Unsupervised Domain Deep Transfer Learning Approach for Rolling Bearing Remaining Useful Life Estimation

In recent years, data-driven approaches have shown promise in fault diagnosis. However, in practical industrial applications, challenges such as discrepancies between source domain and target domain data distributions, coupled with limited labeled fault data, have hindered the performance of traditional domain adaptation algorithms in bearing fault diagnosis. To address this issue, this study introduces a novel method based on MDANN (Multi Domain Adaptation Neural Network) for diagnosing rolling bearing faults using unlabeled data. Initially, the raw vibration signals are preprocessed using wavelet packet decomposition and reconstruction (WPT), which helps minimize signal redundancy while preserving critical features. Following this, the multi-kernel maximum mean discrepancy (MK-MMD) algorithm is employed to compute the differences in input feature values, and the MDANN network parameters are adjusted via backpropagation, allowing the network to extract domain-invariant features. To ensure the unlabeled target domain data can be effectively used in training, a pseudo-labeling strategy is applied, where the label with the highest probability is treated as the true label, thereby enabling the model to learn from the target domain data and improve the acquisition of reliable diagnostic knowledge. The method is validated using two publicly available datasets, CWRU and PU, with experimental results demonstrating that it outperforms conventional domain adaptation techniques in terms of diagnostic accuracy. This indicates the proposed approach is highly effective in capturing transferable features and addressing the distributional differences between datasets.

This study seeks to enhance the classification performance of breast ultrasound images, addressing the challenges of difficult and costly collection of breast ultrasound datasets as well as the discrepancies in feature distribution of the collected datasets. Performance is enhanced by using a mix of generative adversarial networks (GAN) and domain adaptive networks. First, an adaptive layer is first added to the basic model of the gradually vanishing bridge (GVB), to better match the feature distributions of the source and target domains of the dataset. The multi-kernel maximum mean discrepancy (MK-MMD), which is the most efficient of existing adaptive approaches, is implemented in the fully connected layer of the original model’s feature extraction network. Finally, through the process of fine-tuning, the model that has the highest level of overall performance is determined. In experiments, the proposed method surpassed the conventional unsupervised domain adaptation (DDC) and adversarial domain adaptation (MK_DAAN, GVB) in performance, achieving 85.11% accuracy, 97.48% recall, and 0.92 F1-score.

Objective. The scarcity of electroencephalogram (EEG) data, coupled with individual and scenario variations, leads to considerable challenges in real-world EEG-based driver fatigue detection. We propose a domain adaptation method that utilizes EEG data collected from a laboratory to supplement real-world EEG data and constructs a cross-scenario and cross-subject driver fatigue detection model for real-world scenarios. Approach. First, we collect EEG data from subjects participating in a driving experiment conducted in both laboratory and real-world scenarios. To address the issue of data scarcity, we build a real-world fatigued driving detection model by integrating the real-world data with the laboratory data. Then, we propose a method named cross-scenario and cross-subject domain adaptation (CS2DA), which aims to eliminate the domain shift problem caused by individual variances and scenario differences. Adversarial learning is adopted to extract the common features observed across different subjects within the same scenario. The multikernel maximum mean discrepancy (MK-MMD) method is applied to further minimize scenario differences. Additionally, we propose a conditional MK-MMD constraint to better utilize label information. Finally, we use seven rules to fuse the predicted labels. Main results. We evaluate the CS2DA method through extensive experiments conducted on the two EEG datasets created in this work: the SEED-VLA and the SEED-VRW datasets. Different domain adaptation methods are used to construct a real-world fatigued driving detection model using data from laboratory and real-world scenarios, as well as a combination of both. Our findings show that the proposed CS2DA method outperforms the existing traditional and adversarial learning-based domain adaptation approaches. We also find that combining data from both laboratory and real-world scenarios improves the performance of the model. Significance. This study contributes two EEG-based fatigue driving datasets and demonstrates that the proposed CS2DA method can effectively enhance the performance of a real-world fatigued driving detection model.

Fault diagnosis plays a critical role in ensuring the safe operation of machinery. Multi-source domain adaptation (DA) leverages rich fault knowledge from source domains to enhance diagnostic performance on unlabeled target domains. However, most existing methods only align marginal distributions, neglecting inter-class relationships, which results in decreased performance under variable working conditions and small samples. To overcome these limitations, two stage multi-source domain adaptation (TSMDA) has been proposed for bearing fault diagnosis. Specifically, wavelet packet decomposition is applied to analyze fault information from signals. For small sample datasets, Diffusion is used to augment the dataset and serve as the source domain. Next, multi-scale features are extracted, and mutual information is computed to prevent the negative transfer. DA is divided into two stages. Firstly, multikernel maximum mean discrepancy is used to align the marginal distributions of the multi-source and target domains. Secondly, the target domain is split into subdomains based on the calculated pseudo-labels. Conditional distributions are aligned by minimizing the distance from samples to the center of the non-corresponding domain. The effectiveness of the proposed method is verified by extensive experiments on two public datasets and one experimental dataset. The results demonstrate that TSMDA has high and stable diagnostic performance and provides an effective method for practical fault diagnosis.

Cross-domain bearing fault diagnosis using dual-path convolutional neural networks and multi-parallel graph convolutional networks

All current deep learning-based prediction methods for remaining useful life (RUL) assume that training and testing data have similar distributions, but the existence of various operating conditions, failure modes, and noise lead to insufficient data with similar distributions during the training process, thereby reducing RUL prediction performance. Domain adaptation can effectively solve this problem by learning the cross-domain invariant features of the source domain and target domain to reduce the distribution difference. However, most domain adaptive methods extract the source domain and target domain features into a single space for feature alignment, which may leave out effective information and affect the accuracy of prediction. To address this problem, we propose a data-driven approach named long short-term memory network and multi-representation domain adaptation (LSTM-MRAN). We standardize and process the degraded sensor data with a sliding time window, use LSTM to extract features from the degraded data, and mine the time series information between the data. Then, we use multiple substructures in multi-representation domain adaptation to extract features of the source domain and target domain from different spaces and align features by minimizing conditional maximum mean difference (CMMD) loss functions. The effectiveness of the method is verified by the CMAPSS dataset. Compared with methods without domain adaptation and other transfer learning methods, the proposed method provides more reliable RUL prediction results under datasets with different operating conditions and failure modes.

Accurately predicting the remaining useful life (RUL) of lithium-ion batteries is crucial for ensuring the safe and reliable operation of battery systems. However, due to the highly nonlinear degradation behavior of lithium-ion batteries under complex working conditions, traditional prediction methods often fail to achieve satisfactory accuracy. This study proposes a hybrid prediction model that integrates an attention (ATT) mechanism, a convolutional neural network (CNN), and a bidirectional long short-term memory (BiLSTM) network, termed ATT-CNN-BiLSTM. First, a one-dimensional CNN is employed to extract local degradation patterns and inter-channel correlated features from multi-dimensional sensor time-series data. Then, the BiLSTM network captures long-term dependencies within the degradation process. Furthermore, an attention mechanism is introduced to dynamically adjust the contribution of key time-step features and suppress noise interference. Experimental results based on the NASA battery aging datasets demonstrate that the proposed ATT-CNN-BiLSTM model reduces RMSE and MAE by 49.08% and 54.67%, respectively, compared with the LSTM model, and by 10.89% and 17.85%, respectively, compared with the ACNN-LSTM-MMD model. Ablation experiments also confirm that further confirm that the ATT-CNN-BiLSTM model decreases RMSE and MAE by 26.45% and 33.33%, respectively, compared with CNN-BiLSTM, and by 21.85% and 23.36%, respectively, compared with ATT-BiLSTM.

Transfer learning and domain adaptation (DA) methods have been utilized in bearing prognostic and health management, but most of the current DA methods do not take into account the feature scale change of degraded features when aligning the feature distribution, and these methods are more suitable for the classification problem, which is more robust to the feature scale change. However, they perform poorly in regression problems. In addition, most of the remaining useful life (RUL) prediction methods require preprocessing such as statistical feature extraction on the signal, which makes the prediction process complicated. To solve the above problems, a DA method based on the representation subspace distance (RSD) is proposed for predicting the bearing RUL under different operating conditions. First, the proposed convolutional neural network (CNN) self-attention (SA) long short term memory network model is utilized to extract the deep features from the original signal, which overcomes the limitations of the CNN in extracting time series. Then, the RSD in the Riemannian geometry of the Grassmann manifold is proposed as a domain transfer loss to learn domain invariant features. The modified method can align the feature distribution of the source domain and the target domain without changing the feature scale. At the same time, the bases mismatch penalization is introduced to avoid destroying the semantic information of the features in the process of domain alignment. Finally, the effectiveness of the proposed method is verified by experiments on four types of transfer tasks, and its superiority is also demonstrated by comparison with other advanced methods.

Achieving accurate and reliable fault diagnosis of rotating machinery under complex conditions remains challenging due to data distribution shifts, class imbalance, and limited labeled samples. Existing domain adaptation methods typically rely on raw vibration signals alone, neglecting the high frequency diagnostic information captured by signal decomposition, while decomposition based approaches often discard the raw signal, overlooking global structural features. To address these dual limitations, we propose a multi-source dual-channel attention transfer (MDAT) method that jointly utilizes the raw vibration signal and its first intrinsic mode function (IMF1), extracted via improved complete ensemble empirical mode decomposition with adaptive noise. The proposed method adopts a multi-source dual-channel architecture, where raw signals and their IMF1 components are encoded separately and fused through a multi-head attention mechanism. To reduce domain discrepancy, feature alignment is performed across domains using multikernel maximum mean discrepancy. Experimental results conducted on multiple representative datasets demonstrate that MDAT consistently achieves superior generalization capability and diagnostic accuracy, achieving improvements of 1.02%, 4.10%, and 9.37% over methods using only raw signals on standard, imbalanced, and small sample datasets, respectively. The rationale of using raw signals and IMF1 as complementary dual-channel inputs has also been validated across diverse scenarios. These results confirm the method’s effectiveness in enhancing accuracy and robustness in rotating machinery fault diagnosis based on vibration measurement.

Air temperature (T, °C), relative humidity (RH, %), and global solar radiation (G, MJ.m-2 ) have been measured in the meteorological station at South Valley University (SVU) at Qena, Egypt, from 2000 to 2013, while the total column ozone (TCO, DU) is downloaded from Giovanni's website. T, RH, and TCO are important meteorological parameters, and they are useful to estimate the missing data of the global solar radiation (G), as global solar radiation is desirable for electricity generation applications and for agriculture. Qena is a subtropical region in Upper Egypt, as it's characterized by clear weather most days of the year, and it's very hot in the summer and cold in the winter. The meteorological station at South Valley University (SVU) stopped measuring the G. The linear regression equation and the most important statistical indices are used in this paper such as, the determination coefficient (R2 ), the mean absolute error (MAE), the mean absolute bias error (MABE), the mean square error (MSE), the root mean square error (RMSE), the mean percentage error (MPE), the mean bias error (MBE), the model efficiency (ME), and the agreement index (d). For verification of the empirical models' efficiency, the data of a new period has used, 2013, and the results of the models were excellent and valid for estimating the missing data, as R2 was more than 0.92 in all models but it was near one in models 1 and 4. MAE was close to zero for all models. MBE, MPE, and MABE were close to zero for all models except model 3. Model 1 was the best one, as, R2 , MAE, MABE, MSE, RMSE, MAPE, d, ME, and MPE were 0.9883, 0.0193, 0.383, 0.2303, 0.4799, 1.927, 0.9979, 0.992, and 0.2932, respectively

Remaining useful life (RUL) prediction is one of the core issues in the equipment maintenance process. It aims to accurately forecast machines’ run-to-failure life span using previous and current state data. As various data-driven models are proving to be effective, because RUL labels for machines in particular conditions are difficult to obtain, domain adaptation approaches begins to be explored in the RUL prediction issue. We propose a novel Dual Mix-up Adversarial Domain Adaptation (DMADA) approach to further improve the RUL forecasting accuracy, building on existing RUL domain adaptation studies. In DMADA, both time-series mix-up and domain mix-up regularization are conducted. Virtual samples generated by linear interpolation lead to enriched and more continuous sample space. The linear interpolation encourages consistency of prediction in-between samples and allows the model to explore the feature space more thoroughly. At the same time, the domain mixup conserves the invariance of learned features. The two mixup regularizations combined promote both the transferability and the discriminability of extracted features, which is essential to satisfactory unsupervised domain adaptation performance. Thorough experiments on the C-MPASS dataset are conducted and satisfactory results prove the proposed approach effective.

With the development of deep learning methods, the data-driven fault diagnosis methods have attracted a great deal of interest. However, as for the data-driven fault diagnosis methods, technology has to overcome various difficulties in the practical industrial scenarios, such as variable working conditions, insufficient effective samples, and environmental noise interference. Combining with the time–frequency analysis of vibration signals, a domain adaptation fault diagnosis model based on ResNet and Transformer (DAFDMRT) is proposed in this work, aiming to solve the problems encountered by current rotating machinery fault diagnosis methods in the field of application. Firstly, the vibration signal is processed by wavelet packet transform and the time–frequency information grayscale maps is constructed. Next, a deep fusion feature extraction network combining ResNet and Transformer encoder, is designed for the extraction and fusion of the local and global features of multi-scale time–frequency information. Finally, the multi-kernel maximum mean discrepancy is applied to measure and minimize the distribution difference between the deep features of source and target domain, thereby improving the diagnostic performance of the diagnosis model in variable working conditions. In this work, comparative experiments are conducted as for bearing and gearbox datasets under variable working conditions. The results indicate that DAFDMRT can show excellent performances in terms of fault diagnosis and generalization ability.

Sustainable tourism, unlike conventional tourism, explicitly strives to maintain economic, environmental, and social sustainability in tourist destinations. Saipan, the capital and the largest island of the chain of Pacific islands known as the Commonwealth of the Northern Mariana Islands (CNMI), is a tropical destination that heavily relies on the tourist industry as a source of income. When analyzing the trends of tourist arrivals from the top three countries that account for the majority of tourist arrivals in Saipan-China, Japan, and Korea-the necessity to implement sustainable tourism in Saipan becomes clear. This article forecasts tourism demand, which is one of the main factors in determining sustainable tourism on a small island. Analyzing both long-term and cyclical trends in tourist arrivals assists budget-constrained policymakers and related organizations in implementing sustainable tourism. Therefore, forecasting accuracy is essential in tourism policy and planning. The tourist arrival data obtained from Marianas Visitors Authority’s official Japanese website are made into a time series for Chinese, Japanese, and Korean visitor data from 2006 to 2016. Given the various time-series forecasting models, we employ simple but powerful forecasting techniques, such as ARIMA and ETS to control for the possible seasonality and trend in our tourism data. Using these forecast methods, the tourist arrival data is used to predict the future trends of tourist arrivals in Saipan for the next twelve months. The performance of our forecasting model in this study is evaluated based on the five measures of accuracy, as measured by comparison with actual tourism flows: mean error, mean absolute scaled error, mean percentage error, mean absolute error, and mean absolute percentage error. A comparison of the two popular forecast techniques in this study shows that the ETS model statistically outperforms the ARIMA model in terms of various measures of accuracy. A practical implication of this finding is that the ETS approach should be used when forecasting tourism demand for small islands. We predict the visitor arrivals from China, Japan, and Korea to Saipan by using the better model. The estimated 80% and 95% intervals for the number of tourist arrivals by the ETS model are presented in this study. The upper (lower) bound is the maximum (minimum) value of the estimated tourist arrivals. We find strong evidence that the tourist arrival trend for Korean visitors to Saipan will increase in the short run; although, the Chinese and Japanese tourist arrival trends seems to be stable, except for Japanese tourist arrivals in August. Nonetheless, with such an intense demand for Saipan’s natural resources, sustainable island tourism becomes necessary. Our estimated result is useful to practitioners in the sense that they are able to prepare for carrying capacities and the means of destination management on Saipan. Proper management for maintaining the island environmentally, economically, and socially requires the preservation of indigenous culture, local quality of life, natural resources, and economic stability while accommodating the demands of arriving tourists. It is crucial to preserve the assets of Saipan in order to maintain and improve its economy. Thus, our time series modeling for forecasting the number of visitors becomes necessary to develop overall economic, environmental, and social sustainability in the island. An interesting addition to implementing sustainable tourism in Saipan is the idea of maintaining a “cap” on the number of incoming tourists visiting the island. Perhaps, with more research on Saipan’s environment, specifically an estimate of how much of the island’s resources tourists exhaust and how much the island can withstand the aforementioned, this “cap” could be practical to prevent perpetual damage to Saipan’s biodiversity, resources, culture, and economy. Along with this idea, knowing the minimum number of tourists needed to sustain Saipan’s necessities is valuable.

In this study, we explore and compare various time series models to forecast OPEC oil prices, aiming to identify the most accurate model for predicting future trends. The methodology involves the application of multiple time series forecasting models, including Random Walk, Linear Trend, Quadratic Trend, Exponential Trend, S-Curve Trend, Simple Moving Average, Simple Exponential Smoothing, Brown's Linear Exponential Smoothing, Holt's Linear Exponential Smoothing, Brown's Quadratic Exponential Smoothing, and ARIMA. Each model is evaluated based on performance metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Error (ME), Mean Percentage Error (MPE), Akaike Information Criterion (AIC), Hannan-Quinn Criterion (HQC), and Bayesian Information Criterion (BIC) to determine its accuracy and reliability. Among the models tested, the ARIMA(1,1,0) model emerged as the best-performing model, providing the most accurate predictions. Using this model, we forecast OPEC oil prices for a 12-month period from July 2024 to May 2025. Our findings include detailed forecasts for each month within this period, offering valuable insights for policymakers and market participants. These results highlight the importance of selecting an appropriate time series model to improve forecasting accuracy and inform decision-making processes in the energy sector.

Motivation: The substantiation of the scientific problem and the practical value of the study are determined by the enhanced cooperation of the BRICS countries, in particular the growing weight of Russia and China on the world stage. The main objective of the study: is to analyze the problems of the development of bilateral trade between China and Russia and to provide a statistically proven forecast for this trade for 2019-2020, as well as to develop recommendations for the improvement of bilateral trade relations. Novelty: The author's statistical model was developed and its testing was presented through confirming the 24-month forecast of bilateral trade between Russia and China. The model involves solving various problems of bilateral trade. Recommendations for improvement of bilateral trade relations are proposed through formation of an investment and innovation model of bilateral trade.Methodology and Methods: The work used the method of forecasting time series, which suggested the use of a model to predict future values based on previously observed values. To evaluate the modern prospects of the trade between the countries, the authors produced a forecast of the goods turnover trend for 2019-2020. The forecast was issued via the software tool Statgraphics Centurion 18. A reasonable model of the 24-month forecast based on the statistical model Random Walk is developed. The adequacy of the proposed forecast model was subjected to statistical tests. To verify the statistical adequacy of the model the relevant tests were done to determine the compliance of the model with the informational criteria ME (Mean Error), MSE (Mean Squared Error), МАЕ (Mean Absolute Error), МАРЕ (Mean Absolute Percentage Error), МРЕ (Mean Percentage Error). However, it should be noted that the forecast was issued in accordance with the trends which had been identified in the preceding periods. Data and empirical analysis: The factors influencing bilateral trade are analyzed, as well as examples of implemented projects of international cooperation between Russia and China are presented. The current dynamics of sales turnover between Russia and China for the period of 2010-2018 with the use of various statistical and analytical methods is studied, and a reasonable model of the 24-month forecast based on the statistical model Random Walk is developed. The adequacy of the proposed forecast model was subjected to statistical tests. The basic hypothesis is suggested for the upward trend based on reference time series. Policy considerations: It can be said with certainty that the level of technological development of BRICS countries will help Russia and China to start building their cooperation in many fields at a completely new level, taking into account their joint experience in overcoming global crises and Western sanctions. International cooperation between Russia and China in the innovation field will help them unite their efforts and achieve significant synergy. Coordination of countries on this issue will help to reduce production costs, cooperation of production, joint research and development, as well as increase bilateral trade turnover.