• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.7 no.5
• /
• pp.92-97
• /
• 1998

The morpholgies of the wear particles are directly indicative of wear processes occuring in machinery and their severity. The neural network was applied to identify wear debris generated from the machine driving system. The four parameters(50% volumetric diameter, aspect, roundness and reflectivity) of wear debris are used as inputs to the network and learned the friction condition of five values(material 3, applied load 1, sliding distance 1). It is shown that identification results depend on the ranges of these shape parameters learned. The three kinds of the wear debris had a different patter characteristic and recognized the friction condition and materials very well by artificial neural network. We discussed how the network determines differencee in wear debris feature, and this approach can be applied to foreseeability and decisio for moving condition of the Machine driving system.

• Journal of Information Processing Systems
• /
• v.18 no.1
• /
• pp.12-25
• /
• 2022

In recent years, deep convolutional neural networks have made significant progress in the research of single image super-resolution. However, it is difficult to be applied in practical computing terminals or embedded devices due to a large number of parameters and computational effort. To balance these problems, we propose CSRNet, a lightweight neural network based on channel split residual learning structure, to reconstruct highresolution images from low-resolution images. Lightweight refers to designing a neural network with fewer parameters and a simplified structure for lower memory consumption and faster inference speed. At the same time, it is ensured that the performance of recovering high-resolution images is not degraded. In CSRNet, we reduce the parameters and computation by channel split residual learning. Simultaneously, we propose a double-upsampling network structure to improve the performance of the lightweight super-resolution network and make it easy to train. Finally, we propose a new evaluation metric for the lightweight approaches named 100_FPS. Experiments show that our proposed CSRNet not only speeds up the inference of the neural network and reduces memory consumption, but also performs well on single image super-resolution.

• Journal of electromagnetic engineering and science
• /
• v.15 no.4
• /
• pp.258-265
• /
• 2015

A de-embedding method for multiport networks, especially for coupled odd interconnection lines, is presented in this paper. This method does not require a conversion from S-parameters to T-parameters, which is widely used in the de-embedding technique of multiport networks based on cascaded simple two-port relations, whereas here, we apply an operation to the S-matrix to generate all the uncoupled and coupled coefficients. The derivation of the method is based on the relations of incident and reflected waves between the input of the entire network and the input of the intrinsic device under test (DUT). The characteristics of the intrinsic DUT are eventually achieved and expressed as a function of the S-parameters of the whole network, which are easily obtained. The derived coefficients constitute ABCD-parameters for a convenient implementation of the method into cascaded multiport networks. A validation was performed based on a spice-like circuit simulator, and this verified the proposed method for both uncoupled and coupled cases.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.3
• /
• pp.130-138
• /
• 1995

In this paper, a Novel Approach is Proposed which Identifies linear system Parameters Using a multilayer feedforward neural network trained with backpropagation algorithm. The parameters of linear system can be represented by x9t)/x(t) and x(t)/u(t). Thud, its parameters can be represented in terms of the derivative of output with respect to input of parameters can be represented in terms of the derivative of output with respect to input of trained neural network which is a function of weights and output of neurons. Mathematical representation of the proposed approach is derived, and its validity is shown by simulation results on 2-layer and 3-layer neural network.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.25 no.8
• /
• pp.808-820
• /
• 2014

In this paper, we fabricated two on-wafer type DUT(Device-Under-Test)s; a 10-dB attenuator and an amplifier using commercially available MMIC and we proposed the measurement method of the noise parameters for the two fabricated DUTs. Since the 10-dB attenuator DUT is a passive device, its noise parameters can be accurately determined when its S-parameters are measured. In the case of the amplifier DUT, its noise parameters are available in the datasheet. Hence, the measured noise parameters using the proposed method can be assessed by comparing with the known noise parameters. The noise parameter measurement method having been presented by the authors requires the S-parameters of the 8-port network used in the measurement and limited to coaxial type DUTs. When on-wafer probes are included in the 8-port network, the 8-port S-parameters requires the measurements with different kinds of connectors. In this paper, we obtained the 8-port S-parameters using the Smart-Cal function in the network analyzer. The measured noise parameters shows about ${\pm}0.2dB$ fluctuations for $NF_{min}$. Other noise parameters with the frequency change show good agreement with the expected results.

• Journal of Biosystems Engineering
• /
• v.37 no.4
• /
• pp.271-278
• /
• 2012

Worldwide trends in animal welfare have resulted in an increased interest in individual management of sows housed in groups within hog barns. Estrus detection has been shown to be one of the greatest determinants of sow productivity. Purpose: We conducted this study to develop a method that can automatically detect the estrus state of a sow by selecting optimal texture parameters from images of a sow's pudendum and by optimizing the number of neurons in the hidden layer of an artificial neural network. Methods: Texture parameters were analyzed according to changes in a sow's pudendum in estrus such as mucus secretion and expansion. Of the texture parameters, eight gray level co-occurrence matrix (GLCM) parameters were used for image analysis. The image states were classified into ten grades for each GLCM parameter, and an artificial neural network was formed using the values for each grade as inputs to discriminate the estrus state of sows. The number of hidden layer neurons in the artificial neural network is an important parameter in neural network design. Therefore, we determined the optimal number of hidden layer units using a trial and error method while increasing the number of neurons. Results: Fifteen hidden layers were determined to be optimal for use in the artificial neural network designed in this study. Thirty images of 10 sows were used for learning, and then 30 different images of 10 sows were used for verification. Conclusions: For learning, the back propagation neural network (BPN) algorithm was used to successful estimate six texture parameters (homogeneity, angular second moment, energy, maximum probability, entropy, and GLCM correlation). Based on the verification results, homogeneity was determined to be the most important texture parameter, and resulted in an estrus detection rate of 70%.

• The KIPS Transactions:PartB
• /
• v.14B no.5
• /
• pp.377-382
• /
• 2007

This paper proposes a training algorithm for global optimization of the parameters of radial basis function networks. Since conventional training algorithms usually perform only local optimization, the performance of the network is limited and the final network significantly depends on the initial network parameters. The proposed hybrid simulated annealing algorithm performs global optimization of the network parameters by combining global search capability of simulated annealing and local optimization capability of gradient-based algorithms. Via experiments for function approximation problems, we demonstrate that the proposed algorithm can find networks showing better training and test performance and reduce effects of the initial network parameters on the final results.

• Proceedings of the IEEK Conference
• /
• 1998.06a
• /
• pp.86-89
• /
• 1998

Multiple types of information in a multimedia service are delivered though multiple virtual connections on ATM network, while each virtual connection may be controlled independently. A multimedia service requires an associative relationship among multiple information streams to provide required harmonization. There may be required additional traffic descriptors to guarantee the required harmonization among multiple information streams in a multimedia service. For buffering of large bandwidth information stream(e.g., video), extremely large buffer size is necessary, but this approach should not be efficient way to compensate a severely delayed cells/blocks experienced at network. The best way to solve this problem will be minimization of relative-delayed-transfer of cells/blocks to application processes through ATM network control. To minimize a delayed transfer the mapping between relative delay parameter(i.e., associative Group QoS parameters) and per-VC traffic descriptor will be necessary. This paper is present additional functions and parameters to interpret the mapping between relative delay parameters(i.e., associative Group QoS parameters) and per-VC traffic descriptors in ATM API for multimedia application services.

• Journal of the Korean Society for Precision Engineering
• /
• v.15 no.5
• /
• pp.43-49
• /
• 1998

Adaptive neural network approach is presented for determining Electrical Discharge Machining (EDM) parameters. Electrical Discharge Machining has been widely used with its capability of machining hard metals and tough shapes. In the past few years, EDM has been established in tool-room and large-scale production. However. in spite of it's wide application, an universal selection method of EDM parameters has not been established yet. No attempt has been tried before to suggest a logical method in determining essential machine parameters considering the machining rate and resulting surface roughness integrity. The paper presents a method, which is focusing on determining appropriate machining parameters. Depending on the electrode wear and surface roughness, an adaptive neural network is designed for providing suitable machining guideline.

• Transactions of Materials Processing
• /
• v.31 no.2
• /
• pp.81-88
• /
• 2022

Numerical analysis and dynamic material properties are required to analyze the behavior of workpiece during an electrohydraulic forming (EHF) process. In this study, EHF experiments were conducted under three conditions (6, 7, 8 kV). Dynamic material properties of Al 5052-H34 were inversely estimated through an ANN (Artificial Neural Network) model constructed based on LS-Dyna analysis results. Parameters of Cowper-Symonds constitutive equation, C and p, were used to implement dynamic material properties. By comparing experimental results of three conditions with ANN model results, optimized parameters were obtained. To determine the reliability of the derived parameters, experimental results, LS-Dyna analysis results, and ANN results of three conditions were compared using MSE and SMAPE. Valid parameters were obtained because values of indicators were within confidence intervals.