Search

In order to further improve the accuracy and efficiency of predicted compressible flowfields arounds airfoils at high Reynolds number, large angle of attack (AoA), and high Mach numbers, a machine learning inference method based on coordinate transformation method and UNet neural network is propesed. Firstly, a novel coordinate transformation method for data pre-processing is developed. This method transforms physical quantities and grid information in computational fluid dynamics into spatial information of neural network, making the distribution of flowfield information more in line with the input requirements of the neural network. Secondly, an improved deep UNet neural network is established, allowing the model to learn the fine and complex localized flow characteristics of the airfoil flowfield. The two innovative methods are combined to establish a machine learning inference method for the compressible flowfield of airfoils, and a fast and high-precision inference model is obtained. Finally, the flowfields and aerodynamic forces of different types of airfoils are predicted and analyzed, and results are compared with those from traditional machine learning method. The results show that the machine learning inference method proposed in this paper can better predict the compressible flowfield of airfoils, and can better capture the complex flow behavior at high Reynolds numbers, and predict the flow separation and shock waves phenomena under high Mach numbers with large AoA. [ABSTRACT FROM AUTHOR]

In wind tunnel experiments, sensors are normally placed on wing surfaces to measure the aerodynamic load at the corresponding positions. Due to the limited number of the placed sensors, it is difficult to directly obtain the holographic aerodynamic load distribution of the whole wing. In this paper, we use machine learning methods to reconstruct the holographic aerodynamic load on the wing surface from the limited sensor data, and propose a method to optimize the placement of sensors using the simulation data. We select limited position data from the wing holographic aerodynamic data calculated by computational fluid dynamics (CFD) to simulate the sensor experimental data, and compare the reconstruction accuracy of the deep learning model, Gaussian process regression (GPR), support vector regression (SVR) and BP neural network (NN) to the aerodynamic load. The sensor layout is optimized by evaluating the holographic load accuracy reconstructed by the sensor data. The M6 wing is taken as an example to verify the proposed method under two given working conditions. The experimental results show that the GPR model achieves the best accuracy of aerodynamic load reconstruction. The optimal section number and single section number of M6 wing under different total number of sensors, the optimal arrangement under the lowest number of sensors, and the arrangement of sensors in the leading edge area and spanwise section where the flow field changes greatly are all given. [ABSTRACT FROM AUTHOR]

To accomplish the task of aerial manipulator’s autonomous grasping, it is very important to recognize and locate the object. At present, most of the recognition algorithms of aerial manipulators use traditional feature extraction methods. In order to improve the accuracy and efficiency of object recognition and positioning, this paper designs a visual recognition and positioning algorithm based on YOLOv5 deep learning object detection algorithm and RGB‑D sensor. The algorithm can detect the object in real‑time and estimate its pose, which serves for the grasping task of aerial manipulator. At the same time, aiming at the problems that the deep learning algorithm has a huge amount of calculation and cannot achieve high‑performance real‑time detection in the embedded system, model quantization is introduced to optimize the algorithm, which can greatly improve the processing speed of the algorithm. This paper introduces the overall framework and implementation process of the algorithm and verifies the effectiveness of some algorithms of object detection and pose estimation by using the COCO dataset and motion capture system. [ABSTRACT FROM AUTHOR]

Meta reinforcement learning（Meta-RL）aims at automatically learning induction bias for a new reinforcement learning task from a set of different but related tasks. It plays an important role in improving the sample efficiency and generalization of reinforcement learning algorithm in difficult scenarios. This paper first introduces a framework in which three key components of Meta-RL are identified，i.e.，learning experience （related tasks），inductive bias and learning objective. Based on this，current research progress in this field is analyzed and reviewed，and the principles and characteristics of several representative algorithms are described. The paper also gives a detailed account of commonly used benchmark environments and performance evaluation methods for meta-RL. The limitation of current research and potential future development directions are also discussed. [ABSTRACT FROM AUTHOR]

Abnormal event detection is a hot topic in the field of intelligent surveillance monitoring technology research currently. As an important research content of computer vision, its main goal is to use computers to automatically detect abnormal events. Traditional methods have limitations in the weakness of low-level video feature description ability, high computational cost of anomaly detection methods and poor robustness in modeling complex scenes. In recent years, how to design a high-level semantic feature extraction method, accelerate the process of abnormal event detection, and model complex scenes such as multiple cameras have become the forefront topic of current research. Based on the research situation at home and abroad and the mainstream methods, this paper introduces the basic techniques involved in abnormal event detection in surveillance videos, and analyzes the advantages and disadvantages of various types of surveillance video feature extraction methods， feature learning models and anomaly detection methods. This paper also summarizes the commonly used benchmark datasets that can be used to abnormal event detection in surveillance videos. Finally, we discuss the difficulties, challenges and future development trends of abnormal event detection in surveillance videos. [ABSTRACT FROM AUTHOR]

Mixed sampling data，whose different features are collected at different sampling frequencies， pervasively exists in the real world. Because the different sampling data set not only has different features，but also has different number of samples，traditional classification methods cannot be used directly. Therefore， this paper proposes a classification method for mixed sampling data based on Fisher discrimination dictionary learning to solve the classification problem of mixed sampling data. Inspired on some classification ideas for multi-view data，the proposed model compares multiple data collected at multiple sampling frequencies to multiple views of multi-view data，and designs a way to learn a sub-dictionary for each class of each view. A structured dictionary whose dictionary atoms have correspondence to the class labels is learned，so that the within-class scatter is less and the between-class scatter is bigger based on the Fisher discrimination criterion. In addition，this paper designs a specific classification scheme for the inconsistent sample size of mixed sampling data. Finally，experimental results demonstrate the effectiveness of the proposed model. [ABSTRACT FROM AUTHOR]

Tool wear prediction plays an important role in ensuring the machining quality and efficiency, and reducing the machining cost, especially in the aerospace manufacturing field where difficult machining materials are widely used. Data and mechanism fusion model can combine the advantages of mechanism model and data-driven model. It is an effective way to realize tool wear prediction. However, the existing fusion methods are difficult to effectively balance the weight of data and mechanism on the model, which makes it difficult to really achieve the expected effect of the fusion model. To solve the above problems, this paper proposes a tool wear prediction method based on meta learning (ML) and PINN (Physics-informed neural network). The solution space of the data-driven model is constrained by the wear mechanism, and the loss function of the fusion model is optimized by the meta learning algorithm to make rational use of the information provided by the data and mechanism. The experimental results show that the proposed method can effectively improve the accuracy and stability of tool wear prediction under varying cutting conditions. [ABSTRACT FROM AUTHOR]

Iterative learning control problem is studied for a class of discrete nonlinear partial difference systems where the trial lengths vary randomly in this paper. The considered systems are composed of parabolic partial difference equations. Firstly, for the trial lengths vary randomly in systems, a stochastic variable satisfying Bernoulli distribution is introduced. Then, according to the properties of partial difference systems and P-type distributed update schemes, iterative learning controller is designed. Through rigorous theoretical analysis, the convergence sufficient condition of output error under λ norm and contractive mapping principle is established to ensure the convergence of tracking error in the sense of norm. The convergence speed of the algorithm is faster than the classical iterative learning algorithm. Finally, the effectiveness of the algorithm is verified by the numerical simulation. [ABSTRACT FROM AUTHOR]

Despite their prior knowledge，a large amount of unsupervised industrial data cannot be effectively exploited in real-world applications regions. In this paper，we propose a semi-supervised learning algorithm based on similarity measurement. This method combines specific prior knowledge and unsupervised learning to calculate the similarity between unlabeled data and a small amount of labeled data to augment the minority samples. By improving the classification effect of the model on the minority class，we can mitigate sample imbalance in training phrases and marking difficulty. Empirically，a good minority recognition effect has been achieved in a series of power-sensor detection classification tasks. Compared with state-of-the-art methods， including the traditional and semi-supervised methods，the recall rate and the F1 value comprehensively exceed the traditional ones. [ABSTRACT FROM AUTHOR]

In global friction measurement, thin oil film technique can well represent the distribution of surface friction. Irradiated with ultraviolet light of a specific wavelength, the oil film with different thickness of the fluorescent color molecules will emit different brightness. Using this principle, the thickness of the oil film can be calculated by detecting the gray value of the excited fluorescent oil film.In this paper, BP neural network and extreme learning machine（ELM）neural network are used to build models to complete the prediction of the relationship between the fluorescence oil film thickness and gray level, and Hopfield neural network is used to complete the identification of the corresponding parameters. The experimental results show that the prediction errors of ELM neural network model, BP neural network model and interpolation model are 5.150%, 5.485% and 5.935%, respectively. Through Hopfield neural network identification, the parameter error rates of the influence factors such as light source power, optical distance and exposure coefficient are controlled at about 1%, which meets the requirements of practical engineering application. Compared with the traditional interpolation method, the higher precision can be obtained by using the neural network, which provides a feasible method for the study of the gray and thickness of the fluorescent oil film. [ABSTRACT FROM AUTHOR]

Aiming at the online learning with outliers， this paper proposes a robust regularized online sequential extreme learning machine（RR-OSELM）. The proposed RR-OSELM is able to learn the newly arrived samples incrementally by a recursive fashion，and assign inverse weights for each example based on the priori error so as to reduce its sensibility to outliers. The Tikhonov regularization technique is incorporated in the RR-OSELM to further enhance the stability of the algorithm in real applications. Experimental results show that the proposed RR⁃OSELM is more robust than its counterparts，and it can be applied to the online modeling and prediction of data streams with outliers. [ABSTRACT FROM AUTHOR]

In the O2O marketing, coupon is an effective marketing tool. However, when we are not clear whether customers are willing to consume, coupons will be in spamming. In order to improve the utilization rate of coupons, in this paper, firstly the three⁃way decision⁃making thought is introduced to forecast the coupons utilization. And combine with the integrated algorithm XGBoost in machine learning, the model for the coupon usage is established. Second, in the process of three⁃way decision⁃making, taking into account misclassification cost and learning cost, the classification process should be more close to the actual classification. Finally, the real consumption data of users'coupon provided in Tianchi platform by Alibaba is experimentally analyzed. The results show that the XGBoost⁃based three⁃classification algorithm could effectively improve the accuracy of classification, so that merchants could not only maintain the regular customers, but also identify potential new customers, thereby reducing their marketing cost. [ABSTRACT FROM AUTHOR]

Aircraft structure damage seriously affects the aircraft flight safety. In order to effectively identify aircraft composite structure damage, a new method combining general regression neural network (GRNN) and extreme learning machine (ELM) of composite structure damage diagnosis is proposed in this paper. Firstly, the data of fiber optic sensor on composite material laminated plates are gathered and pre-processed after striking and stretching on composite laminated plates. Secondly, strain information is decomposed by variational mode decomposition (VMD), and intrinsic mode functions (IMFs) are obtained. Meanwhile, the singular entropy feature of each IMF is derived. Then, a feature vector is built by kernel independent component analysis (KICA). Finally, the fusion feature vector is used to bui_d GRNN-ELM identification model. Experimental data verify the effectiveness of the GRNNELM method, and the result shows that the GRNN-ELM model can realize aircraft composite structure damage identification more effectively compared with ELM and GRNN models, respectively, thus it has a good engineering application value. [ABSTRACT FROM AUTHOR]