From Simple Auditory Inspection to Acoustic Feature Extraction of Drone Brushless Motors

Acoustic Emission Monitoring of Steel Joints Using Cepstral Coefficients

Vibration-based Structural Health Monitoring (SHM) techniques are effective for identifying global dynamic behavior, but local damage detection and source localization remain challenging, particularly when damage initiates at small scales or in confined joint regions. In steel structures, critical deterioration mechanisms often develop at welded or mechanically connected details, where modal vibration signatures can remain largely unaltered during early damage stages, reducing their diagnostic sensitivity. This limitation motivates the integration of monitoring approaches that are driven directly by material-level response. Acoustic Emissions (AE) provide a non-destructive and high-sensitivity monitoring strategy generated by transient elastic waves released from a material when mechanically stressed. AE events originate at the damage source itself, enabling a direct assessment of local condition variations in structural details and joints. AE monitoring is therefore particularly suited for steel connections, where deterioration may remain undetected by global dynamic metrics. Current AE monitoring approaches often rely on elementary signal descriptors—including peak amplitude and waveform duration—which are widely reported in the literature as indicators of critical structural states. On the other hand, dominant frequency tracking can be ambiguous to extract reliably due to noise sensitivity, mode mixing, and signal non-stationarity. To address this challenge, the present work explores the use of Cepstral Coefficients (CC) computed from AE waveforms as a stable and straightforward feature for signal characterization. Unlike spectral peaks, CC are directly and uniquely derived from the input signal through a computationally lightweight procedure. These characteristics make CC attractive for rapid laboratory assessment and future SHM integration. In this context, an experimental study was carried out on welded, full-penetration steel joint specimens tested at University of Trento. The specimens, instrumented with AE sensors, were tested at the University of Trento under controlled low-cycle loading protocols inducing yielding damage. Based on the controlled load history, a comprehensive dataset of AE signals was collected from both undamaged and damaged specimens to assess how local deterioration alters CC features, which were extracted from raw AE waveforms and statistically compared between undamaged and damaged conditions. The analysis highlights differences in CC distributions, showing potential in discriminating between distinct structural damage states in steel joints without requiring complex frequency identification. The results support the CC-based characterization of AE signals as a complementary monitoring layer for SHM systems targeting critical steel joint regions.

Vibration sensor based on the optical fiber ring resonator for intelligent speech recognition empowered by AI

As the primary medium of human communication, speech demands high-performance acquisition and analysis. Conventional sensors often lack sufficient sensitivity, signal-to-noise ratio, and immunity to electromagnetic interference, limiting the capture of subtle acoustic features. To address this, we propose a Fiber Ring Resonator (FRR) as the sensing element for speech-induced vibrations. Leveraging its high quality factor and strong field enhancement, the FRR significantly improves sensitivity. By integrating the Pound–Drever–Hall (PDH) technique, the system locks onto and demodulates minute sound-pressure-induced frequency shifts, enabling robust extraction of multi-frequency signals. Machine learning algorithms then perform deep feature extraction and intelligent classification to suppress background noise and boost recognition accuracy. This fusion offers a promising solution for non-invasive laryngeal diagnosis, secure acoustic monitoring, and intelligent human-computer interaction, paving the way for next-generation speech perception systems.

Propulsion Noise Mitigation for UAV-Based Acoustic Target Detection: Spectral Characterization, Filter Optimization, and Real-Time Embedded Implementation

Acoustic sensing enables unmanned aerial vehicles (UAVs) to detect sound-emitting targets beyond the visual field, particularly in environments where visual sensing is degraded by darkness, fog, occlusion, or camouflage. However, strong self-generated propulsion noise produced by onboard motors and propellers often masks external acoustic signals, limiting the applicability of UAV-based acoustic perception systems. This study investigates the spectral characteristics of UAV propulsion noise and evaluates a filtering-based approach for improving the observability of external acoustic targets. Acoustic measurements were conducted using a hexacopter UAV under controlled indoor conditions. The experimental dataset comprised recordings collected under motor-only, propeller-attached, and helicopter-noise mixture scenarios across multiple flight modes and source-distance conditions in which helicopter sounds were introduced through a loudspeaker. Signal analysis was performed using waveform inspection, Fourier-based spectral analysis, and spectrogram-based time–frequency representations. The results show that the dominant energy of UAV propulsion noise is concentrated in the low-frequency region of the spectrum. Based on this observation, a fourth-order Butterworth high-pass filter architecture was implemented and systematically evaluated using a benchmark sweep from 250 Hz to 400 Hz in 10 Hz increments. Within the evaluated range, the 340 Hz configuration achieved the highest helicopter top-1 classification accuracy of 62.86%, whereas unfiltered recordings and generic baseline denoising methods yielded 0.00% accuracy. Additional experiments conducted under multiple source-to-UAV distance conditions further revealed the influence of acoustic propagation on propulsion-noise masking behavior. Overall, the findings demonstrate that spectral characterization combined with computationally efficient high-pass filtering can effectively mitigate low-frequency UAV propulsion noise and enhance acoustic target observability. The proposed framework provides an effective preprocessing stage for UAV-based acoustic perception systems and establishes a practical foundation for future UAV-based acoustic target detection and multimodal perception systems. Although helicopter signatures were used as the target class in this study, the proposed preprocessing framework is not fundamentally limited to helicopter detection and may be applicable to other acoustically detectable targets. While the present study identifies an empirically optimal fixed cutoff frequency of 340 Hz for the evaluated platform and operating conditions, future work will investigate real-time adaptive filtering strategies capable of dynamically adjusting filter parameters according to propulsion-noise characteristics and flight conditions.

Use of Principal Component Analysis and Autoencoder Neural Networks in Characterization of Acoustic Emission Waveforms from Composite Coupon Specimens

Acoustic emission (AE) monitoring is a method of structural health monitoring that relies on the detection of elastic waves generated by the release of concentrated strain energy when damage is created in a structural material. One of its strengths is that registered waveforms can, in theory, be used to draw conclusions about the origin of the signals, allowing to estimate damage type, location, and severity. One of the problems, however, is how to automate the interpretation and reliably retain useful information from AE waveforms, which are often thousands of samples long. In this work, a method of waveform analysis in the frequency domain is presented that combines principal compo- nent analysis and an autoencoder to reduce the dimensionality of the problem to a pair of parameters, capturing the spectrum shape and the spectrum en- ergy content. Tensile tests were carried out on composite coupon specimens while monitoring AEs, and the progression of damage was tracked by X-ray scanning them in two locations before and after tensile testing. Locations and types of damage in the scans are in good agreement with the results of the AE monitoring and analysis framework. The method is proposed as a tool for automated interpretation of AE signals with the potential to be generalized to other material, layup, and sensor setups.

PMMDA based on the fusion of acoustic and vibration signals under time-varying speed conditions for bearing fault diagnosis

In engineering applications, mechanical equipment must adapt to complex and dynamic working environments, where the rotational speed often varies over time, resulting in significant distribution discrepancies across different operating conditions. Meanwhile, information obtained from a single vibration signal is often insufficient and susceptible to external interference. Traditional single-source domain adaptation methods may suffer from negative transfer and fail to effectively exploit complementary knowledge from multiple source domains for target-domain fault diagnosis, resulting in reduced reliability and generalization performance of diagnostic models. To address these limitations, this paper proposes a Progressive Multi-Dimensional Multi-Source Domain Adaptation (PMMDA) method. From the perspective of collaborative utilization of multi-source data, the proposed method integrates multimodal information from vibration and acoustic signals and employs a multi-level feature alignment strategy to achieve progressive alignment between source and target domains. Additionally, an adaptive weighting mechanism is introduced to dynamically balance the contributions of different source domains during model training, thereby enhancing the overall learning performance. Experimental results on two sets of bearing fault diagnosis tasks under time-varying rotational speed conditions demonstrate that the proposed method can effectively mitigate the impact of distribution discrepancies, significantly improving the accuracy and generalization capability of the diagnostic model, and verifying its potential and reliability in complex operating conditions.

Classification of Helicopter Transmission Condition in the Relative Spectral Domain under Unknown Sampling Frequency