Aug 2026· Sound & Vibration· 0 citations· 61 references
TL;DR
A convolutional neural network transformer-based prediction method for vehicle interior wind noise by integrating vehicle styling features and acoustic technical parameters is proposed, which was compared with convolutional neural network (CNN), long short-term memory (LSTM), and transformer models.
Abstract
The rapid evaluation of interior aerodynamic noise during the Concept A Surface design stage is important for vehicle acoustic development, but conventional methods are limited by high cost and low efficiency. This study proposes a convolutional neural network transformer-based prediction method for vehicle interior wind noise by integrating vehicle styling features and acoustic technical parameters. An optimal Latin hypercube sampling method was used to generate design combinations, and wind tunnel tests were conducted at 120 km/h. Key vehicle styling parameters, including A-pillar geometry, side mirror dimensions, front windscreen angle, side mirror-to-body spacing, and side window inclination, together with glazing material properties, glass thickness, acoustic transfer function, and interior reverberation time, were selected as input features to predict the driver’s left-ear wind noise spectrum. Based on five-fold cross-validation, the proposed model was compared with convolutional neural network (CNN), long short-term memory (LSTM), and transformer models. The CNN-Transformer model achieved the best performance, with mean absolute percentage error (MAPE) and root mean square error (RMSE) values of 2.23% and 0.94 dB, respectively. Compared with the transformer, CNN, and LSTM models, the proposed method reduced MAPE by 24.91%, 36.29%, and 49.59%, and reduced RMSE by 22.95%, 35.17%, and 48.07%, respectively. The model also maintained reliable performance on an independent test set, with MAPE and RMSE values of 4.82% and 1.44 dB. The mean impact value method was further applied to identify the influence of design parameters on interior wind noise, guiding for early vehicle acoustic optimization.
A temperature-perceptive SFANC (TP-SFANC) approach is proposed that employs a lightweight one-dimensional convolutional neural network (1D CNN) trained using a multi-task learning strategy to dynamically select the optimal control filter.
Boxiang Wang, M. Misol, Zheng-wu Luo et al.· 2 citations
This paper introduces a framework for airfoil inverse design, using a deep-learning approach and incorporating both the parametric section (PARSEC) and class–shape transformation (CST) airfoil parameterization techniques for comparative evaluation. A predictive model has been established to estimate the PARSEC and CST parameters that characterize airfoil shapes based on a range of aerodynamic coefficients as inputs. This model is constructed through the training of a neural network using an aerodynamic dataset generated from XFOIL simulations. The developed model demonstrates its capacity to predict airfoil geometries with reasonable accuracy, effectively linking aerodynamic coefficients with corresponding shapes. A multihead self-attention block is incorporated as a feature-interaction module to process correlated aerodynamic input descriptors before mapping them to geometry parameters. The findings, derived from both the PARSEC and CST parameterization techniques, indicate that the framework reconstructs airfoil geometries with good agreement across diverse airfoil families within the considered Reynolds number and angle-of-attack envelope.
G. Eriş, Ahmet Can Özgören, O. Uzol· Journal of Aircraft· 0 citations
This work proposes an efficient multilayer neural network framework to predict the pitching moment coefficient of canard-controlled missiles, significantly reducing the need for costly CFD data and providing a powerful surrogate tool for rapid design optimization based on trim angle of attack and geometric parameters.
M. Shojaeefard, Masoud Nobakhti· Proceedings of the Instituti...· 0 citations
Vehicle vibration and noise are predominantly driven by road-surface excitation, making robust prediction of these phenomena based on pavement roughness a central challenge in automotive development. This paper presents a fully automated pipeline that begins with high-resolution laser-triangulation scans of actual test tracks to produce centerline elevation profiles. These profiles are processed and classified by a MATLAB routine using ISO 8608-based power-spectral-density analysis to extract the Gh0 roughness coefficient. Concurrently, in-service acoustic and chassis-vibration data, collected at two representative speeds, are transformed into feature vectors comprising statistical PSD descriptors. A regression model then learns the mapping from these features to Gh0, evaluating the feasibility of mapping vehicle-borne signatures to roughness metrics. Predicted Gh0 values drive a profile-synthesis algorithm to generate two-dimensional height grids, which are exported as CRG files and imported into a multibody simulation software (MSC ADAMS) as well as driver-in-the-loop platforms. Simulation results closely reproduce the primary excitation characteristics of the physical tracks, demonstrating a preliminary proof-of-concept pipeline for virtual road surface generation. While the cross-validated regression model indicates limited generalization on the current small dataset (R2=−0.2783), the end-to-end workflow establishes the baseline integration required for future data-driven NVH simulation. To extend applicability beyond a single test vehicle, a set of Vehicle Calibration Transforms is proposed to adapt power-spectral-density features from arbitrary vehicles into the calibrated feature domain. The complete workflow promises to streamline virtual NVH validation, reduce prototype testing, and support full NVH simulator engineering in future research.
Christopher Pfeifer, Gerd Manthei· Applied Sciences· 0 citations
Shear-wave velocity (Vs), together with compressional wave velocity, provides a crucial source of information for both geomechanical and geophysical studies. Vs data are often unavailable. Moreover, direct measurement of Vs remains relatively costly. Four machine learning algorithms were created to predict Vs from traditional well logs in order to get around these restrictions: Probability Neural Network (PNN), Multilayer Feed-Forward Neural Network (MLFFNN), Deep Feed-Forward Neural Network (DFFNN), one-dimensional Convolutional Neural Network (1D-CNN), and an Integrated Convolutional Neural Network (I-CNN). The dataset consists of two wells (19,121 data points were gathered) of authentic industrial wireline logs from two anonymized wells, provided with formal authorization from the data owner exclusively for academic research purposes, including model training, testing, and validation. There were three primary parts in the methodology: (1) pre-processing the data to get rid of noise and change it into the right format; (2) using domain knowledge to drive feature engineering and selection; and (3) training, testing, and optimizing the model. The results demonstrated that the I-CNN model in RCW-1 well achieved the best performance, with an R2 value of 0.971. When applied to the blind well (RCW-2), the I-CNN model maintained strong generalization capability, achieving an average R2 value of 0.956. These findings indicate that the I-CNN outperforms other methods in handling complex, nonlinear relationships in Vs prediction. Overall, this study contributes to the growing body of literature on machine learning applications in petrophysical analysis by introducing an integrated deep learning framework that surpasses traditional approaches.
Rahmat Catur Wibowo, I. S. Yogi, Indra Arifianto et al.· Rudarsko-geološko-naftni Zbo...· 0 citations
High-precision acoustic propagation field prediction plays a crucial role in supporting decision-making for underwater sonar systems. Neural networks offer advantages in acoustic field forecasting due to their ability to adaptively adjust weights based on real experimental data. Simulation results have demonstrated that integrating physical computation into the forward propagation of neural networks can significantly improve prediction accuracy. Building on this foundation, the present study further introduces a terrain feature clustering module to mitigate abrupt changes in the predicted acoustic field caused by sharp variations in topographic features, thereby enhancing the stability of the training process. Additionally, this study addresses acoustic field prediction under uncertainty in sound speed profiles (SSPs). An empirical orthogonal function is employed to extract features from SSPs, which are then incorporated as input to the neural network model. Expanding upon prior work, a dedicated analysis module for SSP features is introduced, and a staged training strategy is adopted for model optimization. Furthermore, the statistical uncertainty in acoustic field predictions is examined. Finally, model training and testing are further refined using experimental data collected from the South China Sea. Experimental results indicate that the inclusion of the terrain feature clustering module improves the average prediction accuracy of the final acoustic field by 0.3 dB. Compared with experimental data, the proposed neural network model achieves an accuracy improvement of 1.78 dB over numerical solutions based on coupled normal modes.
Xiao Feng, Kunde Yang, Minghui Li· IEEE Signal Processing Lette...· 0 citations