Browsing by Author "Tsai, Y. H."

Now showing 1 - 6 of 6

A new indexing motion program for optimum designs of Geneva mechanisms with curved slots
(2017-11-01)
Lin, W. Y.
;
Tsai, Y. H.
;
Hsiao, K. M.
;
機械工程學系
;
Department of Mechanical Engineering
A curved slotted Geneva mechanism can eliminate the adversely infinite angular jerks of the Geneva wheel and might reduce the peak angular acceleration of the Geneva wheel by using a proper indexing motion program. In the literature, the cycloidal, fifth-order polynomial and modified sine indexing motion programs are frequently used for curved slotted Geneva mechanisms. To achieve the better kinematic performance of the curved slotted Geneva wheel than that obtained using the above-mentioned indexing motion programs, a new indexing motion program based on the Hermite interpolating polynomial is proposed for an optimum design with the goals of minimizing the peak angular acceleration and eliminating the adversely infinite angular jerks. The domain of the indexing position function is divided into several segments. Each segment is termed an element, and both ends of each segment are termed nodes. The nodal values of the indexing position function and its derivatives are used as design variables. The position function for each element can be described using the Hermite interpolating polynomial and the design variables. The reason behind the use of the Hermite interpolating polynomial is that the design variables have the clear physical meanings. The four-level Hermite interpolating polynomial is used and two elements are sufficient to obtain the optimum results. In addition, the constraint regarding the radius of curvature of the profile of the inner slot is proposed to prevent sharp curvature of the profile of the inner slot. The findings show that there is a decline in the peak acceleration of the Geneva wheel with six curved slots for the optimum results obtained using the proposed indexing motion program by 33.4% and 24.3%, respectively, as compared with the cycloidal and modified sine indexing motion programs.
A scalable lossy substrate model for nanoscale RF MOSFET noise extraction and simulation adapted to various pad structures
(2007)
Guo, J. C.
;
Tsai, Y. H.
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
A broadband and scalable lossy substrate model is developed and validated for nanoscale RF MOSFETs of different finger numbers and adopting various pad structures such as lossy, normal, and small pads. The broadband accuracy is justified by good match with S- and Y-parameters up to 40 GHz. The measured noise characteristics in terms of four noise parameters can be accurately simulated up to 18 GHz. The scalable lossy substrate model can consistently predict the abnormally strong ringer number dependence and nonlinear frequency response of noise figure (NF(min)) revealed by the devices with lossy pads. Furthermore, the scalable model can precisely distribute the substrate loss between the transmission line (TML) and pads of various metal topologies and the resulted excess noises. The enhanced model provides useful guideline for appropriate layout of pads and TML to effectivelv reduce the excess noises. The remarkably suppressed noise figure to ideally intrinsic performance can be approached by the small pad in this paper.
Deep Learning Based AOI System with Equivalent Convolutional Layers Transformed from Fully Connected Layers
(2019-01-01)
Tsai, Y. H.
;
Lyu, N. Y.
;
Jung, S. Y.
;
Chang, K. H.
;
Chang, J. Y.
The rise of deep learning, especially in the realm of computer vision, paves ways of leveraging automatic optical inspection systems to a higher level. Convolutional neural networks and its derivatives might be the most widely used architectures for defect inspection tasks. In real cases the amount of collected data is often not large, so transferring learning and data augmentation are necessary. In this paper, we explain some details how we implement the deep learning based AOI system where fully connected layers are replaced by convolutional layers, then a classification heat map is output after post-processing. We examine the performance of our model with two data sets collected in industrial manufacturing cases. We further propose an idea to transfer models pretrained on augmented data of different sizes cropped from original image to the present classification task for possible improvements of the performance.
Defect Detection on Randomly Textured Surfaces by Convolutional Neural Networks
(2018-01-01)
Jung, S. Y.
;
Tsai, Y. H.
;
Chiu, W. Y.
;
Hu, J. S.
;
Sun, C. T.
Automatically detecting the defects on the randomly textured surfaces for industrial purpose is a demanding procedure due to the ambiguity between defects and textures, lack of defect-labeled data and the must-have extreme accuracy. In this paper we proposed a procedure as the beginning of automating the defect detection on woods with randomly textured surfaces by employing 3 different architectures of convolutional neural networks. The deep convolutional neural network resulted in 99.80% accuracy, discriminating among normal wood and the other 4 types of defects images. The models were evaluated and understood by visualizing the saliency maps. The results from our work implies that other industrial images with defects on randomly textured surfaces may apply the similar procedures to accelerate the automating of defect detection and progressing of industry 4.0.
Industrial Anomaly Detection and One-class Classification using Generative Adversarial Networks
(2018-01-01)
Lai, Y. T. K.
;
Hu, J. S.
;
Tsai, Y. H.
;
Chiu, W. Y.
;
資訊工程學系
Industrial image datasets for quality inspection are mostly sparse in defects. It is then hard for both automated optical inspection (AOI) machines and simple neural network classifiers to inspect all defects effectively. In this work, we develop a novel framework for industrial anomaly detection in one-class classification manner, which utilized pre-trained generative adversarial networks (GANs) as the rule of thumb to perform anomaly detection. Our results show that GANs are able to capture arbitrary and structural industrial images and can effectively discern defects when the query images are defective.
Optimum variable input speed for kinematic performance of Geneva mechanisms using teaching-learning-based optimization algorithm
(2017-05-01)
Lin, W. Y.
;
Tsai, Y. H.
;
Hsiao, K. M.
;
機械工程學系
;
Department of Mechanical Engineering
An optimum design of variable input speed for the Geneva mechanism is aimed at improving the kinematic performance of the traditional Geneva mechanism by eliminating infinite angular jerks and reducing the peak angular acceleration of the Geneva wheel during the indexing motion. The normalized angular velocity and acceleration of the Geneva wheel corresponding to the normalized time are introduced. A polynomial function of the normalized time is used to describe the normalized angular position of the crank, and therefore, the corresponding polynomial coefficients are considered as the design variables. The optimum design task is very specialized and difficult to solve with some evolutionary and swarm optimization methods because of the extremely large range for the value of the design variable, arising from the utilization of a higher order polynomial for the normalized time parameter with a value between 0 and 1. A new evolutionary algorithm termed teaching-learning-based optimization comprises a teacher phase and a learner phase. In the teacher phase, the entire population can be gradually shifted to a more promising region, which may be very far from the relatively small initial region. The obtained optimal results are compared with those obtained using the length-adjustable deriving link method discussed in the literature. The findings show that the difference in the effectiveness of the variable input speed method and the length-adjustable driving link method for the reduction of the peak angular acceleration of the Geneva wheel is small.