Papers by Keyword: Genetic Neural Network

Abstract: Aimed at the fault diagnosis and prediction of automobile engine, firstly designed a framework structure of automobile engine fault diagnosis and prediction system, and built a hardware platform; Secondly adopted the genetic algorithm neural network to fault prediction and diagnosis reasoning; Finally after analyzing automobile exhaust components, engine vibration, engine abnormal sound parameters, inferred the appeared and impending fault of automobile then made the tips for users on the screen. The results show that the performance of system is well, the accuracy of diagnosis and prediction is 95% in different conditions of experiment and debugging.

Abstract: Based on 33-year typhoon information of South China Sea (SCS) in 1980-2012 and NCEP/NCAR reanalysis data, taking Climatology and Persistence (CLIPER) and earlier physical quantities predictors selected by Stepwise Regression (SWR) and Multidimensional Scaling (MDS) methods as model inputs, the Genetic Algorithm-Artificial Neural Network (GA-ANN) forecast model was built for typhoon gale. The forecast verification results for independent samples in MDS-GA-ANN model show that mean absolute error of 24h forecast for wind velocities at 36 grid points around typhoon centers from July to September is 1.6m/s. Using the same samples, the prediction results of MDS-GA-ANN models for independent samples were compared with that of traditional SWR models. Taking July as example, prediction abilities for 29 MDS-GA-ANN models (81%) among 36 grid points around typhoon centers are superior to that of SWR models; only 2 grid points of MDS-GA-ANN models are worse than that of SWR models (6%). Therefore, prediction ability for most of 36 grid points using MDS-GA-ANN models is superior to that of SWR models and can meet business requirements of meteorological stations at present.

Abstract: This Paper mainly studies the application of the BP neural network in the research of speech recognition. BP neural network can get higher identification precision, but its training speed is very low. a new recognizing algorithm based on BP algorithm by combining with the good effect method in ANN which named genetic algorithm (GA) was proposed and used to improve the BP neural network . Experiment results show that the training speed can be accelerated by the method and the recognition performance is also promoted.

Abstract: With the increasing height of building, high-rise building structure selection becomes more and more important. This paper presents the application of genetic neural network method to study high-rise building structure selection and uses the MATLAB neural network toolbox with a combination of genetic algorithm toolbox to develop a genetic neural network expert system for high-rise building structure selection to make the selection process simple.

Abstract: In this paper, genetic neural network is applied to forecast the short-term traffic flow and traffic guidance. Because of the factors of time correlation and spatial correlation, we construct the short-term traffic flow forecasting model using back-propagation neural network that has the function of arbitrary nonlinear function approximation. In order to find proper initial values of the neural network weights and threshold quickly, a combination of neural network prediction method is presented. This method utilizes genetic algorithm to choose the initial weights and threshold, and uses L-M algorithm to train sample, which can enhance the global convergence rate. Trained network is used for short-term traffic flow prediction with mean square error as the forecast performance evaluation. The results show that the performance of genetic neural network is better than a separate BP neural network for short-term traffic flow prediction.

Abstract: Based on the former calibration data of error coefficient of inertial navigation instrumentation and environment parameters, a three-layer BP neural network is founded to predict the calibration value in current circumstances. To confirm the structure of neural network, genetic algorithm is used to seek the optimal solution, and hence the concept of genetic neural network is introduced. Also since the calibration data has small-sample characteristics; Bayes regularization method is adopted to improve the network generalization ability and predicted performance. In the end the simulated results show that it is reasonable and effective to accomplish the prediction in genetic neural network.

Abstract: In order to estimate the urban economic losses after earthquakes, builds the prediction model based on grey relation and genetic neural network. According to the grey correlation analysis, screens out the major factor effecting the earthquake losses as input vectors of neural network first, adopts genetic algorithm to optimize the initial weights and bias of BP network and then, solves the problem that the traditional neural network has slow convergence velocity and can not easy to get the global optimum. According to the example, we can get the model that has good forecast accuracy and convergence rate and checks the rationality and effectiveness at last.

Abstract: A control method based on genetic neural network is presented to deal with the nonlinear object of the high-power PS-FB-ZVS PWM DC/DC converter. The control system optimizes the initial weight of the BP neural network and PID parameters tuning on line utilizing the genetic algorithm, which directly controls the object in closed-loop and has solved the problem that the controller network initial weight coefficient influences the control effect, thus, the optimal dynamic and steady state performance of the system is ensured. In the MATLAB environment, the control systems consisting of different controllers are simulated, and the output voltage and output current waveforms are obtained when the system is loaded by experiment. The results show that the controller has strong robustness, fast response speed, and small output voltage fluctuations with load changes.

Abstract: Compared to the neural network BP algorithm, the optimized model of genetic neural network based on the genetic algorithm has a more close assessed result to the expected one and smaller relative mistakes. Practical applications show that the new assess way of enterprise intellectual capital is rational and accessible, and it provides as an important tool to enterprise for intellectual capital decision.

Abstract: The inversion method combining the genetic neural network and the discrete element simulation of triaxial tests is firstly described for determining the discrete element model parameters of the conditioned soil. The purpose is to make the error of the simulation curves and the laboratory curves of the triaxial test minimum. The solve approach is the parameters identification based on the genetic neural network. The network training sample is provided by the discrete element simulation. The input sample is the simulation curves of triaxial test, and the output sample is the model parameters. The laboratory triaxial test curves of the conditioned soil are used to determine its model parameters. The simulation curves calculated with the inversed parameters match the laboratory curves well, which illustrate that the discrete element model can accurately predict the deformation characteristics and flow patterns of conditioned soils.