Nonlinear Mapping Relationship Research Articles

Intelligent electromagnetic mapping via physics driven and neural networks on frequency selective surfaces

The high mapping efficiency between various structures and electromagnetic (EM) properties of frequency selective surfaces (FSSs) is the state-of-the-art in the EM community. The most straightforward approaches for beam analysis depend on measurements and conventional EM calculation methods, which are inefficient and time-consuming. Equivalent circuit models (ECMs) with excellent intuitiveness and simplicity have been put forward extensively. Despite several applications, bottlenecks in ECM still exist, i.e. the application scope is restricted to narrow bands and specific structures, which is triggered by the ignorance of EM nonlinear coupling. In this study, for the first time, a lightweight physical model based on neural network (ECM-NN) is proposed , which exhibits great physical interpretability and spatial generalization abilities. The nonlinear mapping relationship between structure and beam behavior is interpreted by corresponding simulations. Specifically, two deep parametric factors obtained by multi-layer perceptron networks are introduced to serve as the core of lightweight strategies and compensate for the absence of nonlinearity. Experimental results of single square loop (SL) and double SL indicate that compared with related works, better agreements of the frequency responses and resonant frequencies are achieved with ECM-NN in broadband (0–30 GHz) as well as oblique incident angles (0°–60°). The average accuracy of the mapping is higher than 98.6%. The findings of this study provide a novel strategy for further studies of complex FSSs.

Bionics structural modification and collaborative optimization design of a waste crossbeam based on an RBF neural network and NSGA-II

ABSTRACT In this work, a bionic structure was redesigned for a specific type of crossbeam and the corresponding modification strategies were formulated. Three modification schemes were put forward and the corresponding parameterized analysis models established. Sensitivity analysis results guided the variable parameter selection and the total design of an orthogonal test. Data obtained were utilized to establish a neural network model that reflects the dynamic nonlinear mapping relation. Then, a collaborative optimization design method was proposed that was based on the neural network and the genetic algorithm for secondary design of the rib thickness. Through optimizing the neural network model by NSGA-II, Pareto optimal results were obtained. The results showed that the mass of the newly designed crossbeam had been reduced by 3.74%, while the first three orders of natural frequency had been increased by 6.67%, 3.47% and 14.20%, respectively. The bionic modification structure and the collaborative optimization design method were proved to be valuable for redesign in the remanufacturing field.

Rock brittleness index inversion method with constraints of seismic and well logs via a CNN-GRU fusion network based on the spatiotemporal attention mechanism

Rock brittleness is an important parameter for evaluating the fracability of the unconventional oil and gas reservoirs such as tight sandstone and shale oil. Compared with the laboratory measurement, rock brittleness in 3D stratum using the joint inversion of seismic and well logs is more significant for oil exploration and development, but the resolution of the inversion results is low due to the narrowly effective bandwidth of seismic data. Deep neural network (DNN) can establish the complex nonlinear mapping relationship between input and output data to improve the resolution of inversion results, but it has the problem of unstable prediction results in the case of insufficient training samples in the geophysical field. Aiming at the problem, firstly, the ratio of Young ‘s modulus to Poisson ‘s ratio is defined as the rock brittleness index, and an AVO approximation equation with brittleness index is derived, thus establishing the convolution model between the rock brittleness index and seismic data. Secondly, an objective function containing the loss of logging data, seismic data and low-frequency components obtained by the inverse distance weighted interpolation of seismic and well logs is proposed to improve the spatial constraints of seismic and well logs. Then, the spatiotemporal attention mechanism is fused with the convolutional neural network (CNN) and gate recurrent unit (GRU) network to improve the sensitivity of the network proposed to important spatiotemporal characteristics. Finally, a DNN based on the double constraints of convolution model and data is established to solve the inversion stability in the case of small samples. The inversion results of the geological model with sand body superposition and pinchout and the tight sandstone reservoir in Junggar Basin verify that the generalization and prediction accuracy of the network proposed.

Soft sensor method of multimode BOF steelmaking endpoint carbon content and temperature based on vMF-WSAE dynamic deep learning

Abstract The difficulty of endpoint determination in basic oxygen furnace (BOF) steelmaking lies in achieving accurate real-time measurements of carbon content and temperature. For the characteristics of serious nonlinearity between process data, deep learning can perform excellent nonlinear feature representation for complex structural data. However, there is a process drift phenomenon in BOF steelmaking, and the existing deep learning-based soft sensor models cannot adapt to changes in the characteristics of samples, which may lead to their performance degradation. To deal with this problem, considering the characteristics of multimode distribution of process data, an adaptive updating deep learning model based on von-Mises Fisher (vMF) mixture model and weighted stacked autoencoder is proposed. First, the stacked autoencoder (SAE) and vMF mixture model are constructed for complex structural data, which can initially establish nonlinear mapping relationships and division of different distributions. Second, for each query sample, the basic SAE network will perform online adaptive fine-tuning according to its data with the same distribution to achieve dynamic updating. Moreover, each sample is assigned a weight according to its similarity with the query sample. Through the designed weighted loss function, the updated deep network will better match the working conditions of the query sample. Experimental studies with numerical examples and actual BOF steelmaking process data are provided to demonstrate the effectiveness of the proposed method.

An intelligent approach for flight risk prediction under icing conditions

Flight risk prediction is significant in improving the flight crew's situational awareness because it allows them to adopt appropriate operation strategies to prevent risk expansion caused by abnormal conditions, especially aircraft icing conditions. The flight risk space representing the nonlinear mapping relations between risk degree and the three-dimensional commanded vector (commanded airspeed, commanded bank angle, and commanded vertical velocity) is developed to provide the crew with practical risk information. However, the construction of flight risk space by means of computational flight dynamics suffers from certain defects, including slow computing speed. Accordingly, an intelligent approach for flight risk prediction is proposed to address these defects based on neural networks. Radial Basis Function Neural Network (RBFNN) is optimized using Adaptive Particle Swarm Optimization (APSO). To optimize both the parameters and the structure of APSO-RBFNN, a fitness function containing the training accuracy and network structure size is proposed. Extensive experimental results demonstrate that the flight risk predicted by APSO-RBFNN is very close to that obtained via computational flight dynamics. The average error (RMSE) is less than 10-1. The approach achieves a speedup close to 1000x compared with computational flight dynamics. In addition, some flight upset and recovery cases are presented to illustrate the efficiency of the intelligent approach for flight risk prediction.

Progressive multitask learning for high-resolution prediction of reservoir elastic parameters

Reservoir elastic parameters play an important role in resource exploration; however, the band-limited characteristics of seismic data and the ill-posed nature of seismic inversion significantly affect inversion accuracy. To alleviate this problem, a high-resolution prediction method for reservoir elastic parameters based on the progressive multitask learning network (PMLN) is proposed. Our network consists of three parts: network 1 for low-frequency extension (LFE), network 2 for reservoir parameter inversion, and network 3 for image superresolution (SR). Taking the seismic frequency band as the link, network 1 is first used to predict the low-frequency information of seismic data. Then, the nonlinear mapping relationship between the high-pass-filtered seismic data (and its envelope) and full-frequency seismic data is established. Second, network 2 directly predicts the reservoir elastic parameters using the seismic data after LFE. Finally, the SR of the inversion results is achieved from the image perspective based on network 3. The three networks have a progressive relationship and can share network features, which is beneficial for improving computing efficiency. As the features extracted by the network represent different contributions to the prediction target, a channel attention mechanism is introduced. Furthermore, the loss function of network 2 is improved using dip constraints to obtain high-precision reservoir parameters. Synthetic and field data analyses find that all three networks are competent for their respective tasks, and the PMLN can obtain high-resolution prediction results of reservoir elastic parameters. Compared with traditional full-waveform inversion, the PMLN effectively improves prediction accuracy. Therefore, the PMLN is expected to become a powerful tool for predicting the elastic parameters of reservoirs.

Multiobjective Optimization of a 3D Laser Scanning Scheme for Engineering Structures Based on RF-NSGA-II

To promote precision installation accuracy of engineering structural components and improve construction quality, point cloud data obtained by 3D laser scanning are widely used in construction quality monitoring. Reasonable parameters are very important to control errors and improve efficiency during the process of 3D laser scanning. Therefore, this paper proposes a random forest-nondominated sorting genetic algorithm (RF-NSGA-II) multiobjective optimization model with an elite strategy. To study the relationship between 3D laser scanning parameters and measurement errors or efficiency, the best scanning parameters are identified. In this paper, high-precision prediction of the relative error and scanning time by 3D laser scanning parameters is achieved by using an RF, and the nonlinear mapping relationship function is obtained, which is used as the objective optimization function. The RF-NSGA-II multiobjective optimization algorithm is developed to optimize the relative error and scanning time, and the scanning time is the shortest under the condition of reasonable relative error. Through this study, we can mainly draw the following conclusions: (1) based on the RF, we can obtain a prediction model of relative error and scanning time with high accuracy, in which the R2 (determination coefficient) value and root mean square error (RMSE) of the relative error prediction model are 0.967 and 0.0277, respectively, and the R2 value and RMSE of the scanning time prediction model are 0.978 and 0.0243, respectively; (2) optimized design parameters for 3D laser scanning of structural components are obtained, including a horizontal incident angle of 90°, an inclination angle of 90°, point cloud density of 3.2 mm, measurement distance of 3 m, resolution of 0.456, and visibility of 8.67 km; and (3) the RF-NSGA-II model developed can effectively reduce the relative error of the 3D laser (by 2.16 mm) and shorten the scanning time (by 148.926 s) compared with the average value. The structural components also meet the requirements in the deviation test. Therefore, the application of the RF-NSGA-II model in the assembly and fabrication of structural components can realize the intellectualization of the production process and improve the precision of the prefabrication of structural components, which has high engineering application value.

Engineering Safety Evaluation of the High Rock Slope of a Hydropower Project: A Case Study of 684 m-High Slope Related to Lianghekou Hydropower Project at Yalong River

The stability of the slope is a very important topic in the construction of hydropower projects, especially the slope engineering in the dam area, as its stability will directly affect the safety of engineering. Taking the inlet slope of the flood discharge structure of Lianghekou Hydropower Project as the research object, based on the analysis and exploration of the geological condition of the slope and the field monitoring data, GA-LSSVM is used to establish the non-linear mapping relationship, and the BP neural network is used to establish the mechanical parameters back analysis of the slope at different water impoundment stages. A numerical simulation model is also established to set up different reservoir impoundments to study the stability and sensibility of the slope and provide guidance for slope operation. This case study shows that there is a hysteresis in the response of slope deformation to reservoir impoundment. At the same time, the mechanical parameters of the slope will be weakened by the seepage. In the process of water level changes, the stability of the slope decreases due to the decrease in mechanical parameters. The study will be practically useful for engineering applications.

Real‐time risk assessment of cascading failure in power system with high proportion of renewable energy based on fault graph chains

AbstractThe access of a high proportion of renewable energies has deepened the randomness and complexity of cascading failures (CFs) in power systems. In this regard, a real‐time risk assessment method for CFs in power systems with high proportion of new energy is proposed. First, combined with historical statistical data and relevant national standards, a CF simulation model that considers the off‐grid protection action of renewable energy units in the event of a power grid fault is proposed. The model is based on the continuous steady‐state power flow model, which simulates the spread of CFs via continuous power flow calculations. Second, via introducing the concept of a fault graph chain, the electrical and topological characteristics of the continuous dynamic of the power system in the process of CFs can be described. Then, through continuous CF simulation and replay buffer, a data‐driven method is used to calculate the CF risk index corresponding to the fault graph. Finally, a cascaded graph neural network is employed to fit the nonlinear mapping relationship between fault graphs and CF risk indicators. The simulation results in the IEEE 39‐bus system show that the proposed method can accurately and real‐time evaluate the risk of CFs.

An improved TOPSIS-ANP composite shielding material performance evaluation method based on gray relational projection algorithm

The performance of shielding materials directly affects the radiation protection effect and plays a very important role in the process of ensuring the safety of nuclear energy. Therefore, this paper introduces the performance evaluation of composite shielding materials, which firstly points out the disadvantages of the traditional TOPSIS method, proposes a weighted projection model of composite shielding materials under extended TOPSIS theory, and clarifies the principle of projection dimensional reduction and algorithm implementation. Secondly, this paper also introduces the basic assumption of non-linear mapping relationship between index dimensions, and scientifically determines the weight of index system based on ANP structural model, so as to form an improved TOPSIS-ANP composite shielding material performance evaluation method based on gray relational projection algorithm with coupling characteristics. The empirical results show that the improved TOPSIS-ANP composite shielding material evaluation method proposed in this paper is consistent with the conclusion of the ratio of lead-boron-polyethylene shielding materials optimized based on genetic algorithm, which proves the effectiveness of the evaluation method proposed in this paper. Meanwhile, the evaluation index system of this method is more comprehensive, and the evaluation method is more efficient and scientific as well, which has a good promotion prospects and application advantages.

Intelligent Feedback Analysis of Fluid–Solid Coupling of Surrounding Rock of Tunnel in Water-Rich Areas

To realize parameter feedback optimization of tunnel construction in water-rich areas, a feedback analysis method for tunnel parameters under fluid–solid coupling conditions was established based on an intelligent optimization algorithm. Firstly, the numerical calculation model was established and solved using the fluid–solid coupling model. In orthogonal design analysis, the displacement of surrounding rock and pore water pressure distribution in different rock mass parameter combinations were obtained, and the learning samples needed for machine learning were established. The input group was surrounding rock displacement and pore water pressure, and the output was rock mass parameters. Then, the Gaussian process algorithm was used to obtain the nonlinear mapping relationship contained in the learning samples. A differential evolution algorithm was used to optimize the critical parameters involved in this process. Furthermore, according to the established regression model and the measured displacement and pore water pressure in the research area, differential evolution was used again to optimize the rock mass parameters and obtain the parameter feedback analysis results. Finally, the inversion values were compared with the actual measured values, and the reliability of the surrounding rock parameters obtained from the feedback analysis was verified, providing an effective method for obtaining surrounding rock parameters for similar projects.

Lateral semi-trailer truck control using a parameter self-learning MPC method in urban environment

This paper analyzes the lateral control technology of autonomous semi-trailer trucks. Existing researches on the lateral control algorithm of semi-trailer trucks focus on making the head-truck or trailer follow a track well while ignoring the motion characteristics during the turning process, leading to specific security issues. Meanwhile, it is difficult to cope with the complex and uncertain factors influencing lateral control effect, such as the curvature of the desired trajectory, the load, and the velocity of the semi-trailer truck. This paper proposes a parametric self-learning model predictive control (MPC) based on the Proximal Policy Optimization of One Step (OSPPO) method to solve these problems. After modeling the kinematics of the semi-trailer truck, a lateral motion controller for the relationship between the head-truck and trailer based on the MPC method is established. The traditional MPC method has difficulty in adapting to the changeable influencing factors. Thus, a deep reinforcement learning algorithm named OSPPO is introduced to improve the flexibility of the MPC method. OSPPO establishes the nonlinear mapping relationship between the critical parameter of the MPC method and the factors influencing control effect by self-learning, avoiding a large amount of labeled data for training. In simulations, Trucksim and Matlab were used to conduct co-simulation to verify the usefulness of the control method. The method was implemented on an autonomous semi-trailer truck for many outdoor scenes. The actual experimental results showed the validity and advantages of the proposed method.

An inverse design paradigm of multi-functional elastic metasurface via data-driven machine learning

Elastic metasurfaces have become one of the most promising platforms for manipulating mechanical wavefronts with the striking feature of ultra-thin geometry. The conventional design of mechanical metasurfaces significantly relies on numerical, trial-and-error methods to identify structural parameters of the unit cells, which requires huge computational resources and could be extremely challenging if the metasurface is multi-functional. Machine learning technique provides another powerful tool for the design of multi-functional elastic metasurfaces because of its excellent capability in building nonlinear mapping relation between high-dimensional input data and output data. In this paper, a machine learning network is introduced to extract the complex relation between high-dimensional geometrical parameters of the metasurface unit and its high-dimensional dynamic properties. Based on a big dataset, the well-trained network can play the role of a surrogate model in the inverse design of a multi-functional elastic metasurface to significantly shorten the time for the design. Such method can be conveniently extended to design other multi-functional metasurfaces for the manipulation of optical, acoustical or mechanical waves.

Hyperspectral Image Classification Based on Unsupervised Regularization

Due to the powerful feature expression ability of deep learning and its end-to-end non-linear mapping relationship, deep learning-based methods have become the mainstream method for hyperspectral image classification tasks. However, the accuracy of deep learning methods greatly depends on the use of a large number of labeled samples to train the model. Also, hyperspectral images have few labeled samples and unbalanced categories, which makes the depth model prone to overfitting, which seriously affects the classification accuracy. Therefore, how to alleviate the overfitting phenomenon caused by small samples in the classification problem based on deep learning is still a problem that needs to be solved. Considering that it is relatively easier to obtain a large number of unlabeled samples in the field of remote sensing, making full use of the unsupervised information learned from unlabeled data can regularize the supervised classification model, which can effectively alleviate the overfitting phenomenon caused by the small samples problem. In the supervised training process, unsupervised information from the overall distribution of the sample is introduced to guide the regularization of the model, so as to realize the effective classification of the data in the case of a small number of labeled samples. Experimental results demonstrate the effectiveness of the proposed method in terms of hyperspectral image (HSI) classification with few training samples.

Porosity and permeability prediction using a transformer and periodic long short-term network

Effective reservoir parameter prediction is important for subsurface characterization and understanding fluid migration. However, conventional methods for obtaining porosity and permeability are based on either core measurements or mathematical/petrophysical modeling, which are expensive or inefficient. In this study, we develop a reliable and low-cost deep learning (DL) framework for reservoir permeability and porosity prediction from real logging data at different regions. We leverage an advanced learning architecture (i.e., the transformer model) and design a new regression network (RPTransformer) that is sensitive to the depth period change of the logging data. The RPTransformer is composed of 1D convolutional, long short-term memory (LSTM), and transformer layers. First, we use a 1D convolutional layer for the first layer of the network to extract significant features from the logging data. Then, the nonlinear mapping relationships between logging data and reservoir parameters are established using several LSTM layers with a period parameter. Afterward, we use the encoder in the vision transformer with the self-attention mechanism to further extract logging data features. The developed network is a data-driven supervised learning framework and indicates highly accurate and robust prediction results when applied to different geographic regions. To demonstrate the reliable prediction performance of our network, we compare it with several classic machine learning and state-of-the-art DL methods, e.g., random forest, multilayer LSTM, and long short-term time-series network (LSTNet). More importantly, we find the generalization and uncertainty of the network in real-world applications through comprehensive numerical experiments.

Optimization of micro-channel heat sink based on genetic algorithm and back propagation neural network

In order to efficiently solve the problem of optimization of the micro-channel heat sink, an optimization strategy combining intelligent algorithms and CFD was proposed. The micro-channel heat sink with the trapezoidal cavity and solid/slotted oval pins was proposed to enhance heat transfer. The aspect ratio, distance from the center of the oval pin to the center of the cavity, and slot thickness were design variables. The thermal resistance and pumping power of the micro-channel heat sink were objective functions. Within the selected range of design variables, thirty groups of uniformly sampled sample points were obtained by the Latin hypercube experiment. The 3-D model was established by SOLIDWORKS software, and the numerical simulation was carried out by using FLUENT soft-ware. The genetic algorithm optimized back propagation neural network to construct the prediction model, and the simulated data of Latin hypercube sampling were trained to obtain the non-linear mapping relationship between design variables and objective functions. The optimal combination of structural parameters of the micro-channel heat sink was obtained by optimization of the genetic algorithm, which was verified by numerical simulation. The results show that the optimization scheme was suitable for getting the optimal value of the structural parameters of the micro-channel heat sink, which provided a reference for the optimal design of the micro-channel heat sink.

Brain-Inspired Spike Echo State Network Dynamics for Aero-Engine Intelligent Fault Prediction

Aero-engine fault prediction aims to accurately predict the development trend of the future state of aero-engines, so as to diagnose faults in advance. Traditional aero-engine parameter prediction methods mainly use the nonlinear mapping relationship of time series data but generally ignore the adequate spatio-temporal features contained in aero-engine data. To this end, we propose a brain-inspired spike echo state network (Spike-ESN) model for aero-engine intelligent fault prediction, which is used to effectively capture the evolution process of aero-engine time series data in the framework of spatio-temporal dynamics. In the proposed approach, we design a spike input layer based on Poisson distribution inspired by the spike neural encoding mechanism of biological neurons, which can extract the useful temporal characteristics in aero-engine sequence data. Then, the temporal characteristics are input into a spike reservoir through the current calculation method of spike accumulation in neurons, which projects the data into a high-dimensional sparse space. In addition, we use the ridge regression method to read out the internal state of the spike reservoir. Finally, the experimental results of aero-engine states prediction demonstrate the superiority and potential of the proposed method.

House Price Forecasting in Ames Based on Bayesian Regularized BP Neural Network

Housing has always been an important issue related to the national economy and people's livelihood. House prices not only affect people's welfare, but also have a significant impact on the national economy and social stability. Therefore, the prediction of house prices is also a necessary means. This paper forecasts the housing price data of Ames City in the United States, and establishes a Bayesian regularized BP neural network model to solve the nonlinear mapping relationship between housing prices and indicators. Through curve fitting analysis of samples, the overall R value is 0.9667, the overall result is better, and the error histogram also conforms to normal distribution. The experimental results show that the BP neural network model based on Bayesian regularization is very effective in dealing with the problem of house price prediction, and can better analyze and predict the trend of house price. This study provides a basis for real estate developers to develop and position products, and also provides model support for buyers to better judge the real price of houses.

Magnetotelluric noise suppression via convolutional neural network

It is well known that a magnetotelluric (MT) signal with high signal-to-noise ratio is an important prerequisite for correct interpretation of subsurface structures. However, MT signals collected in the environment of strong cultural noise often are of low data quality due to noise pollution, which seriously affects the accuracy of interpretation. As can be seen from the MT time-domain waveform, the noise is highly energetic, diverse, and random. This means MT denoising methods should have strong applicability to guarantee accurate and effective separation of MT signal from noise data. Therefore, we propose a deep-learning-based data nonlinear mapping method for MT signal-to-noise separation. First, this method focuses on learning the nonlinear mapping relationship between a large amount of noise data and the corresponding noise contour by using the convolutional neural network (CNN) in advance. Then, the mapping transformation of noise data to noise contour in the measured data is realized by CNN model. Specifically, based on the features of MT noise data, we construct a large amount of training data very close to it by mathematical functions. At the same time, we also select some of the measured data to be added to the training set. This not only expands the amount and diversity of the training set but also improves the adaptability of CNN when dealing with complex data. Finally, we evaluate the denoising performance of the proposed method in terms of time-domain waveforms, apparent resistivity-phase curves, and polarization directions before and after denoising. The processing results of the simulated data and the measured data collected in Luzong area have verified the feasibility and effectiveness of the proposed method in MT data denoising.

Sphere Gap Breakdown Voltage Prediction Based on ISSA Optimized BP Neural Network and Effective Electric Field Feature Set

Electric field distribution is a determinant factor of air gap breakdown voltage. This paper defined an effective electric field zone between the sphere gaps for feature extraction, and 29 parameters were extracted from this zone to reflect the electric field distribution. A data mining model was established by back propagation neural network (BPNN) for sphere gap breakdown voltage prediction, taking the electric field features as inputs. The maximum information coefficient (MIC) method was used to select features strongly correlated to the breakdown voltage, and a multi‐strategy improved salp swarm algorithm (ISSA) was applied to optimize the BPNN parameters to enhance the prediction performance. The standard sphere gaps provided in IEC 60052 were taken as training set and test set, which were divided according to the electric field nonuniform coefficients. The breakdown voltage prediction results of test sample sphere gaps demonstrate good agreements with the experimental values. With input features selected by the MIC method, the ISSA optimized BPNN model has a mean absolute percentage error of only 0.83% and a maximum relative error of 5.86%. By constructing the nonlinear mapping relationship between the effective electric field features and the insulation strength, this study provides a new approach to predict air gap breakdown voltage. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.