• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.13 no.5
• /
• pp.81-86
• /
• 2004

A fundamental study for developing a system of fault diagnosis of a pump is performed by using neural network. The acoustic signals were obtained and converted to frequency domain for normal products and artificially deformed products. The signals were obtained in various driving frequencies in order to obtain many types of data from a limited number of pumps. The acoustic data in frequency domain were managed to multiples of real driving frequency with the aim of easy comparison. The neural network model used in this study was 3-layer type composed of input, hidden, and output layer. The normalized amplitudes at the multiples of real driving frequency were chosen as units of input layer. Various sets of teach signals made from original data by eliminating some random cases were used in the training. The average errors were approximately proportional to the number of untaught data. The results showed neural network trained by acoustic signals can be used as a simple method for a detection of machine malfuction or fault diagnosis.

• New & Renewable Energy
• /
• v.4 no.4
• /
• pp.23-29
• /
• 2008

Wind farm development project contains high business risks because that a wind farm, which is to be operating for 20 years, has to be designed and assessed only relying on a year or little more in-situ wind data. Accordingly, long-term correction of short-term measurement data is one of most important process in wind resource assessment for project feasibility investigation. This paper shows comparison of general Measure-Correlate-Prediction models and neural network, and presents new method using neural network for increasing prediction accuracy by accommodating multiple reference data. The proposed method would be interim step to complete long-term correction methodology for Korea, complicated Monsoon country where seasonal and diurnal variation of local meteorology is very wide.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.5
• /
• pp.261-268
• /
• 2020

Genetic algorithms find the optimal solution by mimicking the evolution of natural organisms. In this study, the genetic algorithm was used to enable Pac-Man's reinforcement learning, and a simulator to observe the evolutionary process was implemented. The purpose of this paper is to reinforce the learning of the Pacman AI of the simulator, and utilize genetic algorithm and artificial neural network as the method. In particular, by building a low-power artificial neural network and applying it to a genetic algorithm, it was intended to increase the possibility of implementation in a low-power embedded system.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10a
• /
• pp.436-440
• /
• 1992

As an approach to design the intelligent controller, this paper proposes a new FNN(Fuzzy Neural Network) control method using the hybrid combination of fuzzy logic control and neural network. The proposed FNN controller has two important capabilities, namely, adaptation and learning. These functions are performed by the following process. Firstly, identification of the parameters and estimation of the states for the unknown plant are achieved by the MNN(Model Neural Network) which is continuously trained on-line. And secondly, the learning is performed by FNN controller. The error back propagation algorithm is adopted as a learning technique. The effectiveness of the proposed method will be demonstrated by computer simulation of a two d.o.f. robot manipulator.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10a
• /
• pp.607-612
• /
• 1992

For hydroforming process, determination of back-up fluid pressure in chamber is one of the most essential tasks. In this paper, we present a back-up pressure estimation system which estimates the back-up pressure of hydroforming process utilizing a multi-layered neural network. The neural network learns the nonlinear relation ship between the back-up pressure and the geometric state variables of hydroforming process. The proposed method does not necessitate sophisticated analysis on hydroforming process but some geometric intuition. The experimental results show that the neural network well approximates the nonlinear relationship between the back-up pressure and the geometric state variables of hydroforming process, thus giving the good estimation of back-up pressure vs punch stroke curve.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10b
• /
• pp.558-563
• /
• 1992

In this paper an identification of nonlinear continuous systems by using neural network is considered. The nonlinear continuous system is identified by two steps. At first, a linear approximate model of the continuous system with nonlinearity is obtained by IIR filtering approach. Then the modeling error due to the nonlinearity is reduced by a neural network compensator. The teaching signals to train the neural network is gotten by smoothing the measurement data corrupted by noise. An illustrative example is given to demonstrate the effectiveness of the proposed approach.

• 제어로봇시스템학회:학술대회논문집
• /
• 1995.10a
• /
• pp.75-78
• /
• 1995

In this paper it is concerned to develop control method using jar-test results in order to predict the optimum dosage of coaglant, PAC(PoliAluminum Chloride). Considering the relations with the reactions with the reaction of coagulation and flocculation, the five independent variables ( e, g, turbidity of raw water, water turbidity in flocculators, temperature, pH, and alkalynity) are selected out of parameters and they are put into calculation to develop a neural network model for PAC dosing process in water purification system. This model is utilized to predict optimum dosage of PAC. That is, the optimum dosage of PAC is searched in neural network model for PAC dosing process to minimize the water turbidity in flocculators. This searching is implemented by means of expert heuristics. The efficacy of the proposed contorl schemem and feasibility of acquired neural network model for PAC dosing contorl in water purification system is evaluated by means of computer simulation.

• Journal of Institute of Control, Robotics and Systems
• /
• v.4 no.5
• /
• pp.651-657
• /
• 1998

In this study, an intelligent scheduling with hybrid architecture, which integrates expert system and neural network, is proposed. Neural network is trained with the data acquired from simulation model of FMC to obtain the knowledge about the relationship between the state of the FMC and its best dispatching rule. Expert system controls the scheduling of FMC by integrating the output of neural network, the states of FMS, and user input. By applying the hybrid system to a scheduling problem, the human knowledge on scheduling and the generation of non-logical knowledge by machine teaming, can be processed in one scheduler. The computer simulation shows that comparing with MST(Minimum Slack Time), there is a little increment in tardness, 5% growth in flow time. And at breakdown, tardness is not increased by expert system comparing with EDD(Earliest Due Date).

• Journal of Institute of Control, Robotics and Systems
• /
• v.6 no.4
• /
• pp.305-312
• /
• 2000

This paper presents a new control scheme using a sliding mode controller with a multilayer neural network for the robot manipulator operating under the sea which has large uncertainties such as the buoyancy and the added mass/moment of inertia. The multilayer neural network using the error back propagation loaming algorithm acts as a compensator of the conventional sliding mode controller to improve the control performance when the initial assumptions of uncertainty bounds are not valid. Computer simulation results show that the proposed control scheme gives an effective path way to cope with the unexpected large uncertainties in the underwater robot manipulator.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.672-675
• /
• 1999

In this paper, we propose a MLP neural network architecture and feature extraction for Korean syllable recognition. In the proposed syllable recognition system, firstly onset is classified by onset classification neural network. And the results information of onset classification neural network are used for feature selection of imput patterns vector. The feature extraction of Korean syllables is based on sonority. Using the threshold rate separate the syllable. The results of separation are used for feature of onset. nucleus and coda. ETRI's SAMDORI has been used by speech DB. The recognition rate is 96% in the speaker dependent and 93.3% in the speaker independent.