Original Model Parameters Research Articles

Linear Logistic Scoring Equations for Latent Class and Latent Profile Models: A Simple Method for Classifying New Cases

Researchers are often interested in using latent class or latent profile parameter estimates to obtain posterior class membership probabilities for observations other than those of the original sample. In this paper, we demonstrate that these probabilities typically take on the form of linear logistic equations with coefficients which are functions of the original model parameters. In other words, the posterior class membership probabilities can be computed with a prediction formula similar to that of a multinomial logistic regression model. We derive the scoring equations for nominal, ordinal, count, and continuous indicators, as well as investigate models with missing values on class indicators, local dependencies, covariates, or multiple latent variables. In addition to the mathematical derivations of the scoring equations, we describe how either exact or approximate scoring equations can be obtained by estimating a multinomial regression model using a weighted data set.

Efficient Tobacco Pest Detection in Complex Environments Using an Enhanced YOLOv8 Model

Due to the challenges of pest detection in complex environments, this research introduces a lightweight network for tobacco pest identification leveraging enhancements in YOLOv8 technology. Using YOLOv8 large (YOLOv8l) as the base, the neck layer of the original network is replaced with an asymptotic feature pyramid network (AFPN) network to reduce model parameters. A SimAM attention mechanism, which does not require additional parameters, is incorporated to improve the model’s ability to extract features. The backbone network’s C2f model is replaced with the VoV-GSCSP module to reduce the model’s computational requirements. Experiments show the improved YOLOv8 model achieves high overall performance. Compared to the original model, model parameters and GFLOPs are reduced by 52.66% and 19.9%, respectively, while mAP@0.5 is improved by 1%, recall by 2.7%, and precision by 2.4%. Further comparison with popular detection models YOLOv5 medium (YOLOv5m), YOLOv6 medium (YOLOv6m), and YOLOv8 medium (YOLOv8m) shows the improved model has the highest detection accuracy and lightest parameters for detecting four common tobacco pests, with optimal overall performance. The improved YOLOv8 detection model proposed facilitates precise, instantaneous pest detection and recognition for tobacco and other crops, securing high-accuracy, comprehensive pest identification.

Ophthalmic artery Doppler at 36 weeks' gestation in prediction of pre-eclampsia: validation and update of previous model.

To validate and extend a model incorporating maternal ophthalmic artery Doppler at 35-37 weeks' gestation in the prediction of subsequent development of pre-eclampsia (PE). This was a prospective validation study of screening for PE (defined according to the 2019 American College of Obstetricians and Gynecologists criteria) by maternal ophthalmic artery peak systolic velocity (PSV) ratio in 6746 singleton pregnancies undergoing routine care at 35 + 0 to 36 + 6 weeks' gestation (validation dataset). Additionally, the data from the validation dataset were combined with those of 2287 pregnancies that were previously used for development of the model (training dataset), and the combined data were used to update the original model parameters. The competing-risks model was used to estimate the individual patient-specific risk of delivery with PE at any time and within 3 weeks from assessment by a combination of maternal demographic characteristics and medical history with PSV ratio alone and in combination with the established PE biomarkers of mean arterial pressure (MAP), uterine artery pulsatility index (UtA-PI), serum placental growth factor (PlGF) and serum soluble fms-like tyrosine kinase-1 (sFlt-1). We evaluated the predictive performance of the model by examining, first, the ability to discriminate between the PE and non-PE groups using the area under the receiver-operating-characteristics curve and the detection rate (DR) at fixed screen-positive (SPR) and false-positive rates of 10% and, second, calibration by measuring the calibration slope and calibration-in-the-large. McNemar's test was used to compare the performance of screening by a biophysical test (maternal factors, MAP, UtA-PI and PSV ratio) vs a biochemical test (maternal factors, PlGF and sFlt-1), low PlGF concentration (< 10th percentile) or high sFlt-1/PlGF concentration ratio (> 90th percentile). In the validation dataset, the performance of screening by maternal factors and PSV ratio for delivery with PE within 3 weeks and at any time after assessment was consistent with that in the training dataset, and there was good agreement between the predicted and observed incidence of PE. In the combined data from the training and validation datasets, good prediction for PE was achieved in screening by a combination of maternal factors, MAP, UtA-PI, PlGF, sFlt-1 and PSV ratio, with a DR, at a 10% SPR, of 85.0% (95% CI, 76.5-91.4%) for delivery with PE within 3 weeks and 65.7% (95% CI, 59.2-71.7%) for delivery with PE at any time after assessment. The performance of a biophysical test was superior to that of screening by low PlGF concentration or high sFlt-1/PlGF concentration ratio but not significantly different from the performance of a biochemical test combining maternal factors with PlGF and sFlt-1 for both PE within 3 weeks and PE at any time after assessment. Maternal ophthalmic artery PSV ratio at 35-37 weeks' gestation in combination with other biomarkers provides effective prediction of subsequent development of PE. © 2023 International Society of Ultrasound in Obstetrics and Gynecology.

Discover, Explain, Improve: An Automatic Slice Detection Benchmark for Natural Language Processing

Abstract Pretrained natural language processing (NLP) models have achieved high overall performance, but they still make systematic errors. Instead of manual error analysis, research on slice detection models (SDMs), which automatically identify underperforming groups of datapoints, has caught escalated attention in Computer Vision for both understanding model behaviors and providing insights for future model training and designing. However, little research on SDMs and quantitative evaluation of their effectiveness have been conducted on NLP tasks. Our paper fills the gap by proposing a benchmark named “Discover, Explain, Improve (DEIm)” for classification NLP tasks along with a new SDM Edisa. Edisa discovers coherent and underperforming groups of datapoints; DEIm then unites them under human-understandable concepts and provides comprehensive evaluation tasks and corresponding quantitative metrics. The evaluation in DEIm shows that Edisa can accurately select error-prone datapoints with informative semantic features that summarize error patterns. Detecting difficult datapoints directly boosts model performance without tuning any original model parameters, showing that discovered slices are actionable for users.1

Experimental design and re-parameterization of the Neue-Kuss model for accurate and precise prediction of isocratic retention factors from gradient measurements in reversed phase liquid chromatography

The present work describes a re-parameterization of the Neue Kuss (NK) model for describing retention in liquid chromatography, and this re-parameterized model is used to fit a large set of isocratic retention measurements with improved convergence properties relative to the original parameterization of the model. Next, an experimental design for retention measurements using mobile phase gradient elution conditions is proposed for the purpose of obtaining accurate and precise NK parameters. Simulated retention data for mobile phase gradient elution conditions with two different levels of noise, as well as an essentially zero noise level were fit with the re-parameterized model. The results showed that the re-parameterized fits yielded average (absolute value) prediction errors for the parameters at the highest noise level of 7.2 % for S1,ref, 18 % for S2,ref and 6.2 % for kref (the re-parameterized NK model parameters). These errors were significantly smaller than those for the original parameterization of the NK model, where the errors were 23 % for S1, 25 % for S2 and 160 % for kw (the original NK model parameters). Furthermore, isocratic retention factors predicted using these model parameters were found to have an average magnitude of error of 0.51 % for the re-parameterized model, as opposed to 6800 % for the model with the original parameterization. A further test of this approach was carried out for independent experimental measurements for five solutes on a C18 column. The average magnitude of error of the isocratic retention factors predicted from parameters obtained from fits of gradient data was 1.6 %, provided that the range of organic solvent compositions that the solute sampled in the mobile phase gradient experiments was consistent with the isocratic experiments. These results indicate that the re-parameterization of the NK model allows for significant improvements in the fitting process, and that the proposed experimental design allows for NK parameters to be extracted from mobile phase gradient experiments, with prediction accuracies of isocratic retention factors on the order of 1-2 %.

FedSIGN: A sign-based federated learning framework with privacy and robustness guarantees

Federated learning enables clients to train a global model jointly without sharing their private local datasets. Despite its benefits, due to the untrustworthiness of clients and the server, traditional federated learning faces the risk of privacy leakage and poisoning attacks. Privacy-preserving methods change the original model parameters whereas robust aggregation algorithms required accurate parameters. To solve such a dilemma, we propose a new framework named FedSIGN. On the one hand, it utilizes the sign of local model update to update the global model to protect privacy and improve efficiency. On the other hand, focus on the Sybil-based poisoning attack: malicious client controlled by a single adversary who directs those clients to launch a poisoning attack, we design a Poisoning Attack Detector to identify malicious clients based on the similarity between sign vectors. Experimental results show that FedSIGN resists privacy and poisoning attacks while maintaining better global model performance. Especially, FedSIGN is not affected by the number of malicious clients and is effective in both the IID and non-IID scenarios.

Identification of Two FitzHugh-Nagumo Neuron Models Based on the Speed-Gradient and Filtering

The paper is devoted to the problem of parameter identification of two FitzHugh-Nagumo neuron models. The FitzHugh-Nagumo model is a simplification of the Hodgkin-Huxley model and it is very valuable for using on practice thanks to its simplicity. However, within an experiment only one variable of the FitzHugh-Nagumo model, the membrane potential, is measured, while another variable of cumulative effects of all slow ion currents responsible for restoring the resting potential of the membranes and both variables’ derivatives cannot be measured. This circumstance brings additional difficulties to the parameters estimation problem and, therefore, this case needs special attention. Firstly, the model was transformed to more simple form without unmeasured variables. Variables obtained from applying second-order real filter-differentiator were used instead of unmeasured derivatives in model’s equations. As a result, a linear equation was gotten and for this equation the identification goal, which guarantees correct parameters’ adjustment, was formulated and an adaptive system, parameters of which are estimations of original system’s parameters and an output of which estimates the output of the linear equation, was constructed. Then, the integral objective function was defined and the algorithm for the original model parameters identification was designed with the speed-gradient method. The results of computer simulation in the Simulink environment are presented. These results demonstrate that estimates of the model’s state and parameters converge to their true values rather fast. Unlike existing solutions of the FitzHugh-Nagumo identification problem, we propose a much easier deterministic algorithm. Moreover, the parameters are estimated for a system collected from two FitzHugh-Nagumo models, which opens perspectives for using the proposed method in modeling neuron population activity.

Knowledge Distillation for Optical Flow-Based Video Super-resolution

Recently, deep learning-based super-resolution (SR) models have been used to improve SR performance by equipping preprocessing networks with baseline SR networks. In particular, in video SR, which creates a high-resolution (HR) image with multiple frames, optical flow extraction is accompanied by a preprocessing process. These preprocessing networks work effectively in terms of quality, but at the cost of increased network parameters, which increase the computational complexity and memory consumption for SR tasks with restricted resources. One well-known approach is the knowledge distillation (KD) method, which can transfer the original model’s knowledge to a lightweight model with less performance degradation. Moreover, KD may improve SR quality with reduced model parameters. In this study, we propose an effective KD method that can effectively reduce the original SR model parameters and even improve network performance. The experimental results demonstrated that our method achieved a better PSNR than the original state-of-the-art SR network despite having fewer parameters.

Enhanced models with time‐series coefficients used for electric arc furnace

AbstractElectric arc furnace (EAF) is one of the most disturbing loads in the power system as it is highly time‐varying and nonlinear. In this paper, new models based on the instantaneous EAF voltage and current, recorded from EAFs in Mobarakeh steel company in Isfahan/Iran are proposed. The focus of this paper is on developing efficient models based on the Schwarz model where three modified types are proposed. The original Schwarz model has four constant parameters. Unlike the original Schwarz model which uses invariant parameters, the time‐varying nature of the arc's parameters is considered in the proposed models. The first modified model has two constant model orders and two time‐varying coefficients. Additionally, to further enhance the model's accuracy, in the second modified model, six parameters include three constant model orders, and three time‐varying coefficients exist. The third model uses all the original model's parameters as time‐series parameters. The parameters of the time‐varying model are derived using autoregressive moving average (ARMA) models. The accuracy of the proposed model is evaluated by calculating the active power, harmonics, instantaneous flicker, and short‐term flicker. Finally, three MATLAB codes are provided to generate the model parameters time series.

Risk of Cytokine Release Syndrome with Glofitamab Is Predicted By an Updated Model with a Potential Clinical Application

Background: Cytokine release syndrome (CRS) is a potentially life-threatening toxicity caused by immune activation. CRS can be triggered non-specifically by T-cell engaging therapies. Komanduri et al. (ASH 2021) outlined a model based on eight baseline factors (CRS-risk score; CRS-RS) which allowed accurate classification of risk for Grade ≥2 CRS upon treatment with glofitamab. Glofitamab is a CD20xCD3 T-cell-engaging bispecific monoclonal antibody with a novel 2:1 (CD20:CD3) configuration that has shown promising efficacy with a manageable safety profile in NP30179 (NCT03075696), an ongoing Phase I/II dose-finding study of glofitamab in pts with relapsed or refractory (R/R) non-Hodgkin lymphoma (NHL; Dickinson et al. ASCO 2022). While CRS is observed with glofitamab, cases typically occur with Grade 1/2 severity (per ASTCT grading; Lee et al. Biol Blood Marrow Transplant 2019) mostly in the first cycle (C). Here, we introduce a streamlined model based on five baseline factors (CRS-RS.5p) derived from the original CRS-RS, an ongoing non-interventional study (GO43921) to assess the performance of CRS-RS.5p, and the potential application of the risk score in a clinical setting. Methods: In both models, using a weighted sum of baseline clinical, laboratory and imaging parameters, pts are assigned a risk score as a continuous measure for developing CRS after the first dose of glofitamab. Using a defined cut-off, pts are classified as low or high risk of CRS. The original CRS-RS model parameters were: age >64 years at study entry, lactate dehydrogenase pre-anti-CD20 treatment >280 U/L, white blood cell count pre-anti-CD20 treatment >4.5x109 cells/L, Ann Arbor Stage III/IV at study entry, sum of the product of the perpendicular diameters at study entry ≥3000mm2, cardiac comorbidity, bone marrow infiltration, and atypical or leukemic cells in peripheral blood at study entry. The simplified CRS-RS.5p concentrates on the first five parameters utilizing the same weighting scheme as in CRS-RS. The primary objective of the GO43921 study is to determine if a five-baseline-parameter model can predict Grade ≥2 CRS events after the first dose of glofitamab. Seven studies were identified in which pts with aggressive NHL (first line and R/R, excluding mantle cell lymphoma) were treated with glofitamab step-up dosing as either monotherapy or part of combination therapy. In most studies, pts are pretreated with 1000mg obinutuzumab, with intravenous glofitamab being administered on Days (D) 1 (2.5mg) and 8 (10mg) of C1, then at the target dose (30mg) from C2D1 onwards. Several combination studies started glofitamab step-up dosing after multiple cycles of R-CHOP (glofitamab D8 [2.5mg], D15 [10mg], target dose D8 [30mg] in the following cycle). Prospective data has been collected since December 2021, and retrospective data from included studies were analyzed to increase the sample of Grade ≥2 CRS events. Here, CRS-RS.5p is compared with CRS-RS for the NP30179 dataset, and CRS-RS.5p performance in the ongoing GO43921 study is analyzed. Results: In NP30179, data acquired since the introduction of CRS-RS (n=44) showed that CRS-RS and CRS-RS.5p identified as high risk the same pts who experienced Grade ≥2 events (Figure A; four Grade 2, one Grade 3). In the whole NP30179 validation cohort, both CRS-RS and CRS-RS.5p achieved negative predictive values of 0.97 (n=148, Standard error=0.02). For GO43921, at the clinical cut-off date (June 6, 2022), the CRS-RS.5p model had predicted all 11 of the Grade ≥2 CRS events in the second-line or later setting for pts with aggressive NHL (n=123, with data collected retrospectively for 82 pts; Figure B). The association of the CRS-RS.5p score with the ASTCT CRS Grade after the first dose of glofitamab was similar across all available data sets from NP30179 and GO43921 (logistic regression; p<0.001 for NP30179, p=0.004 for GO43921). No influence of steroid premedication class on CRS-RS performance was observed. A low-risk group (CRS-RS.5p <4.0) comprised 48% of pts in NP30179 and 46% in GO43921. Conclusions: Data from the CRS-RS.5p model showed identical performance with the previous 8-parameter CRS-RS model in predicting which pts with aggressive NHL are most at risk of Grade ≥2 CRS after the first dose of glofitamab. With reduced complexity, the CRS-RS.5p has a potential application as a practical, clinical decision tool to guide hospitalization for pts at high risk of Grade ≥2 CRS. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

The in-vivo dynamics of Plasmodium falciparum HRP2: implications for the use of rapid diagnostic tests in malaria elimination

BackgroundRapid diagnostic tests (RDTs) that rely on the detection of Plasmodium falciparum histidine-rich protein 2 (PfHRP2) have become key tools for diagnosing P. falciparum infection. The utility of RDTs can be limited by PfHRP2 persistence, however it can be a potential benefit in low transmission settings where detection of persistent PfHRP2 using newer ultra-sensitive PfHRP2 based RDTs can serve as a surveillance tool to identify recent exposure. Better understanding of the dynamics of PfHRP2 over the course of a malaria infection can inform optimal use of RDTs.MethodsA previously published mathematical model was refined to mimic the production and decay of PfHRP2 during a malaria infection. Data from 15 individuals from volunteer infection studies were used to update the original model and estimate key model parameters. The refined model was applied to a cohort of patients from Namibia who received treatment for clinical malaria infection for whom longitudinal PfHRP2 concentrations were measured.ResultsThe refinement of the PfHRP2 dynamic model indicated that in malaria naïve hosts, P. falciparum parasites of the 3D7 strain produce 33.6 × 10−15 g (95% CI 25.0–42.1 × 10−15 g) of PfHRP2 in vivo per parasite replication cycle, with an elimination half-life of 1.67 days (95% CI 1.11–3.40 days). The refined model included these updated parameters and incorporated individualized body fluid volume calculations, which improved predictive accuracy when compared to the original model. The performance of the model in predicting clearance of PfHRP2 post treatment in clinical samples from six adults with P. falciparum infection in Namibia improved when using a longer elimination half-life of 4.5 days, with 14% to 67% of observations for each individual within the predicted range.ConclusionsThe updated mathematical model can predict the growth and clearance of PfHRP2 during the production and decay of a mono-infection with P. falciparum, increasing the understanding of PfHRP2 antigen dynamics. This model can guide the optimal use of PfHRP2-based RDTs for reliable diagnosis of P. falciparum infection and re-infection in endemic settings, but also for malaria surveillance and elimination programmes in low transmission areas.

Mathematical modeling of respiratory viral infection and applications to SARS-CoV-2 progression.

Viral infection in cell culture and tissue is modeled with delay reaction‐diffusion equations. It is shown that progression of viral infection can be characterized by the viral replication number, time‐dependent viral load, and the speed of infection spreading. These three characteristics are determined through the original model parameters including the rates of cell infection and of virus production in the infected cells. The clinical manifestations of viral infection, depending on tissue damage, correlate with the speed of infection spreading, while the infectivity of a respiratory infection depends on the viral load in the upper respiratory tract. Parameter determination from the experiments on Delta and Omicron variants allows the estimation of the infection spreading speed and viral load. Different variants of the SARS‐CoV‐2 infection are compared confirming that Omicron is more infectious and has less severe symptoms than Delta variant. Within the same variant, spreading speed (symptoms) correlates with viral load allowing prognosis of disease progression.

A Combined Detection Algorithm for Personal Protective Equipment Based on Lightweight YOLOv4 Model

Improper wearing of personal protective equipment may lead to safety incidents; this paper proposes a combined detection algorithm for personal protective equipment based on the lightweight YOLOv4 model for mobile terminals. To ensure high detection accuracy, a channel and layer pruning method (CLSlim) to lightweight algorithm is used to reduce computing power consumption and improve the detection speed on the basis of the YOLOv4 network. This method applies L1 regularization and gradient sparse training on the scaling factor of the BN layer in the convolutional module: global pruning threshold and local safety threshold are used to eliminate redundant channels, the layer pruning threshold is used to prune the structure of the shortcuts in the Cross Stage Partial (CSP) module for inference speed improvement, and finally, a lightweight network model is obtained. The experiment improves the YOLOv4 and YOLOv4-Tiny models for CLSlim lightweight separately in GTX2080ti environment. Results show that (1) CLSlim-YOLOv4 compresses the YOLOv4 model parameters by 98.2% and increases the inference speed by 1.8 times with mAP loss of only 2.1% and (2) CLSlim-YOLOv4-Tiny compresses the original model parameters by 74.3% and increases the inference speed by 1.1 times with mAP increase of 0.8%, which certificates that this improved lightweight algorithm serves better for the real-time ability and accuracy of combined detection on PPE with mobile terminals.

FedCVT: Semi-supervised Vertical Federated Learning with Cross-view Training

Federated learning allows multiple parties to build machine learning models collaboratively without exposing data. In particular, vertical federated learning (VFL) enables participating parties to build a joint machine learning model based upon distributed features of aligned samples. However, VFL requires all parties to share a sufficient amount of aligned samples. In reality, the set of aligned samples may be small, leaving the majority of the non-aligned data unused. In this article, we propose Federated Cross-view Training (FedCVT), a semi-supervised learning approach that improves the performance of the VFL model with limited aligned samples. More specifically, FedCVT estimates representations for missing features, predicts pseudo-labels for unlabeled samples to expand the training set, and trains three classifiers jointly based upon different views of the expanded training set to improve the VFL model’s performance. FedCVT does not require parties to share their original data and model parameters, thus preserving data privacy. We conduct experiments on NUS-WIDE, Vehicle, and CIFAR10 datasets. The experimental results demonstrate that FedCVT significantly outperforms vanilla VFL that only utilizes aligned samples. Finally, we perform ablation studies to investigate the contribution of each component of FedCVT to the performance of FedCVT.

Precision Detection of Dense Plums in Orchards Using the Improved YOLOv4 Model.

The precision detection of dense small targets in orchards is critical for the visual perception of agricultural picking robots. At present, the visual detection algorithms for plums still have a poor recognition effect due to the characteristics of small plum shapes and dense growth. Thus, this paper proposed a lightweight model based on the improved You Only Look Once version 4 (YOLOv4) to detect dense plums in orchards. First, we employed a data augmentation method based on category balance to alleviate the imbalance in the number of plums of different maturity levels and insufficient data quantity. Second, we abandoned Center and Scale Prediction Darknet53 (CSPDarknet53) and chose a lighter MobilenetV3 on selecting backbone feature extraction networks. In the feature fusion stage, we used depthwise separable convolution (DSC) instead of standard convolution to achieve the purpose of reducing model parameters. To solve the insufficient feature extraction problem of dense targets, this model achieved fine-grained detection by introducing a 152 × 152 feature layer. The Focal loss and complete intersection over union (CIOU) loss were joined to balance the contribution of hard-to-classify and easy-to-classify samples to the total loss. Then, the improved model was trained through transfer learning at different stages. Finally, several groups of detection experiments were designed to evaluate the performance of the improved model. The results showed that the improved YOLOv4 model had the best mean average precision (mAP) performance than YOLOv4, YOLOv4-tiny, and MobileNet-Single Shot Multibox Detector (MobileNet-SSD). Compared with some results from the YOLOv4 model, the model size of the improved model is compressed by 77.85%, the parameters are only 17.92% of the original model parameters, and the detection speed is accelerated by 112%. In addition, the influence of the automatic data balance algorithm on the accuracy of the model and the detection effect of the improved model under different illumination angles, different intensity levels, and different types of occlusions were discussed in this paper. It is indicated that the improved detection model has strong robustness and high accuracy under the real natural environment, which can provide data reference for the subsequent orchard yield estimation and engineering applications of robot picking work.

Nonlinear Vibration Characteristics Identification of Drum - Rocker System of Shearer Based on Artificial Intelligence

As the core component of coal interception and coal loading, the spiral drum of coal shearer directly affects the working efficiency of coal shearer and the quality of coal. In the past research on shearing force of shearer drum, there are many problems and deficiencies in research methods and theoretical analysis. Therefore, based on the actual production conditions and the original drum model parameters, a simulation model of shearer drum cutting coal wall is established to simulate the dynamic process of coal cutting, and the research on the nonlinear vibration performance of the shearer rocker system affects its stability. The key issue of the performance and service life is therefore important to study the nonlinear vibration characteristics of the rocker system. This paper mainly studies the nonlinear vibration characteristics of the rocker system. In this paper, based on ANSYS software, the model analysis of the rocker arm shell and the traction part shell is carried out. Harmonic response analysis was carried out on the housing of the traction part, and its response under sinusoidal excitation of different frequencies was obtained. The lateral vibration acceleration and swing angle acceleration of the shearer body are 0.0098 and 0.0074 , respectively. The larger the average cutting impedance of coal rock, the greater the lateral acceleration of each part of the shearer under the oblique cutting condition, the greater the fluctuation, and the direction changes with time. The average cutting impedance of coal rock has a great influence on the front drum and front rocker of the shearer. The experimental results show that the study of nonlinear vibration characteristics provides a reference for the optimization of the shearer's structural strength and the improvement of its resistance to shock and vibration.

Real-Time Identification of Rice Weeds by UAV Low-Altitude Remote Sensing Based on Improved Semantic Segmentation Model

Real-time analysis of UAV low-altitude remote sensing images at airborne terminals facilitates the timely monitoring of weeds in the farmland. Aiming at the real-time identification of rice weeds by UAV low-altitude remote sensing, two improved identification models, MobileNetV2-UNet and FFB-BiSeNetV2, were proposed based on the semantic segmentation models U-Net and BiSeNetV2, respectively. The MobileNetV2-UNet model focuses on reducing the amount of calculation of the original model parameters, and the FFB-BiSeNetV2 model focuses on improving the segmentation accuracy of the original model. In this study, we first tested and compared the segmentation accuracy and operating efficiency of the models before and after the improvement on the computer platform, and then transplanted the improved models to the embedded hardware platform Jetson AGX Xavier, and used TensorRT to optimize the model structure to improve the inference speed. Finally, the real-time segmentation effect of the two improved models on rice weeds was further verified through the collected low-altitude remote sensing video data. The results show that on the computer platform, the MobileNetV2-UNet model reduced the amount of network parameters, model size, and floating point calculations by 89.12%, 86.16%, and 92.6%, and the inference speed also increased by 2.77 times, when compared with the U-Net model. The FFB-BiSeNetV2 model improved the segmentation accuracy compared with the BiSeNetV2 model and achieved the highest pixel accuracy and mean Intersection over Union ratio of 93.09% and 80.28%. On the embedded hardware platform, the optimized MobileNetV2-UNet model and FFB-BiSeNetV2 model inferred 45.05 FPS and 40.16 FPS for a single image under the weight accuracy of FP16, respectively, both meeting the performance requirements of real-time identification. The two methods proposed in this study realize the real-time identification of rice weeds under low-altitude remote sensing by UAV, which provide a reference for the subsequent integrated operation of plant protection drones in real-time rice weed identification and precision spraying.

Numerical Stability of Modified Lorentz FDTD Unified From Various Dispersion Models.

The finite-difference time-domain (FDTD) method has been widely used to analyze electromagnetic wave propagation in complex dispersive media. Until now, there are many reported dispersion models including Debye, Drude, Lorentz, complex-conjugate pole-residue (CCPR), quadratic complex rational function (QCRF), and modified Lorentz (mLor). The mLor FDTD is promising since the mLor dispersion model can simply unify other dispersion models. To fully utilize the unified mLor FDTD method, it is of great importance to investigate its numerical stability in the aspects of the original dispersion model parameters. In this work, the numerical stability of the mLor FDTD formulation unified from the aforementioned dispersion models is comprehensively studied. It is found out that the numerical stability conditions of the original model-based FDTD method are equivalent to its unified mLor FDTD counterparts. However, when unifying the mLor FDTD formulation for the QCRF model, a proper Courant number should be used. Otherwise, its unified mLor FDTD simulation may suffer from numerical instability, different from other dispersion models. Numerical examples are performed to validate our investigations.

Scale electro-thermal model of a lithium-ion battery for time-accelerated experiments in a hardware in the loop process

A scale electro-thermal model has been developed for LiFePO4/graphite lithium-ion battery. Such model is appropriate in order to develop a physical emulation of a battery in the context of a hardware in the loop process, especially for testing energy management strategies of microgrids under the same conditions (solar irradiation for PV arrays, wind speed for wind turbine, state of health for storage device) and potentially by compressing testing time. Classically, physical emulation allows achieving laboratory size-scaled analysis but one major originality of the proposed approach deals with the “time-compressed experimental analysis”. The electro-thermal model is based on the extended modified shepherd model coupled with a 1D thermal model. The model parameters are estimated through a sequential characterization approach from several input profiles such as the hybrid pulse power characterization protocol and the open circuit voltage measurement. Both the dimensional analysis and the Vaschy-Buckingham theorem are used to obtain the scaling factors (voltage, current, “but also” time) which are applied on the original model parameters. The accuracy of the scale electro-thermal model is validated on a robustness analysis (the battery current profile is based on a specific energy management strategy for a typical microgrid application) with voltage, current and time scaling. The simulation results presented in the paper show that the reduction of the time horizon of experimental tests for HIL process is possible by means of an appropriate dimensional analysis (scaling) on the model parameters.

Entrainment of Weakly Coupled Canonical Oscillators with Applications in Gradient Frequency Neural Networks Using Approximating Analytical Methods

Solving phase equations for systems with high degrees of nonlinearities is cumbersome. However, in the case of two coupled canonical oscillators, that is, a reduced model of translated Wilson–Cowan neuronal dynamics, under slowly varying amplitude and rotating wave approximations, we suggested a convenient way to find their average relative phase evolution. This approach enabled us to find an explicit solution for the average relative phase of the two coupled canonical oscillators based on the original neuronal model parameters, and importantly, to find their phase-locking constraint. This methodology is straightforward to implement in any Wilson–Cowan-type coupled oscillators with applications in gradient frequency neural networks (GFNNs).