Traditional Back Propagation Neural Network Research Articles

CONSTRUCTION OF REGIONAL WEIGHTED MEAN TEMPERATURE MODEL BASED ON OPTIMIZATION BP NEURAL NETWORK

Abstract. The tropospheric weighted mean temperature (Tm) is one of the key characteristic parameters in the troposphere, which plays an important role in the conversion of Zenith Wet Delay (ZWD) to atmospheric Precipitable Water Vapor (PWV). The precision of Global Navigation Satellite System (GNSS) inversion of PWV can be significantly improved with the accurate calculation of Tm. Due to the strong nonlinear mapping ability of Back Propagation (BP) neural network, the algorithm can be used to excavate the law with massive data. In term of the nonlinear and non-stationary characteristics of GNSS precipitable water vapor, in this paper, we proposes a forecast method of GNSS precipitable water vapor based on BP neural network, which can modelling the weighted mean temperature of troposphere. The traditional BP neural network has some shortcomings, such as large amount of calculation, long training time and easy to appear “over-fitting” phenomenon and so on. In order to optimize the deficiency and numerical simulation, the three characteristic values include water vapor pressure, surface pressure and surface temperature provided are selected as input parameters, named as BP_Tm. The optimal initialization parameters of the model were obtained from the 2016 radiosonde data of 89 radiosonde stations in China, and the modeling and accuracy verification were conducted with the 2017 radiosonde data，and the accuracy of the new model was compared with the common regional Tm model. The results show the BP_Tm model has good simulation accuracy, the average deviation is −0.186K, and the root mean square error is 3.144K. When simulating the weighted mean temperature of a single station, the accuracy of the four models to simulate Tm is compared and analyzed, which the BP_Tm model can obtain good accuracy and reflect better stability and reliability.

Parameter Estimation for PMSM based on a Back Propagation Neural Network Optimized by Chaotic Artificial Fish Swarm Algorithm

Permanent Magnet Synchronous Motor(PMSM) control system with strong nonlinearity makes it difficult to accurately identify motor parameters such as stator winding, dq axis inductance, and rotor flux linkage. Aiming at the premature convergence of traditional Back Propagation Neural Network(BPNN) in PMSM motor parameter identification, a new method of PMSM motor parameter identification is proposed. It uses Chaotic Artificial Fish Swarm Algorithm(CAFSA) to optimize the initial weights and thresholds of BPNN, and then strengthens training by BPNN algorithm. Thus, the global optimal network parameters are obtained by using the global optimization of CAFSA and the local search ability of BPNN. The simulation results and experimental data show that the initial value sensitivity of the network model optimized by CAFS-BPNN Algorithm is weak, the parameter setting is robust, and the system stability is good under complex conditions. Compared with other intelligent algorithms, such as RSL and PSO, CAFS-BPNNA has high identification accuracy and fast convergence speed for PMSM motor parameters.

Probabilistic wind power forecasting based on spiking neural network

Accurate and reliable quantification of uncertainty in wind power forecasting is critical to the economic operation and real time control of the electric power and energy system. To this end, this paper proposes a novel direct method for probabilistic wind power forecasting based on spiking neural network. In this method, a new forecasting framework is firstly formulated to simultaneously calculate the coverage probability and sharpness with associated confidence levels. Then, group search optimizer is introduced and re-designed to optimize the parameters of the forecasting framework and directly generate the prediction intervals, so that the prediction reliability and stability are ensured. The main advantage of the proposed probabilistic forecasting method is that it does not involve any distribution assumption of the prediction errors required by most existing forecasting methods. The wind power datasets from real wind farms in Belgium and China are used in the case studies. Traditional back-propagation neural network (BPNN), support vector machine (SVM) and extreme learning machine (ELM) are selected as the benchmarking algorithms. The simulation results show that the average coverage of the proposed method is improved by 72.0%, 54.9% and 51.3% respectively, when compared to BPNN, SVM and ELM. The improvement rates of sharpness index are 43.1%, 28.1% and 21.0%, respectively. These statistical results show that the proposed method outperforms BPNN, SVM and ELM in terms of forecasting accuracy, demonstrating that this method has high practical applications in real power systems.

Discriminative Neural Network for Coronary Heart Disease Detection

Cardiovascular disease is one of the commonest diseases and main causes of death over the world. As the major type of cardiovascular diseases, correct and timely diagnosis of coronary heart disease (CHD) is very essential. Traditional back-propagation (BP) neural network aims to train a multilayer feedforward neural network which transforms data into the feature space to learn good decision boundaries. However, the performance of BP neural network tends to deteriorate when dealing with complexity medical diagnostic tasks. To improve the detection of CHD, this study proposes a discriminative neural network called DNN. DNN explores discriminative information by maximizing the difference of the compactness within each class and separability between different classes. DNN integrates the discriminative information into the framework of BP neural network, and can be easily implemented by the existing neural network software. Experimental results on Z-Alizadeh Sani dataset show that DNN achieves satisfactory performance in sensitivity, specificity, accuracy and receiver operating characteristic (ROC) curve.

Recognition of different Longjing fresh tea varieties using hyperspectral imaging technology and chemometrics

AbstractHyperspectral imaging technology was applied to detect and recognize six different varieties of Longjing fresh tea. The data contained image and spectral information at 370–1042 nm; color and texture features were the foci of the image research. Spectral pre‐processing was performed by multiplicative scatter correction (MSC) and standard normal variate (SNV), and then, we selected the corresponding position variable and vegetation indexes as spectral features. Representative features including the most information were chose by principal component analysis (PCA). A novel back propagation (BP) neural network, with a self‐generated number of hidden layer neurons, was proposed. Using spectral features, image features, and spectral image fusion features as input, three fresh tea recognition models were established: the improved BP neural network, traditional BP neural network, and support vector machine (SVM). Results suggested that the improved BP neural network could promote performance of the model, especially for the spectral pre‐processed data. Mixed‐feature models did better than individual feature models, with 100% accuracy of the predictive set. This study shows that hyperspectral imaging technology can be a potential rapid and nondestructive approach to identify different varieties of Longjing fresh teas.Practical applicationsThis article introduced application of hyperspectral imaging technology to identify Longjing fresh tea of different origins and varieties. Samples were analyzed using spectral and image characteristics. We provided a basis for full utilization of tea characteristics. At the same time, an improved BP neural network, with less calculation complexity and workload than the traditional BP neural network, was proposed. In summary, we outlined a convenient and reliable method for differentiation of Longjing fresh teas. Furthermore, we established a theoretical foundation for development of portable instruments to be used in similar studies.

Economic benefit of shale gas exploitation based on back propagation neural network

Under the influence of COVID-19, the economic benefits of shale gas development are greatly affected. With the large-scale development and utilization of shale gas in China, it is increasingly important to assess the economic impact of shale gas development. Therefore, this paper proposes a method for predicting the production of shale gas reservoirs, and uses back propagation (BP) neural network to nonlinearly fit reservoir reconstruction data to obtain shale gas well production forecasting models. Experiments show that compared with the traditional BP neural network, the proposed method can effectively improve the accuracy and stability of the prediction. There is a nonlinear correlation between reservoir reconstruction data and gas well production, which does not apply to traditional linear prediction methods

A novel Monte Carlo-based neural network model for electricity load forecasting

The ongoing rapid growth of electricity over the past few decades greatly promotes the necessity of accurate electricity load forecasting. However, despite a great number of studies, electricity load forecasting is still an enormous challenge for its complexity. Recently, the developments of machine learning technologies in different research areas have demonstrated their great advantages. General vector machine (GVM) is a new machine learning model, which has been proven very effective in time series prediction. In this article, we apply it in electricity load forecasting. A detailed comparison with traditional back-propagation neural network (BP) is presented in this paper. To improve the load forecasting accuracy, we propose many methods to train the GVM model. Analysis of our approach on the historical Queensland electricity load dataset has demonstrated that GVM could achieve better forecasting results, which shows the strong potential of GVM for general electricity load forecasting.

Design of Multi-Stage Gear Modification for New Energy Vehicle Based on Optimized BP Neural Network

The NVH (Noise, Vibration, and Harshness) characteristics of new energy vehicles are the key indexes to measure interior comfort. The multi-stage gear reducer in the transmission system is the primary source of vibration and noise. The parameterized 3D model of the multi-stage gear transmission system of the new energy vehicle was established through Romax software, and the comprehensive gear modification method of the tooth direction combined with the tooth profile was built. Then a complete simulation analysis process is established to solve the maximum vibration acceleration of the multi-stage gear transmission system under constant speed condition, to obtain the simulation data of two-stage gear set under different modification parameters. The traditional BP (Back Propagation) neural network is optimized and improved through the optimal selection of network parameters combined with Bayesian regularization. Based on the optimized BP neural network, a modified parameter-vibration noise prediction model is constructed. Finally, the GA (Genetic Algorithm) optimization algorithm is used to solve the prediction model to obtain the optimal combination of modification parameters aiming at the minimum vibration acceleration, the effectiveness and reliability of the modified design are verified through actual simulation. It provides ideas and a basis for the research on vibration and noise reduction of multi-stage gears.

Beetle antennae search for neural network model with application to population prediction: An intelligent optimization algorithm

Traditional back-propagation (BP) neural networks can implement complex nonlinear mapping relationships, and solve internal mechanism problems. However, as number of samples increases, training BP neural networks may consume a lot of time. For this reason, to improve the efficiency as well as prediction accuracy of the neural network model, in this paper, we propose an intelligent optimization algorithm, by leveraging the beetle antennae search (BAS) strategy to optimize the weights of neural network model, and apply it to the population prediction. A series of experiments demonstrate the improved accuracy of the proposed algorithm over BP neural networks. In particular, the calculation time spent of neural network model via the proposed algorithm is only 20% of the one of BP neural network model. Finally, we present a reasonable trend of population growth in China, and analyze the causes of changes in population trends, which may provide an effective basis for the department to adjust population development strategies

Urban cold-chain logistics demand predicting model based on improved neural network model

With the popularity of the Internet and mobile terminals, the development of e-commerce has become hotter. Therefore, e-commerce research starts to focus on the statistics and prediction of the cargo volume of logistics. This study briefly introduced the back-propagation (BP) neural network model and principal component analysis (PCA) method and combined them to obtain an improved PCA-BP neural network model. Then the traditional BP neural network model and the improved PCA-BP neural network model were used to perform the empirical analysis of the cold chain logistics demand of fruits and vegetables in city A from 2010 to 2018. The results showed that the main factors that affected the local cold chain logistics demand were the growth rate of GDP, the added value of primary industry, the planting area of fruits and vegetables, and the consumption price index of fruits and vegetables; both kinds of neural networks model could effectively predict the cold chain logistics demand, but the predicted value of the PCA-BP neural network model was more fitted with the actual value. The prediction error of the BP neural network model was larger, and the fluctuation was obvious within the prediction interval. Moreover, the time required for the prediction by the PCA-BP neural network model was less than that by the BP neural network model. In summary, the improved PCA-BP neural network model is faster and more accurate than the traditional BP model in predicting the cold chain logistics demand.

Periodic-corrected data-driven coupling of blood flow and the vessel wall for virtual surgery

Fast and realistic coupling of blood flow and the vessel wall is of great importance to virtual surgery. In this paper, we propose a novel data-driven coupling method that formulates physics-based blood flow simulation as a regression problem, using an improved periodic-corrected neural network, estimating the acceleration of every particle at each frame to obtain fast, stable, and realistic simulation. We design a particle state feature vector based on smoothed particle hydrodynamics, modeling the mixed contribution of neighboring proxy particles on the blood vessel wall and neighboring blood particles, giving the extrapolation ability to deal with more complex couplings. We present a semi-supervised training strategy to improve the traditional back propagation neural network, which corrects the error periodically to ensure long-term stability. Experimental results demonstrate that our method is able to implement stable and vivid coupling of blood flow and the vessel wall while greatly improving computational efficiency.

A novel compound wind speed forecasting model based on the back propagation neural network optimized by bat algorithm.

Wind power, a clean and renewable resource, is regarded as one of the most promising and economical resources during the transformation from fossil fuels to new energy resources. Thus, the accuracy of wind speed forecasting work is very important to integrate the wind resource into electrical power system on a large scale. To improve the short-term wind speed forecasting accuracy, a novel compound model is introduced in this paper. For the proposed model, the fast ensemble empirical mode decomposition method was employed to do the data preprocessing. After the data preprocessing, phase space reconstruction was used for choosing each sub-series' input and output vectors for the forecasting model dynamically. Then, the bat algorithm was applied to optimize the connection weights and thresholds of the traditional back propagation neural network. The forecasting results can be obtained through the aggregation of sequential prediction. The performance evaluation of this proposed model indicates that it can capture the nonlinear characteristics of the wind speed signal efficiently. The proposed model shows better performance when being compared with the parallel models.

Deep Learning Application for Predicting Soil Organic Matter Content by VIS-NIR Spectroscopy

Deep learning is characterized by its strong ability of data feature extraction. This method can provide unique advantages when applying it to visible and near-infrared spectroscopy for predicting soil organic matter (SOM) content in those cases where the SOM content is negatively correlated with the spectral reflectance of soil. This study relied on the SOM content data of 248 red soil samples and their spectral reflectance data of 400–2450 nm in Fengxin County, Jiangxi Province (China) to meet three objectives. First, a multilayer perceptron and two convolutional neural networks (LeNet5 and DenseNet10) were used to predict the SOM content based on spectral variation and variable selection, and the outcomes were compared with that from the traditional back-propagation neural network (BPN). Second, the four methods were applied to full-spectrum modeling to test the difference to selected feature variables. Finally, the potential of direct modeling was evaluated using spectral reflectance data without any spectral variation. The results of prediction accuracy showed that deep learning performed better at predicting the SOM content than did the traditional BPN. Based on full-spectrum data, deep learning was able to obtain more feature information, thus achieving better and more stable results (i.e., similar average accuracy and far lower standard deviation) than those obtained through variable selection. DenseNet achieved the best prediction result, with a coefficient of determination (R2) = 0.892 ± 0.004 and a ratio of performance to deviation (RPD) = 3.053 ± 0.056 in validation. Based on DenseNet, the application of spectral reflectance data (without spectral variation) produced robust results for application-level purposes (validation R2 = 0.853 ± 0.007 and validation RPD = 2.639 ± 0.056). In conclusion, deep learning provides an effective approach to predict the SOM content by visible and near-infrared spectroscopy and DenseNet is a promising method for reducing the amount of data preprocessing.

Retrieval of Chemical Oxygen Demand through Modified Capsule Network Based on Hyperspectral Data

This study focuses on the retrieval of chemical oxygen demand (COD) in the Baiyangdian area in North China, using a modified capsule network. Herein, the capsule model was modified to analyze the regression relationship between 1-D hyperspectral data and COD values. The results indicate there is a statistically significant correlation between COD and the hyperspectral data. The accuracy of the capsule network was compared with the results obtained from using a traditional back-propagation neural network (BP) method. The capsule network achieved superior accuracy with fewer iterations, compared with the BP algorithm. An R2 value of 0.78 was obtained against measured COD values retrieved using the capsule network method, compared with a value of 0.42 for the BP algorithm retrievals. This suggests the capsule network method has great potential to solve regression problems in the field of remote sensing.

An intelligent bearing fault diagnosis method based on the AFEEMD and 1D CNNs

To process the non-stationary vibration signals and improve accuracy of bearing fault diagnosis, this paper presents a novel intelligent fault diagnosis method based on the adaptive fast ensemble empirical mode decomposition (AFEEMD) and one-dimensional convolutional neural networks (1D CNNs). First, the AFEEMD algorithm is utilized to decompose the raw signals into intrinsic mode functions (IMFs). Then, the time and frequency statistic features of the first several IMFs are analyzed to form feature vector, which are used as the input of 1D CNNs to identify the bearing fault. The performance of the proposed method is validated using the dataset from the Case Western Reserve University (CWRU). Compared with the traditional back propagation neural network (BPNN), the results show that the proposed AFEEMD-1D CNNs method not only can obtain higher accuracy and achieve more reliable performance, but also can improve the generalization performance. Due to the end-to-end feature learning capacity, it can be extended to other machinery for fault diagnosis.

Research on the damage prediction method of offshore wind turbine tower structure based on improved neural network

Tower structure is a basic part of offshore wind power system, and how to ensure the stability and safety of tower structure in wind turbine operation and develop the efficient and accurate damage prediction method to judge the damage location and degree of tower structure has become an important research topic in the offshore wind power industry. With high adaptability, nonlinearity and strong function approximation ability, Artificial neural network (ANN) has been widely used in structural damage prediction and showed better performance than other diagnostic methods, however, traditional Back Propagation neural network (BPNN) has shortcomings of complex training parameters and low accuracy. To address this issue, this paper considered an improved neural network damage prediction method based on step-by-step identification for offshore wind turbine tower structure, an experimental research was carried out to simplify wind turbine model and experimentally measure the intact steel cylinders with different damage locations and different damage degrees. Then, the modal analysis was carried out by using the numerical analysis method and the results were compared with those of experimental measurement to modify the model. The training data are generated by Abaqus software to verify the proposed method. Results showed that the step-by-step prediction method could effectively reduce the complexity of the network, improve the prediction accuracy of the neural network, and save the training time.

Color prediction of mushroom slices during drying using Bayesian extreme learning machine

Color is an important appearance attribute of fruits and vegetables during drying processing, as it influences consumer’s preference and acceptability. Establishing color change kinetics model is an effective way for better understanding the quality changes and optimization of drying process. However, it is difficult to quickly and accurately predict color change kinetics during drying as it is highly nonlinear, complex, dynamic, and multivariable. To alleviate this problem, a new model based on extreme learning machine integrated Bayesian methods (BELM) has been developed for the prediction of color changes of mushroom slices during drying process. The effects of drying temperature (55, 60, 65, 70, and 75 °C) and air velocity (3, 6, 9, and 12 m/s) on color change kinetics of mushroom slices during hot air impingement drying were firstly explored and the experimental results indicated that both drying temperature and air velocity significantly affected the color attributes. Then, to validate the robustness and effectiveness of BELM, the basic extreme learning machine (ELM) and traditional back-propagation neural network (BPNN) models have also been employed to predict the color quality. In terms of prediction accuracy and execution time, BELM could achieve least similar or even better performance than ELM and BPNN. It overcame the overfitting problems of ELM. The test results of optimal BELM model by two new cases revealed that the lowest R2 and highest RMSE of BELM model were 0.9725 and 0.0563, respectively. The absolute values of relative errors between the actual and predicted values were lower than 8.5%.

Reliability Analysis of Complex Systems with Failure Propagation

Failure propagation is a critical factor for the reliability and safety of complex systems. To recognise and identify failure propagation of systems, a deep fusion model based on deep belief network (DBN) and Bayesian structural equation model (BSEM) is proposed. The deep belief network is applied to extract features between status monitoring data and the performance degradation in different failure components. To calculate the path weight of failure propagation, the Bayesian structural equation model is proposed to study the relationship among different fault modes. After getting the performance degradation of each fault through DBN and calculating the path weight of fault propagation by BSEM, it is available to get the overall reliability of the system. The aircraft landing gear system with 19 fault patterns is selected to evaluate the feasibility of the proposed deep fusion model. The results demonstrate that the overall reliability of the system can be obtained by analysing the fault propagation of multiple fault patterns, and the proposed model has a lower deviation than traditional back propagation neural network.

Performance Sensing Data Prediction for an Aircraft Auxiliary Power Unit Using the Optimized Extreme Learning Machine.

The aircraft auxiliary power unit (APU) is responsible for environmental control in the cabin and the main engines starting the aircraft. The prediction of its performance sensing data is significant for condition-based maintenance. As a complex system, its performance sensing data have a typically nonlinear feature. In order to monitor this process, a model with strong nonlinear fitting ability needs to be formulated. A neural network has advantages of solving a nonlinear problem. Compared with the traditional back propagation neural network algorithm, an extreme learning machine (ELM) has features of a faster learning speed and better generalization performance. To enhance the training of the neural network with a back propagation algorithm, an ELM is employed to predict the performance sensing data of the APU in this study. However, the randomly generated weights and thresholds of the ELM often may result in unstable prediction results. To address this problem, a restricted Boltzmann machine (RBM) is utilized to optimize the ELM. In this way, a stable performance parameter prediction model of the APU can be obtained and better performance parameter prediction results can be achieved. The proposed method is evaluated by the real APU sensing data of China Southern Airlines Company Limited Shenyang Maintenance Base. Experimental results show that the optimized ELM with an RBM is more stable and can obtain more accurate prediction results.

Modal Strain Energy-Based Structural Damage Detection Using Convolutional Neural Networks

In this paper, a convolutional neural network (CNN) was used to extract the damage features of a steel frame structure. As structural damage could induce changes of the modal parameters of the structure, the convolution operation was used to extract the features of modal parameters, and a classification algorithm was used to judge the damage state of the structure. The finite element method was applied to analyze the free vibration of the steel frame and obtain the first-order modal strain energy for various damage scenarios, which was used as the CNN training sample. Then vibration experiments were carried out, and modal parameters were obtained from the modal analysis of the vibration signals. The experimental data were inputted into the CNN to verify its damage detection capability. The result showed that the CNN was effective in detecting the intact structure, single damage, and multi damages with an accuracy of 100%. For comparison, the same samples were also applied to the traditional back propagation (BP) neural network, which failed to detect the intact structure and multiple-damage cases. It was found that: (1) The proposed CNN could be trained from finite element simulation data and used in real frame structure damage detection, and it performed better in structural damage detection than BP neural networks; (2) the measured data of a real structure could be supplemented by numerical simulation data, and satisfactory results have been demonstrated.