Airborne Acoustic Emission Research Articles

Unsupervised quality monitoring of metal additive manufacturing using Bayesian adaptive resonance

Metal additive manufacturing is a recent breakthrough technology that promises automated production of complex geometric shapes at low operating costs. However, its potential is not yet fully exploited due to the low reproducibility of quality in mass production. The monitoring of parts quality directly during manufacturing promises to solve this problem, while machine learning showed efficient performance correlating versatile manufacturing measurements with different quality grades. Today, most monitoring algorithms are based on semi- or supervised learning, thus, requiring a collection and ground-truth validation of training sets. This is costly and time consuming in real-life conditions. Our work is a feasibility study of the application of unsupervised machine learning to monitor different manufacturing regimes and quality in metal additive manufacturing. The algorithm combines the kernel Bayes rule for inference and Bayesian adaptive resonance for structuring the incoming data. Airborne acoustic emission from laser powder bed fusion is used as an algorithm input. The recognition of the main manufacturing regimes (conduction mode, stable, and unstable keyholes) are shown on real-life data, while the self-learning accuracy of developed algorithm exceeds 88 %. Our approach promises future development of plug-and-play quality monitoring systems for laser processing technology, requiring minimum modifications of the existing machines, reducing time/cost for algorithm preparation and providing continuous data driven adaptation of the algorithm to changes in manufacturing conditions.

Inference of highly time-resolved melt pool visual characteristics and spatially-dependent lack-of-fusion defects in laser powder bed fusion using acoustic and thermal emission data

With a growing demand for high-quality fabrication, the interest in real-time process and defect monitoring of laser powder bed fusion (LPBF) has increased, leading manufacturers to incorporate a variety of online sensing methods including acoustic sensing, photodiode sensing, and high-speed imaging. However, real-time acquisition of high-resolution melt pool images in particular remains computationally demanding in practice due to the high variability of melt pool morphologies and the limitation of data caching and transfer, making it challenging to detect the local lack-of-fusion (LOF) defect occurrences. In this work, we propose a new acoustic and thermal information-based monitoring method that can robustly infer critical LPBF melt pool morphological features in image forms and detect spatially-dependent LOF defects within a short time period. We utilize wavelet scalogram matrices of acoustic and photodiode clip data to identify and predict highly time-resolved (within a 1.0 ms window) visual melt pool characteristics via a well-trained data-driven pipeline. With merely the acoustic and photodiode-collected thermal emission data as the input, the proposed pipeline enables data-driven inference and tracking of highly variable melt pool visual characteristics with R2≥0.8. We subsequently validate our proposed approach to infer local LOF defects between two adjacent scanlines, showing that our proposed approach can outperform our selected baseline theoretical model based on previous literature. Revealing the physical correlation between airborne acoustic emission, thermal emission, and melt pool morphology, our work demonstrates the feasibility of creating an efficient and cost-effective acoustic- and thermal-based approach to facilitate online visual melt pool characterization and LOF defect detection. We believe that our work can further contribute to the advances in quality control for LPBF.

Process comparison of laser deep penetration welding in pure nickel using blue and infrared wavelengths

AbstractLaser sources with wavelengths in the visible blue spectrum are suitable for heat conduction mode welding of materials like copper and nickel due to the significantly increased Fresnel absorption compared to infrared laser radiation. Recently, blue laser sources with 445 nm wavelength have become available with sufficient power and beam parameters to exceed the intensity threshold for laser deep penetration welding. In laser beam deep penetration welding, the total absorption is significantly increased due to the multiple reflections in the keyhole compared to heat conduction mode welding. However, since the absorbed energy per reflection inside the keyhole is wavelength-dependent, it can be hypothesized that the choice of laser wavelength causes changes in the local energy distribution inside the keyhole, changing the keyhole dynamics. To investigate this, laser beam deep penetration welding experiments were carried out on 2.4068 pure nickel using an infrared laser source and a blue laser source with comparable beam properties. The experiments were monitored and compared by a multi-sensor setup and metallographic analyses. This setup included measurements of airborne acoustic emissions and two high-speed video cameras for spatter tracking and tracking of the keyhole area. The use of a blue laser beam led to a lower spatter quantity, an increase of porosity and a significant change of acoustic emissions, proving the hypothesis for pure nickel.

Localized defect frequencies for Laser Metal Deposition processes

Laser Metal Deposition (LMD) uses laser energy and powder material to create structures on existing components. It is capable of producing cost-effective multi-material compositions, such as reinforcing metals with ceramic particles for improved wear resistance. However, the use of dissimilar materials often leads to defects, particularly delamination. Previous studies have found a connection between these defects and specific airborne acoustic emissions (AE).To mitigate the impact of defects, extensive optimization of process parameters and real-time process monitoring are recommended. For AE, precise localization of defects is crucial besides to time- and frequency-resolved information, especially while producing multiple components on a substrate material.This study evaluates multi-sensor arrays for the localization of delamination defects. The research investigates the influence of localization algorithms and array patterns on the accuracy and reliability of defect localization. Experiments were conducted on a test platform with simulated acoustical events to determine the most suitable localization setup.

Self-Supervised Bayesian representation learning of acoustic emissions from laser powder bed Fusion process for in-situ monitoring

This study presents a self-supervised Bayesian Neural Network (BNN) framework using air-borne Acoustic Emission (AE) to identify different Laser Powder Bed Fusion (LPBF) process regimes such as Lack of Fusion, conduction mode, and keyhole without ground-truth information. The proposed framework addresses the challenge of labelling datasets with semantic complexities into discrete process dynamics. This novel AE-based in-situ monitoring approach provides a promising alternative to quantify part density in LPBF process. The study demonstrates the effectiveness of a Bayesian encoder backbone for learning the manifold representations of LPBF regimes, which were visually separable in a lower-dimensional representation using t-distributed stochastic neighbour embedding. The generalized representations learned by the Bayesian backbone allowed traditional classifiers trained on smaller datasets to exhibit high classification accuracy. The feature map computed using pre-trained Bayesian encoder on other datasets was also effective in anomaly detection, achieving 92% accuracy with one-class Support Vector Machine. Additionally, the representation learned by the BNN facilitates transfer learning, where it can be fine-tuned for classification tasks on different process maps, which is also demonstrated in this work. Our proposed framework improves the generalization and robustness of the LPBF monitoring, particularly in the face of varying data distribution across multiple process parameter spaces.

Monitoring of Joint Gap Formation in Laser Beam Butt Welding using Neural Network-Based Acoustic Emission Analysis

This study aimed to explore the feasibility of using airborne acoustic emission in laser beam butt welding for the development of an automated classification system based on neural networks. The focus was on monitoring the formation of joint gaps during the welding process. To simulate various sizes of butt joint gaps, controlled welding experiments were conducted, and the emitted acoustic signals were captured using audible-to-ultrasonic microphones. To implement an automated monitoring system, a method based on short-time Fourier transformation was developed to extract audio features, and a convolutional neural network architecture with data augmentation was utilized. The results demonstrated that this non-destructive and non-invasive approach was highly effective in detecting joint gap formations, achieving an accuracy of 98%. Furthermore, the system exhibited promising potential for the low-latency monitoring of the welding process. The classification accuracy for various gap sizes reached up to 90%, providing valuable insights for characterizing and categorizing joint gaps accurately. Additionally, increasing the quantity of training data with quality annotations could potentially improve the classifier model’s performance further. This suggests that there is room for future enhancements in the study.

A novel approach of online monitoring for laser powder bed fusion defects: Air-borne Acoustic Emission and Deep Transfer Learning

Defects in laser powder bed fusion (L-PBF) are a serious constraint on the application of the technology. Developing a real-time monitoring technology to guide the production process of parts can solve this problem. Currently, the quality of L-PBF with different scanning strategies varies greatly. Therefore, the online monitoring methods for a specific scanning strategy cannot be effectively generalized to other scanning strategies. This is a rarely investigated topic in L-PBF, and defects monitoring mechanism research is insufficient to effectively support the study of L-PBF defects monitoring technology utilizing advanced sensing technologies. Towards this end, we explored the acoustic source generation mechanism, analyzed the acoustic monitoring principle of L-PBF defects, and hence, we propose a novel online monitoring method for L-PBF defects based on air-borne acoustic emission (ABAE) and deep transfer learning (DTL). The method uses time-frequency spectrograms of acoustic signals as the input to the network, and a method of deep transfer learning with multi-source domains knowledge fusion (DTL-MDKF) is proposed to realize the classification of defects. The proposed method was compared with the traditional transfer learning method based on single-source domain knowledge. The results showed that the classification accuracy of the proposed method for L-PBF defects is 98.2 %. In addition, the feature mining capability of the DTL-MDKF is demonstrated by visualizing the features of transfer learning models with different knowledge. Looking at the results, the proposed method can be considered a promising L-PBF defects online monitoring method for complex and changeable working conditions.

Improved Water Sound Synthesis using Coupled Bubbles

We introduce a practical framework for synthesizing bubble-based water sounds that captures the rich inter-bubble coupling effects responsible for low-frequency acoustic emissions from bubble clouds. We propose coupled-bubble oscillator models with regularized singularities, and techniques to reduce the computational cost of time stepping with dense, time-varying mass matrices. Airborne acoustic emissions are estimated using finite-difference time-domain (FDTD) methods. We propose a simple, analytical surface-acceleration model, and a sample-and-hold GPU wavesolver that is simple and faster than prior CPU wavesolvers. Sound synthesis results are demonstrated using bubbly flows from incompressible, two-phase simulations, as well as procedurally generated examples using single-phase FLIP fluid animations. Our results demonstrate sound simulations with hundreds of thousands of bubbles, and perceptually significant frequency transformations with fuller low-frequency content.

Tool Health Monitoring of a Milling Process Using Acoustic Emissions and a ResNet Deep Learning Model

In the industrial sector, tool health monitoring has taken on significant importance due to its ability to save labor costs, time, and waste. The approach used in this research uses spectrograms of airborne acoustic emission data and a convolutional neural network variation called the Residual Network to monitor the tool health of an end-milling machine. The dataset was created using three different types of cutting tools: new, moderately used, and worn out. For various cut depths, the acoustic emission signals generated by these tools were recorded. The cuts ranged from 1 mm to 3 mm in depth. In the experiment, two distinct kinds of wood-hardwood (Pine) and softwood (Himalayan Spruce)-were employed. For each example, 28 samples totaling 10 s were captured. The trained model's prediction accuracy was evaluated using 710 samples, and the results showed an overall classification accuracy of 99.7%. The model's total testing accuracy was 100% for classifying hardwood and 99.5% for classifying softwood.

Airborne acoustic emission of an abrasive waterjet cutting system as means for monitoring the jet cutting capability

Abrasive waterjet cutting is a manufacturing technology making use of a high-speed waterjet with abrasive particles in suspension, for cutting materials with different mechanical properties. Product quality requirements are pushing towards an improvement of tracking and stabilization methods of the relevant process variables. Amongst those, the jet kinetic power defines the cutting capability and has a significant impact on the final cut features. This variable is subject to relevant fluctuations versus time. Besides, the current state of the art does not provide means for its in-line monitoring. The aim of this contribution is to monitor the airborne acoustic emission of an abrasive waterjet cutting head and investigate its correlation with the jet kinetic power. The investigation is carried out by means of factorial studies, in which the jet is fired at various water pressures and abrasive feed rates, providing different kinetic powers. The acoustic emission is synchronously monitored by means of a condenser microphone, installed on the cutting head. Data at frequencies above 40 kHz is found to constitute a robust and selective acoustic signature of the airborne jet. The acoustic signature is proven to be an effective in-line indicator of the jet kinetic power and its pressure-induced variations, whilst abrasive-induced variations remain undetected. A calibration procedure is presented, for translating the acoustic data into a jet kinetic power. The method is validated by means of further experiments that envisage its deployment in a real scenario. Overall, the presented method constitutes a robust tool for monitoring pressure-induced variations of the jet cutting capability.

Differentiation of materials and laser powder bed fusion processing regimes from airborne acoustic emission combined with machine learning

ABSTRACT This study investigates the use of a low cost microphone combined with state-of-the-art machine learning (ML) algorithms as online process monitoring to differentiate various materials and process regimes of Laser-Powder Bed Fusion (LPBF). Three processing regimes (lack of fusion pores, conduction mode and keyhole pores) and three alloys (316L stainless steel, bronze (CuSn8), and Inconel 718) were selected. Three conventional ML algorithms and a Convolutional Neural Network (CNN) were chosen to perform the classification tasks resulting in five main findings. First, we proved that the AE features are related to the laser-material interaction and not from undesired machine or environmental noise. Second, the process regimes are classified with high accuracy (> 87%) regardless of the algorithms and materials. Third, it is possible to build a single model from the three materials and still reach high classification accuracy (>86%) of the different regimes. Forth, the AE features used for the classifications are material and regime dependent. Finally, with LPBF processing of multi-materials on the rise, a strategy for classifying the material and the process regimes simultaneously using a CNN multi-label architecture reached a very high classification accuracy (≈ 93%). The results demonstrate the potential of our approaches for online LPBF process monitoring of different materials and regimes.

Signal processing of airborne acoustic emissions from laser metal deposited structures

Laser Metal Deposition (LMD) is an additive manufacturing process that enables the production of complex structures on existing parts. To reduce the use of costly and energy-intensive produced materials like titanium, these structures can be selectively deposited on favorable but differing substrate materials. However due to differing material characteristics, this results in a high defect potential.Extensive process parameter development and online process monitoring helps to minimize the impact of defect formation. Monitoring of airborne acoustic process emissions can be used to identify critical defects early during the process and gain information about process stability.Two approaches for in-process time-frequency monitoring of the acquired acoustic data were compared within this work: The short time Fast Fourier Transformation and the Continuous Wavelet Transformation. Performance criteria based on detected distinct defect events were defined to evaluate both approaches and define specifications for an Acoustic Emission In-Process Monitoring system.

A Hybrid Approach for Noise Reduction in Acoustic Signal of Machining Process Using Neural Networks and ARMA Model.

Intelligent machining has become an important part of manufacturing systems because of the increased demand for productivity. Tool condition monitoring is an integral part of these systems. Airborne acoustic emission from the machining process is a vital indicator of tool health, however, it is highly affected by background noise. Reducing the background noise helps in developing a low-cost system. In this research work, a feedforward neural network is used as an adaptive filter to reduce the background noise. Acoustic signals from four different machines in the background are acquired and are introduced to a machining signal at different speeds and feed-rates at a constant depth of cut. These four machines are a three-axis milling machine, a four-axis mini-milling machine, a variable speed DC motor, and a grinding machine. The backpropagation neural network shows an accuracy of 75.82% in classifying the background noise. To reconstruct the filtered signal, a novel autoregressive moving average (ARMA)-based algorithm is proposed. An average increase of 71.3% in signal-to-noise ratio (SNR) is found before and after signal reconstruction. The proposed technique shows promising results for signal reconstruction for the machining process.

Resonant Airborne Acoustic Emission for Nondestructive Testing and Defect Imaging in Composites

A new version of an acoustic emission mode which is different from its traditional counterpart is discussed in view of applications for nondestructive testing. It is based on the effect of acoustic waves generation from the defect area in ambient air by local standing wave vibration developed in this area at the defect resonant frequency. Another approach which does not require preliminary knowledge of local defect-resonance frequency is one that uses wideband acoustic activation by a noise-like input signal. The acoustic emission field from the defect area is a “fingerprint” of the radiation source, and thus is applicable to defect detection and imaging. This enables the use of commercial microphone scanning for detecting and imaging various defects in composites. An improvement in the acoustic-emission scanning mode based on a multiple-axis robot is studied to applications to complex shape components. A rapid, full-field imaging of the acoustic-emission field is implemented by means of an array of microphones (acoustic camera). Numerous case studies validate the potential of the resonant acoustic-emission modes for integration in the defect imaging system based on inexpensive, fully acoustic instrumental components.

Mutual comparison of acoustic, pyrometric and thermographic laser powder bed fusion monitoring

This paper compares three distinctive sensors for laser powder bed fusion metal additive manufacturing process monitoring. A microphone for airborne acoustic emissions, an on-axis two-colour pyrometer for melt pool temperature measurement and an off-axis thermographic camera are simultaneously applied. They are challenged with a large build area to investigate their robustness and sensitivity. This paper does not assess the sensors’ ability to detect specific process flaws, but instead gives a common ground comparison of general sensor characteristics. The camera provides a descriptive result in form of a heat-map, while it exhibits a lack of sensitivity. In contrast, the microphone presents a sensitivity up to 40 times higher than the camera and is still 15 times more sensitive than the pyrometer. However, with this comes increased susceptibility; its signal strength is strongly dependent on the distance to the melt pool as a result of frequency dependent dissipation. The pyrometer’s signal is sensitive enough for relevant process deviations to be uncovered, while being robust towards different sensing distances. Recommendations are given for successful implementation of the sensors. Additionally, novel process phenomena were uncovered: an interaction of the scanning direction with the shielding gas is discussed, plus insights regarding overhang scanning are acquired.

Detection and analysis of photo-acoustic emission in Direct Laser Interference Patterning

Functional laser texturing by means of Direct Laser Interference Patterning is one of the most efficient approaches to fabricate well-defined micro textures which mimic natural surfaces, such as the lotus effect for self-cleaning properties or shark skin for reduced friction. While numerous technical and theoretical improvements have been demonstrated, strategies for process monitoring are yet to be implemented in DLIP, for instance aiming to treat complex and non-plane surfaces. Over the last 35 years, it has been shown that the sound pressure generated by a laser beam hitting a surface and producing ablation can be detected and analysed using simple and commercially available transducers and microphones. This work describes the detection and analysis of photo-acoustic signals acquired from airborne acoustic emission during DLIP as a direct result of the laser–material interaction. The study includes the characterization of the acoustic emission during the fabrication of line-like micro textures with different spatial periods and depths, the interpretation the spectral signatures deriving from single spot and interference ablation, as well as a detailed investigation of the vertical extent of the interference effect based on the ablated area and its variation with the interference period. The results show the possibility to develop an autofocusing system using only the signals from the acoustic emission for 3D processing, as well as the possibility to predict deviations in the DLIP processing parameters.

Tool Health Monitoring Using Airborne Acoustic Emission and Convolutional Neural Networks: A Deep Learning Approach

Tool health monitoring (THM) is in great focus nowadays from the perspective of predictive maintenance. It prevents the increased downtime due to breakdown maintenance, resulting in reduced production cost. The paper provides a novel approach to monitoring the tool health of a computer numeric control (CNC) machine for a turning process using airborne acoustic emission (AE) and convolutional neural networks (CNN). Three different work-pieces of aluminum, mild steel, and Teflon are used in experimentation to classify the health of carbide and high-speed steel (HSS) tools into three categories of new, average (used), and worn-out tool. Acoustic signals from the machining process are used to produce time–frequency spectrograms and then fed to a tri-layered CNN architecture that has been carefully crafted for high accuracies and faster trainings. Different sizes and numbers of convolutional filters, in different combinations, are used for multiple trainings to compare the classification accuracy. A CNN architecture with four filters, each of size 5 × 5, gives best results for all cases with a classification average accuracy of 99.2%. The proposed approach provides promising results for tool health monitoring of a turning process using airborne acoustic emission.

Classification of specimen density in Laser Powder Bed Fusion (L-PBF) using in-process structure-borne acoustic process emissions

Currently, the laser powder bed fusion (L-PBF) process cannot offer a reproducible and predefined quality of the processed parts. Recent research on process monitoring focuses strongly on integrated optical measurement technology. Besides optical sensors, acoustic sensors also seem promising. Previous studies have shown the potential of analyzing structure-borne and air-borne acoustic emissions in laser welding. Only a few works evaluate the potential that lies in the usage during the L-PBF process.This work shows how the approach to structure-borne acoustic process monitoring can be elaborated by correlating acoustic signals to statistical values indicating part quality. Density measurements according to Archimedes’ principle are used to label the layer-based acoustic data and to measure the quality. The data set is then treated as a classification problem while investigating the applicability of existing artificial neural network algorithms to match acoustic data with density measurements. Furthermore, this work investigates the transferability of the approach to more complex specimens.

Aeroacoustics and Artificial Neural Network Modeling of Airborne Acoustic Emissions During High Kinetic Energy Thermal Spraying

This work describes an online, non-destructive monitoring technology for thermal spray coating processes based on the airborne acoustic emissions (AAE) in the booth. First, numerical simulations were carried out to probe into the relationship between AAE signals and the frequency spectrum generated during high velocity-oxy-fuel thermal spray. The experimental part consisted of spraying a plane substrate. The torch was traversed in front of the substrate at a constant speed, 90° impact angle and for different combinations of standoff distance and powder feed rate. The AAE signals were acquired using a broadband piezoelectric sensor positioned at a fixed point near the torch, and the experimental power spectrum of the signal was processed and compared with model predictions. A neural network-based model was implemented capturing and representing the complex relationships between the power spectrum of the AAE and the resulting coating microhardness. The research outcomes demonstrate that the sound contains detectable information associated with spray parameters such as powder feed rate, spray distance and the resulting coating microhardness. The proposed technology can be used to detect process flaws so that deviations from the optimum spraying conditions can be detected and corrected promptly.

A neural network based approach for background noise reduction in airborne acoustic emission of a machining process

Tool wear prediction has become an indispensable technique to prevent downtime in manufacturing and production processes. Airborne emission from a machining process using a low-cost microphone may provide a vital signal of tool health. However, the effect of background noise results in anomaly in data that may lead to wrong prediction of tool health. The paper presents an adaptive approach using neural networks for background noise filtration in acoustic signal for a turning process. Acoustic signal of a turning process is mixed with background noise from four different machines and introduced at different RPMs and feed-rate at a constant depth of cut. A comparison of Backpropagation neural network (BPNN), Self-organizing map and k-means clustering algorithm for noise filtration is investigated in this paper. In this regard, back-propagation neural network showed better performance with an average accuracy for all the four sources. It shows 100 % accuracy for grinding machine signal, 94.78 % accuracy for background signal from 3-axis milling machine, 45.57 % and 12.69 % for motor and 4-axis milling machine, respectively. Signal reconstruction is then done using Discrete cosine transform (DCT). The proposed technique shows a promising future for noise filtration in airborne acoustic data of a machining process.