Taiwan Association of Engineering and Technology Innovation: E-Journals
Not a member yet
887 research outputs found
Sort by
Evaluation of Time-Frequency Representations for Deep Learning-Based Rotating Machinery Fault Diagnosis
Full text linkThis study evaluates and compares five time-frequency representation (TFR) methods for fault diagnosis in rotating machinery, aiming to ensure operational reliability and reduce unexpected downtime. The methods—short-time Fourier transform (STFT), continuous wavelet transform (CWT), modified S-transform (MS-transform), smoothed pseudo Wigner-Ville distribution (SPWVD), and Hilbert-Huang transform (HHT)—are investigated. Vibration signals from benchmark bearing and gearbox datasets are converted into two-dimensional TFR data and classified using a convolutional neural network (CNN). Results show that MS-transform achieves the highest accuracy (up to 99.87%) under ideal conditions. STFT and CWT demonstrate better robustness in noisy environments, maintaining over 99% accuracy at 15 dB signal-to-noise ratio (SNR). SPWVD is computationally intensive with moderate performance, while HHT performs poorly under noise. Renyi entropy, energy conservation, and training time are also used to assess TFR quality. These findings support selecting appropriate TFR methods for industrial fault diagnosis
System-Theoretic Analysis of Hazard Causal Factors and Scenario Development for Complex Systems
Full text linkThis paper proposes a system-theoretic process and analysis (STPA) for complex systems (STPA_CS) to analyze the hazard factors arising from interactions between multiple components in a system. STPA_CS is realized by adding the following features to STPA: (1) using class diagrams to define control signal inputs and outputs of components, (2) defining component configurations by using composite configuration diagrams, (3) clarifying hazard occurrence processes by tracing the control signal exchange paths, and (4) facilitating seamless hazard analysis by detailing the class diagrams and repeating the hazard analysis. When STPA_CS and STPA are applied to a chiller system, STPA_CS clarifies fifty-one hazard factors and scenarios, while STPA clarifies thirty hazard factors and scenarios. This represents an improvement of 60% in the analysis results and demonstrates the advantages of STPA_CS
Multi-Objective Optimization of EV Charging for Cost and Loss Minimization Under TOU Tariff
Full text linkThis study proposes an optimal electric vehicle (EV) charging (OEVC) management methods to minimize electricity costs and energy losses in the distribution system, which arise from the growing demand for EV charging. a multi-objective particle swarm optimization (MOPSO) algorithm is used to solve the OEVC multi-objective optimization (MOO). Additionally, the time-of-use (TOU) tariff is used to coordinate between the distribution system operator and EV users, which can help increase the efficiency of the charging schedule. Monte Carlo Simulation (MCS) is used to model virtual EV user behavior and create EV charging load profiles. The proposed MOPSO-based OEVC approach is verified on the modified IEEE 33-bus distribution test system, using MATLAB software, under both uncontrolled and controlled charging case studies. The simulation results demonstrate that the proposed method optimizes EV charging efficiently, achieving reductions of approximately 7.60% in electricity costs and 28.73% in energy losses compared to the uncontrolled charging case
3D Printer-Based PCR Reagent Dispenser with Syringe Pump and Three-Way Valve for Rapid Nucleic Acid Diagnostics
Full text linkThis study presents a cost-effective automation solution for preparing polymerase chain reaction (PCR) reagent cartridges used in automated nucleic acid analyzers. Manual preparation is labor-intensive and error-prone, often causing inaccurate volumes and reagent mismatches. To address this, a dispensing system based on open-source 3D printer technology is developed. It incorporates a motion platform and a syringe-based pump, and precisely dispenses reagents into cartridge chambers designed for magnetic DNA extraction and real-time PCR. The system is evaluated for manual inefficiency and error. Dispensing accuracy, assessed gravimetrically using 500 μL of distilled water, shows a relative accuracy of 0.30% and a coefficient of variation (CV) of 2.64%, both within ISO 8655 limits. In terms of efficiency, the system fills a single cartridge chamber in 13.57 seconds, much faster than the approximately 3 minutes required for manual reagent injection. These results highlight the system’s potential to improve throughput and precision in cartridge preparation
Optimizing Ammonia Removal from Secondary Aluminum Dross as a Potential Raw Material Substitution in the Cement Industry
Full text linkSecondary aluminum dross (SAD), produced by small and medium-sized enterprises in Jombang Regency, Indonesia, is a hazardous waste with high ammonia content that threatens the environment and human health. Although SAD has potential as an alumina source for cement production, ammonia emissions restrict its safe use. This study applies a simultaneous heat-stirred alkaline leaching method to optimize ammonia removal for use as raw material in cement manufacturing. It addresses gaps in single-factor studies by optimizing multiple factors (NaOH concentration, temperature, reaction time, and stirring speed) using the Box–Behnken Design within Response Surface Methodology. Temperature and reaction time are the most influential, while interactions between NaOH and temperature, and between temperature and stirring speed, are critical for maximizing removal. The optimized process removed 98.81% ammonia, while an alternative yielded 98.34% with lower chemical and energy inputs. It enables safe SAD reuse and promotes the circular economy through waste valorization
Improving the Vehicle Small Object Detection Algorithm of Yolov5
Full text linkTo address the problems of low accuracy and poor robustness in vehicle small object detection for autonomous driving tasks, this study aims to propose an improved vehicle small object detection algorithm model based on YOLOv5. Firstly, some convolutions in the backbone network are replaced with receptive field attention convolutions, and the weights of the convolution kernels are dynamically assigned based on the importance of image features to ensure the extraction of important features. Secondly, adding a channel attention mechanism to the backbone network enhances the attention to small target features. Finally, the Focal-EIoU loss function is introduced to increase the attention on high-quality samples in the regression stage of object detection boxes. When the model is applied to the small object test set of the KITTI dataset, the precision rate, recall rate and mean average precision are 88.5%, 82.8%, and 84.9%, respectively, and the frame processing rate reaches 87.83FPS
Modeling and Suppression of Inhomogeneous Intensity Edible Bird Nest Images for Impurity Segmentation Using β-Variational Autoencoder
Full text linkThis study proposes a β-variational autoencoder (β-VAE) method to address intensity inhomogeneity (IIH) in edible bird’s nest (EBN) images, which creates an uneven intensity that obscures fine impurity details and reduces segmentation accuracy. First, the β-VAE is used to learn the feature distribution of EBN images by mapping them into a latent space. This latent space is then disentangled through selective filtering and penalization of specific latent dimensions. This unsupervised learning approach effectively captures and isolates IIH in EBN images. Additionally, enabling precise segmentation of EBN and impurities without requiring annotated datasets. It also enhances robustness in handling unseen image instances. The proposed method achieves an intersection over union of 73.08% (equivalent to a Dice coefficient of 84.44%), surpassing existing segmentation techniques. By resolving IIH, this method improves the reliability and adaptability of automated EBN inspection systems for practical applications
A CAD-Driven and Cloud-Based Autonomous Process Planning Framework for Reconfigurable Bending Press Machines
Full text linkSheet metal part manufacturers are increasingly under pressure to meet highly variable consumer demands. As product customization increases, the production process for sheet metal bending parts becomes more complex. This article proposes a fully integrated cloud-based system for sheet metal process planning. The system is developed based on a computer-aided design application and has the capability to rapidly convert a standard for the exchange of product data (STEP) file manufacturing instructions. A new mathematical model for calculating the overall production cycle time is also formulated. Two sheet metal components are used to test the system. The results demonstrate that the proposed cloud-based framework can display the 3D model, its face relationships, and a table containing the manufacturing information
Deep Learning-Based Smart Invigilation System for Enhanced Exam Integrity
Full text linkThis study proposes a smart invigilation system to preserve exam integrity by detecting suspicious student behaviors using deep learning. The model consists of three phases, i.e., student identity verification using face recognition, behavioral sampling for model training utilizing gesture analysis and convolutional 3D networks for emotion analysis, and live video analysis of suspicious activities integrating gesture, emotional analysis, and face recognition. The model is evaluated using 4,000 training and 1,000 test images and identifies non-cheating activities with 99% accuracy and cheating activities with 97.6% accuracy. The proposed model outperforms other methods, achieving accuracies of 98.4% for identifying cheating behaviors and 99.2% for non-cheating behaviors, resulting in an overall accuracy of 98.8% and a low misclassification rate of 1.2%. While the system demonstrates high accuracy, challenges remain in scaling to larger classrooms due to increased computational demand and the need for additional hardware to ensure comprehensive monitoring
Investigation of Affordable Technologies for Real-Time See-Through Various Indoor Surfaces and Walls
Full text linkWireless scanning for detecting objects behind various surfaces or walls in indoor settings has garnered significant interest recently. This study presents experimental results on several widely accessible, affordable, and portable see-through technologies. The technologies evaluated include a radio frequency (RF) device, a chip-sized multiple-input and multiple-output (MIMO) radar, an ultra-wideband sensor, and a motion sensor. These can be used either as standalone transceivers or mounted on unmanned aerial vehicles (UAVs) to extend their range, particularly for emergencies in high-rise buildings. Tests on various wall and surface materials show that RF and Wi-Fi devices can detect objects through wood, glass, and plasterboard, but metal and concrete significantly block or limit signal penetration. The results suggest that affordable see-through technologies need to improve their performance against concrete and metals