum allowable time step. With this approach, just doubling the used time step can halve the number of iterations in time. This can reduce substantially the process power used and the overall runtime (given the numerical complexity of the scheme doesn’t increase equally). We study the stability of the scheme analytically and use the results to validate the numerical simulations that we implement.
Once that part is done, we test a mixed time scheme approach: using a standard time scheme over most of the elements of the mesh and using a high order time scheme on the smaller elements that limit the largest allowable time step used for a given simulation. We investigate test cases numerically in 1D with reference to the obtained results analytically and presented in [1]. Using the valid script, we test the mixed time scheme for higher orders trying to establish numerically their stability limits since we do not have any analytical values for reference. To further confirm the findings with the 1D test case, we test the mixed time scheme approach on a larger 2D test case closer to the 3D mesh of a dam.
The overall purpose of the internship is to confirm an improvement of the time step using the mixed time scheme and implement it on a large model of the dam under a seismic load. Simultaneously, we need to keep the computational cost stays as low as possible.

Juliette Anna Naumann

Vestas Deutschland GmbH, under the supervision of Patrick Cordes and Robert Wilsch.

Transformer-induced tonal noise in wind turbines has emerged as a critical issue in both regulatory compliance and public acceptance, particularly in countries like Germany where tonal penalties of up to 6 dB are applied under the TA-L ̈arm guidelines. Unlike broadband aerodynamic noise, tonal components are perceived as more disturbing and are subject to stricter evaluation. Field measurements revealed persistent tonal emissions suspected to originate from the transformer, prompting a comprehensive root cause analysis and mitigation study.
These tonal emissions stem from magnetostrictive effects in the transformer’s laminated steel core, which cause periodic dimensional changes under alternating magnetic fields. The resulting mechanical vibrations typically manifest at 100 Hz which is twice the grid frequency in Europe, with higher harmonics becoming prominent under non-linear excitation or structural resonance.
This thesis investigates the origin of transformer-related tonal noise and evaluates mitigation strategies. A detailed vibroacoustic measurement campaign confirmed the transformer as the dominant source, with peaks at 200 Hz, 300 Hz, and 600 Hz. Subsequent transfer path analysis revealed both structure-borne and airborne transmission, with the nacelle wall potentially acting as a radiator.
A review of conventional approaches, including damping mats or foam in the nacelle as well as mechanical damping or decoupling strategies, was conducted. Of these, vibroacoustic damping mats were tested experimentally and showed limited effectiveness at critical tonal frequencies.
A novel mitigation concept based on the Vibroacoustic Black Hole (VABH) principle was developed and evaluated through Finite Element (FE) simulations in ANSYS. The VABH concept relies on a tapered thickness profile that slows down and concentrates flexural wave energy toward a thin tip region, where it can be effectively dissipated. The VABH plate was designed to target frequencies above a cut-on threshold of92 Hz. Simulations demonstrated that the VABH-treated plate significantly reduced structural vibrations at key tonal frequencies compared to a flat reference plate. Furthermore, the concept was applied to a specific use case by modeling the VABH plate attached to a simplified transformer structure. While this configuration achieved notable vibration attenuation without increasing far-field acoustic radiation, the flat reference plate also showed improvements, suggesting that a significant portion of the effect may be attributed to the added mass rather than the VABH geometry alone.
The findings suggest that, although conventional damping methods provide only partial relief, decoupling the transformer mechanically could interrupt structural transmission paths and alter resonance frequencies. Additionally, the VABH concept presents a promising, low-invasiveness solution for targeted vibration and noise reduction. Future work should focus on prototyping, experimental validation, and integration of these strategies into existing turbine systems to improve acoustic performance and ensure regulatory compliance.

Anthia Patsinakidou

University of Stuttgart, under the supervision of P. Cillo, I. Wenger, P. Ziegler and E.P. Eberhard.

Achieving consistent reproducibility when building stringed instruments is essential in the music industry, since even seemingly identical designs can exhibit audible tonal differences. Particularly in the case of the classical guitar, its vibrational behavior, which has been proven to be highly sensitive to its geometric parameters, can be accurately analyzed using the finite element method (FEM). Although FEM, when coupled with numerical optimization, has proven effective for replicating the vibrational behavior of a guitar soundboard, the prohibitive computational expense associated with the numerous parametric evaluations required, limits its applicability in design exploration.
The present thesis addresses this challenge by proposing a surrogate modeling framework based on an artificial neural network (ANN), which serves as a predictive tool for estimating the guitar’s modal frequencies based on its geometric characteristics, thereby reducing computational time. A fully parametric FEM model that encompasses the complete assembly of the guitar is initially developed to generate the corresponding training dataset. A subsequent sensitivity analysis identifies the most influential geometric parameters, which serve as the primary inputs to the network. The surrogate model is evaluated with respect to its predictive accuracy and computational efficiency, while its utility is demonstrated through a practical application, where the aim is to achieve vibrational consistency between two virtual prototypes, which differ in their material properties. This material variability is compensated by applying the necessary geometric modifications based on the surrogate model.

Fiona Power

University of Coimbra, under the supervision of Luis Godinho and Paulo Amado Mendes, and Bart Van Damme (EMPA, Switzerland)

In the realm of railway transport and mobility, the investment and expansion of transport infrastructure increases the need for effective noise and vibration control. Emitted noise from railway vehicles poses an environmental concern as noise and vibration impacts the health, safety, and wellness of people and the environment. The dynamic action of the train wheel-rail interaction causes vibration of the rail itself which is a significant contributor to the overall emitted railway noise. It is necessary to model vibroacoustic behaviors with the purpose of quantifying the contribution of the rail vibration with the addition of rail track components such as the sleeper, railpads, and ballast.
The demand for accurate and accessible track dynamic models with the development of vibration mitigation strategies has become increasingly important. A Finite Element Method rail track model was implemented considering the rail as an infinite structure in the longitudinal direction coupled with Bloch-Floquet periodicity. The accuracy of this periodic model was validated in cooperation with a 3D FEM model. The periodic model utilized a unit cell, coupling mechanical tuned resonators to the rail model to mitigate vibration in the low frequency regime. An optimization of resonator parameters, such as stiffness, and placement along therail track was implemented to propose various resonator-beam models including elements such as the sleeper, railpads, and ballast.
Modal analysis and beam mechanics were used to extract eigenvalues and mode shapes to characterize the dynamic behavior of the structure under free and forced vibration. The first bending mode was targeted to apply resonators and reduce vibration. Finally, the mobility of the formulated beam-resonator models was analyzed, considering the effects of the sleeper, railpads and ballast. The addition of tuned resonators was found to reduce the peak mobility across several rail trackmodels. The mobility response of the periodic unit cell model was validated and compared with a 3D FEM rail track model. Valuable information can be retrieved through the unit cell approach when comparing the mobility frequency response function (FRF) results of both the 3D and 1D FEM models.

Aleksandra Romakhova

Boskalis, under the supervision of Pauline Levassor

This thesis presents a numerical investigation of bubble-based noise mitigation for offshore impact pile driving, a major source of low-frequency underwater noise. Using the Finite Element Method (FEM) in COMSOL Multiphysics, complemented by MATLAB post-processing, the study examines the acoustic reflection and transmission mechanisms of large, isolated air bubbles in water. The work focuses on two key phenomena—acoustic impedance mismatch and bubble resonance—identified as primary contributors to underwater noise reduction.
A two-dimensional axisymmetric waveguide model was developed to simulate Gaussian modulated sinusoidal pulse representative of pile-driving noise, with central frequency at 100 Hz. The Minnaert resonance theory was first validated through eigenfrequency analyses for bubble radii of 33–80 mm, showing excellent agreement with numerical results. Time-domain simulations were then conducted to quantify reflection and transmission coefficients, sound pressure levels, and frequency-dependent transmission loss for varying bubble radii, constant air fractions, and different source frequencies.
Results demonstrate that larger bubbles produce stronger reflections and lower transmission, with optimal noise mitigation occurring near the bubble’s resonance frequency. Transmission loss spectra reveal that maximum attenuation aligns with Minnaert resonance and decreases away from it. The findings provide insight into the physical principles governing bubble-based mitigation and highlight design considerations for optimizing bubble curtains in offshore applications.
The developed modeling framework offers a computationally efficient and physically accurate tool for exploring bubble–wave interactions in the low-frequency regime, supporting future research and engineering design of environmentally compliant pile-driving operations.

Sara Sberro

Amadeus, under the supervision of Adrien de Giovanni

This dissertation explores the acoustic propagation and optimization of line arrays, a subject that remains insufficiently documented in academic literature. Existing research is largely driven by major industry players, where many details are either confidential or based on broad simplifying assumptions that are not explicitly documented for competitive reasons. For modern applications such as sound spatialization, however, line arrays represent a critical topic: their behavior must be modeled with accuracy to ensure reliable prediction, simulation, and ultimately system design. Addressing this need is therefore both academically and industrially relevant.
The primary objectives of this work are to characterize the propagation behavior of line arrays and to develop computationally efficient methods for their simulation and optimization. More specifically, the research aims to analyze the construction of a representative line array model and assess its compliance with established acoustic criteria, implement a simulation framework capable of describing near-, transition-, and far-field behaviors with a balance between accuracy and computational cost, and apply an optimization strategy to improve sound pressure level (SPL) coverage while ensuring physical feasibility within real loudspeaker constraints.
To achieve these objectives, the methodology integrates theoretical analysis, numerical simulation, and algorithmic optimization. The study begins with an examination of the construction principles of a specific line array, comparing its transducer spacing, active radiation factor, and curvature angles against the requirements defined in classical line array theory. Building on this foundation, wave propagation is modeled using the Complex Directivity Point Source (CDPS) method, which allows each loudspeaker module to be represented as a source with complex directivity data. This approach captures phase and amplitude information, modeling interference phenomena while reducing the computational burden compared to fully discretized line-source simulations. Analytical criteria are employed to establish validity zones, distinguishing near-, transition-, and far-field regimes, and to determine the limits within which CDPS provides reliable predictions.
The simulation framework is then used to test a range of parameters, including module length, number of discretization points, and frequency, in order to evaluate both accuracy and computational efficiency. Special attention is given to the transition zone, where line arrays deviate most significantly from the idealized cylindrical or spherical propagation models, and where reliable prediction is most challenging. A correction function within the discretization framework is also explored as a means of improving agreement between analytical and simulated results, the findings indicate that more complex computational methodology is required to accurately model changes in the geometric factors that dominate propagation behavior rather than function correction terms.
On the optimization side, an interpretation of the Polygonal Audience Line Curving (PALC) algorithm is implemented to refine array curvature and cabinet splay angles for improved audience coverage. The algorithm operates geometrically, assigning each cabinet a target zone along a polygonal representation of the audience area. Iterative adjustment of coverage angles ensures that the entire audience space is uniformly addressed. Importantly, the algorithm is constrained by the discrete hinge angles available in real loudspeaker models, ensuring that the optimized configurations correspond to physically realizable systems.
The results demonstrate that the CDPS framework significantly reduces computational cost compared to high-density discretization, while preserving predictive accuracy across most of the frequency range. The PALC optimization produces arrays that achieve noticeably more uniform SPL distribution across the listening area compared to straight, vertically stacked configurations. Furthermore, validation against commercial simulation software such as EASE Focus confirms the robustness and applicability of the proposed methods. The framework not only reproduces key propagation characteristics but also highlights the practical limits of current models, thereby offering a bridge between theoretical acoustics and real-world system design.
Overall, this dissertation contributes both theoretical and practical insights into the understanding of line array behavior. It clarifies the propagation characteristics of line sources in different acoustic regimes, develops a computationally efficient simulation method grounded in CDPS theory, and demonstrates the effectiveness of algorithmic optimization under real-world constraints. By advancing the state of simulation capabilities, the work breaks down necessary principles for developing the necessary tools to better predict, design, and deploy line arrays in performance spaces.