BP Neural Network Training Research Articles

Application of BP Neural Network Based on Immune Particle Swarm Optimization for Fault Diagnosis of Power Transformer

Cloing and hypermutation of immune theory were used in optimization on particle swarm optimization (PSO), an immune particle swarm optimization (IPSO) algorithm was proposed , which overcome the problem of premature convergence on PSO. IPSO was used in BP Neural Network training to overcome slow convergence speed and easily getting into local dinky value of gradient descent algorithm. BP Neural Network trained by IPSO was used to fault diagnosis of power transformer, it has high accuracy after experimental verification and to meet the power transformer diagnosis engineering requirements.

Reach on Micro-Motor Acoustics Fault Diagnosis Based on Loose Wavelet Neural Network

Nicro-motor voice signal contains abundant running status information as well as vibration signal, aiming at the problem that it is difficult to obtain vibration signal in the production line of micro-motor, this paper proposes a micro-motor acoustic fault diagnosis methods based on loose wavelet neural network. Wavelet packet decomposition and reconstruction algorithm is utilized to extract micro-motor voice signals in each frequency band energy as the characteristic parameters of fault characteristic parameter samples will input to improve the BP neural network for training, build up the fault type of classifier, the realization of fault intelligent diagnosis. Application results show that a reasonable design of neural network has strong ability of fault identification; use loose micro-motor acoustic wavelet neural network fault diagnosis is feasible.

The Application of BP Neural Network Learning Algorithm Based on the Particle Swarm Optimization

The development of real estate has been affected by various social factors, including economic factors. BP neural network can more accurately forecast the trend of real estate industry according to economic development indicators. But BP neural network is slow convergence in the training process, and easily falls into local optimum. The BP neural network learning algorithm based on the particle swarm optimization (PSO) optimizes the weights and thresholds of the network by PSO algorithm, then to train BP neural network. The experimental results show that the performance of this new algorithm is better than BP neural network, but also has good convergence.

A Method for Damage Detection Based on Correlation Characteristic of Acceleration Response

A damage detection method using BP neural network based on a novel damage index, the correlation characteristic of the acceleration response, is proposed, and is evaluated through the FEM simulation and experiment verification. On the basis of achievements in existence, the feasibility of using the correlation characteristic as damage index is validated theoretically. The damage detection for a simple-supported beam using the proposed method was FEM simulated. The results showed that the trained BP neural network can correctly detect the location and extent of damages in both single damage case and multi-damage case. A model test of a reinforced concrete simple-supported beam was performed to verify the validity and efficiency of the damage detection method. From the results of the model test, it is shown that the trained BP neural network can correctly locate the damage mostly detect the extent of damage. A conclusion is given that the novel damage detection method using the correlation characteristic of the acceleration response as damage index is feasible and efficient.

Research on the Optimized Algorithms on Neural Network

In order to overcome the disadvantage of neural networks that their structure and parameters were decided stochastically or by one’s experience, an improved BP neural network training algorithm based on genetic algorithm was proposed.In this paper,genetic algorithms and simulated annealing algorithm that optimizes neural network is proposed which is used to scale the fitness function and select the proper operation according to the expected value in the course of optimization,and the weights and thresholds of the neural network is optimized. This method is applied to the stock prediction system.The experimental results show that the proposed approach have high accuracy,strong stability and improved confidence.

Based on the EMD and PSO-BP Neural Network of Short-Term Load Forecasting

This paper proposes that based on the EMD and PSO-BP neural network of short-term load forecasting. This method will be automatically historical load sequence into several independent intrinsic mode functions (IMF) by using EMD. As the BP neural network training for a long time and easy to fall into local minimum of the shortcomings, we use genetic algorithm to optimize BP neural network algorithm replaces the traditional BP algorithm. Finally, we use the BP neural network optimized separately for each IMF component for training and prediction. We should add the component to the final prediction forecast. This method has higher precision than the EMD-BP model prediction by Simulation results show.

Defects Rapid Identification of Marking Region Based on Binarization and BP Neural Network

The status that domestic technology forming the high-pressure gas bottles surface leading to the poor marking quality is analyzed. The rapid image processing method based on Binarization is presented to identify the defects position and profile from the marking region quickly, and it is designed with four steps: CCD capture, pixel enhancement, edge identification and feature extractions. By the statistical analysis of the project practice, the defects is defined as four typical types in shape, and then through the BP neural network training to identify the defects type effectively and driving the machine pre-set coping strategies to make appropriate responses automatically without manual interaction. The final test shows that the automatic high-efficiency marking defects identification is achieved

Prediction of Concrete Strength Based on BP Neural Network

Based on BP neural network, this paper had a prediction analysis on concrete strength at the 28th day. Taking the influences of application amount of cement, application amount of scoria, application amount of fly-ash, water-ash ratio into account, the prediction model of concrete strength at the 28th day based on BP neural network was obtained. It was found that, the average value of absolute value for the relative error of fitting value of concrete strength at the 28th day compared with the observed value for 43 groups of independent variables training BP neural network model was 0.64372%; And the relative error of prediction value of concrete strength at the 28th day was -2.0392%~0.964668% compared with the observed value for 5 groups of independent variables validating BP neural network model, and the average value of absolute value for the relative error was 0.95467%. The following conclusion can be drawn that, the prediction model is accurate and credible, and the BP neural network is a excellent method, and it is worthy to spread its application in the prediction analysis of concrete strength at the 28th day.

Quantum hybrid genetic algorithm based on simulated annealing and its application

Quantum genetic algorithm (QGA) is firstly improved for numerical optimization with real coding, where populations are updated by a simple rotation method which inspires a real quantum genetic algorithm (RQGA), then simulated annealing (SA) is reasonably introduced in the optimizing process of RQGA, and a hybrid quantum genetic algorithm (HQGA) is presented, which could not only effectively avoid the premature phenomenon but also accelerate the search efficiency under the introduction of SA. Besides HQGA is applied to numerical optimization and the training of BP neural network, and through a comparison among QGA, RQGA and HQGA, it is obviously shown that HQGA performs better on running speed and optimizing capability. Key words: Quantum genetic algorithm simulated annealing, hybrid algorithm, real coding.

Prediction of Ultimate Bearing Capacity of Prestressed Pipe Pile Based on BP Neural Network

Based on BP neural network, this paper had a prediction on ultimate bearing capacity of prestressed pipe pile. Taking pile diameter, effective pile length, ultimate average value of friction standard value, ultimate average value of end resistance standard value as influences factors, the prediction model of pile bearing capacity based on BP neural network was obtained. It was found that, the average value of absolute value for the relative error of fitting value of pile bearing capacity compared with the observed value for 70 groups of independent variables training BP neural network model was 3.1498%; And the average value of absolute value for the relative error of prediction value of pile bearing capacity compared with the observed value for 10 groups of independent variables validating BP neural network model was 3.50126% whose precision was better than ANFIS’5.32293%. The following conclusion can be drawn that, the prediction model of ultimate bearing capacity of prestressed pipe pile based on BP neural network is feasible.

Microwave Power Real-Time Soft-Measuring Based on Improved BP Neural Network

There are many faults in microwave power measurement in the working spots. In this paper, the novel technique of microwave power measurement and Schottky detector and the technique of soft-measuring based on improved BP neural network are presented. Making use of directional coupler, the part of power of microwave signal in transmission line was taken in the certain proportion. Then, the microwave power meter indicated the magnitude of power. The microwave power measurement system was composed of tuneable attenuator, directional coupler, matching load, peak-peak value detector, oscillograph, thermocouple, soft-measuring algorithm and so on. The low microwave power meter obtained the sample data of training BP neuron network. Then, the output of trained BP neural network may represent the result of high microwave power measurement. The experiments and applications in practice project prove that this new method has many advantages.

Matlab for Forecasting of Energy Consumption Based on BP Neural Network

The paper used Matlab to program for single-layer BP neural network, selected the annual energy consumption data as the training sample and inspection sample to train BP neural network, then, predict the future China's energy consumption quantity.

Mountain ground movement prediction caused by mining based on BP-neural network

Six main influencing factors: slope, aspect, distance, angle, angle of coal seam, and the ratio of depth and thickness, were selected by Grey correlation theory and Grey relational analysis procedure programmed by the MATLAB software package to select the surface movement and deformation parameters. On this basis, the paper built a BP neural network model that takes the six main influencing factors as input data and corresponding value of ground subsidence as output data. Ground subsidence of the 3406 mining face in Haoyu Coal was predicted by the trained BP neural network. By comparing the prediction and the practices, the research shows that it is feasible to use the BP neural network to predict mountain mining subsidence.

Research of PID Control Algorithm Based on Neural Network

Through the research on the existing BP network and PID control technology, for the defects of general BP algorithms such as slow convergence and easy to fall into local minimum, this paper presents an improved BP algorithm, and based on this algorithm proposes an implementation scheme of PID control system, the results show that the scheme can not only improve the convergence speed of the algorithm in the training process, and the trained BP neural network also has strong adaptive and self-learning capability, which further improves the performance of PID controllers. © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of ESEP 2011

Improved Wavelet Neural Network Based on Hybrid Genetic Algorithm Applicationin on Fault Diagnosis of Railway Rolling Bearing

The method of improved wavelet transform neural network based on hybrid GA(genetic algorithm) is presented to diagnose rolling bearings faults in this paper. Genetic Artificial Neural Networks(GA-ANN) overcomes BP neural network’s fault of slow convergence, long hours of training, and falling into the local minimum point. And First, the signal is processed through the wavelet deoising, Then, three— layer wavelet packet is adopted to decompose the denoising signal of rolling beatings, and constructs the wavelet packet energy eigenvector, then takes those wavelet packet energy eigenvectors as fault samples to train BP neural network. Genetic algorithm is used to optimize the training process of BP network. The experimental results show that the optimized BP network by genetic algorithm can diagnose bearing faults, and it is superior to the BP network without optimization, the method has fair prospects of application for the rotary machine fault diagnosis.

Uniform Wearing Optimum Design of Extrusion Die Profile

Die wear is regarded as a crucial factor which affects die life and quality of products. In hot extrusion process, finite-element method (FEM), BP neural network and genetic algorithm were combined together to optimize extrusion die profile which yielded more uniform wear depth distribution on die profile. A method of B-spline function interpolation was used to describe extrusion die profile. The temperature, pressure and velocity field of nodes that lied on extrusion die profile were gained by FEM simulation. Wear depth of extrusion die profile was calculated by modified Archard theory. The results were used to train BP neural network, so that nonlinear mapping relations between reference points of die profile and wear depth were obtained. In order to gain uniform wear depth, genetic algorithm was applied to optimize extrusion die profile. Optimum result, compared with common conical die profile, reduced wear depth of extrusion die and improved service life. At the same time, the optimal result accorded with practical conditions.