Search

Remaining useful life (RUL) prediction of aircraft engine (AE) is of great importance to improve its reliability and availability, and reduce its maintenance costs. This article proposes a novel deep bidirectional recurrent neural networks (DBRNNs) ensemble method for the RUL prediction of the AEs. In this method, several kinds of DBRNNs with different neuron structures are built to extract hidden features from sensory data. A new customized loss function is designed to evaluate the performance of the DBRNNs, and a series of the RUL values is obtained. Then, these RUL values are reencapsulated into a predicted RUL domain. By updating the weights of elements in the domain, multiple regression decision tree (RDT) models are trained iteratively. These models integrate the predicted results of different DBRNNs to realize the final RUL prognostics with high accuracy. The proposed method is validated by using C-MAPSS datasets from NASA. The experimental results show that the proposed method has achieved more superior performance compared with other existing methods.

Rheumatoid arthritis (RA) is a chronic systemic autoimmune disease with predominant synovitis that has no complete cure or preventive treatment. Citrus essential oils, used in natural fragrances, contain a variety of functional ingredients that are worthy of investigation for their potential as natural anti-inflammatory drug sources. In this study, essential oils were hydro distilled from the peels of four citrus species: Citrus sinensis (L.) Osbeck (CSEOs), Citrus paradisi Macfad. (CPEOs), Citrus limon (L.) Osbeck (CLEOs) and Citri Reticulatae Pericarpium (CREOs). Altogether, 81 compounds were identified using gas chromatography-mass spectrometry (GC-MS), of which d-limonene (17.96%-94.66%) was an abundant component of all four oils. The stable 1,1-diphenyl-2-pyrrole hydrazine (DPPH) free radical test showed that all four essential oils had excellent antioxidant properties (IC

Remaining useful life (RUL) prediction is a key solution to improve the reliability, availability and maintainability of engineering systems. Long short-term memory (LSTM) and convolution neural networks (CNN) are the current hotspots in the field of RUL prediction. However, the LSTM-based prognostic approach has a slow loop step to process large-scale time-series data since the dependence of the data processing process at each time on the output of the previous time limits parallelism, and the CNN-based prognostic approach is not fit for time-series data although it can process the data in parallel. In this paper, a new auto-encoder quasi-recurrent neural networks (AEQRNN) based prognostic approach is proposed for RUL prediction of the engineering systems. The AEQRNN contains convolution components that can process input data in parallel, and pooling components which has two LSTM-like gate structures to process time-series data. In addition, the AEQRNN can automatically extract hidden features from monitoring signals without manual feature design. The effectiveness of the proposed prognostic approach is validated by three prognostic benchmarking datasets, including a turbofan engine dataset, a rolling bearing dataset, and a machining tool dataset. Experimental results demonstrate that this approach has both superior prognostic performance and training speed in comparison with other kinds of recurrent neural network-based approaches and various state-of-the-art approaches in the recent literature.

Remaining useful life prediction of rolling element bearings is significant to improve the safety and reliability of engineering systems. It is a vital issue to perceive abnormal symptoms for remaining useful life prediction so as to set the occurrence time as the first prediction time. Traditional methods need to define indicators manually to determine the abnormal symptoms, which rely on a lot of domain knowledge and expert experience. This paper proposes a novel abnormal symptoms-triggered remaining useful life prediction approach for rolling element bearings. An adaptive kernel spectral clustering model is constructed in the abnormal symptoms-triggered remaining useful life prediction approach to adaptively detect the abnormal symptoms in real time from multi-dimensional degradation features extracted by integrating time-domain, frequency-domain, time-frequency, and time-series analysis. The occurrence time of abnormal symptoms is set as the first prediction time of the remaining useful life predictor, which is a new type of particle swarm optimization-quantile regression neural network. A benchmarking dataset of rolling element bearings is adopted to evaluate the proposed abnormal symptoms-triggered remaining useful life prediction approach. Experimental results show that this abnormal symptoms-triggered remaining useful life prediction approach is superior to many other prognostic approaches.

The discrete element method has been used to simulate the particle flow in a ball mill under different operating conditions. The model was validated by comparing the simulated results of the flow pattern and input power with those measured from a same-scale laboratory mill. The impact energy of the particles under different operating conditions was analysed in detail. The results showed that the impact energy was affected by the operating conditions of the mill and can be linked to the grinding rate for a given material. The correlation between impact energy and grinding rate follows first-order grinding kinetics. Mill performance decreases with increasing mill size. Furthermore, a data-driven machine learning framework has been proposed to predict the impact energy for different operating conditions. It was found that the prediction for mills with diameters of 2000 and 3000 mm based on the training model developed by mills with diameter less than 254 mm could be achieved with an accuracy of 80% and a correlation coefficient of 0.9. Through the combination of DEM simulation and data-driven approach, the computing time required in the determination of impact energy for large scale mills can be dramatically reduced.

Remaining useful life (RUL) prediction is of great importance to improve the reliability and availability of machinery. Traditional rolling bearing RUL prediction methods ignore the impact of bearing degradation states recognition on accurate the RUL prediction. A novel deep learning-based two-stage RUL prognostic approach is presented in this paper by using fast search and find of density peaks clustering (FSFDPC) and multi-dimensional deep neural network (MDDNN). In the first stage, health states of the rolling bearing are automatically perceived by the FSFDPC, where no manual parameters need to be preset. In the second stage, MDDNN is constructed with parallel BLSTM and BGRU channels to extract the multi-dimensional feature maps from the input for accurate RUL prediction. The effectiveness of the proposed two-stage prognostic approach is validated by rolling bearing experimental data of three cases. The comparisons with other existing approaches show that the proposed approach has superior prediction performance under different working conditions.

This paper presents a study on the Brazil-nut segregation of binary granular mixtures under vibration by the discrete element method. The ascending velocity of large particles (Va) and the overall segregation degree (Dsep) are correlated to the vibration velocity amplitude (Vb) and the dimensionless frequency (fd). Both Dsep and the average Va increase linearly with Vb under a given fd, while the coefficient shows a critical change at fd =1. Va conforms to a single-peak distribution, and the shape parameters of the distribution converge after Vb > 1.7. Both Va and Dsep can be correlated to (A/ds)‧fd0.5 universally, where A is amplitude and ds is the small particles size. The increase of the volume fraction of large particles leads to the increase of Dsep and Va, which is probably linked with the increase in granular temperature and the difference between the vibration patterns of large and small particles.

Deep learning, due to its excellent feature-adaptive capture ability, has been widely utilized in the fault diagnosis field. However, there are two common problems in deep-learning-based fault diagnosis methods: (1) many researchers attempt to deepen the layers of deep learning models for higher diagnostic accuracy, but degradation problems of deep learning models often occur; and (2) the use of multiscale features can easily be ignored, which makes the extracted data features lack diversity. To deal with these problems, a novel multiscale feature fusion deep residual network is proposed in this paper for the fault diagnosis of rolling bearings, one which contains multiple multiscale feature fusion blocks and a multiscale pooling layer. The multiple multiscale feature fusion block is designed to automatically extract the multiscale features from raw signals, and further compress them for higher dimensional feature mapping. The multiscale pooling layer is constructed to fuse the extracted multiscale feature mapping. Two famous rolling bearing datasets are adopted to evaluate the diagnostic performance of the proposed model. The comparison results show that the diagnostic performance of the proposed model is superior to not only several popular models, but also other advanced methods in the literature.

A solution in finite form for the screw dislocation interacting with the nanocrack incorporating surface piezoelectricity under anti-plane loads is developed. By the boundary integration and the Cauchy’s integral formula to convert the problem into a first-order differential equation, the finite-form solution in physical region is established via the mapping technique. At the crack tip, the results from the present solution indicate that the electric displacements and the stresses are finite and the singularities are removed when the surface piezoelectricity is incorporated, which is against those in the classic solution and also different from other reported solutions where the singularities are weakened but non-trivial. The numerical results shows that the stresses, the electric displacements and the dislocation forces are strongly size dependent and their magnitudes reduce with the decrease in crack length, and the results approach to those by the classic solution with an increasing in the crack length.

Remaining useful life (RUL) prognosis is of great significance to improve the reliability, availability, and maintenance cost of an industrial equipment. Traditional machine learning method is not fit for dealing with time series signals and has low generalization and stability in prognostic. In this article, a novel ensemble long short-term memory neural network (ELSTMNN) model for RUL prediction is proposed to enhance the RUL prognosis accuracy and improve the adaptive and generalization abilities under different prognostic scenarios. The ELSTMNN contains a series of long short-term memory neural networks (LSTMNNs), each of which is trained on a unique set of historical data. A novel ensemble method is first proposed using Bayesian inference algorithm to integrate multiple predictions of the LSTMNNs for the optimal RUL estimation. The effectiveness of the ELSTMNN-based RUL prognosis method is validated using two characteristically different turbofan engine data sets. The experimental results show a competitive performance of the ELSTMNN in comparison with other prognostic methods.

Deep learning has been widely used in fault diagnosis, especially the convolutional neural network (CNN). However, traditional CNN models usually use single-scale kernels to extract features, which ignores the multiscale features of input data. This article develops a novel residual neural network named multiscale feature fusion deep residual networks for fault diagnosis of aerostat. The designed multiscale feature fusion block (MFF Block) realizes automatic extraction, fusion and compression of multiscale features. The series connection of multiple MFF Blocks makes the proposed model able to extract deeper and wider features from raw signal segments. Then a multiscale pooling layer is developed to extract the most effective features for accurate fault diagnosis. The proposed model is evaluated on an exclusive aerostat strain signal dataset. The comparison results illustrate that the proposed model achieves superior diagnostic performance than several popular models.

Accurate health evaluation is crucial to reliable operation of complex degradation systems. Although traditional machine learning methods such as artificial neural network (ANN) and support vector machine (SVM) have been used widely, state assessment schemes based on a single classification model still suffer from low multiclass classification efficiency, high variance, and deviation. To solve these problems, this article proposes a novel health evaluation method based on stacking ensemble learning and generalized multiclass support vector machine (GMSVM) algorithm. The proposed health evaluation framework includes three parts: 1) abnormal value elimination and missing value processing are applied for multiple sensor data; 2) statistical features are extracted from the observed data and the Pearson correlation coefficient is applied for feature selection; and 3) ensemble generalized multiclass support vector machines (EGMSVMs) are utilized to evaluate the health situation of a degradation system. Unlike the binary classifiers and deep-learning-based classifiers, EGMSVMs utilize the stacking-based method to combine several GMSVMs as submodels and random forest as a metamodel, and the metamodel ensembles the results of submodels to reach a satisfied performance. Compared to traditional SVM- and ANN-based algorithms, EGMSVMs, in processing multiclass problems, achieve high efficiency and, meanwhile, low variance and deviation. The proposed method is verified using a hydraulic test rig. The experimental results show the feasibility and applicability of the proposed health evaluation framework.

China is the leading source of international students for the UK. However, there are many differences between Chinese and UK classroom culture; for example, Chinese students may do well in tests an...

The purpose of this article is to balance product quality requirements and maintenance costs. Considering that manufacturing system is deteriorating and its maintenance situations are complex, main...

Understanding and quantification of the interactions between gold nanoparticles (AuNPs) allows us to explore their potentials for a wide range of applications. In present work, the interactions between AuNPs of size ranging from 3 nm to 20 nm are investigated by molecular dynamics (MD) simulation and their interparticle force models are developed from a quick and effective approach. Specifically, an external force is initially applied to a pair of spherical AuNPs at certain surface separation to accelerate their approaching toward each other. Such applied force is then removed at a certain distance, which can avoid the potential impact of such external force on the calculated interaction forces between AuNPs. Our results show that the attractive force dominates the interactions between spherical AuNPs. Such force increases exponentially with the decrease of particle surface separation prior to particle collision, which can be described quantitatively by simple mathematical force models. In addition, simulation on two AuNPs with different sizes shows that the smaller particle tends to merge onto the surface of larger one upon their collision. The reported approach provides a quick way to developing the interparticle force models between nanoparticles of different materials and shape, which is of importance in exploring nanoparticle systems and their self-assembly.

This paper presents a method to establish explicit force models for identical superellipses. Currently for two superellipses in contact, the overlap parameters must be solved by using numerical iteration but cannot be explicitly calculated. Our method is firstly to numerically solve these solutions in the comprehensive range of the orientation angles with the angles varying discretely. Then using two-dimensional (2D) discrete Fourier transform (DFT), the discrete solutions as a function of two angles are transformed to the frequency domain. The transformed terms are sorted according to their magnitudes, and those below a certain value are filtered in the inverse discrete Fourier transform (IDFT). This finally results in simplified Fourier series that can approximate overlap parameters explicitly. The method was realized for a wide range of superellipse shapes and the errors between the approximate Fourier series and the original solutions were analyzed and shown to be small. In addition, the established explicit force model was applied to the simulation of the packing of superellipses. The results show that the explicit force model established by this method can be effectively used in DEM simulations. And the packing of superellipses is complicatedly dependent on the squareness, aspect ratio and friction. These findings provide a novel way to establish the explicit force models for non-spherical particles and also improve our understanding on the packing of non-spherical particles.

The performance of medical image classification has been enhanced by deep convolutional neural networks (CNNs), which are typically trained with cross-entropy (CE) loss. However, when the label presents an intrinsic ordinal property in nature, e.g., the development from benign to malignant tumor, CE loss cannot take into account such ordinal information to allow for better generalization. To improve model generalization with ordinal information, we propose a novel meta ordinal regression forest (MORF) method for medical image classification with ordinal labels, which learns the ordinal relationship through the combination of convolutional neural network and differential forest in a meta-learning framework. The merits of the proposed MORF come from the following two components: a tree-wise weighting net (TWW-Net) and a grouped feature selection (GFS) module. First, the TWW-Net assigns each tree in the forest with a specific weight that is mapped from the classification loss of the corresponding tree. Hence, all the trees possess varying weights, which is helpful for alleviating the tree-wise prediction variance. Second, the GFS module enables a dynamic forest rather than a fixed one that was previously used, allowing for random feature perturbation. During training, we alternatively optimize the parameters of the CNN backbone and TWW-Net in the meta-learning framework through calculating the Hessian matrix. Experimental results on two medical image classification datasets with ordinal labels, i.e., LIDC-IDRI and Breast Ultrasound Dataset, demonstrate the superior performances of our MORF method over existing state-of-the-art methods.

Brassica napus L. is a vital plant oil resource worldwide. The fatty acid biosynthesis and oil accumulation in its seeds are controlled by several genetic and environmental factors, including daytime and nighttime temperatures. We analyzed changes in oleic and erucic acid content in two double haploid (DH) lines, DH0729, a weakly temperature-sensitive line, and DH0815, a strongly temperature-sensitive line, derived from B. napus plants grown at different altitudes (1600, 1800, 2000, 2200, and 2400 m a.s.l., 28.85° N, 112.35° E) and nighttime temperatures (20/18, 20/16, 20/13 and 20/10 °C, daytime/nighttime temperature). Based on medium- and long-chain fatty acid metabolites, the total oleic acid content 35 and 43 days after flowering was significantly lower in low nighttime temperature (LNT, 20/13 °C) plants than in high nighttime temperature (HNT, 20/18 °C) plants (HNT: 58–62%; LNT: 49–54%; an average decrease of 9%), and the total erucic acid content was significantly lower in HNT than in LNT plants (HNT: 1–2%; LNT: 8–13%; an average increase of 10%). An RNA-seq analysis showed that the expression levels of SAD (LOC106366808), ECR (LOC106396280), KCS (LOC106419344), KAR (LOC106367337), HB1(LOC106430193), and DOF5 (LOC111211868) in STSL seeds increased under LNT conditions. In STSL seeds, a base mutation in the cis-acting element involved in low-temperature responsiveness (LTR), the HB1 and KCS promoter caused loss of sensitivity to low temperatures, whereas that of the KCS promoter caused increased sensitivity to low temperatures.

Composite structures are widely used due to their excellent performance. To improve their safety and reliability, non-destructive testing (NDT) methods are implemented to achieve efficient damage detection. In this paper, a novel stress–strain-based damage detection approach is proposed for composite structures by using continuous wavelet transform (CWT) and selective kernel convolutional network (SKNet), which exhibit good robustness when dealing with the stress–strain signals collected from different positions of composite structures. First, a stress–strain-based measuring scheme is designed for vibration response monitoring of the composite structures. Second, the collected stress–strain signals are converted into two-dimensional time–frequency images by the CWT. Then, the SKNet is constructed to classify these images for damage detection. Finally, an experimental test is conducted on a composite capsule, data from which is employed to evaluate the proposed method. The comparison results of our proposed method and other state-of-the-art approaches demonstrate its superiority.

A finite-form solution for the problem of an elastic body weakened by a straight nano-crack and disturbed by a screw dislocation in the mode-III deformation is proposed. The boundary equations incorporating the surface effects are established and reduced to single one via the integration over the crack boundary in the physical region. The cracked physical region is then mapped onto an image region with circular hole. By using Muskhelishvili (Some basic problems of the mathematical theory of elasticity, Noordhoff, Groningen, 1953) method and Cauchy’s integral formula, the reduced boundary equation is further converted into a first-order differential equation. Consequently, the solution of the differential equation is mapped back to the physical region and the complex potential representing the displacement is formulated in finite form. The results show that the stresses at the crack tip are finite and the stress singularity is eliminated completely by the incorporation of the surface effects. The non-singularity of the stresses at the crack tip derived from the present article radically differs from the square-root singularity obtained from the classic theory of the elastic fracture and, however, is reasonable from the physical essence. The numerical results shows that when the surface effects are incorporated, the magnitudes of the stresses and the image forces near the crack tip are reduced, which are strongly dependent on the crack size. The corresponding results approach those from the classic theory as the crack length increases. The results by the present solution are also compared with those from other solutions.

Identifying at-risk students is one of the most important issues in online education. During different stages of a semester, students display various online learning behaviors. Therefore, we propose a phased prediction model to predict at-risk students at different stages of a semester. We analyze students’ individual characteristics and online learning behaviors, extract features that are closely related to their learning performance, and propose combined feature sets based on a time window constraint strategy and a learning time threshold constraint strategy. The results of our experiments show that the precision of the proposed model in different phases is from 90.4 to 93.6%.

The molecular signaling mechanisms of Coenzyme Q10 (CoQ10) in diabetic nephropathy (DN) remain poorly understood. We verified that mitochondrial abnormalities, like defective mitophagy, the generation of mitochondrial reactive oxygen species (mtROS) and the reduction of mitochondrial membrane potential, occurred in the glomerulus of db/db mice, accompanied by reduced PINK and parkin expression and increased apoptosis. These changes were partially reversed following oral administration of CoQ10. In inner fenestrated murine glomerular endothelial cells (mGECs), high glucose (HG) also resulted in deficient mitophagy, mitochondrial dysfunction and apoptosis, which were reversed by CoQ10. Mitophagy suppression mediated by Mdivi-1 or siPINK abrogated the renoprotective effects exerted by CoQ10, suggesting a beneficial role for CoQ10-restored mitophagy in DN. Mechanistically, CoQ10 restored the expression, activity and nuclear translocation of Nrf2 in HG-cultured mGECs. In addition, the reduced PINK and parkin expression observed in HG-cultured mGECs were partially elevated by CoQ10. CoQ10-mediated renoprotective effects were abrogated by the Nrf2 inhibitor ML385. When ML385 abolished mitophagy and the renoprotective effects exerted by CoQ10, mGECs could be rescued by treatment with mitoTEMPO, which is a mtROS-targeted antioxidant. These results suggest that CoQ10, as an effective antioxidant in mitochondria, exerts beneficial effects in DN via mitophagy by restoring Nrf2/ARE signaling. In summary, CoQ10-mediated mitophagy activation positively regulates DN through a mechanism involving mtROS, which influences the activation of the Nrf2/ARE pathway.

Machine health monitoring is of great importance in industrial informatics field. Recently, deep learning methods applied to machine health monitoring have been proven effective. However, the existing methods face enormous difficulties in extracting heterogeneous features indicating the variation until failure and revealing the inherent high-dimensional features of massive signals, which affect the accuracy and efficiency of machine health monitoring. In this paper, a novel data-driven machine health monitoring method is proposed using adaptive kernel spectral clustering (AKSC) and deep long short-term memory recurrent neural networks (LSTM-RNN). This method include three steps: First, features in the time domain, frequency domain, and time-frequency domain are, respectively, extracted from massive measured signals. And, an Euclidean distance based algorithm is designed to select degradation features. Second, the AKSC algorithm is introduced to adaptively identify machine anomaly behaviors from multiple degradation features. Third, a new deep learning model (LSTM-RNN) is constructed to update and predict the failure time of the machine. The effectiveness of the proposed method is validated using a set of test-to-failure experimental data. The results show that the performance of the proposed method is competitive with other existing methods.

Health degradation monitoring of rolling element bearing (REB) is of great significance to ensure safety and availability of mechanical equipment. This paper presents a new online health degradation monitoring method of REB based on growing self-organizing mapping (GSOM) and clustered support vector machine (CSVM). In the proposed method, multidimensional health degradation features of the REB are extracted to reflect health degradation process, including time-domain statistical features, frequency spectrum features, intrinsic mode function energy features, wavelet packet frequency band energy features. Multiple GSOMs are utilized to adaptively fuse each kind of the extracted features for the health indices. CSVM is constructed to achieve accurate health status identification of the REB. A health degradation experimental case of the REB is analyzed to demonstrate the effectiveness of the proposed method. The results show that the proposed method has obvious superior performance compared to other existing methods.

BACKGROUND Endometrial carcinoma (EC) is the most common gynecological malignancy worldwide, and 15-20% of patients with EC have a rapid relapse within 3 years. This study aims to develop an autophagy-related genes (ARGs) signature to predict the prognosis of EC. MATERIAL AND METHODS In our study, differentially expressed ARGs were identified by "edgeR" package in R and pathway enrichment analysis was performed to explore biological functions. Univariate and multivariate Cox regression analyses were employed to build autophagy signature. Gene set enrichment analysis (GSEA), Kaplan-Meier curve analysis, and ROC curve analysis were conducted to compare the differences between the high- and low-risk groups. RESULTS A total of 60 differentially expressed ARGs (DEARGs) including 34 upregulated and 26 downregulated DEARGs were identified from the TCGAUCEC dataset, with the adjusted P0.05 and |Fold Change|1.5. By using univariate and multivariate Cox regression analyses, ERBB2, PRKAB2, GRID2, NRG3, CDKN2A were identified to construct a prognostic signature with AUC 0.673, 0.719, and 0.791, at 1-, 3- and 5- years, respectively. Patients with EC were divided into low- or high-risk group by median risk score, and GSEA showed that low-risk group was enriched in adjacent cells communication pathways while high-risk group was involved in metabolism and immune pathways. The nomograms could also help to guide personal prognostic prediction and therapeutic strategies in EC. CONCLUSIONS Our study not only determine 5 ARGs signature that could predict the prognosis of EC but also provide novel insights into the underlying mechanisms of autophagy.

The purpose of this study was to explore the value of comprehensive nursing intervention in the perioperative period of ruptured bleeding of ectopic pregnancy.From January 2015 to January 2020, 164 patients with rupture and bleeding of ectopic pregnancy who needed laparoscopic treatment in the department of gynecology at our hospital were selected and randomly divided into the basic nursing group and the comprehensive nursing group, with 82 cases each. During the perioperative period, comprehensive nursing intervention or basic nursing intervention were performed, and the nursing effects of the two nursing interventions were compared.The disappearance time of abdominal pain, the time to get out of bed, and the length of hospitalization in the comprehensive nursing group were significantly shorter than those in the basic nursing group (P0.05). After surgery, blood sugar levels, aldosterone, cortisol, C-reactive protein (CRP), and IL-6 in the two groups were significantly higher than those before surgery (P0.05), but there was no statistically significant difference between the groups (P0.05). After the operation, the proportion of patients with Visual Analogue Scale (VAS) scores of 7-10 in the comprehensive nursing group was significantly lower than that in the basic nursing group (P0.05). Before the intervention, the Hamilton Anxiety Scale (HAMA) and Hamilton Depression Scale (HAMD) scores of the two groups were compared, and the difference was not statistically significant (P0.05). After the intervention, the HAMA and HAMD scores of the comprehensive nursing group were significantly lower than those of the basic nursing group (P0.05). The fallopian tube recanalization rate of patients in the comprehensive care group was significantly higher than that of the basic care group (P0.05), and the complication rate was significantly lower than that of the basic care group (P0.05).In summary, a comprehensive nursing program during the perioperative period can improve the treatment effect and significantly shorten the recovery time of patients, which is worthy of clinical promotion.

Deep learning-based methods have achieved promising performance in early detection and classification of lung nodules, most of which discard unsure nodules and simply deal with a binary classification -- malignant vs benign. Recently, an unsure data model (UDM) was proposed to incorporate those unsure nodules by formulating this problem as an ordinal regression, showing better performance over traditional binary classification. To further explore the ordinal relationship for lung nodule classification, this paper proposes a meta ordinal regression forest (MORF), which improves upon the state-of-the-art ordinal regression method, deep ordinal regression forest (DORF), in three major ways. First, MORF can alleviate the biases of the predictions by making full use of deep features while DORF needs to fix the composition of decision trees before training. Second, MORF has a novel grouped feature selection (GFS) module to re-sample the split nodes of decision trees. Last, combined with GFS, MORF is equipped with a meta learning-based weighting scheme to map the features selected by GFS to tree-wise weights while DORF assigns equal weights for all trees. Experimental results on the LIDC-IDRI dataset demonstrate superior performance over existing methods, including the state-of-the-art DORF.

For flexible job shop scheduling, an improved migratory bird migration algorithm is proposed to minimize the maximum completion time. Firstly, global, local and random selection methods were used to initialize the population, which improved the quality of the initial solution; The structure of migratory bird group was changed by increasing leading birds, and the crossover and mutation operators were introduced to enhance the global optimization ability. And different variable neighborhood search algorithms on critical path were introduced among different bird groups, which improved the local optimization ability of and the diversity of migratory bird groups; The jumping and competition mechanisms were proposed to improve the ability of the algorithm to jump out of the local optimal solution and convergence. Finally, the simulation results of the benchmarks verified the algorithm’s searching ability and stability of the algorithm.

The purpose of this study was to investigate the influence of smartphone reading on the ocular surface and to compare the various effects of different screens and light conditions on the ocular surface. One hundred nineteen volunteers were randomly divided into: light + organic light‐emitting diode (OLED), light + electronic ink (eINK), dark + OLED, and dark + eINK. Ocular surface examinations, including noninvasive break‐up time (NIBUT), noninvasive keratograph tear meniscus height (NIKTMH), ocular redness, fluorescein break‐up time (FBUT), corneal fluorescein staining, meibomian gland assessment, Schirmer I Test, and blinking frequency, were performed before and after a reading task. Symptoms were evaluated using the Ocular Surface Disease Index (OSDI) and Computer Vision Syndrome Questionnaire (CVS‐Q). NIBUT and FBUT were decreased statistically significantly after participants read on an OLED screen for 2 hours compared with the baseline in light and dark environments, whereas no statistically significant decrease was observed on an eINK screen. NIKTMH was statistically significantly decreased after reading on an OLED screen in light and dark settings, and the eINK screen had a lesser effect on NIKTMH. An obvious increase in the ocular redness, OSDI and CVS‐Q scores was observed after reading on an OLED screen, whereas the eINK screen had a lesser effect on these indicators. Blink rate increased gradually in OLED subgroups during the reading task, whereas no statistically significant difference was observed in the eINK subgroups. Our research suggested that reading on an OLED screen can cause ocular surface disorder and obvious subjective discomfort, whereas reading on an eINK screen can minimize ocular surface disorder in both dark and light environments.

The user’s behavior sequence can well reflect the user’s interest and intention, but there is less research on the use of user behavior sequence in the field of CTR prediction. Therefore, introducing the idea of sequence recommendation into CTR prediction is an exciting idea. Aimless browsing by users is a common phenomenon in many recommended scenarios (e-commerce, music, video streaming), which has not been paid attention to in previous research. To this end, this paper introduces the concept of user browsing status, and divides it into Discover and Intent, which respectively represent the user’s unintentional status and intentional status. A new framework named User Status Recognition Framework (USRF) is proposed to solve this problem. USRF can perform both CTR prediction and next-item recommendation. The framework captures the weights of two user status from the current user historical behavior sequence, models the two different status separately to mine user interests, and combines the captured user status to make more accurate recommendations. In addition, in order to solve the problem that less attention has been paid to the complex connection between candidate items and interacted items in the previous sequence recommendation research, this paper uses the attention mechanism to model the relationship between candidate items and interacted items, implements a simple model for CTR prediction based on USRF. Experiments on three different scene datasets show good results on both the AUC and F1-score, proving the advantages of the framework.

By numerically solving the Bogoliubov-de Gennes equations for the single vortex state in a two-band superconductor, we demonstrate that the disparity between the healing lengths of two contributing condensates is strongly affected by the band Fermi velocities, even in the presence of the magnetic field and far beyond the regime of nearly zero Josephson-like coupling between bands. Changing the ratio of the band Fermi velocities alters the temperature dependence of the condensate lengths and significantly shifts parameters of the ``length-scales locking" regime at which the two characteristic lengths approach one another., Comment: 10 pages, 9 figures

As sequential recommendation mainly obtains the user preference by analyzing their transactional behavior patterns to recommend the next item, how to mine real preference from user’s sequential behavior is crucial in sequential recommendation, and how to find the user long-term and short-term preference accurately is the key to solve this problem. Existing models mainly consider either the user short-term preference or long-term preference, or the relationship between items in one session, ignoring the complex item relationships between different sessions. As a result, they may not adequately reflect the user preference. To this end, in this paper, a Long- and Short-Term Preference Network (LSPN) based on graph embedding for sequential recommendation is proposed. Specifically, item embedding with a complex relationship of items between different sessions is obtained based on graph embedding. Then this paper constructs the network to obtain the user long- and short-term preferences separately, combing them through the fuzzy gate mechanism to provide the user final preference. Furthermore, the results of experiments on two datasets demonstrate the efficiency of our model in Recall@N and MRR@N.

Image ordinal estimation is to predict the ordinal label of a given image, which can be categorized as an ordinal regression problem. Recent methods formulate an ordinal regression problem as a series of binary classification problems. Such methods cannot ensure that the global ordinal relationship is preserved since the relationships among different binary classifiers are neglected. We propose a novel ordinal regression approach, termed Convolutional Ordinal Regression Forest or CORF, for image ordinal estimation, which can integrate ordinal regression and differentiable decision trees with a convolutional neural network for obtaining precise and stable global ordinal relationships. The advantages of the proposed CORF are twofold. First, instead of learning a series of binary classifiers \emph{independently}, the proposed method aims at learning an ordinal distribution for ordinal regression by optimizing those binary classifiers \emph{simultaneously}. Second, the differentiable decision trees in the proposed CORF can be trained together with the ordinal distribution in an end-to-end manner. The effectiveness of the proposed CORF is verified on two image ordinal estimation tasks, i.e. facial age estimation and image aesthetic assessment, showing significant improvements and better stability over the state-of-the-art ordinal regression methods., Comment: Accepted by IEEE TNNLS

Various simulations have been conducted to understand the macroscopic behavior of particles in the solid-gas flow in rotating drums in the past. In these studies, the no-slip wall boundary condition and fixed restitution coefficient between particles were usually adopted. The paper presents a numerical study of the gas-solid flow in a rotating drum to understand the effect of the specularity coefficient and restitution coefficient on the hydrodynamic behavior of particles in the segregation process. The volume fraction, granular pressure, granular temperature and their relationships are examined in detail. The boundary conditions of the no-slip and specularity coefficient of 1 are compared. In the simulations, two different sizes of particles with the same density are considered and the Eulerian–Eulerian multiphase model and the kinetic theory of granular flow (KTGF) are used. The results reveal that the hydrodynamical behavior of the particles in the rotating drum is affected by the boundary condition and restitution coefficient. In particular, the increase of specularity coefficient can increase the active region depth, angle repose, granular pressure for both small and large particles and granular temperature for large particles. With increasing restitution coefficient, the angle of repose decreases and granular pressure and temperature increase at the same volume fraction for both small and large particles.

In order to accurately identify whether there are damages and damage types on the surface of the aerostat capsule, combined with the powerful data processing capabilities and abnormal pattern recognition capabilities of deep learning, this paper proposes a continuous wavelet transform (CWT) and deep convolution. The diagnosis method of capsule surface damage combined with convolutional neural network (Convolutional Neural Networks, CNN). First, use CWT to convert the collected original stress and strain signals into time-frequency domain images, and then use CNN to classify and identify the time-frequency domain images to determine the damage category of the capsule surface. The CWT-CNN method is different from the traditional fault diagnosis method, it needs to go through the traditional feature extraction process, and the pros and cons of the extracted features often determine the final recognition accuracy. This method effectively overcomes the traditional fault diagnosis method that requires a large amount of signal processing technology. And rich engineering practice experience to extract the shortcomings of fault experience. The experimental results show that the CWT-CNN method can achieve an accuracy of more than 95% in the recognition of the surface damage of the capsule.

The capsule of the buoy is usually woven with composite materials. The existence of material anisotropy makes it difficult to establish the degradation mechanism model of the capsule, which affects the defect detection and health management of the capsule. To this end, this paper proposes a data-driven airbag defect detection method based on stress wave measurement. Firstly, strain signals under acoustic excitation are collected by sticking strain gauges on the surface of the capsule. Then, VMD is used to decompose the signal and extract the energy characteristics of each modal signal component to construct the feature vector. Finally, the feature vector is input into the GenSVM model for defect type recognition. The example analysis shows the effectiveness and applicability of the proposed method.

The problem of the nonlinear thermal buckling and post-buckling of magneto-electro-thermo-elastic functionally graded porous nanobeams is analyzed based on Eringen’s nonlocal elasticity theory and by using a refined beam model. The beams with immovable clamped ends are exposed to the external electric voltages, magnetic potentials, a uniform transverse load and uniform temperature change. For the first time, the four types of porosity distribution in the nanobeam are considered and compared in complex electric–magnetic fields. Besides, the new formula of the effective material properties is proposed in this paper to simultaneously estimate the material distribution and porosity distribution in the thickness direction. The generalized variation principle is used to induce the governing equations, then the approximate analytical solution of the METE-FG nanobeams based on physical neutral surface is obtained by using a two-step perturbation technique. Finally, detailed parametric analyses are performed to get an insight into the effects of different physical parameters, including the slenderness ratio, small-scale parameter, volume fraction index, external electric voltages, magnetic potentials, porosity coefficient and different porosity distributions, for providing an effective way to improve post-buckling strength of porous beams.

A theoretical model is developed that examines the effect of grain boundary (GB) sliding near a nanovoid on crack growth in deformed nanocrystalline materials. By combining complex variable method of Muskhelishvili, superposition principle of elasticity and distributed dislocation technique, the singular integral equations are solved numerically and then the stress intensity factors (SIFs) near the left crack tip are obtained. The influences of the location of the disclination dipole, the dipole arm, the orientation and the relative crack length on the SIFs near the left crack tip are evaluated in detail. The results indicate that the wedge disclination dipole produced by the GB sliding shields mode I crack tip, but anti-shields mode II crack tip. At the same time, surface stresses of the nanovoid characterized by positive surface elasticity and negative surface elasticity both show shielding effect to mode I SIFs of the crack tip, yet have negligible effect on mode II SIFs of the crack tip. Meanwhile, the mode I crack tip is more significantly shielded by the negative residual surface stresses than that of the positive residual surface stresses.