Papers by Keyword: Elman Network

Abstract: Hydraulic servo system is highly nonlinear. Building an accurate model of the system and predicting its remaining life are difficult. Thus, this study focuses on the prediction of the Hydraulic servo System based on Support vector regression (SVR). Elman neural network is utilized to build an observer to estimate the normal state output. The residual that contains a large amount of fault information is obtained, by calculating the difference between the estimated and actual values. Then we defined degradation index (DI) value which reflect the health of the system to normalize the residual. Lastly, a prediction model based on SVR established. The algorithm is verified by experiment.

Abstract: With the development of digital campus network users, web users exhibit scale up, the campus network users to use different computer level each are not identical, uneven, a potential threat to network is more serious, campus network security has become an urgent need to solve the problem. In this paper, based on the neural network, the concept of Elman memory, and proposed an improved algorithm of Elman neural network, the realization of network intrusion detection. The experimental results show that, the algorithm can effectively improve the accuracy of network intrusion detection algorithm.

Abstract: In order to quickly determine the performance of a transient voltage suppressor (TVS), improve time domain identification capability of Elman network, the simulation of electromagnetic pulse (EMP) inject effects based on improved Elman network is proposed. Derivation proved that improved Elman network trained by standard BP algorithm has a similar form with the basic Elman network trained dynamic BP algorithm. We establish and improve its Elman network predictive modeling based on the measured parameters of TVS and then demonstrate that improved Elman network has the characteristics of quick speed, high precision, good performance and strong generalization ability, and broad use of prospects.

Abstract: Aiming at the problem of modeling and compensation of the fiber optic gyroscope (FOG) drift caused by temperature, a novel compensation method for FOG temperature drift based on transformed unscented Kalman filter (TUKF) is proposed. Elman network with faster convergence speed is used to modeling and TUKF algorithm is adopted to train the weights of Elman network, which effectively solves the problem of numerical instability. The results prove that the proposed method has higher precision compared with Elman network and IUKF network models. By using the TUKF algorithm, the root mean square errors (RMSE) are improved by 60%  in temperature rise period and 50.5% in fall period.

Abstract: In order to improve the modeling ability for nonlinear system, an Elman modeling method based on Particle Swarm Optimization (PSO) algorithm is proposed. It uses PSO algorithm to optimize the parameters of Elman network. The simulation result shows that the proposed hybrid method combined Elman with PSO algorithm has a good modeling performance with fast training rate for complex nonlinear system.

Abstract: This paper presents Feedforward Neural network (FFNN) and Elman network controllers to control the maximum power point tracking (MPPT) of photovoltaic (PV). MPPT is a method used to extract the maximum available power from photovoltaic module by designs them to operate efficiently. Thus, cell temperatures and solar irradiances are two critical variable factors to determine PV output powers. The performances of the controller is analyzed in four conditions which are i) constant irradiation and temperature, ii) constant irradiation and variable temperature, iii) constant temperature and variable irradiation and iv) variable temperature and irradiation. The proposed systems are simulated by using MATLAB-SIMULINK. Based on the results, FFNN controller has shown the better performance compare to the Elman network controller during partial shading conditions.

Abstract: According to pipe racking system exist nonlinear characteristics and in order to get smooth velocity of the racking in moving process. This paper structures a new kind of fuzzy neural network PID which identifies the target model and also provides a non-linear relation model for dynamic programming. In addition, by adopting robust feedback controller, the stability of the closed-loop system and satisfactory control results in initial stage of fuzzy neural network learning are also guaranteed. And we analyze the error response curve of sine signal tracking, the experimental results show that the improved fuzzy neural network PID controller has a higher control performance. The control method has fast response speed, less overshoot and error, strong robust and can meet the requirements of the nonlinear system.

Abstract: In order to improve work stability and measurement accuracy of drilling inclinometer, and overcome the poor stability of Elman networks and lower compensation precision of genetic Elman neural networks, we combined ant colony algorithm and neural networks, using the Adaptive Ant Colony Algorithm that its pheromone evaporation factorand pheromone update strategy adjust adaptively to optimize Elman neural network weights and thresholds, and applied it to drilling inclinometer sensor compensation. Simulation results show that the compensation effect of adaptive ant colony Elman neural networks is better than that of Elman networks and genetic Elman networks, the compensation accuracy is 10^-8.

Abstract: A dynamic size control model during cylindrical grinding is built. The model consists of Elman neural network, fuzzy control subsystem and deformation optimal adaptive control subsystem. To improve the size prediction accuracy, the first and the second derivative of the actual amount removed from the workpiece are added into the Elman network input; To self-adapt and adjust the quantification factor and scale factor in the fuzzy control, the flexible factor is introduced to the fuzzy control model. Simulation and experiment verify that the developed prediction control model is feasible and has high prediction and control precision.

Abstract: A dynamic intelligent prediction control system is built in slender cylindrical grinding. Elman network is used in the dynamic size prediction control model, and the first and the second derivative of the actual amount removed from the workpiece are added into the network input, which can greatly improve the size dynamic prediction accuracy. Moreover, a surface roughness equation with vibration data is proposed. Based the equation, the surface roughness dynamic fuzzy neural network prediction subsystem is built. Experiment verifies that the developed prediction control system is feasible and has high prediction and control accuracy.