Taiwan Association of Engineering and Technology Innovation: E-Journals
Not a member yet
887 research outputs found
Sort by
A Multiple-Parameter Optimization of Darrieus Vertical Axis Wind Turbine for Enhanced Performance
Full text linkThis paper aims to optimize a small vertical-axis wind turbine (VAWT) by analyzing chord length, hub radius, and circular angle, and establishing a relationship between design parameters and performance. The approach involves evaluating five airfoils and identifying the best airfoil using QBlade, based on comparisons of the power coefficient (CP), tip speed ratio (TSR), and the power output (P). Mathematical relationships are developed through extrapolation using polynomial, logarithmic, and modified Avrami equations. The result shows that the 'S1046 17%' is the best-performing airfoil among the five. The optimum chord length to hub radius ratio yields the highest power output, and the impact of circular angle on performance is negligible. The Avrami equation shows better fitness to the original data than the polynomial equation. The Troposkien variant shows a higher CP over a wide range of TSR, while the straight blade produces higher power output across varying wind speed (V)
Pest-YOLO: A YOLOv5-Based Lightweight Crop Pest Detection Algorithm
Full text linkTraditional crop pest detection methods face the challenge of numerous parameters and computations, making it difficult to deploy on embedded devices with limited resources. Consequently, a lightweight network is an effective solution to this issue. Based on you only look once (YOLO)v5, this paper aims to design and validate a lightweight and effective pest detector called pest-YOLO. First, a random background augmentation method is proposed to reduce the prediction error rate. Furthermore, a MobileNetV3-light backbone replaces the YOLOv5n backbone to reduce parameters and computations. Finally, the Convolutional Block Attention Module (CBAM) is integrated into the new network to compensate for the reduction in accuracy. Compared to the YOLOv5n model, the pest-YOLO model’s Parameters and Giga Floating Point Operations (GFLOPs) decrease by about 33% and 52.5% significantly, and the Frames per Second (FPS) increase by approximately 11.1%. In contrast, the Mean Average Precision (mAP50) slightly declines by 2.4%, from 92.7% to 90.3%
Dynamic Allocation Strategy for the Car Rental Industry with Multiple Rental Channels: Dynamic Allocation Strategy for the Car Rental
No full textThis paper addresses the rental car allocation problem in which a car rental company rents its cars through various channels. This study proposes a constrained model based on the nested-allocation structure to allocate the rental cars to the car rental system to maximize the total rental revenues. The nested allocation strategy uses the concept of whether the number of sales accumulated to a certain level exceeds a threshold corresponding to that sharer. This approach allows the quota of low-profit channels to support the leasing demand of high-profit channels at any time, thereby increasing revenue. The proposed model can dynamically adjust the nested strategy as the dynamic programming model. However, it avoids the issues of ample storage space and long calculation time commonly encountered by the dynamic programming model. Computational results show that the proposed nested approach substantially outperforms the traditional exclusive allocation approach
Performance Trade-Offs in AI-Based Speed Control of PMDC Motors: A Comparative Study of Fuzzy Logic and Neural Network Controllers
Full text linkThis paper aims to compare the speed control of DC motors using two distinct artificial intelligence controllers: fuzzy logic controllers (FLC) and artificial feedforward neural networks (AFFNN). This study presents the first comprehensive comparison of FLC and ANN under identical test conditions, offering actionable guidelines for industrial applications. The driving system has been designed and tested using MATLAB/Simulink. Simulation results show that the AFFNN controller’s rise time at 130 (rad/s) is 14.9 ms, whereas the fuzzy logic controller’s is 32.4 ms. Furthermore, the neural network controller and fuzzy logic controller overshoot by 2.6e-06% and 0%, respectively. However, the neural controller takes 213.5 ms to reach its peak, whereas the fuzzy controller achieves this level earlier, at 94.6 ms. AFFNN gives a faster rise time and minimal settling time. However, FLC gives a faster peak response with zero overshoot for effective PMDC motor control
Harmonic Elimination Method for Permanent Magnet Synchronous Motor Utilizing Active Disturbance Rejection Control
Full text linkThis study investigates the control of permanent magnet synchronous motor (PMSM) as the core component of speed control systems. The extensive employment of PMSMs in electric cars is ascribed to their efficiency, simplicity, dependability, and stable performance. Therefore, this article suggests an active disturbance rejection control (ADRC) strategy for electric motor control systems to improve the system’s robustness, accelerate response speed, and enhance tracking performance. The implementation of ADRC reduces the speed response time from 0.14 s to 0.04 s and the three-phase current drop time from 0.03 s to 0.01 s. The total harmonic distortion (THD) under the existing proportional-integral (PI) speed controller and the suggested ADRC speed controller are 10.58% and 5.73%, respectively, with the proposed ADRC having a reduced THD
A Wide-Band Millimeter-Wave On-Chip Six-Port Reflectometer
Full text linkFollowing the previous success of measuring the reflection coefficients of devices under test at 20 GHz, this paper proposes a new six-port reflectometer (SPR) chip that aims to work at 40 GHz. The new SPR is implemented with the 0.13-μm IBM BiCMOS-8HP technology, and the overall chip area is 1.5 mm in width and 1 mm in height. To demonstrate the SPR’s excellent performance over a wide band, this study utilizes a programmable tuner to create fifteen different loads for the SPR to measure at 30 GHz, 40 GHz, and 50 GHz, respectively. Among the loads, the programmable tuner serves as an important instrument for producing various sliding terminations, which are essential for calibrating the SPR. Compared with the measurement results of a vector network analyzer, the SPR displays maximum measurement errors of -28.6 dB, -32.4 dB, and -27.7 dB while operating at 30 GHz, 40 GHz, and 50 GHz
Multifunctional Intelligent Helmet to Enhanced Safety and Comfort of Laborers in the Mining Industry
Full text linkThis study aims to enhance miner safety through real-time monitoring and emergency responses. To achieve this, a multi-functional mining helmet (MFMH) is designed with location tracking via a global system for mobile communications (GSM) and global positioning system (GPS), hazardous gas detection, lighting, and temperature regulation, along with vibration-based alerts for emergency notification. The helmet is tested in simulated mining environments to assess its performance. The system successfully detected hazardous gases at concentrations of 41.23 ppm, triggered automatic lighting when luminosity dropped below 35 lux, and maintained internal temperatures between 26 ℃ and 27 ℃, demonstrating its effectiveness in safety
Comparative Analysis of Facial Expression Recognition Using Image-Based and Landmark-Based Methods
Full text linkThis study compares the effectiveness of image-based and landmark-based methods for facial expression recognition (FER) in classifying hurt and normal facial expressions, utilizing datasets from the Delaware Pain Database and UTKFace. Five machine learning models are assessed, including convolutional neural networks (CNN), support vector machines (SVM), random forest classifier (RFC), logistic regression classifier (LRC), and gradient boosting classifier (GBC). The findings indicate that CNN achieves the highest accuracy at 95% when using landmark-based features, while SVM and GBC also perform admirably with these features. Conversely, LRC exhibits inconsistent results, especially when relying on image-based features. These findings offer valuable insights into the strengths and weaknesses of each approach, guiding the selection of effective FER techniques
Analysis of Characteristics of Complaints on Parenting Q&A Sites Using pLSA and Data Augmentation
Full text linkThis study investigates the classification and clustering of complaints on a Japanese parenting Q&A site, aiming to identify meaningful patterns from limited labeled data. To address data scarcity, generative AI was utilized for data augmentation through prompts that reflected authentic parenting frustrations, with synthetic data validated by comparing classification performance under varying proportions of generated content. Complaint texts were vectorized using Bag-of-Words, Doc2Vec, and Sparse Composite Document Vectors, providing multiple levels of semantic representation. LightGBM was used as the classifier, and F1 scores measured performance. Clustering of predicted complaints employed probabilistic Latent Semantic Analysis, with topic numbers selected via Bayesian Information Criterion. Six distinct themes emerged, including childcare stress and family conflict. Incorporating generated data improved the F1 score from 0.824 to 0.865. The findings highlight the potential of generative AI to augment low-resource datasets and demonstrate the effectiveness of context-aware embeddings and probabilistic clustering in structuring real-world text data
An Enhanced BiLSTM-Based Model with Bidirectional Attention and Ant Colony Optimization for English NLP
Full text linkThis study aims to overcome limitations in traditional natural language processing (NLP) models, particularly in network structure and hyperparameter tuning, which often hinder optimal performance across diverse tasks. To address these issues, the ant colony optimization (ACO) algorithm is introduced. This paper optimizes the layer count and other training hyperparameters of the Bidirectional Long Short-Term Memory (BiLSTM) network, enhancing both its flexibility and classification accuracy. To further enhance BiLSTM’s bidirectional selectivity, a bidirectional attention mechanism (BAM) is incorporated, strengthening the model’s capacity to integrate historical and future contextual information. The proposed ACO-BiLSTM-BAM model is validated on the Internet Movie Database (IMDb) movie review dataset, where it achieves a classification accuracy of 92.74%, marking a significant 12.05% improvement over the base BiLSTM model, particularly in discriminating sentiment at varied levels