Remaining useful life prediction of bearings based on self-attention mechanism, multi-scale dilated causal convolution, and temporal convolution network

To improve the prediction accuracy of the performance degradation trend of proton exchange membrane fuel cell (PEMFC), this paper proposes a temporal convolutional network (TCN) model based on genetic algorithm (GA) optimization to predict the performance degradation trend of PEMFC. Firstly, variational mode decomposition (VMD) and wavelet threshold denoising (WTD) algorithms are used to denoise the original data. Then the hyperparameters of the TCN model are optimized by GA, and the GA-TCN model for predicting the performance degradation trend of PEMFC is constructed. Finally, this paper uses the PEMFC stack degradation experimental dataset disclosed in the IEEE PHM 2014 Data Challenge to verify, and compares the proposed model with the backpropagation neural networks model (BP), the long short-term memory model (LSTM) and the classical temporal convolutional network model. The results show that the proposed method has the highest performance degradation trend prediction accuracy. In particular, when the training dataset accounts for 30%, i.e. the training samples are small, the root mean square error (RMSE) and mean absolute error (MAE) of the GA-TCN model are 0.004726 and 0.003119, respectively, which are 14.48% and 20.05% lower than that of the classical TCN model. Consequently, this methodology can forecast the degradation trend of PEMFC with high accuracy.

“…To overcome the above problems, TCN uses dilated convolutions to realize the receiving domain of finite layer network with exponential size [25]. For one-dimensional time series input…”

Section: Dilated Convolutionsmentioning

confidence: 99%

Performance degradation trend prediction of proton exchange membrane fuel cell based on GA-TCN

Zhao,

Shen,

Wang

2024

“…The attention mechanism is a structure specifically designed to handle the interdependence between input and output data, which can automatically learn and calculate the weight of each input data's contribution to the output data [35]. In the field of deep learning, attention mechanisms have become an important technology widely used in various tasks, especially in natural language processing.…”

Section: Mhamentioning

confidence: 99%

Remaining useful life prediction for bearing based on automatic feature combination extraction and residual multi-Head attention GRU network

He,

Zhang,

Zhang

et al. 2023

Rolling bearings are indispensable parts in mechanical equipment, and predicting their remaining useful life is critical to normal operation and keep equipment in good repair. However, the complex characteristics of bearings make it difficult to describe their degradation characteristics. To address this issue, a novel method that combines an automatic feature combination extraction mechanism with a gated recurrent unit (GRU) network that has a residual multi-head attention mechanism for rolling bearing life prediction is proposed. Firstly, the automatic feature combination extraction mechanism is used to learn the degradation representation of the bearing vibration signal in the time domain, frequency domain, and time–frequency joint domain, and automatically extract the optimal bearing degradation feature combination. Then, the GRU network with residual multi-head attention mechanism is developed to weight and distinguish the learned degradation features, thereby improving the network’s attention to important bearing degradation features. In the end, the proposed method is validated on the prediction and the health management of systems dataset and compared to other advanced approaches. The experimental results show that the proposed method can effectively capture the complex and dynamic features of rolling bearings and has high accuracy and generalization ability in rolling bearing life prediction.

“…Table 5 presents the influence of different cores and ranks on prediction accuracy. Employing the TT-SVD in Algorithm 1, when the input data consists of 14 dimensions, setting the approximate accuracy ε = 0.05 enables us to determine the range of TT rank and TT core that impact model parameters and training time which are (1,4) and (1,5), respectively. Based on the experimental findings, it can be concluded that Tensor-Train decomposition significantly contributes to parameter reduction and shorter training times.…”

Section: Ablation Analysis the Ablation Experiments In Tablementioning

confidence: 99%

“…Data-driven modelling methods utilize available data to unveil concealed patterns, facilitating accurate time series forecasting [5]. Time series forecasting techniques can be broadly categorized into traditional machine learning and deep learning models.…”

Section: Introductionmentioning

confidence: 99%

Multi-level deep domain adaptive adversarial model based on tensor-train decomposition for industrial time series forecasting

Yang,

Peng,

Chen

et al. 2023

The polyester industry is a complex process industry, building a time series prediction model for new production lines or equipment with new sensors can be challenging due to a lack of historical data. The time-series data collected from sensors cross-production-line often exhibit varying distributions. Current domain adaptation (DA) approaches in data-driven time series forecasting primarily concentrate on adjusting either the features or the models, neglecting the intricacies of industrial time series data. Furthermore, constructing deep neural networks for industrial data necessitates substantial computational resources and runtime due to their large and high-dimensional nature. In order to tackle these obstacles, we propose a novel Multi-level deep domain adaptive adversarial model based on tensor-train decomposition (TT-MDAM). Our model aims to strike a dynamic balance between prediction accuracy and runtime efficiency. By integrating multiple perspectives at the feature, trend, and model levels, we leverage DA to enhance the prediction accuracy of our model in the target domain. Additionally, by analyzing the weight matrix of the neural network, we generate a low-rank model to improve operational efficiency. The application of the proposed TT-MDAM approach to both the three-phase flow facility process (TPFF) dataset and a real-world polyester esterification process dataset reveals promising results, outperforming state-of-the-art methodologies in terms of prediction performance. The results indicate that the approach provides a viable solution for building time series prediction models in industrial processes with new equipment or production lines.