查看更多>>摘要：? 2022 Elsevier LtdMany key quality variables are difficult to measure in complex industrial processes for various reasons, such as working conditions or economic costs, leading to inefficient production monitoring. In recent years, soft sensors with outstanding performance in variable estimation have been widely used. However, quality samples collected from industrial sites are often limited, which results in incomplete datasets that cannot meet the training requirements of soft sensors and poor performance in model learning and prediction. In this paper, a new virtual sample generation method DA-GAN based on generative adversarial network (GAN) is proposed to provide extra training samples for soft sensors. Adversarial net-and adversarial sample-based dual adversarial learning is implemented to reduce the adversarial noise in the discriminator gradient, which can improve the convergence speed and learning stability of the generator and obtain virtual samples with higher similarity to the real data. Furthermore, a sample screening method based on asymmetric acceptable domain range expansion is introduced to choose high-quality virtual samples. Experimental results of two industrial case studies show that the virtual samples provided by DA-GAN are closer to real samples than several other widely used generation methods. The performance of the prediction model trained with the dataset added by the virtual samples yielded from DA-GAN can be better improved.
查看更多>>摘要：? 2022 Elsevier LtdThe opening area error of the gas valve is mainly caused by machining, assembling errors and transmission clearances. Accurate measurement of the opening area is important for the production quality inspection of gas valves. This paper proposes a novel solution with a specific measurement system. Due to the reflective effect of the valve surface, the valve opening's saliency is enhanced by the combination of top lighting and camera exposure. Then, a partial image is used to extract the single closed edge of the valve opening. To obtain the precise edge position, a subpixel location based on the rotating area effect is applied. Finally, the area computation based on the minimum circumscribed circle and the maximum inner circle of the edge is provided. The experiment results prove the system reliability, and the general significances of the proposed algorithms.
查看更多>>摘要：? 2022This study proposes a valley-positioning-assisted discrete cross-correlation algorithm for the fast cavity length interrogation of fiber-optic Fabry–Perot (FP) sensors. Because the number of template cavity lengths that require to be cross-correlated with the spectral function of the FP sensor is dramatically decreased by considering the interference order and discretization of the template function, the computation amount can be effectively reduced without sacrificing the interrogating resolution. The feasibility and performance of the proposed algorithm were successfully verified through simulations and experiments. In the cavity length range of 20–130 μm, fiber-optic FP sensors were successfully interrogated with high performance. The maximum error is 3.97 nm, and the resolution reaches a value of 2.62 nm.
查看更多>>摘要：? 2022 Elsevier LtdThis paper classifies non-ectopic (N), supraventricular ectopic (S), ventricular ectopic (V), and fusion (F) beats in the MIT-BIH arrhythmia database. The classification encounters serious class imbalance since the number of beats in N (majority class) with sample number above the average per class is heavily outnumbered than that in S, V, and F (minority classes) with sample number below the average per class. To address the class imbalance, a novel model based on active training subset selection and modified broad learning system (MBLS) is proposed. In each iteration, the MBLS trained with the current training subset is used to predict the class label of the test sample and actively select a new training subset for the next iteration. Finally, the class of the test sample is determined by voting on the predictions of all iterations. The experimental results show that our method has excellent performance and outperforms the existing methods.
查看更多>>摘要：? 2022The cutting temperature is essential for phenomena understanding and quality improvement in metal cutting while its in-situ online measurement is still a challenge. This paper presents a near-infrared fiber-optic multi-spectral method for in-situ online cutting temperature measurement. Using thermal radiation spectrum for temperature measurement, the method optimizes the lower limit of temperature measurement to 150 °C while improving accuracy. The calibration shows that in the range of above 250 °C, the average relative error of temperature measurement is stable below 0.5%. The titanium alloy cutting experiments are carried out. In-situ online measurement of tool temperatures in dry/wet cuttings are realized using the self-developed system. The influence of cutting parameters on cutting temperature is studied, and the real-time response of the temperature measurement system to the cutting state is verified. As for industrial application, the capability of the system in heavy-duty turning is proved by railway wheelsets turning experiments. Tool wear experiments are conducted, and a positive correlation between the cutting temperature and tool wear is revealed. Tool wear status recognition is realized based on cutting temperature by sparse autoencoder and k-means clustering, and a recognition accuracy of 97.3% is achieved. These results indicate promising prospects in cutting mechanism research, machining status monitoring and industrial applications, empowering the advancement of intelligent manufacturing and industry 4.0.
查看更多>>摘要：? 2022 Elsevier LtdThis paper presents an image segmentation algorithm by combining Manifold Projection and Persistent Homology (MP_PH). First, for a given image, the spectral measures of each pixel and its neighbor pixels are modeled with Gaussian Probability Distribution Function (GPDF) in an exponential family fashion. The Riemannian manifold, i.e. the data sub-manifold for the pixel, is built by taking the parameters of the GPDF exponential family model as its coordinates to depict the statistical characteristics of the original image. By Legendre transformation, the data sub-manifold is transformed into a parameter sub-manifold to depict all possible segmentation results. Only points representing classes of current segmentation results are activated on the parameter sub-manifold. Then, simplicial complexes constructed from the original image are used to compute persistent homology. The optimal scale can be obtained from persistent homology to compute the optimal homology group generated by homology generators which are referred to some pixels belonging to the same class. Finally, the segmentation is performed by projecting points of the data sub-manifold belonging to the same homology generator to the nearest activated point of the parameter sub-manifold, and updating all the activated points according to the projection results. As a result, all the activated points tend to be optimal segmentation. The experiments for synthetic and real images show that the proposed algorithm has high segmentation accuracy.
查看更多>>摘要：? 2022 Elsevier LtdIn this paper, YOLOv4 algorithm based on deep learning is used to detect coal gangue. Firstly, the data set of coal gangue was made, which provides sufficient data for the training and verification of the detection algorithm model. Then, the coal gangue data set was used to test the influence of the combined use of optimization methods on the YOLOv4 detection algorithm. Finally, the performance of YOLOv4, SSD and Faster R-CNN detection algorithms combined with optimization methods in the field of coal gangue detection was compared through the coal gangue test data sets and the detection experiments. According to the coal gangue test data sets and coal gangue detection experiments, the combined use of optimization methods results in the mAP value of the YOLOv4 detection algorithm reaching 97.52%, which is 40.70% and 43.81% higher than those of the SSD and Faster R-CNN detection algorithms, respectively. Moreover, the accuracy, recall rate, and real-time performance of the YOLOv4 detection algorithm with the optimization methods are also better than those of the SSD and Faster R-CNN detection algorithms.
Rashedul Islam M.Tariqul Islam M.Bais B.Jit Singh M....
11页
查看更多>>摘要：? 2022 Elsevier LtdIn this research paper, a new reflected mirror rectangular split-ring resonator-shaped metamaterial sensor is proposed for the detection of materials and thickness of the materials. The metamaterial structure is designed in the same dimension as the opening of the X-band waveguide as if a total guided electromagnetic wave penetrates the structure, and its size is 22.86 × 10.16 mm2. Computer simulation technology (CST) microwave studio is used to design the metamaterial structure and the obtained reflection coefficient is validated with the advanced design system (ADS) results. Various parametric analysis has been done to optimize the design and size of the structure. Three materials FR-4, Rogers RO4350B, Rogers, and RT5880 have been attached to the metamaterial sensor and have shown the overall results both experimentally and theoretically. The resonance frequency shifted 120 MHz between FR-4 and Rogers RO4350B, and 250 MHz shifted between FR-4 and Rogers RT5880. Three different thicknesses of the FR-4 have been used to see the response of the metamaterial sensor. The resonance frequency shifted 80 MHz between 1.6 and 1.3 mm thickness of the FR-4, and 110 MHz shifted between 1.6 and 1 mm thickness of the FR-4. The simulated and measured outcomes are quite similar. The sensitivity of the structure is 1.29 and the quality factor (Q-factor) is 435, the figure of merit (FOM) is also analyzed, and its value is 561.15. Since the proposed sensor has high sensitivity, high Q-factor, and shows better performance, hence it can be used in industry to detect the various materials and thickness of the materials.
查看更多>>摘要：? 2022 Elsevier LtdA machine vision system was developed to discriminate in-demand and unwanted Baijiu brewing-sorghum at single kernel sample level. Three types of in-demand sorghum and seven types of unwanted sorghum were detected. Xception was employed to build classification model, reaching 89.08% and 88.21% correct classification rate for training and validation set, respectively. To achieve higher performance, two types of anti-aliased networks (anti-aliased max pooling (AntiMaxP) and anti-aliased convolutional (AntiConV)). Compared with the baseline Xception, the AntiMaxP and AntiConV both achieved higher overall accuracy. The AntiConV model obtained the best result, with accuracy of 89.22% and 89.15% for training and validation set, respectively. In view of practical application, the AntiConV model also obtained the most satisfactory result. Thus, AntiConV mdoel was integrated in the system. Adulterated samples were prepared to test the whole system. The results showed feasibility of the intelligent vision-based system to meet the practical application demands of Baijiu industry.
查看更多>>摘要：? 2022 Elsevier LtdA novel transient heat flux sensor (THFS) was developed to measure the heat flux of an explosive driven shock tube. The working principle of the THFS is based on the transverse Seebeck effect (TSE) of a single crystal Bi2Te3. The sensor has a rise time of 51 μs and sensitivity of up to 69.8 μV/(kW/m2), with no signal amplifier required for the transient high heat flux test. This sensor allows for highly time-resolved measurements of heat flux in the explosion field, thereby supporting the study of the thermal damage effect of an explosion.
NSTL
Elsevier