Abstract:Brain-computer interfaces (BCI) based on steady-state visual evoked potentials (SSVEP) have attracted great interest because of their higher signal-to-noise ratio, less training, and faster information transfer. However, the existing signal recognition methods for SSVEP do not fully pay attention to the important role of signal phase characteristics in the recognition process. Therefore, an improved method based on extended Canonical Cor-relation Analysis (eCCA) is proposed. The phase parameters are added from the stimulus paradigm encoded by joint frequency phase modulation to the reference signal constructed from the training data of the subjects to achieve phase constraints on eCCA, thereby improving the recognition performance of the eCCA method for SSVEP signals, and transmit the collected signals to the robotic arm system to achieve control of the robotic arm. In order to verify the effectiveness and advantages of the proposed method, this paper evaluated the method using SSVEP signals from 35 subjects. The research shows that the proposed algorithm improves the average recognition rate of SSVEP signals to 82.76%, and the information transmission rate to 116.18 bits/min, which is superior to TRCA and traditional eCAA-based methods in terms of information transmission speed and accu-racy, and has better stability.

Abstract:As the growing requirements for the stability and safety of process industries, the fault detection and diagnosis of pneumatic control valves have crucial practical significance. Many of the approaches were presented in the literature to diagnose faults through the comparison of residual sequences with thresholds. In this study, a novel hybrid neural network model has been developed to address the issue of pneumatic control valve fault diag-nosis. First, the feature extractor automatically extracts in-depth features of the signals through multi-scale convolutional neural networks with different kernel sizes, which not only adequately explores the local dis-tinguishable features, but also takes into account the global features. The extracted features are then fused by the feature fusion layer to reduce redundant features. Finally, the long short-term memory for fault identification and the dense layer for fault classification. Experimental results demonstrate that the average test accuracy is above 94% and 16 out of the 19 conditions can be successfully detected in the simulated actual industrial en-vironment. The effectiveness and practicability of the proposed method have been verified through a com-parative analysis with existing intelligent fault diagnosis methods, and the results suggest that the developed model has better robustness.

Abstract:With the increasing popularity of 3D sensors (e.g., Kinect) and light field cameras, technologies such as driv-erless, smart home and virtual reality have become hot spots for engineering applications. As an important part of 3D vision tasks, point cloud semantic segmentation has received a lot of attention from researchers. In this work, we focus on realistically collected indoor point cloud data and propose a point cloud semantic segmen-tation method based on PAConv and SE_variant. The SE_variant module captures global perception from a broad perspective of feature space by fusing different pooling methods, which fully utilize the channel in-formation of point clouds. The effectiveness of the method is verified by comparing with other methods on S3DIS and ScanNetV2 semantic tagging benchmarks, and achieving 65.3% mIoU in S3DIS, 47.6% mIoU in ScanNetV2. The results of the ablation experiments verify the effectiveness of the key modules and analyze how to use the attention mechanism to improve the 3D semantic segmentation performance.

Abstract:To address the alignment and measuring force problem in the segmenting-stitching technique for the circularity metrology of small cylindrical workpieces (diameter less than 1.5 mm and length less than 10 mm), a magnet combination jig method is proposed. A small round magnet is attached between the round magnetic jig and small cylinder, and the other end of the small cylindrical workpiece is attached to some cylindrical magnets. Thus, the smaller cylinder can be put in the V-groove and measured successfully with the magnet combination. For ver-ifying the advantage of the magnet combination jig, four measurement quality evaluations are proposed: the circumferential deviation of neighbor arc contours, radial deviation of neighbor arc contours, angle of inclina-tion between the V-groove and small cylinder, and curvature of the obtained arc. The results show that the matching coefficient is enhanced by 98%, the Euclidean distance of overlap parts of neighbor arc contours is reduced by 68%, the position error is reduced 27%, and the average curvature of the arc contours is improved. It can be concluded that the measuring quality can be enhanced prominently by this magnet combination method for the segmenting-stitching method.

Abstract:Mach number is a key metric in the evaluation of wind tunnel flow field performance. This complex process of wind tunnel test mainly has the problems of nonlinearity and time lag. In order to overcome the problems and control the Mach number stability, this paper proposes a new method of Mach number prediction based on a nonlinear autoregressive exogenous-genetic algorithm-Elman (NARX-GA-Elman) model, which adopts NARX as the basic framework, determines the order of the input variables by using the false nearest neighbor (FNN), and uses the dynamic nonlinear network Elman to fit the model, and finally uses the global optimization algo-rithm GA to optimize the weight thresholds in the model to establish the Mach number prediction model with optimal performance under single working condition. By comparing with the traditional algorithm, the predic-tion accuracy of the model is improved by 61.5%, and the control accuracy is improved by 55.7%, which demonstrates that the model has very high prediction accuracy and good stability performance.

Abstract:The test section’s Mach number in wind tunnel testing is a significant metric for evaluating system performance. The quality of the flow field in the wind tunnel is contingent upon the system's capacity to maintain stability across various working conditions. The process flow in wind tunnel testing is inherently complex, resulting in a system characterized by nonlinearity, time lag, and multiple working conditions. To implement the predictive control algorithm, a precise Mach number prediction model must be created. Therefore, this report studies the method for Mach number prediction modelling in wind tunnel flow fields with various working conditions. Firstly, this paper introduces a continuous transonic wind tunnel. The key physical quantities affecting the flow field of the wind tunnel are determined by analyzing its structure and blowing process. Secondly, considering the nonlinear and time-lag characteristics of the wind tunnel system, a CNN-LSTM model is employed to establish the Mach number prediction model by combining the 1D-CNN algorithm with the LSTM model, which has long and short-term memory functions. Then, the attention mechanism is incorporated into the CNN-LSTM predic-tion model to enable the model to focus more on data with greater information importance, thereby enhancing the model's training effectiveness. The application results ultimately demonstrate the efficacy of the proposed approach.