Springer Science and Business Media LLC

In order to strengthen the management of English classroom discipline and improve the efficiency of students’ English classroom learning, students’ English classroom behavior based on intelligent image recognition is analyzed in IoT (Internet of things). The working scenes and practical significance of deep learning and IoT are analyzed and then the effects of four models on students' behavior analysis in English classroom are discussed. The results show that the classroom behavior analysis model proposed is feasible. The recognition system judges whether the students are listening seriously from three aspects, namely students' side face, head up and down, and their eyelid opening. The comparison of the four models of VGG16, ResNet18, ResNet50 and AlexNet shows that the accurate recognition rate of VGG16 for students' behavior in English classroom reaches 94.15%. Experiments show that the method provides a more objective evaluation of students’ classroom behavior. As a whole, students’ classroom behavior analysis based on IIRT (intelligent image recognition technology) in IOT is practicable for improving English classroom efficiency.

Software fault prediction techniques use previous software metrics and also use the fault data to predict fault-prone modules for the next release of software. In this article we review the literature that uses machine-learning techniques to find the defect, fault, ambiguous code, inappropriate branching and prospected runtime errors to establish a level of quality in software. This paper also proposes a hybrid technique for software fault prediction which is based on genetic programming and ensemble learning techniques. There are multiple software fault prediction (machine-learning) techniques available to predict the occurrence of faults. Our experiments perform a comparative study of the performance achieved by simple ensemble methods, simple genetic programming based classification and the hybrid approach. We find that machine learning techniques have different learning abilities that can be exploited by software professionals and researchers for software fault prediction. We find that the performance obtained by this proposed approach is superior to the simple statistical and ensemble techniques used in the automated fault prediction system. However, more studies should be performed on lesser used machine learning techniques.

The availability and throughput of offshore oil and gas plants operating in the Arctic are adversely influenced by the harsh environmental conditions. One of the major challenges in quantifying such effects is lack of adequate life data. The data collected in normal-climate regions cannot effectively reflect the negative effects of harsh Arctic operating conditions on the reliability, availability, and maintainability performance of the facilities. Expert opinions, however, can modify such data. In an analogy with proportional hazard models, this paper develops an expert-based availability model to analyse the performance of the plants operating in the Arctic, while accounting for the uncertainties associated with expert judgements. The presented model takes into account waiting downtimes and those related to extended active repair times, as well as the impacts of operating conditions on components’ reliability. The model is illustrated by analysing the availability and throughput of the power generation unit of an offshore platform operating in the western Barents Sea.

The Internet of Things (IoT) models are getting more complicated day by day with the rising demand in IoT automated network system. As the devices use wireless medium for broadcasting the data, it is easy to target for an attack. Machine Learning based solution is more promising to protect and detect the scheme that present in the abnormal state. This paper aims to implement a new attack detection system in IoT using KDD cup dataset. Initially, the possible paths from node to destination are created based on the Euclidean distance and connectivity between the nodes. Further, the path with minimum distance is chosen as the shortest path and data transmission takes place accordingly. Two phases of work is done, initial one is finding the presence of attacker by a pre-trained Optimized Deep belief network (DBN). Subsequently, if the presence of attacker is detected by DBN, the control is given to the bait process, which removes the corresponding attacker node. To ensure the precise detection process, the weights in DBN will be optimally tuned by a new Whale with Distance based Update (W-DU) algorithm. Finally, the performance of proposed system is evaluated over other traditional schemes with respect to parameters accuracy, specificity, precision, FPR, FDR and FOR.

This study proposes a two-degree-of-freedom PID controller based on the moth flame optimizer (MFO) algorithm for the load frequency management issue in a three-area unequal linked power system. Load frequency control is use to control the frequency of the grid to its scheduled value in the power system. A objective function is formulated in the LFC which will be utilized by the optimization techniques for the tuning of the parameter of the controller. The proposed controller’s efficiency is tested by contrasting it’s response with outcome of PID and fractional order PID (FOPID) for various scenario. The suggested controller’s parameters were concurrently tuned using a meta-heuristic approach called moth flam optimizer (MFO). The simulation result with MFO appraised with other optimizer like SCA (Sine–cosine algorithm), SSA (Slap swarm algorithm), PSO (Particle swarm optimization algorithm), ALO (Ant-lion optimization algorithm) for the various scenarios. The superiority of proposed techniques is further examine by including system non-linearity like governor Dead Band, generation Rate Constraint, and communication delay. Furthermore, to validate the supremacy of the suggested method, the statistical analysis with the help of Wilcoxon Sign Rank Test has been performed on 20 independent runs. The result gained through broad simulation states that the proposed tactic undoubtedly intensify the system performance compare to prevailing controllers and optimization technique in the existing literature.

This paper aims to improve the accuracy and efficiency of image aesthetic classification in environmental art design and provide a more professional and convenient method. Based on the Mobile Edge Computing (MEC) and Convolution Neural Network (CNN) Deep Learning (DL) algorithm, the current situation and shortcomings of the existing image aesthetic classification are analyzed. Thereupon, the MEC technology is combined with CNN, and the MEC-based Image Recognition (IR) architecture and parallel Deep CNN-based aesthetic evaluation method are proposed. Then, the environmental art design images are analyzed and classified using the proposed method. Experiments are designed to verify the performance of the proposed methods. The results show that the average Response Time (RT) of different images recognition under the mobile cellular network and WiFi network is more than 2000 ms and less than 1000 ms, respectively. Comparison of the transmission speeds given different images indicates that the IR RT under the proposed MEC Hierarchical Discriminant Analysis (MECHDA) architecture is faster than the other three: the MECHDA architecture occupies a smaller bandwidth, with a constant 1kB image transmission traffic. Additionally, the aesthetic classification accuracy of the proposed model in A large-scale database for aesthetic visual analysis data set has reached 85.3%. In the data set of the CUHK Photo Quality Dataset (CUNK-PQ), the accuracy of aesthetic classification has reached 93%. The environmental art design pre-analysis and beauty collection classification method proposed enables the computer to help humans make more accurate image aesthetic analysis and classification. It improves the efficiency and accuracy of image aesthetic classification and provides a reference for research in related fields.

Năng lực năng lượng gió đang mở rộng toàn cầu. Hầu hết các tuabin gió hoạt động trong môi trường khắc nghiệt, đòi hỏi các phương pháp điều khiển chống lỗi. Dự án này nhằm cung cấp một hệ thống điều khiển năng lượng gió bền vững, chống lỗi. Một biện pháp bền vững xác định độ ổn định của cảm biến và các khuyết điểm có thể bị tổn thương về hiệu suất theo những khả năng xảy ra lỗi. Phương pháp được đề xuất đảm bảo hiệu suất cao nhất có thể trong trường hợp cảm biến gặp sự cố và các điều kiện hoạt động không bình thường. Các cảm biến tốc độ máy phát điện bị lỗi làm giảm hiệu suất và độ an toàn của hệ thống điều khiển năng lượng gió, từ đó ảnh hưởng đến hiệu quả của hệ thống. Bài báo này trình bày một hệ thống điều khiển cảm biến tốc độ máy phát điện chống lỗi dựa trên bộ quan sát. Đây là một hệ thống điều khiển chống lỗi thụ động mà không cần tái cấu hình bộ điều khiển. Thay vì nhận tín hiệu đầu ra từ cảm biến, bộ điều khiển nhận tốc độ máy phát điện dự đoán. Mô hình toán học được sử dụng để xây dựng một bộ quan sát chống lỗi tính toán các vấn đề về hộp số và ổ trục của tuabin gió. Kết quả mô phỏng chứng minh tính khả thi của phương pháp.

In order to improve the effect of power distribution network inspection and reduce the hidden dangers and operating costs of the power distribution network inspection, this paper combines the bionic vision image processing technology to construct an intelligent power distribution network inspection vision system, and proposes a bionic model based on the principle of biological visual distance that takes the Kinect Depth information value as a parameter. Moreover, this paper uses the model in the vision system to improve its real-time performance. In addition, with the support of intelligent algorithms, this paper constructs the structure model of the power distribution network inspection vision system, and proposes a new type of intelligent inspection system design for power distribution network to achieve a good effect of improving the efficiency of power distribution network inspection and management level in production practice. Finally, this paper combines experiments to prove the reliability of this system.

Due to the exponential growth in the automobile industry, we need intelligence transportation system. Vehicular ad-hoc network (VANET), a part of the intelligence transportation system is the network created by vehicles. Security is the main issue in vehicular ad-hoc network. Many intruders try to use the vulnerability presents in the vehicular network. In VANET communication between two nodes, may involves multiple intermediate nodes to forward the data due to low transmission range. The intermediate nodes must be trustworthy enough to be a part of the communication process. Rogue or malicious nodes can accept the data and drop the data in between source to destination. In this paper, we proposed a trust based model to detect rogue nodes in a vehicular network. The proposed model first estimates the trust value of the nodes and based on that identifies the rogue nodes in the network. We select only trustworthy nodes to relay the data in the routing process. The simulation and performance evaluation of the proposed model performed with the help of network simulator (NS-2). We evaluate the performance of the network based on the four performance matrices i.e. successful packet delivery fraction, throughput, routing load and end to end delay. The simulation result shows that the proposed model enhances network performance significantly.