• Journal of Mechanical Science and Technology
• /
• v.14 no.6
• /
• pp.590-596
• /
• 2000

Cause of excessive vibration with twice the rotational speed of a two-pole generator rotor for the fossil power plants was investigated. The two-pole generator rotor, treated as a typically asymmetric rotor in vibration analysis, produces asynchronous vibration with twice the rotational speed, sub-harmonic critical speeds, and potentially unstable operating zones due to its own inertia and/or stiffness asymmetry. This paper introduces a practical balancing procedure, and presents the results of the investigation on sources of the excessive vibration based on the experimental vibration data of the asymmetric two-pole rotor in balancing.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.10 no.4
• /
• pp.157-162
• /
• 2010

The cart-pole balancing problem is a pseudo-standard benchmark problem from the field of control methods including genetic algorithms, artificial neural networks, and reinforcement learning. In this paper, we propose a novel approach by using online reinforcement learning(OREL) to solve this cart-pole balancing problem. The objective is to analyze the learning method of the OREL learning system in the cart-pole balancing problem. Through experiment, we can see that approximate faster the optimal value-function than Q-learning.

• Journal of KSNVE
• /
• v.9 no.4
• /
• pp.835-840
• /
• 1999

Cause of excessive vibration signals with twice the rotational speed of a 2-pole generator rotor in balancing for fossil power plants was investigated. The 2-pole generator rotor is treated as a typically asymmetric rotor in vibration analysis, and produces asynchronous vibration with twice the rotational speed for its own inertia and stiffness asymmetry. This paper introduces practical balancing procedure and experimental vibration data of the asymmetric 2-pole rotor in balancing, and presents the results of investigation into sources of the excessive vibration signals.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.40 no.2
• /
• pp.217-223
• /
• 1991

Most common applications of neural networks to control problems are the automatic motor controls using the artificial perceptual function. These control mechanisms are similar to those of the intelligent and pattern recognition control of an adaptive method frequently performed by the animate nature. In this paper, the pole-balancing problem is selected as the control object and an actual cart-pole controller is implemented by a computer interfacing and demonstrated as motor control using the reinforcement learning rule. In the experiment, given a change of the main parameters of cart-pole dynamics, a comparison is made between the LQR scheme and neural network method. The neural network method exhibits a more effecftive control action in a real situation having a large uncertainty than the LQR scheme.

• Journal of Power System Engineering
• /
• v.11 no.1
• /
• pp.16-19
• /
• 2007

The large generator rotor used in fossil power plant has the possibility of high 2X vibration due to asymmetric shaft stiffness. The generator rotor is machined into pole faces to reduce stiffness difference and then is tested through 2X vibration measurement when the balancing works are performed in the balancing shop. However, there are many cases of large difference values between 2X vibration in the balancing shop and 2X vibration in site. This paper presents a new method to estimate 2X vibration in site with more accuracy and applied for the retrofit of a fossil 400 MW class deteriorated generator. It shows that the new generator rotor is manufactured with a good 2X vibration characteristics and is operated in a low 2X vibration level although the generator rotor has high 2X vibration in the balancing shop.

• Journal of the Korean Society for Precision Engineering
• /
• v.14 no.7
• /
• pp.59-66
• /
• 1997

A general procedure for a motion capture and mimic system has been delineated. Utilizing sensors operated in the magnetic fields, complicated and optimized movements are easily digitized to analyze and repreduce. The system consists of a motion capture module, a motion visualization module, a motion plan module, a motion mimic module, and a GUI module. Design concepts of the system are modular, open, and user friendly to ensure the overall system performance. Custom-built and/or off-the-shelf modules are ease- ly integrated into the system. With modifications, this procedure can be applied for complicated motion controls. This procedure is implemented on tracking a head and balancing a pole. A neural controller based on this control scheme dtilizing human motions can easily evolve from a small amount of learning data.

• Proceedings of the KIPE Conference
• /
• 2015.07a
• /
• pp.177-178
• /
• 2015

This paper presents a neutral point deviation compensating control algorithm applied to a 3-level NPC converter. The neutral point deviation is analyzed with a focus on the current flowing out of or into the neutral point of the dc link. Based on the zero sequence components of the reference voltages, this paper analyzes the neutral point deviation and balancing control for 3-level NPC converter. An analytical method is proposed to calculate the injected zero sequence voltage for NP balancing based on average neutral current. This paper also proposes a control scheme compensating for the neutral point deviation under generalized unbalanced grid operating conditions. The positive and negative sequence components of the pole voltages and ac input currents are employed to accurately explain the behavior of 3-level NPC converter. Simulation and experimental results for a test set up of 30kW are shown to verify the validity of the proposed algorithm.

• 제어로봇시스템학회:학술대회논문집
• /
• 1999.10a
• /
• pp.73-78
• /
• 1999

This paper presents an effective construction method of adaptive multiple control systems utilizing some knowledge upon the plants. The adaptive multiple control system operates plants un-der widely changing environmental conditions. The adaptive multiple control system is composed of a family of candidate controllers together with a supervisor. The system does not require any identification schemes of environmental conditions. Monitoring outputs of the plant, the supervisor switches from one candidate controller to another, The basic ideas of adaptation are as follows: (1)each candidate controller is prepared for each environmental condition in advance; (2)the supervise. applies a sequence of speculative controls to the plant with candidate controllers just after the start of control or just after the detection of a change in the environmental condition. Each candidate controller can keep the system stable during one-step period of the speculative control and the most appropriate candidate controller for the environmental condition to which the system is exposed can be selected before the last trial of speculative control step comes to an end. We proposed a construction method of adaptive multiple control system without any knowledge of plant dynamics and applied the method to a cart-pole balancing problem and a vehicle anti skid braking system. In real applications, as we can often easily obtain a piece of knowledge upon plant dynamics beforehand, we intend to extend the method such that multiple control systems can be efficiently designed using the knowledge. We apply the new idea to the cart-pole balancing problem with variable length of the pole. The simulation experiments lead us to the conclusion that the new attempt can reduce the manpower to design the candidate controllers for adaptive multiple control systems.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2000.11a
• /
• pp.163-167
• /
• 2000

The eligibility is used to solve the credit-assignment problem which is one of important problems in reinforcement learning. Conventional eligibilities which are accumulating eligibility and replacing eligibility make ineffective use of rewards acquired in learning process. Because only an executed action in a visited state is learned by these eligibilities. Thus, we propose a new eligibility, called the weighted eligibility with which not only an executed action but also neighboring actions in a visited state are to be learned. The fuzzy Q-learning algorithm using proposed eligibility is applied to a cart-pole balancing problem, which shows improvement of learning speed.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.531-531
• /
• 2000

The challenge is to control unstable nonlinear dynamic systems using only sparse feedback from the environment concerning its performance. The design of such controllers can be achieved by evolving neural networks. An evolutionary approach to train neural networks in realtime is proposed. Evolutionary strategies adapt the weights of neural networks and the threshold values of neuron's synapses. The proposed method has been successfully implemented for pole balancing problem.