Despite their high prediction accuracy, deep learning-based soft sensor (DLSS) models face challenges related to adversarial robustness against malicious adversarial attacks, which hinder their ...widespread deployment and safe application. Although adversarial training is the primary method for enhancing adversarial robustness, existing adversarial-training-based defense methods often struggle with accurately estimating transfer gradients and avoiding adversarial robust overfitting. To address these issues, we propose a novel adversarial training approach, namely domain-adaptive adversarial training (DAAT). DAAT comprises two stages: historical gradient-based adversarial attack (HGAA) and domain-adaptive training. In the first stage, HGAA incorporates historical gradient information into the iterative process of generating adversarial samples. It considers gradient similarity between iterative steps to stabilize the updating direction, resulting in improved transfer gradient estimation and stronger adversarial samples. In the second stage, a soft sensor domain-adaptive training model is developed to learn common features from adversarial and original samples through domain-adaptive training, thereby avoiding excessive leaning toward either side and enhancing the adversarial robustness of DLSS without robust overfitting. To demonstrate the effectiveness of DAAT, a DLSS model for crystal quality variables in silicon single-crystal growth manufacturing processes is used as a case study. Through DAAT, the DLSS achieves a balance between defense against adversarial samples and prediction accuracy on normal samples to some extent, offering an effective approach for enhancing the adversarial robustness of DLSS.
In recent years, strain sensors have penetrated various fields. The capability of sensors to convert physical signals into electrical signals is of great importance in healthcare. However, it is ...still challenging to obtain sensors with high sensitivity, large operating range and low cost. In this paper, a stretchable strain sensor made of a double-layer conductive network, including a biomimetic multilayer graphene-Ecoflex (MLG-Ecoflex) substrate and a multilayer graphene-carbon nanotube (MLG-CNT) composite up-layer was developed. The combined action of the two layers led to an excellent performance with an operating range of up to 580% as well as a high sensitivity (gauge factor (GF
) of 1517.94). In addition, a pressure sensor was further designed using the bionic vein-like structure with a multi-layer stacking of MLG-Ecoflex/MLG-CNT/MLG-Ecoflex to obtain a relatively high deformation along the direction of thickness. The device presented a high sensing performance (up to a sensitivity of 0.344 kPa
) capable of monitoring small movements of the human body such as vocalizations and gestures. The good performance of the sensors together with a simple fabrication procedure (flip-molding) make it of potential use for some applications, for example human health monitoring and other areas of human interaction.
Recently, erdafitinib (Balversa), the first targeted therapy drug for genetic alteration, was approved to metastatic urothelial carcinoma. Cancer genomics research has been greatly encouraged. ...Currently, a large number of gene regulatory networks between different states have been constructed, which can reveal the difference states of genes. However, they have not been applied to the subtypes of Muscle-invasive bladder cancer (MIBC). In this paper, we propose a method that construct gene regulatory networks under different molecular subtypes of MIBC, and analyse the regulatory differences between different molecular subtypes. Through differential expression analysis and the differential network analysis of the top 100 differential genes in the network, we find that SERPINI1, NOTUM, FGFR1 and other genes have significant differences in expression and regulatory relationship between MIBC subtypes. Furthermore, pathway enrichment analysis and differential network analysis demonstrate that Neuroactive ligand-receptor interaction and Cytokine-cytokine receptor interaction are significantly enriched pathways, and the genes contained in them are significant diversity in the subtypes of bladder cancer.
Strain sensors that can rapidly and efficiently detect strain distribution and magnitude are crucial for structural health monitoring and human–computer interactions. However, traditional electrical ...and optical strain sensors make access to structural health information challenging because data conversion is required, and they have intricate, delicate designs. Drawing inspiration from the moisture-responsive coloration of beetle wing sheaths, we propose using Ecoflex as a flexible substrate. This substrate is coated with a Fabry–Perot (F–P) optical structure, comprising a “reflective layer/stretchable interference cavity/reflective layer”, creating a dynamic color-changing visual strain sensor. Upon the application of external stress, the flexible interference chamber of the sensor stretches and contracts, prompting a blue-shift in the structural reflection curve and displaying varying colors that correlate with the applied strain. The innovative flexible sensor can be attached to complex-shaped components, enabling the visual detection of structural integrity. This biomimetic visual strain sensor holds significant promise for real-time structural health monitoring applications.
Deep learning is thought of as a promising mean to identify maize diseases. However, the drawback of deep learning is the huge sample data and low accuracy. In this paper, we proposed a multi-scale ...convolutional global pooling neural network to improve the accuracy of maize diseases identification. Firstly, on the basis of the AlexNet model, a convolutional layer and new Inception module are added to enhance the ability of AlexNet features extraction. Then, in order to avoid the over-fitting problem caused by too many parameters, we use the global pooling layer to replace the original fully-connected layer. Besides, we also adopt the transfer learning method to solve the over-fitting problem caused by insufficient sample data. The improved model can reduce over-fitting and epochs to enhance the performance of maize diseases recognition. From the considerable experimental results, we can conclude that the proposed model has better performance compared with convolutional neural network models VGGNet-16, DenseNet, ResNet-50 and AlexNet in recognition accuracy.
Effectively enhancing oil recovery can be achieved by reducing the viscosity of crude oil. Therefore, this paper investigated the viscosity reduction behavior of carbon nanotube viscosity reducers ...with different molecular structures at the oil–water interface, aiming to guide the synthesis of efficient viscosity reducers based on molecular structure. This study selected carbon nanotubes with different functional groups (NH2-CNT, OH-CNT, and COOH-CNT) for research, and carbon nanotubes with varying carbon chain lengths were synthesized. These were then combined with Tween 80 to form a nanofluid. Scanning electron microscopy analysis revealed an increased dispersibility of carbon nanotubes after introducing carbon chains. Contact angle experiments demonstrated that -COOH exhibited the best hydrophilic effect. The experiments of zeta potential, conductivity, viscosity reduction, and interfacial tension showed that, under the same carbon chain length, the conductivity and viscosity reduction rate sequence for different functional groups was -NH2 < -OH < -COOH. The dispersing and stabilizing ability and interfacial tension reduction sequence for different functional groups was -COOH < -OH < -NH2. With increasing carbon chain length, conductivity and interfacial tension decreased, and the viscosity reduction rate and the dispersing and stabilizing ability increased. Molecular dynamics simulations revealed that, under the same carbon chain length, the diffusion coefficient sequence for different functional groups was -NH2 < -OH < -COOH. The diffusion coefficient gradually decreased as the carbon chain length increased, resulting in better adsorption at the oil–water interface. This study holds significant importance in guiding viscosity reduction in heavy oil to enhance oil recovery.
A new type of chitosan-modified hyperbranched polymer (named HPDACS) was synthesized through the free-radical polymerization of surface-modified chitosan with acrylic acid (AA) and acrylamide (AM) to ...achieve an enhanced oil recovery. The optimal polymerization conditions of HPDACS were explored and its structure was characterized by Fourier-transform infrared spectroscopy, hydrogen nuclear magnetic resonance, and environmental scanning electron microscopy. The solution properties of HPDACS in ultrapure water and simulated brine were deeply studied and then compared with those of partially hydrolyzed polyacrylamide (HPAM) and a dendritic polymer named HPDA. The experimental results showed that HPDACS has a good thickening ability, temperature resistance, and salt resistance. Its viscosity retention rate exceeded 79.49% after 90 days of aging, thus meeting the performance requirements of polymer flooding. After mechanical shearing, the viscosity retention rates of HPDACS in ultrapure water and simulated brine were higher than those of HPAM and HPDA, indicating its excellent shear resistance and good viscoelasticity. Following a 95% water cut after preliminary water flooding, 0.3 pore volume (PV) and 1500 mg/L HPDACS solution flooding and extended water flooding could further increase the oil recovery by 19.20%, which was higher than that by HPAM at 10.65% and HPDA at 13.72%. This finding indicates that HPDACS has great potential for oil displacement.
The selective polarizers play an important role in silicon-based integrated circuits. The previous polarizers based on silicon waveguides have the defects of large scale and low extinction ratio. In ...this work, TM- and TE-pass polarizers only 10 μm long were developed based on phase-change material of Sc0.2Sb2Te3 (SST) hybrid silicon waveguide, where several SST bars with a varied distance was designed. Because of the excellent characteristics of the refractive index of the material, ultra-high extinction ratios (ERs) were achieved. A 3-D finite element simulation was carried out to optimize the structure of the polarizers, and the distribution of light field, as well as the transmission behavior of TE and TM modes in the two polarizers, was further demonstrated in detail. When the SST is crystalline, the unwanted mode can be attenuated, while the wanted mode can pass through with low loss. Compared with the GST-based polarizers, the proposed ones achieved high extinction ratios of ~43.12 dB (TM-pass one) and ~44.21 dB (TE-pass one), respectively; at the same time, ILs for the wanted modes could be negligible. The design of high-performance polarizers paves a new way for applications of all-optical integrated circuits.
A rapid screening method for 84 pesticide residues in dendrobium perfringens parent material with different polarities was developed using a Sin-QuEChERS Nano clean-up column combined with gas ...chromatography-tandem mass spectrometry (GC-MS/MS). The differences in extraction efficiency of the targets were compared with different extraction solvents (acetonitrile containing 1% acetic acid, acetone) and methods (immersion with or without water). The purification effect and extraction recoveries of Sin-QuEChERS Nano method and classical dispersive solid-phase extraction (dSPE), solid-phase extraction (SPE) and QuEChERS were systematically compared using
samples. The differences in matrix effects between the Sin-QuEChERS Nano method, which was more effective in purification, and the dSPE method were also analyzed. The purification effects of three commercially available Sin-QuEChERS Nano purification columns (simple matrix purification column, complex matrix purification column and herbal purification column) were
To solve the problem of the low efficiency of traditional lettuce freshness classification methods and sample damage, we proposed an automatic lettuce freshness classification method based on ...improved deep residuals convolutional neural network (Im-ResNet). We built an image acquisition system to obtain the freshness classification dataset of lettuce leaves. For improving the classification accuracy, we developed an image acquisition system for curating the freshness of lettuce leaves. Then, we proposed a novel method that was derived from the existing ResNet-50 (which uses ReLU activation function) known as Improved Residual Networks (Im-ResNet): the new method factored extra convolutional layer, pooling layer, fully-connected layers, and a random ReLU (RReLU) activation function. We also performed the corresponding experiments using the Im-ResNet network compared with four network architectures (AlexNet, GoogleNet, VGG16 and ResNet50). The experimental results showed that the proposed network had more significant advantages in the recognition accuracy and loss value of lettuce freshness compared with the traditional deep networks. The recognition accuracy of the validation set of the proposed model can reach to 95.60%. Different from the physical and chemical methods, our scheme can automatically and non-destructively classify the freshness of lettuce.