Evaluation of vehicle vibration comfort using deep learning

Abstract Traditional objective methods for vehicle vibration comfort evaluation are insufficient as they use a simple calculated objective evaluation index for evaluating the degree of discomfort. The detailed vibration contents, for example, the frequency spectrum, have been proven to be critical but are not considered comprehensively by these methods. In this study, by using deep learning, a new method is developed to assess passenger car vibration comfort based on the detailed contents of vibration signals. Four passenger cars are tested under different velocities and road types to generate an initial dataset. To overcome the limitation of dataset size, a new data augmentation method is proposed and verified for deep learning, which includes the data segmentation and resampling. Two neural networks, namely the feed-forward neural network and gated recurrent units, are used to generate two basic models, where the gated recurrent units take the time sequence characteristic of vibration data into consideration. Besides, the frequency content is considered by adding a fast Fourier transform layer to the basic models to generate two new architectures. The results suggest that the proposed data augmentation method enlarges the dataset efficiently and makes it feasible to apply deep learning to ride comfort evaluation. Also, the time sequence and frequency contents both improved the model prediction accuracy significantly, while the former manifested a greater improvement. The gated recurrent units with the fast Fourier transform layer presents the highest performance to predict the degree of comfort for new vibration scenarios. To further improve the prediction accuracy and understand the rules to apply deep learning in ride comfort evaluation, the hyperparameters of the superior architecture are tuned via parametric study. After that, the prediction loss is reduced by 76.9%. The superior performance of the final model is also demonstrated by comparing it with a traditional method.

[1]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[2]  P. E. Uys,et al.  The ride comfort vs. handling compromise for off-road vehicles , 2007 .

[3]  Lukás Burget,et al.  Recurrent neural network based language model , 2010, INTERSPEECH.

[4]  Michael J. Griffin,et al.  Response of the seated human body to whole-body vertical vibration: biodynamic responses to sinusoidal and random vibration , 2014, Ergonomics.

[5]  Chong Wang,et al.  Deep Speech 2 : End-to-End Speech Recognition in English and Mandarin , 2015, ICML.

[6]  Michael J Griffin,et al.  Response of the seated human body to whole-body vertical vibration: discomfort caused by mechanical shocks , 2017, Ergonomics.

[7]  P. S. Els,et al.  The applicability of ride comfort standards to off-road vehicles , 2005 .

[8]  Wolfgang Lehner,et al.  Feature-based comparison and generation of time series , 2018, SSDBM.

[9]  Rakesh Chandmal Sharma.  Evaluation of Passenger Ride Comfort of Indian Rail and Road Vehicles with ISO 2631-1 Standards Part 2 - Simulation , 2016 .

[10]  Saeid Nahavandi,et al.  Directional and sectional ride comfort estimation using an integrated human biomechanical-seat foam model , 2017 .

[11]  Thomas Plagemann,et al.  Augmenting Physiological Time Series Data: A Case Study for Sleep Apnea Detection , 2019, ECML/PKDD.

[12]  Robert X. Gao,et al.  Deep learning and its applications to machine health monitoring , 2019, Mechanical Systems and Signal Processing.

[13]  Ying Huang,et al.  Study on the Evaluation Model of Vehicle Comfort Based on the Neural Network , 2018 .

[14]  Se Jin Park,et al.  Evaluating Methods of Vibration Exposure and Ride Comfort in Car , 2013 .

[15]  Joaquim R. R. A. Martins,et al.  Efficient Aerodynamic Shape Optimization with Deep-Learning-Based Geometric Filtering , 2020 .

[16]  Jie Liu,et al.  Mini-Batch Semi-Stochastic Gradient Descent in the Proximal Setting , 2015, IEEE Journal of Selected Topics in Signal Processing.

[17]  Mohd Jailani Mohd Nor,et al.  Evaluation of Whole-Body Vibration and Ride Comfort in a Passenger Car , 2009 .

[18]  Gianpiero Mastinu,et al.  Road and Off-Road Vehicle System Dynamics Handbook , 2013 .

[19]  Maria Lúcia Machado Duarte,et al.  Correlation between weighted acceleration, vibration dose value and exposure time on whole body vibration comfort levels evaluation , 2018 .

[20]  K Kato,et al.  Evaluation of seatback vibration based on ISO 2631-1 (1997) standard method: The influence of vehicle seat structural resonance , 2017, Ergonomics.

[21]  SchmidhuberJürgen.  Deep learning in neural networks , 2015 .

[22]  Jorge Nocedal,et al.  Optimization Methods for Large-Scale Machine Learning , 2016, SIAM Rev..

[23]  Tara N. Sainath,et al.  Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups , 2012, IEEE Signal Processing Magazine.

[24]  Shuai Wang,et al.  Deep learning for sentiment analysis: A survey , 2018, WIREs Data Mining Knowl. Discov..

[25]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[26]  Xiaogang Wang,et al.  Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identification , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[27]  Michael J. Griffin,et al.  Handbook of Human Vibration , 1990 .

[28]  Tara N. Sainath,et al.  Deep Neural Networks for Acoustic Modeling in Speech Recognition , 2012 .

[29]  Feng Li,et al.  GRATIS: GeneRAting TIme Series with diverse and controllable characteristics , 2019, Stat. Anal. Data Min..

[30]  Xiaodong Cui,et al.  Data Augmentation for Deep Neural Network Acoustic Modeling , 2015, IEEE/ACM Transactions on Audio, Speech, and Language Processing.

[31]  Neil J. Mansfield,et al.  Evaluation of subjective responses to whole-body vibration exposure: Effect of frequency content , 2008 .

[32]  Genki Yagawa,et al.  Computational mechanics enhanced by deep learning , 2017 .

[33]  Justin Salamon,et al.  Deep Convolutional Neural Networks and Data Augmentation for Environmental Sound Classification , 2016, IEEE Signal Processing Letters.

[34]  Chen Lu,et al.  Fault diagnosis of rotary machinery components using a stacked denoising autoencoder-based health state identification , 2017, Signal Process..

[35]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[36]  Setsuo Maeda,et al.  Evaluation of whole-body vibration by the category judgment method. , 2005, Industrial health.