Papers by Keyword: BP Neural Network

Abstract: Aiming at the shortcomings of large volume, high cost and long detection cycle of traditional solid propellant aging detection methods, a solid propellant aging detection method based on impedance spectroscopy is proposed. Firstly, the internal impedance of the solid propellant changes after aging, and a portable solid propellant impedance spectrum acquisition system based on impedance spectroscopy is designed based on the principle of electrochemical impedance spectroscopy, and the real and imaginary parts of the impedance spectrum are obtained. Secondly, in order to reduce the data dimension of the classification algorithm, the KPCA (Nuclear Principal Component Analysis) feature extraction algorithm is used to extract the impedance spectrum features of the solid propellant. Then, according to the impedance spectrum characteristics, the BP neural network is used for classification training, and the correspondence between the impedance spectrum and the aging time is obtained. Finally, the feasibility and effectiveness of the proposed method are verified on the physical platform, and the results show that the proposed method has the advantages of high precision and accurate classification, and can effectively predict the aging degree of solid propellant.

Abstract: Through the long-term load creep test of CE131 geonet and SD L25 retaining wall foundation, which are widely used in reinforced earth engineering, a large number of experimental data are obtained. On this basis, the least-squares and BP neural network are used to predict its creep variables. The principle of least squares is to find a curve in the curve family to fit the experimental data. From the sum of the squared errors σ = 0. 001 16, the fitting accuracy is higher. The BP neural network has adaptive learning and memory capabilities, especially the three-layer BP neural network model. The maximum error between the predicted value and the actual value is 0.91%, which is a lot better than the error of the least square 3.4%. This method Found a new way for creep prediction.

Abstract: The paper will use BP neural network analysis method to study the thermal conductivity of bentonite and its influencing factors as a system. The heat conduction of bentonite was used as the output of the system, and its influencing factors were used as the system input to simulate. The corresponding simulation model was established to verify the thermal conductivity data. In addition, the analysis of the mechanical properties of the bentonite-PVA fiber cement-based composite materials for construction has not only laid a theoretical and realistic foundation for the prediction and simulation of the thermal conductivity of bentonite, but also has opened up the mechanical properties of the bentonite-PVA fiber cement-based composite materials a new path.

Abstract: Since the rapid development of the construction industry, the production of construction waste has also multiplied, and the construction waste has caused tremendous pressure on the environment. Therefore, the main research of this subject is that the waste concrete is formed into a recycled material after a certain treatment--concrete powder. And the cement in the dry-mixed mortar is replaced by 0-30% concrete powder. The compressive strength of recycled concrete powder under different dosages was tested by experimental method. The compressive strength is then applied to the artificial neural network to establish a predictive model. Taking time as a variable, the feasibility and the best dosage of the 28-day compressive strength method for the 3d compressive strength during the test are discussed. In order to reduce the test cycle, improve work efficiency, and ultimately achieve the purpose of improving construction waste utilization.

Abstract: As structural health problems are becoming more and more important, a neural networks model is introduced to detect structural damage. The structural modal flexibility matrix can be accurately constructed by the natural frequency and modal information. All elements of changes in the modal flexibility matrix are looked on as inputs of the networks. Damage locations and extents are both considered with different outputs in the present study. A simply supported truss structure is studied with different damage cases. To localize damage, one case is chosen as location input/target pairs to train the present BP network model. But to identify damage extent, two cases are chosen as extent pairs to train. Although modals of BP neural networks with different outputs are presented for different damage detecting schemes, it is more difficult to ascertain damage extent than location. The results indicate that the present BP neural network modal can effectively detect damage of structures with changes in the flexibility matrix between the intact and the damaged cases.

Abstract: In the paper, the Metal Magnetic Memory Testing signal of pipeline crack is extracted. The BP neural network is constructed and trained. The experiment shows that the BP neural network can effectively identify the crack parameters of oil and gas pipeline in quantitative.

Abstract: BP neural network is introduced and applied to identify and diagnose both location and extent of bridge structural damage; static load tests and dynamic calculations are also made on bridge structural damage behind abutment. The key step of this method is to design a reasonably perfect BP network model. According to the current knowledge, three BP neural networks are designed with horizontal displacement rate and inherent frequency rate as damage identification indexes. The neural networks are used to identify the measurement of structure behind abutment and the calculation of damage location and extent, at the same time, they can also be used to compare and analyze the results. The test results show that: taking the two factors (static structural deformation rate and the change rate of natural frequency in dynamic response) as input vector, the BP neural network can accurately identify the damage location and extent, implying a promising perspective for future applications.

Abstract: In this paper, a space vector PWM (SVPWM) control algorithm based on BP neural network is proposed to cope with the complex calculation required in SVPWM through analyzing SVPWM for three phase voltage fed inverter. This method uses the strong nonlinear approximation ability of the BP neural network to fit the eqivalent segment SVPWM modulated wave, modulate with the triangular carrier wave, and then get the control signals of the three phase voltage inverter. A simulation model for PMSM control system was developed by MATLAB/Simulink with the neural network Toolbox. The results show that the BP neural network based SVPWM algorithm makes the motor control system has a smaller current harmonics and torque ripple, and reduce the amount of computation in digital control system with strong guidance.

Abstract: s: According to index selection principle, and considering the process of selection as well as the characteristics of the architecture enterprise, this paper establishes a set of comparatively rational index appraisal system. Based on method of primary component analysis, we obtain the components from credit appraisal index of 24 architecture enterprises. It singles out the GA-BP Neural Network as the credit appraisal method of architecture enterprise, sets up a credit appraisal model of architecture enterprise, and finally verifies the practical and scientific attribute of the model.

Abstract: In this paper, the relationship between microstructure, parameters of cyclic loading and high cycle fatigue property of Ti-6Al-4V alloy was established by artificial neural network (ANN) modeling. The back propagation (BP) neural network and radial basis function (RBF) neural network were established by MATLAB. The input parameters of these models were the primary α volume fraction, primary α size, cyclic loading frequency and stress ratio. The output parameter was high cycle fatigue strength. The neural networks were trained with dataset collected from the literature. The prediction results showed that both of the networks have good generalization ability. In addition, the BP neural network with Levenberg-Merquardt (LM) learning algorithm has better fault tolerance and versatility in dealing with high cycle fatigue property, which is able to predict the high cycle fatigue property with a high accuracy.