The early diagnosis of laryngeal cancer (LCA) is crucial for prognosis, driving our search for an accurate, precise, and sensitive deep learning model to assist in LCA detection.
As the key component of industrial robot, harmonic drive is unfortunately also the weakest part of the robot. Therefore, fault diagnosis is needed to find out what causes the problems and hence, find the remedies. Most of the existing mature machine learning approaches can obtain ideal results in the fault diagnosis of rotating machinery. However, the shortcoming of these methods is that the classifier only can output one label when the test sample is a composite fault signal, rather than multiple labels. Consequently, these traditional methods cannot simultaneously identify and output every label in the composite fault signal. To solve this problem, an intelligent method named Multi-Sensors Convolutional Neural Networks with Sigmoid function Classifier (MSCNN-SC) is proposed for compound fault identification. CNN is adopted to effectively obtain the representative features of the original signals. Binary Cross-Entropy (BCE) Loss and Sigmoid functions are then employed to calculate and output the probability of each label. The decision strategies using adjusted thresholds are designed to obtain the decoupling results of compound fault diagnosis. The proposed approach is validated by the multi-axises harmonic drives of the industrial robot. The experimental results demonstrate that the proposed approach can correctly identify the compound fault.
The automobile welding workstation is an important part of vehicle production, and the reliable operation of a large number of vulnerable non-standard parts directly affects the processing quality of the product. Therefore, it is necessary to monitor its real-time status to avoid economic losses caused by unexpected shutdowns. However, the current regular maintenance method relies on the mature experience of the technicians, the cost is high and the maintenance effect is not ideal. In addition, massive amounts of data have also caused difficulties for effective analysis. To solve this problem, an intelligent method named Stacked Autoencoder and Long short-term memory (SAE-LSTM) is proposed for fault detection. First, the time-domain signals, frequency signals, and time-frequency signals are collected by the Visual Components software. Second, the data preprocessing module will be used to initially remove useless system signals. Third, the SAE network is further used to reduce the dimensionality of massive data and maintain the key features of the original system. Finally, the LSTM network is used to predict the upcoming data through real-time collected data and performs fault detection by comparing it with the real data. The proposed method is validated by the real automobile welding workstation. The experimental results demonstrate that the proposed approach can effectively detect equipment failures.
To achieve accurate and completely autonomous navigation for spacecraft, inertial/celestial integrated navigation gets increasing attention. In this study, a missile-borne inertial/stellar refraction integrated navigation scheme is proposed. Position Dilution of Precision (PDOP) for stellar refraction is introduced and the corresponding equation is derived. Based on the condition when PDOP reaches the minimum value, an optimized observation scheme is proposed. To verify the feasibility of the proposed scheme, numerical simulation is conducted. The results of the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) are compared and impact factors of navigation accuracy are studied in the simulation. The simulation results indicated that the proposed observation scheme has an accurate positioning performance, and the results of EKF and UKF are similar.
Image Inpainting is to repair the images with given mask or missing area and produce a realistic fake image similar to the original image. Methods with Generative Adversarial Networks(GANs) and deep convolutional neural network(CNNs) have achieved great performance on generating visual plausible results. However, those results still have blurry issues on restored region and unpleasant boundaries. Based on these problems, a two-phase neural network method is proposed to produce better results on image inpainting in this paper. The proposed method adopts the Shift-Net [1] as the first phase to generate primary images, which are then fed into refinement network (the second phase) for further detailed texture restoring. This model adopts improved Wasserstein GAN(WGAN-GP) to ensure the training stability and performance on generating high-resolution images. Experiments on shows the proposed model can produce clear and realistic plausible face images. Through the comparison with state-of-the-art methods on both PSNR and SSIM metrics, the proposed method has a better performance on recovering the feature information on missing region.
In order to improve the short-term traffic flow prediction precision,this paper proposed a short-term traffic flow forecasting model based on wavelet neural network.Wavelet neural network can describe the short-term traffic flow's stochastic and uncertainty and predict nonlinearly.The experiment was carried out to verify the validity of the model.The results show that the wavelet neural network improves the traffic flow prediction precision and has good application value.
Harmonic drive is the core component of the industrial robot, and its fault diagnosis is crucial to the reliability and performance of the equipment. Most machine learning methods achieve good results based on the assumption of data balance. However, the scarce fault data of harmonic drive is difficult to collect, resulting in various imbalanced health status samples, which has an adverse effect on fault diagnosis. In this article, we propose a data generation method based on generative adversarial networks (GANs) to solve the problem of data imbalance and utilize the multiscale convolutional neural network (MSCNN) to realize the fault diagnosis of the harmonic drive. First, the data collected from three vibration acceleration sensors are preprocessed by fast Fourier transform (FFT) to obtain the frequency spectrum of the vibration signal. Second, multiple GANs were adopted to generate various fault spectrum data and the data selection module (DSM) is elaborately designed to filter and purify these data. Third, the filtered generated data will be combined with the real data to form a balanced dataset, and then the MSCNN is used to achieve multiclassification of the health status of the harmonic drive. Finally, the experiments have been implemented on an industrial robot vibration test bench to validate the effectiveness of our approach. The results have shown the fault multiclassification accuracy as 98.49% under imbalanced fault data conditions, which outperforms that of the other compared methods.
In this paper, we present a novel method to simulate single scattering in homogeneous participating media, which can greatly enhance the immersion of environment simulation. We construct a light volume with a polygonal mesh based on shadow map to restrict the ray-marching segments, place samples at discontinuities along the epipolar lines of the final image to speed up the ray-marching process, and add support for textured light. Our method is implemented on GPUs, and can render a complex scene with participating media in real time.
'/%3e%3cuse xlink:href='%23a' transform='translate(288.889 -214.211)'/%3e%3cuse xlink:href='%23a' transform='translate(108 342.749)'/%3e%3cuse xlink:href='%23a' transform='translate(469.189 342.749)'/%3e%3c/svg%3e)
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
'/%3e%3cuse xlink:href='%23a' transform='rotate(144 450 300)'/%3e%3cuse xlink:href='%23a' transform='rotate(216 450 300)'/%3e%3cuse xlink:href='%23a' transform='rotate(288 450 300)'/%3e%3c/svg%3e)