Название: System Innovation for a Troubled World: Applied System Innovation
Автор: Artde Donald Kin-Tak Lam, Stephen D. Prior, Siu-Tsen Shen
Издательство: CRC Press
Серия: Smart Science, Design and Technology
Год: 2023
Страниц: 179
Язык: английский
Формат: pdf (true)
Размер: 10.2 MB

The major themes on technology included Material Science & Engineering, Communication Science & Engineering, Computer Science & Engineering, Electrical & Electronic Engineering, Mechanical & Automation Engineering, Architecture Engineering, IOT Technology, and Innovation Design.

Sentiment analysis of text data is an important task in natural language processing (NLP). Taking the product review system as an example, proper use of sentiment analysis can reveal the opinions of consumers, and thus adjustments can be made to products. However, most current sentiment analysis methods assume that there is a large amount of training data available for the classification model. The classification accuracy is not ideal when little data are available for analysis.Aiming at sentiment analysis of text data with few data samples, in this paper, we propose an embedded learning model, which combines convolutional neural network and bidirectional long short-term memory learning models to improve the relevance of data by dimensionality reduction and thus strengthen the classification accuracy of few-shot learning. In addition, we segment the training data and use them to train the proposed model in batches to avoid the overfitting problem that is often observed in few-shot learning models.

Volume decomposition aims to split a computer-aided design (CAD) model into sub-volumes that can be meshed with existing hexahedral mesh generation algorithms. The possibility of making automatic decomposition has been investigated, but the challenge remains whenever a CAD model has many extrusion features. Therefore, we propose an approach for separating extrusion features from the main body of a thin-shell model by assigning different face types. In addition, a novel method for classifying thin-shell models into regular and irregular wall types is introduced. This study mainly focuses on the type of irregular wall and divides the face-type recognition algorithm into the following four steps: separation of inner and outer regions, classification of regular and irregular-wall types, further classification of irregular-wall types, and assigning face types on each sub-type of irregular walls. As a result, a set of attributes that are usable for decomposition derivation is available for each face on the model. To verify the feasibility of the proposed method, several realistic thin-shell models are employed to demonstrate the face types recognized. In addition, the application of the face types obtained in volume decomposition is demonstrated.

LoRaWAN (long range wide area network) is widely considered as a promising wireless communication technology that can meet the requirements for thousands or millions of IoT (Internet of Things) end devices to be able to connected to the Internet. In this paper, a scheme is proposed for end devices to select a spreading factor for uplink transmissions in LoRaWANs. In our proposed scheme, connectivity between end devices and gateways is first constructed, which depends on the received signal strength. As an end device may be able to be connected to multiple gateways at the same time, the reliability of uplink transmissions improves. Nonetheless, the traffic load seen at each gateway increases, thereby leading to more packet collisions during uplink transmissions. In order to deal with this tradeoff situation, a systematic probability-based approach is employed in the selection of a spreading factor for each end device, and a water-filling algorithm is utilized to balance traffic loads between spreading factors, from which the proabability distribution for the spreading factor selection of each end deivce can then be determined. Numerical results show that our proposed scheme significantly outperforms ADR (adaptive data rate), which is a scheme recommended by the standard.

Vehicular Ad Hoc Networks (VANETs) have widely been considered as a promising wireless communication technology to offer both vehicle safety and infotainment. Much research attention has been devoted to designing a multichannel MAC protocol in VANETs, with the aim of maximizing data throughput while ensuring collision-free deliveries of safety messages. For instance, in the coordinated multichannel medium access control (C-MAC) protocol, a time division multiple access (TDMA) approach is employed for the latter, and a balance of the rates is stricken for the former between successful requests for data transmissions and data transmissions that can be performed. However, C-MAC may suffer from channel underutilization as a contention-based mechanism is used to make reservations for data transmissions as well as to identify new vehicle arrivals at the coverage of a road side units (RSU). Two designs are thus proposed in this paper to improve channel utilization. Not only is TDMA better leveraged under both designs, but local information is utilized in one of the designs, which is regularly gathered at a vehicle by receiving beacons from neighboring vehicles. Simulations show either design can substantially outperform C-MAC in terms of throughput. Compare further between the two designs, though the one using local information can achieve a higher throughput than another retaining a vehicle identification process when vehicle intensity is low on a road, the reverse is true when it is high. A switching rule is then proposed and its accuracy is confirmed from numerical results.

A 5G millimeter wave network is able to deliver speeds of Gbps through the use of beam-forming and the dense deployment of small cell access points. In this paper, we consider both the associations of clients and allocations of channel resources, where demands from clients are taken into account. The problem can be formulated into an integer nonlinear optimization. To solve this, a dual decomposition method can be utilized, and then a distributed algorithm is developed. Numerical results demonstrate that our proposed algorithm can substantially outperform the previous work in which no client demand is taken into account.

The objective of this paper is to design a new reversed intake for a pusher aircraft having dimension restriction of the nacelle and establish a testing platform to verify the results. For a pusher aircraft, the turboprop engine is mounted on the tail of the aircraft, and hence a reversed intake with U-elbow to guide airflow flowing through a 180◦ bend, maintaining excellent flow quality, prior to entering the engine is required. The U-elbow channel has a twisted angle to connect the throat and the inlet mouth of the engine. In the design and analysis, the ANSYS CFX was employed to examine the influence produced by the variation of intake geometry on two crucial performance parameters: total pressure recovery and flow distortion coefficient. In the experiment, the TPE331-10 engine was employed and a testing platform was built to carry out the test.