Data Fusion Model Research Articles

Quantitative analysis of textile delusterant based on terahertz spectral and data fusion strategies

Titanium dioxide is a delusterant and an important component in the manufacturing of polyester fiber. For the need of fast, accurate and nondestructive detection of matting agents in textiles, a quantitative analysis method based on terahertz absorption spectroscopy and derivative spectroscopy, combined with chemometrics and data fusion strategy is proposed. This experiment was used two spectra for fusion. The terahertz absorption spectra were obtained in the band of 0.2–1.9 THz by optical parameter extraction. The derivative spectrum was derived from the first-order derivative of the absorption spectrum. Partial least squares (PLS) and data fusion were used to construct a prediction model for titanium dioxide concentration in polyester fiber. Low-level data fusion was the direct combination of two spectral data; The successive projections algorithm (SPA) and Monte Carlo uninformative variable elimination (MCUVE) were employed by mid-level data fusion for feature selection, after which the feature variables were fused; multiple linear regression was used for fusion by high-level data fusion. The prediction accuracy of the high-level data fusion model is higher than that of other models, which the correlation coefficient of cross-validation (Rcv) and correlation coefficient of prediction (Rp) are 0.9229 and 0.9227. The mean relative error (MRE) is 0.2654. The results show that terahertz spectroscopy combined with chemometric methods and high-level data fusion strategies can achieve rapid, accurate and non-destructive detection of titanium dioxide in polyester fiber, which can lay the theoretical foundation for terahertz spectroscopy detection methods for textile additives.

Symbolic aggregate approximation based data fusion model for dangerous driving behavior detection

Detecting dangerous driving behavior is of great significance for reducing the occurrence of traffic accidents, and very challenging as it is affected by multiple factors. However, the existing methods are merely suitable for detecting dangerous driving behavior caused by single factor. Therefore, this paper proposes a novel Symbolic Aggregate approXimation based Data Fusion model (SAX-DF) to fuse multi-source data and applies it to accurately detect dangerous driving behavior. The proposed method considers multiple influencing factors at the same time and analyzes the related information between these factors. Further, a Positive Danger Mapping (PDM) algorithm is designed to unify the direction of multiple influencing factors. To verify the performance of SAX-DF, several experiments are carried out on traffic detector data and weather data. The experimental results illustrate that the proposed method can more effectively detect the dangerous driving behavior compared with the benchmark methods.

Multi-Sensor Data Fusion with a Reconfigurable Module and Its Application to Unmanned Storage Boxes.

We present a multi-sensor data fusion model based on a reconfigurable module (RM) with three fusion layers. In the data layer, raw data are refined with respect to the sensor characteristics and then converted into logical values. In the feature layer, a fusion tree is configured, and the values of the intermediate nodes are calculated by applying predefined logical operations, which are adjustable. In the decision layer, a final decision is made by computing the value of the root according to predetermined equations. In this way, with given threshold values or sensor characteristics for data refinement and logic expressions for feature extraction and decision making, we reconstruct an RM that performs multi-sensor fusion and is adaptable for a dedicated application. We attempted to verify its feasibility by applying the proposed RM to an actual application. Considering the spread of the COVID-19 pandemic, an unmanned storage box was selected as our application target. Four types of sensors were used to determine the state of the door and the status of the existence of an item inside it. We implemented a prototype system that monitored the unmanned storage boxes by configuring the RM according to the proposed method. It was confirmed that a system built with only low-cost sensors can identify the states more reliably through multi-sensor data fusion.

MUST: A Multi-source Spatio-Temporal data fusion Model for short-term sea surface temperature prediction

Sea surface temperature (SST) plays an important role in various oceanic applications, including climate prediction, ocean environment monitoring and marine disaster warning. Although many approaches have been developed for predicting SST, most of them conduct the prediction only based on historical SST data. However, SST is essentially affected by many external factors, e.g., the short-wave radiation from the sun and the long-wave radiation from the atmosphere and ground, which are ignored by existing approaches. In this work, we proposed a Multi-source Spatio-Temporal data fusion model (MUST) to fuse multi-source data, including SST data and external factors, to improve the accuracy of short-term SST prediction. Concretely, MUST first introduces Bicubic Convolutional Interpolation (BCI) to address the issue of inconsistent spatial resolutions of multi-source data, then employs the Spatio-Temporal Dilated ConvLSTM (ST-DC) to learn the spatio-temporal features of SST and external factors, and finally fuses the learned features from multi-source data to predict SST with a Cross Data Fusion (CDF) component. As validated on two real datasets, MUST achieves much better performance than existing SST prediction approaches.

Fusion of Gamma-rays and portable X-ray fluorescence spectral data to measure extractable potassium in soils

The detection and mapping of plant-extractable potassium (Ka) using proximal soil sensing and data fusion (DF) techniques are essential to optimise K2O fertiliser application, improve crop yield, and minimise environmental and financial costs. This work evaluates the potential of combined use of portable gamma ray and x-ray fluorescence spectroscopy for in field detection and mapping of Ka. After subjected to various pre-processing methods, spectral data were split into calibration (75%) and validation (25%) sets, and single sensor and DF models were developed using partial least squares regression (PLSR). Maps of Ka were used to present spatial variability of potassium across an 8.4 ha Voor de Hoeves target field, in Flanders, Belgium. Results showed that the gamma-ray PLSR model using wet soils had greater predictive ability with coefficient of determination (R2) = 0.71, ratio of performance deviation (RPD) = 1.89, root mean square error (RMSE) = 31.7 mg kg-1, and ratio of performance to interquartile range (RPIQ) = 2.36 than the corresponding wet-XRF PLSR model (R2 = 0.61, RPD = 1.64, RMSE = 48.8 mg kg-1 and RPIQ = 1.84). The DF PLSR model on wet soils, resulted in a more accurate Ka predictive ability (R2 = 0.75, RPD = 2.03, RMSE = 31.3 mg kg-1 and RPIQ = 2.79), compared to the individual gamma ray or XRF sensors in wet soils. The best accuracy was obtained with XRF spectrometer using dry samples (R2 = 0.77, RPD = 2.14, RMSE = 26.5 mg kg-1 and RPIQ = 3.39). All Ka prediction maps showed spatial similarity to the corresponding measured maps in the target field. In conclusion, since DF increased the Ka prediction accuracy compared to the single sensor models using wet soils, it is recommended to be adopted in future studies as a potential solution for Ka measurement, mapping, and ultimately for site-specific K2O fertilisation management. The XRF analysis of dry spectra is recommended as the best method for accurate measurement of Ka.

A convolutional neural network-based multi-sensor fusion approach for in-situ quality monitoring of selective laser melting

Selective laser melting (SLM) is an emerging and popular metal additive manufacturing (AM) technique to fabricate advanced metal components with complex geometries. However, its broad adoption in industry is still hampered by poor process repeatability and low part consistency. To overcome this limitation, there are growing efforts towards in-situ monitoring and control technologies for quality assessment of parts. In this paper, a convolutional neural network (CNN)-based multi-sensor fusion approach is proposed to integrate layer-wise images, acoustic emission signals, and photodiode signals for in-situ quality monitoring of SLM. An off-axial in-situ monitoring system equipped with three types of sensors, namely, a digital camera, a microphone, and a photodiode is first developed for signatures acquisition. Porosity and density measurements are then carried out to label the layer-based sensing data and measure the quality. Thereafter, an image processing method and a signal-to-image strategy are proposed to merge the multi-source heterogeneous sensor data. Finally, three CNN-based multi-sensor fusion models are developed from data-level fusion, feature-level fusion, and decision-level fusion respectively and compared in quality identification. Data-level fusion model is established through channel fusion, features are extracted and fused in the feature-level fusion model, and the decision-level fusion model is developed by fusing the classification results from individual models. Results showed that the proposed CNN-based multi-sensor fusion approach can significantly improve the classification accuracy compared to the three individual sensor-based models. Furthermore, the feature-level data fusion model exhibits the best classification performance among the three data fusion models.

A Generic Indirect Deep Learning Approach for Multisensor Degradation Modeling

To monitor the degradation status of units and prevent unexpected failures in engineering systems, health index (HI)-based data fusion technologies have been rapidly developed by combining multiple sensor signals, which are helpful to understand the degradation processes of units and predict their remaining useful lifetime (RUL). Although promising, existing HI-based data fusion models for degradation modeling are still limited due to the restrictive assumptions made during the fusion or the degradation modeling processes, e.g., assuming the fusion model as a linear or kernel-based function from multiple sensor signals, or modeling the degradation process by a preselected basis function. Such assumptions are often invalid in industrial practice and may fail to accurately characterize the complicated relationships between multiple sensor signals and the underlying degradation process. To address the issue, this article proposes a generic indirect deep learning method that constructs an HI by combining multiple sensor signals to better characterize the degradation process. In particular, our innovative idea is to seamlessly integrate a deep neural network (DNN) and a long short term memory (LSTM) model to construct the HI by fusing multiple sensor signals and characterize the degradation process, which can be applied to the degradation modeling of various engineering systems. Domain knowledge including the concept of failure threshold and monotonicity of the degradation process is also considered to enhance the interpretability of the proposed method. For parameter estimation, we develop an indirect gradient descent (IGD) algorithm to train the proposed method. Simulation studies and a case study on the degradation of aircraft gas turbine engines are presented to validate the performance of the proposed method. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The article aims to develop a generic health index (HI)-based data fusion method for degradation modeling when multiple sensors are available to monitor the degradation status of a unit. Specifically, the developed method addresses two challenging questions in practice: 1) how to effectively combine multiple sensor signals to construct an HI that accurately characterizes the underlying degradation status and 2) how to flexibly model the degradation evolution based on the constructed HI. To implement this method in practice, four steps are included as follows: <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">First</i> , collecting multiple sensor signals and failure time of historical units. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Second</i> , constructing the HI and describing the underlying degradation process by training a deep neural network (DNN) model and a long short term memory (LSTM) model, respectively. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Third</i> , estimating model parameters using the proposed IGD algorithm. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Fourth</i> , constructing the HI of in-service units and predicting their remaining useful lifetime (RUL) using the constructed HI. The proposed method is expected to be able to characterize various degradation processes and be applied to the degradation modeling of different engineering systems.

Data Fusion Model for Muscle Proteomics in Sports Applications

Proteome is a cell, tissue, or organism to express all the floorboard of the protein; proteomics is the study of proteomics is an emerging discipline; the research on the law of occurrence of sports fatigue and its mechanism is an important and challenging subject in the field of sports medicine. Exercise-induced fatigue refers to the physiological process in which the body’s functional ability or work efficiency declines and cannot be maintained at a specific level during exercise. With the improvement of the modern competitive sports level and the increasingly fierce competition, the athletes have to bear more and more loads in sports training, and the probability of sports fatigue is also higher. Appropriate sports fatigue and reasonable recovery methods can promote the improvement of athletes’ functional level; on the contrary, excessive fatigue not only affects the training effect but may also cause various dysfunctions, which may damage the athletes’ health. Therefore, understanding the mechanism of sports fatigue is of positive significance for accelerating the elimination of sports fatigue. The purpose of this paper is to study the progress of muscle proteomics in sports. Since each athlete has individual differences, different levels of muscle function corresponding to the left lower limb will produce different EMG signals. Static experiments and random visual stimulation evaluation experiments were performed on the left lower limbs of athletes, and they were coordinated with the right lower limbs, and then the left and right lower limbs were compared with the same muscle-weighted RMS (Recipe Management System) and IEMG (Comprehensive electromyography). Experimental results show that using the statistical method of single-factor variance analysis to the same muscles in two states of RMS was analyzed, respectively, under the two states of differences between muscle electromyography IEMG were analyzed, finally, the muscle integral electrical values on the contribution rate of extent of each muscle under the two states are analyzed. For one-way ANOVA, this study defined the significance level of the difference analysis between groups as P &lt; 0.05 . P &lt; 0.05 is a small probability event, indicating that the possibility of an event occurring is very small, so it is considered that the event is unlikely to occur. A low-probability event is an event that has a low probability of occurring. Then, it is almost impossible to happen in one experiment, but it is bound to happen in many repeated experiments. With the continuous improvement of existing technologies and the emergence of more new technologies, proteomics research will make greater contributions to elucidate the mechanism of exercise on skeletal muscle remodeling and its improvement on health.

The FIRST model: Spatiotemporal fusion incorrporting spectral autocorrelation

Over the past decade, spatiotemporal fusion has become an indispensable tool for monitoring land surface dynamics due to its promising ability to produce surface reflectance products with both high spatial and temporal resolutions. However, existing fusion methods usually generate multispectral band products by predicting each spectral band separately, so the useful information of spectral autocorrelation within the spectrum has been ignored and waits to be exploited. To address this issue, we propose a novel spatiotemporal fusion method, the spatiotemporal Fusion Incorrporting Spectral autocorrelaTion (FIRST) model, to fully utilize the multiple spectral bands of surface reflectance products. Compared with other fusion methods, the model has three distinct advantages: (1) it utilizes spectral autocorrelation in a many-to-many regression framework that simultaneously inputs and predicts multispectral bands without the collinearity effect; (2) it maintains high fusion accuracy when the spatiotemporal variation is large with acceptable computational efficiency; and (3) it can produce robust results even with input images contaminated by haze and thin clouds. We tested the FIRST model at several experimental sites and compared it with four typical methods, the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM), Flexible Spatiotemporal DAta Fusion (FSDAF) model, the regression model Fitting, spatial Filtering and residual Compensation (Fit-FC) model and the enhanced STARFM (ESTARFM). The results demonstrate that FIRST yields better overall performance for its simple and effective technical principles. FIRST is thus expected to provide high-quality remotely sensed data with high spatial resolution and frequent observations for various applications.

An IoT-based intelligent irrigation system with data fusion and a self-powered wide-area network

Water resources have a great influence on human society, but saving water in irrigation still remains a challenge. This article proposes an intelligent irrigation system that integrates a data fusion model and a long-rang (LoRa) network for optimizing the watering schedule. A data fusion model is proposed, which first adopts the long short-term memory (LSTM) network to simulate and predict the proper watering demands by integrating multi-source heterogeneous data, that is, historical weather data, user irrigation logs, weather forecasts, and online monitoring sensor data. A self-powered wide-area network is implemented and deployed by using LoRa to facilitate multiple Internet of Things (IoT) application scenarios. It includes a gateway and two types of nodes: a valve node and a sensing node. The node is capable of energy autonomy through the scheme of waterflow-based power generation, thus realizing maintenance-free throughout the life cycle. A cloud platform is designed to provide network management, intelligent irrigation control, and the interface of the mobile application. The proposed system is evaluated through a case study of landscape watering. On average, the proposed system achieves a water-saving efficiency of 94.74% compared with the conventional manual setting solutions.

Marfusion: An Attention-Based Multimodal Fusion Model for Human Activity Recognition in Real-World Scenarios

Human Activity Recognition(HAR) plays an important role in the field of ubiquitous computing, which can benefit various human-centric applications such as smart homes, health monitoring, and aging systems. Human Activity Recognition mainly leverages smartphones and wearable devices to collect sensory signals labeled with activity annotations and train machine learning models to recognize individuals’ activity automatically. In order to deploy the Human Activity Recognition model in real-world scenarios, however, there are two major barriers. Firstly, sensor data and activity labels are traditionally collected using special experimental equipment in a controlled environment, which means fitting models trained with these datasets may result in poor generalization to real-life scenarios. Secondly, existing studies focus on single or a few modalities of sensor readings, which neglect useful information and its relations existing in multimodal sensor data. To tackle these issues, we propose a novel activity recognition model for multimodal sensory data fusion: Marfusion, and an experimental data collection platform for HAR tasks in real-world scenarios: MarSense. Specifically, Marfusion extensively uses a convolution structure to extract sensory features for each modality of the smartphone sensor and then fuse the multimodal features using the attention mechanism. MarSense can automatically collect a large amount of smartphone sensor data via smartphones among multiple users in their natural-used conditions and environment. To evaluate our proposed platform and model, we conduct a data collection experiment in real-life among university students and then compare our Marfusion model with several other state-of-the-art models on the collected datasets. Experimental Results do not only indicate that the proposed platform collected Human Activity Recognition data in the real-world scenario successfully, but also verify the advantages of the Marfusion model compared to existing models in Human Activity Recognition.

Fusion of THz-TDS and NIRS Based Detection of Moisture Content for Cattle Feed

As an essential index to evaluate feed quality, feed moisture content which is too high or too low will impose an adverse impact on feed nutritional value. Therefore, the quantitative analysis of feed moisture content is significant. In this paper, the detection of feed moisture content based on terahertz (THz) and near-infrared (NIR) spectroscopy and data fusion technology of THz and NIR (THz-NIR) was investigated. First, feed samples with different water content (29.46%–49.46%) were prepared, and THz (50–3000 μm) and NIR (900–1700 nm) spectral data of samples was collected and preprocessed, and the feed samples were divided into correction set and verification set by 2:1. Second, the spectral data was fused through the head-to-tail splicing, and the feed moisture content prediction model was established combined with partial least squares regression (PLSR). Third, competitive adaptive reweighting sampling (CARS) was applied to extract spectral characteristic variables for feature layer fusion, and the feed moisture content prediction model in feature level was constructed combined with PLSR. Finally, the evaluation parameters validation set correlation coefficient (Rp), the root mean square error of prediction (RMSEP), and the residual predictive deviation (RPD) were employed to evaluate the prediction effect of the model. The results indicated that THz, NIR spectra, and data fusion technology could quickly and effectively predict feed moisture content. Among them, the characteristic layer spectral data fusion model achieved the optimal prediction effect while Rp, RMSEP, and RPD reached 0.9933, 0.0069, and 8.7386 respectively. In conclusion, compared with the prediction model established by single THz and NIR spectrum, THz-NIR spectrum data fusion could more accurately predict feed moisture content and provide certain theoretical and technical support for inspirations and methods for quantitative analysis of feed moisture content of livestock and poultry.

Teaching Method of Japanese Professional Writing Course Based on Big Data Fusion

In recent years, the number of Japanese learners has increased year by year. When Japanese is learned as a language, knowledge of the language itself is indispensable. This article aims to explore the teaching methods of Japanese professional writing courses based on the integration of big data. This article first introduces big data fusion and introduces the definition and model of data fusion. Generally, there are two types of data fusion models: distributed and centralized. Then, I analyzed the artificial neural network algorithm, which is a computational model that imitates the animal brain, and then researched the context-based teaching of Japanese professional courses and analyzed the context-based teaching of Japanese professional writing. The results of the context-based Japanese course teaching research show that the average Japanese writing score of the experimental class is 72.7833, the control class score is 66.3333, and the composition fluency scores are 156.27 and 119.73. It can be concluded that the composition performance and composition fluency of students in the experimental class are higher than those in the control class, so educators should let students strengthen vocabulary spelling and memory skills in situational teaching.

A data-model fusion method for online state of power estimation of lithium-ion batteries at high discharge rate in electric vehicles

This paper proposes a novel data-model fusion method (DMFM) for online state of power (SOP) estimation of lithium-ion batteries at high discharge rates in electric vehicles. First, battery polarisation characteristics responding for high discharge rates are experimentally investigated through a series of decremental pulse tests. Battery polarisation voltage is observed with diverse growing patterns over a whole battery operation range, and its underlying correlations with state of charge (SOC), discharge rate and pulse runtime are recognised. Second, a feed-forward neural network (FFNN) with SOC, discharge rate and pulse runtime as inputs, is constructed to characterise battery polarisation voltage through modelling the current excited polarisation resistance. Third, a DMFM is proposed to combine the data-driven method and equivalent-circuit model based method for accurate online SOP estimation in a lengthy prediction window ranging from 30 s to 120 s. Moreover, an unscented Kalman filter is devised to filter the estimation outcomes of the DMFM for noise suppression. The experimental results validate the effectiveness of the constructed FFNN in reproducing the nonlinearity of battery polarisation characteristics at high discharge rates and show the significant improvement in SOP estimation accuracy.

Multitarget identification method for dual‐plane detection based on data fusion and correlation analysis

AbstractThe light screen that tests the velocity of a projectile is often affected by indoor illumination changes and interference caused by air disturbance of the projectile, affecting the identification of the projectile target signals by the test system. To solve these problems, based on the detection light screen formed by the array infrared diodes and the array photosensitive tube, this article proposes the method of identifying multiprojectile targets in the dual‐plane detection of data fusion and related analysis. Use the amplitude of signals generated by the projectile moving in the screen and the time width of valid signals as the diagnosis basis to build a multiprojectile signal identification model for a single plane detection light screen. Using the similarity of projectile passing through two light screens, a matching algorithm for the two‐plane detection screen projectile signal is established. By collecting the multiprojectile signals of the dual‐plane detection light screen and following the single detection light screen data fusion model, the velocity measuring system can identify the projectile signals of the dual‐plane detection screen. Based on the fusion and identification result of each detection plane and the established match algorithm related to the shot signal, the test result of the test system with interference is obtained. The result verifies the correctness of the established model and algorithm.

Multi critical quality attributes monitoring of Chinese oral liquid extraction process with a spectral sensor fusion strategy

The traditional Chinese medicine (TCM) extraction process is a complicated dynamic system with many variables and disturbance. Therefore, multi critical quality attributes (CQAs) monitoring is of great significance to understand the whole process. Spectroscopy is a powerful process analytical tool used for process understanding. However, single senor sometimes could not provide comprehensive information. Sensor fusion is a very practical method to overcome this deficiency. In this study, the extraction process of Xiao’er Xiaoji Zhike Oral Liquid (XXZOL) was carried out in pilot scale, where near infrared (NIR) spectroscopy and mid infrared (MIR) spectroscopy were collected to determine the concentrations of seven CQAs (synephrine, arecoline, chlorogenic acid, forsythoside A, naringin, hesperidin and neohesperidin) during extraction process. Based on fused data blocks, fusion partial least squares (PLS) models were established. Two fusion data blocks are obtained from the concatenation of original spectra (low-level data fusion) and the concatenation of characteristic variables based on band selection (mid-level data fusion) respectively. The results indicated that for all seven analytes, the mid-level data fusion models were superior to the single spectral models, with the prediction performance significantly improved. Specifically, the coefficients of determination (Rp2 and Rt2) of NIR, MIR and fusion quantitative models were all higher than 0.95. The relative standard errors of prediction (RSEP) values were all within 10%, except for models of neohesperidin, which were 10.76%, 12.39%, 12.05%, 10.03% for NIR, MIR, low-level and mid-level models respectively. These results demonstrate that it is feasible to monitor the extraction process of Xiao’er Xiaoji Zhike Oral Liquid more accurately and rapidly by fusing NIR and MIR spectroscopy, and the proposed approach also has vital and valuable reference value for the rapid monitoring of the mixed decoction process of other TCM.

Multi-modal fusion of satellite and street-view images for urban village classification based on a dual-branch deep neural network

• A novel dual-branch deep neural network is proposed for urban village classification. • Multi-modal satellite and street-view images are integrated to increase the performance. • The proposed model yields a high accuracy of 92.61% in Jing-Jin-Ji region of China. • The dataset, Satellite & Street-view images for Urban Village classification (S 2 UV), has been released publicly. With the rapid urbanization process in China, numerous urban villages have been appeared, which are surrounded by the newly-built urban blocks. Due to the high population density, poor hygiene, chaotic waste discharge, and inadequate public facilities, urban villages have many negative impacts on both urban environment and management. The objective of this study is to propose a dual-branch deep learning model for multi-modal satellite and street-view data fusion to detect urban villages in Beijing, Tianjin and Shijiazhuang, which are the core cities of Jing-Jin-Ji region of China. Specifically, the proposed model consists of a satellite branch, a street-view branch and a gated-fusion module. As for the satellite branch, a Trans-MDCNN (multi-scale dilated convolutional neural network) is proposed to learn multi-level local features and global contextual features from high resolution satellite imagery, while for the street-view branch, an MVRAN (multi-view recurrent attention network) is constructed to learn and fuse multi-angle features from street-view images. A gated-fusion module is designed to aggregate the important features of the dual-branches. Experimental results indicate that the proposed model has achieved good performance with an overall accuracy (OA) of 92.61%. Ablation study shows that compared with satellite data alone, the integration of street-view images could increase the OA by about 2%. Besides, 1-D feature fusion outperforms its 2-D counterpart and the classic feature concatenation method. The proposed model also yields a better performance than other deep learning models. Finally, the dataset of this study, S 2 UV (Satellite & Street-view images for Urban Village classification), is publicly available: https://doi.org/10.11922/sciencedb.01410.

Thickness-Related Fault Diagnosis of Steel Strip Based on W-KPLS Method Considering Mechanism Weight Optimization

Due to the lack of a reasonable mechanism explanation for the data model used in the process of quality-related fault diagnosis, the diagnosis model has insufficient ability to identify faults, resulting in the phenomenon of failure detection or false positive. Therefore, this paper adopted the method of mechanism and data model fusion to solve the problem of insufficient interpretation of the influence of existing diagnosis methods on rolling process variables. Firstly, the KPLS achieves strip quality-related fault detection for nonlinear processes. In order to find out the abnormal variables, a nonlinear contribution plot was introduced to calculate the contribution value of each variable to the monitoring index. Secondly, based on the bounce equation of the rolling process, the static comprehensive analysis of the steady rolling process was carried out to reveal the influence of various variables on strip thickness. Thirdly, based on the above analysis of the steady rolling process mechanism, the influence weight method and kernel function method were used to reconstruct and map the original input matrix. A kernel partial least squares method based on influence weight W optimization (W-KPLS) was proposed for quality-related fault monitoring and diagnosis. Finally, the model was applied in the cold rolling process of an aluminum alloy sheet, and the validity of the model was further verified by practical industrial data. The results show that the new method improves the fault detection rate by more than 20% compared with the traditional monitoring method, and the proportion of data points reaching the early warning limit was increased to more than 95%.

Multi-Source Data Processing and Fusion Method for Power Distribution Internet of Things Based on Edge Intelligence

With the rapid advancement of the Energy Internet strategy, the number of sensors within the Power Distribution Internet of Things (PD-IoT) has increased dramatically. In this study, an edge intelligence-based PD-IoT multi-source data processing and fusion method is proposed to solve the problems of confusing storage and insufficient fusion computing performance of multi-source heterogeneous distribution data. First, a PD-IoT multi-source data processing and fusion architecture based on edge smart terminals is designed. Second, the multi-source sensor data in the distribution network is unified in dimension and magnitude. By introducing the Box–Cox transform to improve the data offset problem in the Z-score normalization process, a multi-source heterogeneous data processing method for distribution networks based on the Box–Cox transform Z-score is proposed. Then, the conflicting phenomena of DS inference methods in data source fusion are optimally handled based on the PCA algorithm. A multi-source data fusion model based on DS inference with conflict optimization is constructed to ensure the effective fusion of distribution data sources from different domains. Finally, the effectiveness of the proposed method is verified by an experimental analysis of an IEEE39 node system in a regional distribution network in China.

Evaluating alternative ebullition models for predicting peatland methane emission and its pathways via data–model fusion

Abstract. Understanding the dynamics of peatland methane (CH4) emissions and quantifying sources of uncertainty in estimating peatland CH4 emissions are critical for mitigating climate change. The relative contributions of CH4 emission pathways through ebullition, plant-mediated transport, and diffusion, together with their different transport rates and vulnerability to oxidation, determine the quantity of CH4 to be oxidized before leaving the soil. Notwithstanding their importance, the relative contributions of the emission pathways are highly uncertain. In particular, the ebullition process is more uncertain and can lead to large uncertainties in modeled CH4 emissions. To improve model simulations of CH4 emission and its pathways, we evaluated two model structures: (1) the ebullition bubble growth volume threshold approach (EBG) and (2) the modified ebullition concentration threshold approach (ECT) using CH4 flux and concentration data collected in a peatland in northern Minnesota, USA. When model parameters were constrained using observed CH4 fluxes, the CH4 emissions simulated by the EBG approach (RMSE = 0.53) had a better agreement with observations than the ECT approach (RMSE = 0.61). Further, the EBG approach simulated a smaller contribution from ebullition but more frequent ebullition events than the ECT approach. The EBG approach yielded greatly improved simulations of pore water CH4 concentrations, especially in the deep soil layers, compared to the ECT approach. When constraining the EBG model with both CH4 flux and concentration data in model–data fusion, uncertainty of the modeled CH4 concentration profiles was reduced by 78 % to 86 % in comparison to constraints based on CH4 flux data alone. The improved model capability was attributed to the well-constrained parameters regulating the CH4 production and emission pathways. Our results suggest that the EBG modeling approach better characterizes CH4 emission and underlying mechanisms. Moreover, to achieve the best model results both CH4 flux and concentration data are required to constrain model parameterization.