Chatter is a self-excited vibration that frequently occurs in thin-walled parts milling, which has become a major limitation to productivity and quality. Additionally, convolutional neural network ...(CNN) has been widely used in detection and classification, but the accuracy and convergence are affected by the initial weight and hyperparameters. Therefore, based on CNN, a method of chatter detection for the milling of thin-walled parts is proposed, which is realized by recognizing the image of the machined surface. First, aiming at the challenges of neural networks in which the weight is randomly initialized, a weight initialization method is proposed based on an improved magnetic bacteria optimization algorithm. The optimal magnetosome of each generation is used to adjust the magnetic moment of the next generation so that the population approaches the optimal solution direction to search for the global optimal value. Second, an improved genetic algorithm is proposed to optimize the network structure and improve the optimization efficiency of hyperparameters. The tabu list and the hill climbing algorithm are introduced into the improved genetic algorithm to avoid repeatedly counting the fitness of the same point and solving the oscillation problem near the optimal solution. The experimental results show that the accuracy, the value of the Matthew correlation coefficient, and the F
1
Score value of proposed CNN are 98.3%, 95.5%, and 98.8%, respectively. Compared with other algorithms, the proposed method, which has outstanding recognition performance, is competent in contactless chatter detection.
Full text
Available for:
DOBA, EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, IZUM, KILJ, KISLJ, MFDPS, NLZOH, NUK, OBVAL, OILJ, PILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, SIK, UILJ, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
Machine health management is one of the main research contents of PHM technology, which aims to monitor the health states of machines online and evaluate degradation stages through real-time sensor ...data. In recent years, classic sparsity measures such as kurtosis, Lp/Lq norm, pq-mean, smoothness index, negative entropy, and Gini index have been widely used to characterize the impulsivity of repetitive transients. Since smoothness index and negative entropy were proposed, the sparse properties have not been fully analyzed. The first contribution of this paper is to analyze six properties of smoothness index and negative entropy. In addition, this paper conducts a thorough investigation on multivariate power average function and finds that existing classical sparsity measures can be respectively reformulated as the ratio of multivariate power mean functions (MPMFs). Finally, a general paradigm of index design is proposed for the expansion of sparsity measures family, and several newly designed dimensionless health indexes are given as examples. Two different run-to-failure bearing datasets were used to analyze and validate the capabilities and advantages of the newly designed health indexes. Experimental results prove that the newly designed health indexes show good performance in terms of monotonic degradation description, first fault occurrence time determination and degradation state assessment.
Full text
Available for:
EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, KILJ, KISLJ, MFDPS, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
•A metric adversarial domain adaptation approach is proposed to successfully achieve cross-domain RUL prediction.•A feature extraction scheme with a dual self-attention module is developed to learn ...features with multi-scale semantics.•A supervised positive contrastive module is designed to maximize the target-specific mutual information.
Many existing domain adaptation-based methods try to derive domain invariant features to address domain shifts and obtain satisfactory remaining useful life (RUL) of bearings under multiple working conditions. However, most methods may not consider local semantics about degradation features and mutual information from target-specific data when aligning distribution discrepancies, thus resulting in limitations. Additionally, the use of contrastive learning to maintain mutual information may introduce unstable negative samples. To overcome these issues, a metric adversarial domain adaptation approach (MADA) is proposed to evaluate the bearing RULs under multiple working conditions. More specifically, an adversarial domain adaptation architecture with a supervised positive contrastive module is developed to consider mutual information without a negative sample, further learning domain invariant features. Also, the dual self-attention module is designed to extract multi-scale contextual semantics between degradation features. Meanwhile, extensive experiments are conducted in twelve cross-domain scenarios for two bearing cases. The experimental results show that the proposed method is more competitive.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
Data-driven approaches for prognostic and health management (PHM) increasingly rely on massive historical data, yet annotations are expensive and time-consuming. Learning approaches that utilize ...semi-labeled or unlabeled data are becoming increasingly popular. In this paper, a self-supervised pre-training via contrast learning (SSPCL) is introduced to learn discriminative representations from unlabeled bearing datasets. Specifically, the SSPCL employs momentum contrast learning (MCL) to investigate the local representation in terms of instance-level discrimination contrast. Further, we propose a specific architecture for SSPCL deployment on bearing vibration signals by presenting several data augmentations for 1D sequences. On this basis, we put forward an incipient fault detection method based on SSPCL for run-to-failure cycle of rolling bearings. This approach transfers the SSPCL pre-trained model to a specific semi-supervised downstream task, effectively utilizing all unlabeled data and relying on only a little priori knowledge. A case study on FEMTO-ST datasets shows that the fine-tuned model is competent for incipient fault detection, outperforming other state-of-the-art methods. Furthermore, a supplemental case on a self-built fault datasets further demonstrate the great potential and superiority of our proposed SSPCL method in PHM.
•A self-supervised pretraining via contrast learning (SSPCL) is introduced.•SSPCL implementation on vibration signals with 1D data augmentation.•Incipient fault detection framework based on SSPCL is detailed.•Experimental case studies verified validity and superiority.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
The novelty of this research is as follows:(1)A new transportable cross-domain adaptation method is developed.(2)A residual self-attention is proposed to fully consider the contextual degradation ...information.(3)The contrastive loss is introduced to improve the complex transformation invariance of MK-MMD mapping.
Many data-driven models normally assume that the training and test data are independent and identically distributed to predict the remaining useful life (RUL) of industrial machines. However, different failure models caused by variable failure behaviors may lead to a domain shift. Meanwhile, conventional methods lack comprehensive attention to temporal information, resulting in a limitation. To handle the aforementioned challenges, a transferable cross-domain approach for RUL estimation is proposed. The hidden features are extracted adaptively by a temporal convolution network in which residual self-attention is able to fully consider the contextual degradation information. Furthermore, a new cross-domain adaption architecture with the contrastive loss and multi-kernel maximum mean discrepancy is designed to learn the domain invariant features. The effectiveness and superiority of the proposed method are proved by the case study on IEEE PHM challenge 2012 bearing dataset and the comparison with other methods.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
The tremendous success of deep learning and transfer learning broadened the scope of fault diagnosis, especially the latter further improved the diagnosis accuracy under multiple working conditions. ...However, most existing attempts assume that label distribution is domain-invariant despite taking into account the different feature distributions. This does not accommodate the diversity of fault mode distributions under different operating conditions and weakens the generalization to imbalanced domain adaptation (IDA) scenarios. Therefore, this work proposed a novel deep imbalanced domain adaptation (DIDA) framework for fault diagnosis of bearings, aiming at the challenging scenario where feature shift and label shift exist simultaneously under different working conditions. Specifically, DIDA overcomes the class-imbalanced label shift and achieves a fine-grained latent space matching by cost-sensitive learning and categorical alignment. Besides, margin loss regularization is introduced to further optimize classification boundaries and improve cross-domain generalization for IDA fault diagnosis tasks. Finally, we simulated the IDA protocols on experimental datasets and conducted case studies under multiple working conditions, thus validating the effectiveness and superiority of the proposed framework.
•A novel deep imbalanced domain adaptation (DIDA) is proposed.•DIDA narrows both feature shift and label shift.•DIDA broadens fault diagnosis to IDA scenarios.•Experimental case studies verified validity and superiority.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
Most transfer learning-based methods require sufficient data for training, but the target data may not be available. Also, the health prognosis of target data under unknown conditions is a ...challenging online few-shot issue, which is still not effectively addressed. In addition, the limited knowledge learned from a single source domain may further limit the extraction of degradation features. To address these challenges, a multi-source adversarial online regression (MAOR) method considering the pseudo domain extension is proposed to predict the remaining useful life of bearings under online unknown conditions. It can obtain a target data stream for each round and perform an online learning task. Specifically, when generating pseudo-domains, the domain-level adaptation is designed by considering the heterogeneous distribution between pseudo-domains and the similarity of manifold between pseudo and source domains. Also, the feature-level adaptation is embedded in a multi-source adversarial adaptation architecture to learn robust domain-invariant features and build the offline model. An offline-online prediction framework is developed to predict online target data streams and update the online model with adaptive weighting. To validate the superiority of the proposed MAOR, two bearing cases are extensively evaluated. The experiment results show that MAOR can achieve significant outcomes in different online tasks with competitive performance.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
Traditional health assessment models work well under the assumption that the test and training samples obey a similar distribution. However, it is practically impossible to eliminate domain shifts ...between different tasks. Thus, most work tries to establish a data-driven approach via domain adaptation to accomplish transfer learning between different operating conditions. Sufficient target data are needed to participate in the training, which may not normally be available due to most working scenarios being unseen. An adversarial domain generalization framework with regularization learning (ADGR) is proposed for the health assessment to mine latent domains. Also, the latent domain is expanded to the unseen domain as possible. More specifically, the diversity of the sample distribution is augmented by adversarial training and the maximization of the domain discrepancy between the latent and source domains. Meanwhile, self-supervised interdomain regularization and semantical consistent regularization are proposed to mitigate the feature drift of the domain classifier and semantic divergence between source and latent domains. The case study shows that the ADGR-based health assessment approach achieves competitive prediction accuracy under unseen conditions, demonstrating its potential as a diagnostic solution.
Fault diagnosis (FD) of bearings is a research area with great relevance to industrial applications. Also, multi-source domain adaptation-based FD methods have achieved more widely acclaimed ...performance than the single-source domain. However, these methods try to derive domain-invariant features for multiple sources but ignore the semantic information of the target data under multi-source domain adaptation. In addition, aligning the decision boundaries of different classifiers may lead to the confusion of margin features, thus resulting in a limitation. To overcome these challenges, a two-stage transfer alignment (TSTA) method is proposed to complete fault diagnosis of bearings under transfer tasks. Specifically, a novel temporal transformer is developed as a backbone network to extract common and domain-specific fault features at different stages, where the common feature extractor can further reduce the complexity of the model. Meanwhile, a semantic consistency module is designed to consider the target semantics while eliminating the discrepancies in each pair of source and target domains, thus preserving the target information. Also, an entropy loss is developed to eliminate feature confusion in aligning domain-specific decision boundaries of different classifiers. Two bearing cases are extensively evaluated to validate the superiority of TSTA. The experiment results show that TSTA exhibits more competitive outcomes. In future work, a framework will be explored to treat physics-based domain knowledge as a soft constraint to penalize the loss function of deep neural networks.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPUK, ZAGLJ, ZRSKP
Deep learning (DL) techniques have revolutionized the landscape of prognostic and health management (PHM), with the capability to learn discriminative representations from “big data”. However, ...realistic industry data often show imbalanced distributions, which greatly weakens the method relying on manually balanced datasets. To fill the research gap of bearing remaining useful life (RUL) estimation with imbalanced data, a novel framework is proposed to learn imbalanced regression using cost-sensitive learning and deep feature transfer (CSL-DFT), which introduces the idea of discretization and makes full use of techniques of imbalanced learning. Our CSL-DFT includes these main points: discretization & label distribution smoothing, deep feature transfer via CORrelation ALignment (CORAL), and cost-sensitive learning via class-balanced re-weighting. Considering this is the first application of deep imbalanced regression (DIR) in RUL prediction, a variety of imbalanced bearing training sets are designed based on experimental data, and verified the effectiveness of CSL-DFT. Comparison with other methods further shows its superior performance and rationality of design.
•Proposed an imbalanced regression for bearing remaining useful life estimation.•Discretization enables label distribution smoothing and deep feature transfer.•Proposed cost-sensitive learning with class-balanced Huber loss for regression.•Constructed imbalanced run-to-failure bearing data for experimental study.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP