Bearing Fault Diagnosis under Variable Speed Using Convolutional Neural Networks and the Stochastic Diagonal Levenberg-Marquardt Algorithm

Tra, Viet; Kim, Jae Young; Khan, Sheraz Ali; Kim, Jong‐Myon

doi:10.3390/s17122834

Cited by 56 publications

(41 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the recent advances in structural health monitoring (SHM) [2][3][4][5][6][7][8] and nondestructive evaluation (NDE) [9,10], many monitoring and evaluation methods, such as modal characteristics-based methods [11][12][13], active sensing [14][15][16][17][18][19][20], electromechanical impedance [21][22][23][24][25], ultrasonic guided wave [26], active thermography [26][27][28] and vibrothermography [29,30], among others, are available. Among them, acoustic emission technique [31][32][33][34][35][36][37] is an effective nondestructive technique that can characterize the wear process [38,39], and it has been widely used in civil engineering [40][41][42][43], mechanical engineering [44][45][46][47]<...>…”

Section: Introductionmentioning

confidence: 99%

Feasibility Study of Real-Time Monitoring of Pin Connection Wear Using Acoustic Emission

et al. 2018

View full text Add to dashboard Cite

Pin connections are one of the most important connecting forms and they have been widely used in engineering fields. In its service, pin connections are subject to wear, and it will be beneficial if the health condition of pin connections can be monitored in real time. In this paper, an acoustic emission (AE)-based method was developed to monitor wear degree of low rotational speed pin connections in real time in a nondestructive way. Most pin connections are operated at low rotational speed. To facilitate the research, an experimental apparatus to accelerate the wear test of low rotational speed pin connections was designed and fabricated. The piezoceramic AE sensor was mounted on the test apparatus in a nondestructive way, and it was capable of real-time monitoring. Accelerated wear tests of low rotational speed pin connections were conducted. To verify the results of the AE technique, a VHX-600E digital (from Keyence, Osaka, Japan) microscope was applied to observe the micrographs of the tested pins. The experimental results show that AE activity existed throughout the entire wear process, and it was the most prominent in the serious wear phase. The wear degree of the pin connections can be reflected qualitatively by the signal strength and the accumulative signal strength of the AE signals. In addition, two different wear forms can be distinguished by comparing the signal strength values of all specimens. Micrographs of all specimens confirm these results, and determine that the two wear forms include adhesive wear and abrasive wear. Furthermore, AE results demonstrated that adhesive wear is the main mode of wear for the low rotational speed pin connections, and the signal strength of the adhesive wear is around 190 times larger than that of abrasive wear. This feasibility study demonstrated that the developed acoustic emission technique can be utilized in the wear monitoring of pin connections in real time in a nondestructive way.

show abstract

Section: Introductionmentioning

confidence: 99%

Feasibility Study of Real-Time Monitoring of Pin Connection Wear Using Acoustic Emission

et al. 2018

View full text Add to dashboard Cite

show abstract

“…A CNN has a hierarchical construction and collects from several convolutional, subsampling, and fully connected layers. This network makes optimal use of the indigenous connections (instead of fully-connected layers), weight distribution, and spatial or progressive subsampling to achieve invariance of shifting, scaling, and distortion in their inputs [39,40]. In this study, as TL and CNN are designed together, the learning of the well-trained u layers are passed to the v layers of the target network, where u < v, v = u + 1, and layer number u + 1 denotes the output layer or last layer.…”

Section: Transfer Learning With Convolutional Neural Network (Cnn)mentioning

confidence: 99%

Bearing Fault Diagnosis under Variable Rotational Speeds Using Stockwell Transform-Based Vibration Imaging and Transfer Learning

Hasan

Kim

2018

Applied Sciences

Self Cite

View full text Add to dashboard Cite

In this paper, discrete orthonormal Stockwell transform (DOST)-based vibration imaging is proposed as a preprocessing step for supporting load and rotational speed invariant scenarios for signals of various health conditions. For any health condition, features can easily be extracted from its generated health pattern. To automate the feature selection process, a convolutional neural network (CNN)-based transfer learning (TL) approach for diagnosis has also been introduced. Transfer learning allows an established model to use feature knowledge obtained under one set of working conditions through hidden layers to diagnose faults that occur under other working conditions. The network learns from the massive source dataset, and that knowledge is applied to the target data to identify faults. Using the bearing dataset of Case Western Reserve University, the proposed approach yields an average 99.8% classification accuracy and, specifically, 99.99% for healthy condition (HC), 99.95% for inner race fault (IRF), 99.96% for ball fault (BF), 99.68% for outer race fault for 12 o’clock sensor position (ORF@12), 99.93% for outer race fault for 3 o’clock sensor position (ORF@3), and 99.89% for outer race fault for 6 o’clock sensor position (ORF@6). In this paper, the proposed approach is compared with conventional artificial neural networks (ANNs), support vector machines (SVMs), hierarchical CNNs, and deep autoencoders. The proposed approach outperforms these conventional methods in the accuracy under all working conditions.

show abstract

“…Secondly, the convolutional network makes it practical to stack layers, which enables the convolutional network to detect higher-level features. What is more, the recognition of pulse sound signals of different complexity under various working conditions can be realized by adding or reducing the convolutional layers without changing the structure of other parts [13,14,15,16]. …”

Section: The Intelligent Monitoring Networkmentioning

confidence: 99%

An Intelligent Monitoring Network for Detection of Cracks in Anvils of High-Press Apparatus

Tian

Zeng

Jiang

2018

Sensors

View full text Add to dashboard Cite

Due to the endurance of alternating high pressure and temperature, the carbide anvils of the high-press apparatus, which are widely used in the synthetic diamond industry, are prone to crack. In this paper, an acoustic method is used to monitor the crack events, and the intelligent monitoring network is proposed to classify the sound samples. The pulse sound signals produced by such cracking are first extracted based on a short-time energy threshold. Then, the signals are processed with the proposed intelligent monitoring network to identify the operation condition of the anvil of the high-pressure apparatus. The monitoring network is an improved convolutional neural network that solves the problems that may occur in practice. The length of pulse sound excited by the crack growth is variable, so a spatial pyramid pooling layer is adopted to solve the variable-length input problem. An adaptive weighted algorithm for loss function is proposed in this method to handle the class imbalance problem. The good performance regarding the accuracy and balance of the proposed intelligent monitoring network is validated through the experiments finally.

show abstract

Bearing Fault Diagnosis under Variable Speed Using Convolutional Neural Networks and the Stochastic Diagonal Levenberg-Marquardt Algorithm

Cited by 56 publications

References 32 publications

Feasibility Study of Real-Time Monitoring of Pin Connection Wear Using Acoustic Emission

Feasibility Study of Real-Time Monitoring of Pin Connection Wear Using Acoustic Emission

Bearing Fault Diagnosis under Variable Rotational Speeds Using Stockwell Transform-Based Vibration Imaging and Transfer Learning

An Intelligent Monitoring Network for Detection of Cracks in Anvils of High-Press Apparatus

Contact Info

Product

Resources

About