Xiao-Hua Yu

Structural Damage Detection Using Artificial Neural Networks and Wavelet Transform

Arthur Shi et al. — Wed, 30 Jan 2013 11:30:50 PST

With the ever-increasing demand for the safety and functionality of civil infrastructures, structure health monitoring (SHM) has now become more and more important.

Recent developments in computational intelligence and digital signal processing offer great potentials to develop a more efficient, reliable, and robust structure damage identification system. In this paper, the application of artificial neural networks and wavelet analysis is investigated to develop an intelligent and adaptive structural damage detection system. The proposed approach is tested on an IASC (International Association for Structural Control)-ASCE (American Society of Civil Engineers) SHM benchmark problem. Satisfactory computer simulation results are obtained.

Traffic Signal Optimization Using Ant Colony Algorithm

David Renfrew et al. — Wed, 30 Jan 2013 11:30:48 PST

Traffic signal control is an effective way to improve the efficiency of traffic networks and reduce users’ delays. Ant Colony Optimization (ACO) is a meta-heuristic algorithm based on the behavior of ant colonies searching for food. ACO has successfully been employed to solve many complicated combinatorial optimization problems and its stochastic and decentralized nature fits well with traffic networks. This research investigates the application of the ant colony algorithm to minimize user delay at traffic intersections. Various ACO algorithms are discussed and a rolling horizon approach is also employed to achieve real-time adaptive control. Computer simulation results show that this new approach outperforms conventional fully actuated control, especially under the condition of high traffic demand.

Neural Network Based Edge Detection for Automated Medical Diagnosis

Dingran Lu et al. — Mon, 17 Oct 2011 15:29:04 PDT

Edge detection is an important but rather difficult task in image processing and analysis. In this research, artificial neural networks are employed for edge detection based on its adaptive learning and nonlinear mapping properties. Fuzzy sets are introduced during the training phase to improve the generalization ability of neural networks. The application of the proposed neural network approach to the edge detection of medical images for automated bladder cancer diagnosis is also investigated. Successful computer simulation results are obtained.

Modeling of a Gyro-stabilized Helicopter Camera System Using Artificial Neural Networks

Nicholas Layshot et al. — Mon, 17 Oct 2011 15:29:00 PDT

On-board gimbal systems for camera stabilization in helicopters are typically based on linear models. Such models, however, are inaccurate due to system nonlinearities and complexities. As an alternative approach, artificial neural networks can provide a more accurate model of the gimbal system based on their non-linear mapping and generalization capabilities. This paper investigates the applications of artificial neural networks to model the inertial characteristics (on the azimuth axis) of the inner gimbal in a gyro-stabilized multi-gimbal system. The neural network is trained with time-domain data obtained from gyro rate sensors of an actual camera system. The network performance is evaluated and compared with measurement data and a traditional model. Computer simulation results show the neural network model fits well with the measurement data and significantly outperforms the traditional model.

Facility Power Usage Modeling and Short Term Prediction with Artificial Neural Networks

Sunny Wan et al. — Mon, 20 Dec 2010 13:54:18 PST

Residential and commercial buildings accounted for about 68% of the total U.S. electricity consumption in 2002. Improving the energy efficiency of buildings can save energy, reduce cost, and protect the global environment. In this research, artificial neural network is employed to model and predict the facility power usage of campus buildings. The prediction is based on the building power usage history and weather conditions such as temperature, humidity, wind speed, etc. Different neural network configurations are discussed; satisfactory computer simulation results are obtained and presented.

Bioinformatics Data Mining Using Artificial Immune Systems and Neural Networks

Shane Dixon et al. — Mon, 20 Dec 2010 13:54:16 PST

Bioinformatics is a data-intensive field of research and development. The purpose of bioinformatics data mining is to discover the relationships and patterns in large databases to provide useful information for biomedical analysis and diagnosis. In this research, algorithms based on artificial immune systems (AIS) and artificial neural networks (ANN) are employed for bioinformatics data mining. Three different variations of the real-valued negative selection algorithm and a multi-layer feedforward neural network model are discussed, tested and compared via computer simulations. It is shown that the ANN model yields the best overall result while the AIS algorithm is advantageous when only the “normal” (or “self”) data is available.

Ant Colony Optimization Algorithm for Robot Path Planning

Michael Brand et al. — Mon, 20 Dec 2010 13:54:15 PST

Path planning is an essential task for the navigation and motion control of autonomous robot manipulators. This NP-complete problem is difficult to solve, especially in a dynamic environment where the optimal path needs to be rerouted in real-time when a new obstacle appears. The ACO (Ant Colony Optimization) algorithm is an optimization technique based on swarm intelligence. This paper investigates the application of ACO to robot path planning in a dynamic environment. Two different pheromone re-initialization schemes are compared and computer simulation results are presented.

Design and Implementation of a Neural Network Controller for Real-Time Adaptive Voltage Regulation

Xiao-Hua Yu et al. — Mon, 25 Jan 2010 09:53:43 PST

An adaptive controller based on multi-layer feed-forward neural network is developed for real-time voltage regulation of a class of PSFB (phase-shifted full-bridge) DC–DC converters. The controller has the unique advantages of nonlinear mapping and adaptive learning, and performs well over a wide range of input voltages and output load currents. The controller is implemented and tested in hardware using a DSP (digital signal processor) board. Experimental results show that it outperforms conventional controllers in both line regulation and load regulation.

Traffic Signal Control with Swarm Intelligence

David Renfrew et al. — Mon, 25 Jan 2010 09:53:42 PST

Traffic signal control is an effective way to regulate traffic flow to avoid conflict and reduce congestion. The ACO (Ant Colony Optimization) algorithm is an optimization technique based on swarm intelligence. This research investigates the application of ACO to traffic signal control problem. The decentralized, collective, stochastic, and self-organization properties of this algorithm fit well with the nature of traffic networks. Computer simulation results show that this method outperforms the conventional fully actuated control, especially under the condition of high traffic demand.

Electrocardiogram (ECG) signal modeling and noise reduction using wavelet neural networks

Suranai Poungponsri et al. — Tue, 15 Dec 2009 17:03:11 PST

Electrocardiogram (ECG) signal has been widely used in cardiac pathology to detect heart disease. In this paper, wavelet neural network (WNN) is studied for ECG signal modeling and noise reduction. WNN combines the multi-resolution nature of wavelets and the adaptive learning ability of artificial neural networks, and is trained by a hybrid algorithm that includes the adaptive diversity learning particle swarm optimization (ADLPSO) and the gradient descent optimization. Computer simulation results demonstrate this proposed approach can successfully model the ECG signal and remove high-frequency noise.

Actuator Fault Compensation for a Helicopter Model

Xiao-Hua Yu — Fri, 19 Jun 2009 14:45:17 PDT

A fault-tolerant system is the one that can continue its operation without significant impact on performance in the presence of hardware and/or software errors. In this paper, the design of a fault-tolerant flight controller to control UH-60 helicopter is investigated. A 9th-order state space representation of the helicopter model operating at the forward mode with 80 knots is presented; then a fault-tolerant optimal feedback controller is designed and tested.

Optimization of network signal timing

Xiao-Hua Yu et al. — Fri, 19 Jun 2009 14:45:17 PDT

A typical urban traffic network is a complicated large-scale stochastic system which consists of many interconnected signalized traffic intersections. This paper develops a decentralized real-time adaptive control strategy for the traffic networks based on Markov decision theory. Computer simulation results of this new approach on a five intersection traffic network indicate significant improvement over the traditional fully actuated control algorithm.

A Neuromorphic Controller for a Distillation Column

Xiao-Hua Yu — Fri, 19 Jun 2009 14:45:16 PDT

This paper investigates the design of a neural network based controller to control the concentration of the overhead and bottom product in the model of a distillation column. Satisfactory computer simulation results of this approach are obtained.

An Integrated Model for Signalized Traffic Intersection Control

Xiao-Hua Yu — Fri, 19 Jun 2009 14:45:15 PDT

Traffic signal control is an effective way to regulate traffic flow to avoid conflict and reduce congestions. This research investigates a real-time traffic signal control system that integrates a traffic flow prediction model and an adaptive control scheme based on dynamic programming with rolling horizon. The proposed approach estimates the parameter of the arriving traffic flow at the intersection, predicts the state transition probabilities, and then formulates the traffic signal control problem as a decision-making problem of a stochastic system. Two different traffic arrival patterns are considered, including the normal distribution and the Poission distribution.

A neural network pruning algorithm with embedded gradient-conjugate training for the identification of large flexible space structures

Xiao-Hua Yu — Fri, 19 Jun 2009 14:45:15 PDT

The choice of network dimension is a fundamental issue in the design of artificial neural networks. A larger neural network is powerful for solving problems while a smaller neural network is always advantageous in real-time environment where speed is crucial. In this paper, a network pruning algorithm with embedded gradient-conjugate training is investigated and applied to the identification of a large flexible space structure. Computer simulation results show that this approach can dramatically reduce the size of neural network while maintaining compatible identification accuracy.

Arrival rate identification for a class of traffic signal control problem

Xiao-Hua Yu et al. — Fri, 19 Jun 2009 14:45:14 PDT

Setting signals at traffic intersections to reduce congestion is one of the most challenging problems in traffic management. To find the optimal control strategy, specific information of the traffic flows passing through intersections must be provided in advance. It has been shown that the Markovian decision control theory can be successfully applied to solve traffic signal control problems, when both the state transition probabilities and the one-step reward function are known. In this paper, an online parameter identification algorithm is investigated for adaptive Markovian decision control at an isolated traffic intersection with unknown vehicle arrival rates. The authors give a brief introduction to Markovian control processes and a maximum likelihood estimation algorithm, and discuss the traffic dynamic equations and the adaptive Markovian decision control model for an isolated traffic intersection.. The proposed algorithm is then tested by computer simulation and the result is shown.

A new control strategy for a signalized traffic intersection

Allen R. Stubberud et al. — Fri, 19 Jun 2009 14:45:14 PDT

A traffic control problem can be formulated as a decision-making problem for a stochastic dynamic system. Optimal traffic signal settings at intersections can minimize the vehicle delay time or the queue length at a stop line. In this paper, a new adaptive control strategy for signalized intersections is developed and tested by simulation. The simulation results show significant improvement over the traditional fully actuated control algorithm, especially for the case of high volume traffic demand.

Markovian decision control for traffic signal systems

Xiao-Hua Yu et al. — Fri, 19 Jun 2009 14:45:13 PDT

Setting signals at intersections to minimize the queue length and vehicle delay time is a key goal in traffic management. In this paper, a new control strategy for a signalized traffic intersection is developed by applying Markovian decision control theory. Statistical analysis of simulation results with different arrival rates indicate the excellent potential of this approach.

Improving DC Power Supply Efficiency with Neural Network Controller

Weiming Li et al. — Thu, 18 Jun 2009 08:20:23 PDT

DC-DC converters can be found in almost every power electronics device. To improve the efficiency and controller response of a DC-DC converter to dynamical ~stem changes, neural network has been chosen as an alternative to classic methods. However, no prior work has been done in the neural network approach for control of a PSFB (phase-Shifted Full-Bridge) converter yet. In this research, a multi-layer feedforward neural network controller is proposed. The neural network based controller has the advantage of adaptive learning ability, and can work under the situation when the input voltage and load current fluctuate. Levenberg-Marquardt backpropagation training algorithm is used in computer simulation. The neural controller is then implemented on hardware using a DSP (digital signal processor). Satisfactory experimental results are obtained.

Neural Network Dimension Selection for Dynamical System Identification

Devin Sabo et al. — Thu, 18 Jun 2009 08:20:22 PDT

Choosing an appropriate size of a network is an important issue for any neural network applications. The common practice is to start with an “over-sized” network, then gradually reduces its size to find the optimal solution. In this paper, a new hybrid neural network pruning algorithm for multi-layer feedforward neural networks is investigated. Computer simulation results on system identification and pattern classification problems show this algorithm can significantly reduce the network dimension while still maintaining satisfactory identification and classification accuracy.