1. Introduction
Building envelope airtightness plays a significant role in the evaluation of a building’s energy efficiency [
1,
2]. Unintended airflow through building envelopes is responsible for 30% to 50% of a building’s heating and cooling energy, as assessed in studies of housing stocks from Finland [
3], Estonia [
4], and the UK [
5]. Therefore, understanding and controlling unintended infiltration in buildings is important for energy conservation and environmental sustainability.
The prevalent method for assessing a building’s airtightness is the fan pressurization method or “blower door test”, as described in the ISO 9972 standard [
6]. This technique is used to evaluate if buildings meet energy performance standards, compare the overall airtightness of different buildings, or check air permeability reductions post-refurbishment. However, it primarily provides an overall leakage rate, with the quantification and identification of individual leaks being challenging, time-consuming, and heavily reliant on the operator’s experience. Supplementary methods such as smoke stick use, anemometers [
7], infrared thermography (e.g., qualitative [
8,
9], differential [
10] or lock-in [
11] methods), or tracer gases [
12,
13] are used for leak detection but have limitations, including the need for pressure differences or temperature gradients. Moreover, several innovative techniques have emerged as alternatives to the fan pressurization method, broadening the toolkit available for airtightness assessment [
14].
In this context, this paper is located at the intersection of building science and acoustic engineering, addressing the need for efficient and reliable airtightness assessment methods. It explores the potential of acoustic methods as a supplementary approach to identify and quantify individual leaks in building envelopes. Acoustic testing, which is particularly appealing due to its non-invasive nature, does not require significant air movement through the envelope, allowing for assessments under naturally occurring low or no-flow conditions. Additionally, they do not depend on closed volumes, making them particularly suitable for use during the different stages of a building’s lifecycle, including construction and renovation, when traditional methods might be less effective or feasible.
The ASTM E1186 standard [
7] introduces an acoustic approach for detecting leaks, proposing the use of sound detection. Originating from the work of Keast et al. [
15] in the 1970s, this standard highlights the use of acoustic measurements for assessing building envelope integrity. Despite its potential, the efficacy and accuracy of this acoustic method have not yet been comprehensively studied. Early work by Sonoda and Peterson [
16] introduced a sonic method for air leakage measurements, employing single microphones to characterize air leaks. Iordache et al. [
17,
18] introduced an approach correlating sound transmission loss with air infiltration rate to quantify unintended air infiltration through envelope components, which was later discussed by Hassan [
19] and Berardi et al. [
20]. While Graham [
21] and Card et al. [
22] considered measurements of infrasonic impedance for leakage determination, Kölsch et al. [
23,
24] focused on higher ultrasonic frequencies for locating and quantifying air leaks.
Recent advancements have seen the exploration of microphone arrays and beamforming techniques to detect air leakages, a method traditionally used in fields such as aerospace [
25]. This technology, utilizing microphone arrays to detect sound wave patterns, presents a non-invasive and potentially more efficient alternative for identifying air leakages in building envelopes. In this context, Raman et al. [
26] utilized beamforming and nearfield acoustic holography to detect and quantify leaks from a single window and door. Further studies [
27,
28] have focused on investigating single parts of building envelopes (e.g., single windows or doors) using microphone arrays and beamforming.
Our research extends the existing literature by undertaking (to our knowledge) the largest measurement campaign aimed at assessing building envelope airtightness using acoustic methods. This campaign was conducted on a building complex characterized by highly heterogeneous façades, encompassing a wide variety of materials, ages, and construction techniques, where the locations and extents of air leakages were previously unknown. This contribution provides valuable insights into the applicability and effectiveness of this method to real-world environments. Specifically, we expand the analysis of previous work [
29] on pattern recognition in the selected acoustic spectra of leaks.
The main aim of our study is to demonstrate and validate the effectiveness of acoustic beamforming technology in large-scale, real-world scenarios for assessing building envelope airtightness. By conducting field tests on a variety of office buildings with differing ages and construction styles, this study presents an extensive application of this method.
2. Methods
2.1. Test Site
This study was part of a larger research project that also encompasses a UAV (unmanned aerial vehicle)-based infrared thermography assessment for U-value determination of the same building complex [
30]. Within this larger project, three institute buildings were pre-selected. We obtained detailed information about the building’s location, size, and structure from the institutions. Five assessment criteria were used to determine the test site: (1) the feasibility of drone flight accessibility, ensuring safe and unrestricted UAV operations; (2) the heterogeneous age of the structure of the building parts, with a similarity in their usage to allow for comparative analysis; (3) variability in the façade designs among the building parts with comparable use to assess the impact of architectural diversity; (4) an overall size of the building complex that allows for complete capture; (5) a relatively simple geometric structure, facilitating the rapid and reliable creation of 3D models. The selected test site for this study was a research institution located in Villingen-Schwenningen, Germany. This site comprises five distinct buildings, each serving different functions and constructed at various times, thus offering a diverse test environment. For the purposes of this study, our focus was narrowed to the office buildings, designated as parts A, D, and E. These buildings were selected based on their construction or renovation years—approximately 1990, 1998, and 2019, respectively—providing a range of ages and construction styles for analysis. At the forefront of the measurement campaign, the staff working in Building D had often commented on draughts in the building. Building B is utilized for clean rooms, whereas Building C accommodates a cogeneration unit. An aerial photograph (refer to
Figure 1) provides a visual representation of the site, including the labelling of the individual building parts, thereby offering a comprehensive overview of the measurement context.
2.2. Measurement Setup
During the measurement campaign, an acoustic camera was used to capture acoustic data. The acoustic camera is a device engineered to make sound waves visible to the human eye. The primary function of an acoustic camera is to capture sound pressure variations over time and space, converting these into a form that can be visualized and interpreted. By using beamforming, the device synthesizes the data collected by its microphone array, enabling it to pinpoint the direction and relative intensity of sound sources. The resulting visualization takes the form of a sound map that is superimposed on a visual image of the area being surveyed, which highlights the spatial distribution of sound intensity, thereby identifying the locations of sound-emitting objects (
Figure 2, right).
The experimental setup included a pair of speakers placed inside the building (
Figure 2, left) on one side of the wall and the acoustic camera positioned externally (
Figure 2, center) on the opposite side. A high-frequency speaker, operating with a frequency range of 15
to 120
, and a low-frequency dodecahedron speaker, covering 0.05
to 16
, ensured the generation of sound waves that penetrated through wall leaks, detectable by the acoustic camera as discrete sound sources from the wall’s exterior. The computer-generated white noise signal was emitted inside at a sound pressure level of 85 dB for a duration of 4
. The arrangement of the acoustic camera and speakers was kept stationary during the measurement of each room. The measurement campaign encompassed a total of 57 acoustic measurements across 36 investigated rooms. Some rooms featuring façades on different sides of the building necessitated multiple measurements to ensure thorough coverage.
The microphone array used in these experiments is the “Acoustic Camera Array Ring48 AC Pro” from the company gfai tech GmbH. It is characterized by its ring configuration, consisting of 48 equally spaced microphones arranged in a circle with a diameter of
. This particular geometry was selected for its reduced sensitivity regarding the accuracy of the focal distance compared to other array configurations [
31]. Such a feature is helpful for measurements of non-homogeneous outside façades. The array was designed to capture a wide range of frequencies, ranging from 164
to 20.000
. However, for applications where sound source localization is more important than the absolute measurement of sound pressure levels, the manufacturer indicates the feasibility of extending this range up to 60
. This broad frequency range is crucial for accurately identifying and analyzing the diverse acoustic signatures associated with different types of leakages in building envelopes.
An optical camera was placed at the center of the microphone array to capture a visual image of the surroundings being measured. This camera offers a resolution of 1920 × 1080 pixels. Along with the acoustic data, this visual documentation provided a comprehensive dataset for analyzing the presence and characteristics of air leakages in the building complex.
In each room, one person was responsible for setting up the loudspeaker, and the other focused on positioning the acoustic camera towards the room’s façade. Communication between the team members, e.g., signaling the start of the measurements, was managed through hand signals or via mobile phones. Following an initial evaluation of the collected data, obvious causes for the most prominent acoustic signals were sought. Detailed notes on the measurement procedures, the environmental conditions, and any notable findings were meticulously recorded on a protocol sheet. This careful documentation facilitated the seamless transition to the subsequent room’s measurement setup.
All acoustic signals captured during the measurement campaign were sampled at a high frequency of 192
and digitized at a resolution of 32 bits. Data analysis was conducted using the software NoiseImage [
32]. This high-resolution data capture ensures the accuracy and reliability of the measurements, facilitating detailed analysis of the acoustic characteristics of air leakages.
While we were able to reduce the disturbing influence of external sound sources in the past by recording reference signals next to the speaker inside the building [
27], we did not do so in this study for efficiency and time-saving reasons.
2.3. Acoustic Air Leakage Detection
The utilization of beamforming in acoustic air leakage detection shows the integration of advanced signal processing techniques with practical building diagnostics. Beamforming is a signal processing method that differentiates sound sources based on their directional origins by utilizing a microphone array. This technique focuses on scanning a target point (denoted as
) across a predefined grid on the object under inspection. The operation of beamforming in the context of acoustic air leakage detection involves computing a time function,
, for each focus point, as described by the following equation [
33]:
In this process, the time signals from individual microphones, , are superimposed with a specific time delay, , which corresponds to the sound wave’s travel time from the focus point on the building’s façade to the microphone. Subsequently, the time-corrected signals from all microphones are summed and divided by the total number of microphones, M, yielding a time signal for each focus point.
Following the temporal alignment of the microphone signals, the next step involves calculating the effective sound pressure,
, at each focus point:
Here,
n represents the total number of discrete time samples, and
is the time at the sample index
k. The individual sound pressure values, derived from the microphone array signals, are mapped onto the pixels of the optical camera image capturing the same scene. This mapping is achieved by superimposing the acoustic values,
, as colors onto the optical camera image. Such a visualization technique offers a clear and intuitive representation of sound sources. Further information on this operating principle is outlined in Refs. [
27,
33,
34].
This acoustic method does not rely on temperature or pressure differences in the building envelope, unlike the well-established infrared thermography method for visualizing leaks.
2.4. Evaluation and Categorization of Acoustic Signals
The measurement campaign carried out indicates that sound sources signaling potential leaks are usually detected in the spectral range from 800 to 25 with the equipment provided. There are 16 third-octave frequency bands within this range. In each of these bands, only the highest dB of the recorded sound pressure levels (hereafter referred to as peaks) are superimposed on the visual image. Due to the occurrence of sound peaks at different locations in different frequency bands, indicating potential leaks at different locations, it is rarely possible to image all leaks at the same time. Typically, a sequence of images across different frequency bands is required to illustrate all potential leakages found in the building envelope.
This is demonstrated in
Figure 3, which presents Room 206 (Building D, east façade), showcasing 15 out of the 16 examined third-octave frequency bands. In this representation, the highest sound peaks are highlighted for each frequency band, with the
value being individually adjusted to ensure optimal visualization. In order to understand the necessity of depicting the entire range of frequency bands, we will examine the top edge of the first window from the left. There, the sound peaks are only evident in frequency bands 2, 3.2, 6.3, 8, and 10
. Conversely, at the top of the fourth window frame from the left, the sound peaks are only visible at the frequency bands of 0.8, 1.3, 4, 5, 10, 16, and 20
.
Not every sound peak is necessarily indicative of a leak; they can also be caused by sound reflections or structure-borne noise, such as vibration, which can cause locally high sound levels. The sound peak on the pane from the fourth window from the left in the 10
frequency band in
Figure 3 is clearly at an implausible location for a leak and is more likely caused by the vibration of the pane.
Nevertheless, in many instances, visual inspection at the location of the sound peaks confirmed credible causes of air leakage. In some rooms, a blower door and smoke stick were used to definitively confirm a leakage at the location of a sound peak.
However, often, the cause of a sound peak could not be clearly confirmed due to limited time resources. These peaks, therefore, required a subjective assessment of their plausibility as being leakage-related, which is described below.
Assessing individual peaks across all 16 third-octave frequency bands in all 57 measurements (totaling 912 analyzed frequency bands) necessitated the manual adjustment of the signal’s peak
value in the NoiseImage software [
32]. This adjustment enabled the signal peak to be depicted as the most compact area possible, pinpointing the precise source location. It can be assumed that the spatial resolution of this method depends on the wavelength and is, therefore, more reliable at higher frequencies than at lower ones. The peak position is then utilized to assess the plausibility of a leak as the source of the sound at that particular location.
Table 1 outlines the four evaluation categories along with their respective color codes and scores.
The precise localization of a large number of sound peaks is a laborious manual process, and the evaluation categories are based on subjective criteria with fuzzy boundaries. Both tasks are, therefore, currently time-consuming and error-prone. Nevertheless, the method enables the documentation of the rating on the façade representation using the corresponding color code, as shown in
Figure 3. This visualization can be used to identify potential leakage points on the façade for inspection and possible sealing. However, this process currently requires multiple images for different third-octave frequency bands.
The multi-frequency assessment score (MFAS) is introduced as a quantitative measure of the airtightness of a room. It is calculated as the sum of the acoustic assessment scores (AASs) (see
Table 1) corresponding to the highest color codes in each third-octave frequency band. This calculation rule is admittedly arbitrary. Other options, such as adding the number of peaks occurring for each third-octave band rather than the maximum color code occurrence, seemed to give less reproducible results. Therefore, the simplest possible definition of “multi-frequency score” was chosen.
2.5. Evaluation of Spectral Properties
In order to obtain the spectral characteristics of a point of interest (e.g., a potential leak) within the acoustic image, an iterative process is required. The first step is to analyze the sound pressure level across the entire recorded image. This involves examining individual frequency ranges, usually in one-third octave band intervals. After the frequency range has been found in which the sound pressure level at the location of a point of interest is the highest or can be best distinguished from the rest, the beamforming method is used to obtain its spectrum. The next step is to adjust the examined frequency range so that the peak sound pressure level of this point falls exactly within the limits of the frequency. With the new frequency range, the location of the highest sound pressure level may also change. If this happens, the beamforming method is applied to the new point. The resulting spectrum will then be looked at once again, and the frequency limits will be adjusted accordingly. This iterative process is stopped whenever the point with the highest sound pressure level changes by less than 5 pixels as a result of the frequency range adjustment; otherwise, the process will get caught in an infinite loop. The whole process is graphically illustrated in
Figure 4.
4. Conclusions
To our knowledge, this measurement campaign is the most extensive field study of the acoustic detection of air leakage in building envelopes. The method was successfully demonstrated on the façades of multi-story buildings of different ages and heterogeneous envelope structures, confirming the basic functionality of the method for large buildings and proving the significant potential of this methodology.
Furthermore, this field study provides valuable insights into the practicality, speed, and clarity of acoustic signal interpretation alongside the wider applicability of the method. A considerable number of potential leaks were localized and visualized over large areas, many of which were confirmed as plausible by visual inspection. Specifically, in selected rooms where smoking sticks were deployed, some of the identified leaks were verified. The results indicate that a significant number of potential leaks can be detected, confirming the basic functionality of the method for large buildings. By using this method, it was also possible to distinguish between the airtightness characteristics of the different buildings. In addition, an examination of the spectra at leakage points suggests that spectral information can be used to infer leakage characteristics such as type or severity. For the first time, this work demonstrates the potential of the acoustic beamforming method to detect unknown leaks on large façades without the use of the fan pressurization method.
Further research is needed to improve the reliability of leak detection, especially in distinguishing leaks from non-leak noises such as reflections and vibrations. In addition, our aim is a more detailed analysis of leak size and nature of the leaks found, thus enabling the comprehensive (and possibly an automatable leakage) assessment and quantifiability of detected leaks. For this reason, the ongoing joint research project Q-Leak (see funding section below) involves the development of a dedicated test facility for analyzing leaks under controlled laboratory conditions, intending to systematically examine how the acoustic signatures of different leaks vary with their type and size. Key areas of interest include (a) the impact of measurement configurations (such as the viewing angle and air exchange rate), (b) the role of operating across different spectral ranges (audible to ultrasonic, narrowband, or broadband), and the exploration of diverse loudspeaker signals (including noise, frequency sweeps, and music), and (c) the possible correlations of the acoustic signature (spectrum and sound directionality) of leakages with different leakage paths, materials, and sizes. In addition to the greater automation and simpler visualization of leaks, improvements to the measurement technology are also planned in the project. This includes the development of a suitable ultrasonic transmitter to extend the applicable frequency range, as well as an investigation of a combination of infrared thermography with the existing acoustic method to improve the reliability of leak detection.
With further development, the acoustic beamforming method promises to be used effectively and in a time-saving manner in several areas of the construction industry. This technology holds significant potential for a wide array of stakeholders focused on the energy optimization and renovation of buildings, including the housing industry, building owners, service providers, energy consultants, and manufacturers of airtightness measurement equipment. One of the key advantages of acoustic methods over traditional blower door tests is their independence from closed volumes, enabling their application in both new construction and the refurbishment of existing buildings. For new buildings, this method could serve as a tool for ensuring the airtightness of prefabricated cladding elements and conducting quality testing prior to building acceptance. Furthermore, the technology offers the ability to assess the need for the refurbishment of existing buildings, allowing for the prioritization of leaks according to their potential energy impact and facilitating cost-effective upgrades. The acoustic method has the potential to reduce lengthy pressurized leak detection to a series of much less invasive short-term measurements. This is particularly advantageous for renovations in occupied buildings. Moreover, in the case of serial refurbishment, there is also the advantage that the airtightness of façcade elements can be checked prior to final installation, thereby saving time and costs while minimizing inconvenience to residents.