Data-driven Method Research Articles

Uncertainty Optimization of Vibration Characteristics of Automotive Micro-Motors Based on Pareto Elliptic Algorithm

The NVH (Noise, Vibration, and Harshness) characteristics of micro-motors used in vehicles directly affect the comfort of drivers and passengers. However, various factors influence the motor’s structural parameters, leading to uncertainties in its NVH performance. To improve the motor’s NVH characteristics, we propose a method for optimizing the structural parameters of automotive micro-motors under uncertain conditions. This method uses the motor’s maximum magnetic flux as a constraint and aims to reduce vibration at the commutation frequency. Firstly, we introduce the Pareto ellipsoid parameter method, which converts the uncertainty problem into a deterministic one, enabling the use of traditional optimization methods. To increase efficiency and reduce computational cost, we employed a data-driven method that uses the one-dimensional Inception module as the foundational model, replacing both numerical models and physical experiments. Simultaneously, the module’s underlying architecture was improved, increasing the surrogate model’s accuracy. Additionally, we propose an improved NSGA-III (Non-dominated Sorting Genetic Algorithm III) method that utilizes adaptive reference point updating, dividing the optimization process into exploration and refinement phases based on population matching error. Comparative experiments with traditional models demonstrate that this method enhances the overall quality of the solution set, effectively addresses parameter uncertainties in practical engineering scenarios, and significantly improves the vibration characteristics of the motor.

Wind power prediction through acoustic data-driven online modeling and active wake control

Accurate online prediction of wind power is essential for the reliable operation of power systems. However, the precision of wind turbine power forecasts is often hindered by wake effects and inadequate online modeling capabilities. To tackle this issue, this paper proposes a data-driven online prediction method for wind turbine power, leveraging acoustic data for training. The approach involves a neural network that learns acoustic feature changes related to wind output power, by enhancing the kernel extreme learning machine (KELM) technique and applying data augmentation. Additionally, to address the wake effects in real wind fields, active wake control is integrated, considering that most wind turbines operate in the wake of upstream turbines. Experiments using actual wind tunnel measurement data were conducted to assess the performance of the proposed method. The findings demonstrate that this method surpasses other techniques without increasing computational costs. The mean absolute error (MAE) of the proposed method was only 0.4111 and 0.4302, and the root mean square error (RMSE) was just 0.4868 and 0.5195, respectively, in both experiments. Moreover, the proposed acoustic data-driven online power prediction method proves stable in high background noise environments and achieves accurate wind power prediction with just a 10% data sample following the implementation of the active wake control strategy. This work significantly contributes to the efficient utilization of wind energy resources.

Hybrid deep learning approach for rock tunnel deformation prediction based on spatio-temporal patterns

The ability to predict tunnel deformation holds great significance for ensuring the reliability, safety, and sustainability of tunnel structures. However, existing deformation prediction models often simplify or overlook the impact of spatial characteristics on deformation by treating it as a time series prediction issue. This study utilizes monitoring data from the Grand Canyon Tunnel and introduces an effective data-driven method for predicting tunnel deformation based on the spatio-temporal characteristics of the historical deformation of adjacent sections. The proposed model, a combination of graph attention network (GAT) and bidirectional long and short-term memory network (Bi-LSTM), is equipped with robust spatio-temporal predictive capabilities. Additionally, the study explores other possible spatial connections and the scalability of the model. The results indicate that the proposed model outperforms other deep learning models, achieving favorable root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) values of 0.34 mm, 0.23 mm, and 0.94, respectively. The graph structure based on intuitive spatial connections proves more suitable for meeting the challenges of predicting deformation. Integrating GAT-LSTM with transfer learning technology, remains stable performance when extended to other tunnels with limited data.

Integrating terrain structure characteristics into generative adversarial nets for hillshade generation

A hillshade is a visualization technique that represents three-dimensional terrain in a two-dimensional plane by illumination mapping. The digital relief shading promotes the visualization of the terrain efficiently using DEM data. However, compared with manual shading, the digital algorithm-based method still has a gap in the visual effect of illumination strategy and terrain generalization. The typical landform characteristics and micro-geomorphic properties usually are destroyed. The reason is that the complexity of illumination rules is hard to summarize for different terrain phenomena. In this study, namely the data-driven artificial intelligent method, an alternative strategy based on generative adversarial networks (GANs) is proposed rather than the rule-based method. The DEM and the terrain skeleton lines are input to the model and part of the manual relief shading of Swisstopo is used for the training samples. Through the GAN training and learning, the manual skill imbedded in the hillshade is discovered in the generation model. The results show that the proposed model performs better than the digital relief shading on various landforms in aesthetic visualization and geo-scientific representation. Compared to other models, including other convolution neural network (CNN) based methods, terrain structure is maintained more significantly through the proposed model.

Data-Driven Learning of the Generalized Langevin Equation with State-Dependent Memory.

We present a data-driven method to learn stochastic reduced models of complex systems that retain a state-dependent memory beyond the standard generalized Langevin equation with a homogeneous kernel. The constructed model naturally encodes the heterogeneous energy dissipation by jointly learning a set of state features and the non-Markovian coupling among the features. Numerical results demonstrate the limitation of the standard generalized Langevin equation and the essential role of the broadly overlooked state-dependency nature in predicting molecule kinetics related to conformation relaxation and transition.

AutoSamp: Autoencoding k-space Sampling via Variational Information Maximization for 3D MRI.

Accelerated MRI protocols routinely involve a predefined sampling pattern that undersamples the k-space. Finding an optimal pattern can enhance the reconstruction quality, however this optimization is a challenging task. To address this challenge, we introduce a novel deep learning framework, AutoSamp, based on variational information maximization that enables joint optimization of sampling pattern and reconstruction of MRI scans. We represent the encoder as a non-uniform Fast Fourier Transform that allows continuous optimization of k-space sample locations on a non-Cartesian plane, and the decoder as a deep reconstruction network. Experiments on public 3D acquired MRI datasets show improved reconstruction quality of the proposed AutoSamp method over the prevailing variable density and variable density Poisson disc sampling for both compressed sensing and deep learning reconstructions. We demonstrate that our data-driven sampling optimization method achieves 4.4dB, 2.0dB, 0.75dB, 0.7dB PSNR improvements over reconstruction with Poisson Disc masks for acceleration factors of R = 5, 10, 15, 25, respectively. Prospectively accelerated acquisitions with 3D FSE sequences using our optimized sampling patterns exhibit improved image quality and sharpness. Furthermore, we analyze the characteristics of the learned sampling patterns with respect to changes in acceleration factor, measurement noise, underlying anatomy, and coil sensitivities. We show that all these factors contribute to the optimization result by affecting the sampling density, k-space coverage and point spread functions of the learned sampling patterns.

Data-Driven Structure Recognition of Scanning Tunneling Microscopy Images in a Case of Iron Carbide.

Scanning tunneling microscopy (STM) serves as a critical tool for high-resolution surface imaging, yet deciphering the atomic structures from STM images on multielement surfaces, such as oxides and carbides, remains a challenging task that heavily relies on the expertise and intuition of researchers. In this study, we introduce a data-driven method for rapid structural recognition from STM images. This method involves extracting structural features, filtering through a structural database, and matching with simulated STM images and surface energy analyses, thereby providing researchers with several of the most probable structures. We demonstrate the capabilities of this technique using our previously reported iron carbide grown on an Fe(110) crystal. By proposing a candidate structure set and establishing a comprehensive database linking STM images to corresponding structures and surface energies, we selected 6 out of more than 10 000 possible surfaces. On the basis of these 6 recommendations, researchers can conveniently determine the real surface structures. Our work provides an efficient tool for the structure recognition of STM images to construct surface structures, potentially serving as a universal auxiliary tool for STM structural analysis.

Deep multimodal saliency parcellation of cerebellar pathways: Linking microstructure and individual function through explainable multitask learning.

Parcellation of human cerebellar pathways is essential for advancing our understanding of the human brain. Existing diffusion magnetic resonance imaging tractography parcellation methods have been successful in defining major cerebellar fibre tracts, while relying solely on fibre tract structure. However, each fibre tract may relay information related to multiple cognitive and motor functions of the cerebellum. Hence, it may be beneficial for parcellation to consider the potential importance of the fibre tracts for individual motor and cognitive functional performance measures. In this work, we propose a multimodal data-driven method for cerebellar pathway parcellation, which incorporates both measures of microstructure and connectivity, and measures of individual functional performance. Our method involves first training a multitask deep network to predict various cognitive and motor measures from a set of fibre tract structural features. The importance of each structural feature for predicting each functional measure is then computed, resulting in a set of structure-function saliency values that are clustered to parcellate cerebellar pathways. We refer to our method as Deep Multimodal Saliency Parcellation (DeepMSP), as it computes the saliency of structural measures for predicting cognitive and motor functional performance, with these saliencies being applied to the task of parcellation. Applying DeepMSP to a large-scale dataset from the Human Connectome Project Young Adult study (n = 1065), we found that it was feasible to identify multiple cerebellar pathway parcels with unique structure-function saliency patterns that were stable across training folds. We thoroughly experimented with all stages of the DeepMSP pipeline, including network selection, structure-function saliency representation, clustering algorithm, and cluster count. We found that a 1D convolutional neural network architecture and a transformer network architecture both performed comparably for the multitask prediction of endurance, strength, reading decoding, and vocabulary comprehension, with both architectures outperforming a fully connected network architecture. Quantitative experiments demonstrated that a proposed low-dimensional saliency representation with an explicit measure of motor versus cognitive category bias achieved the best parcellation results, while a parcel count of four was most successful according to standard cluster quality metrics. Our results suggested that motor and cognitive saliencies are distributed across the cerebellar white matter pathways. Inspection of the final k = 4 parcellation revealed that the highest-saliency parcel was most salient for the prediction of both motor and cognitive performance scores and included parts of the middle and superior cerebellar peduncles. Our proposed saliency-based parcellation framework, DeepMSP, enables multimodal, data-driven tractography parcellation. Through utilising both structural features and functional performance measures, this parcellation strategy may have the potential to enhance the study of structure-function relationships of the cerebellar pathways.

A physics-informed data-driven fault location method for transmission lines using single-ended measurements with field data validation

Data driven transmission line fault location methods have the potential to more accurately locate faults by extracting fault information from available data. However, most of the data driven fault location methods in the literature are not validated by field data for the following reasons. On one hand, the available field data during faults are very limited for one specific transmission line, and using field data for training is close to impossible. On the other hand, if simulation data are utilized for training, the mismatch between the simulation system and the practical system will cause fault location errors. To this end, this paper proposes a physics-informed data-driven fault location method. The data from a practical fault event are first analyzed to extract the ranges of system parameters such as equivalent source impedances, loading conditions, fault inception angles (FIA) and fault resistances. Afterwards, the simulation system is constructed with the ranges of parameters, to generate data for training. This procedure merges the gap between simulation and practical power systems, and at the same time considers the uncertainty of system parameters in practice. The proposed data-driven method does not require system parameters, only requires instantaneous voltage and current measurements at the local terminal, with a low sampling rate of several kHz and a short fault time window of half a cycle before and after the fault occurs. Numerical experiments and field data experiments clearly validate the advantages of the proposed method over existing data driven methods.

A data-driven method with sample entropy and CEEMDAN for short-term performance degradation prediction of dynamic hydrogen fuel cells

Short-term degradation prediction plays a crucial role in enhancing the reliability, safety, and cost-effectiveness of hydrogen fuel cells. A data-driven method for short-term degradation prediction is proposed based on Sample Entropy (SE), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), and a varying-depth Gated Recurrent Unit (GRU). CEEMDAN decomposes the voltage sequence into Intrinsic Mode Functions (IMFs), which are then reconstructed into three meaningful combined Intrinsic Mode Functions (co-IMFs) using SE. The varying-depth GRU is employed to predict these co-IMFs. Results show that under dynamic conditions, the proposed method achieves an RMSE of 0.0012 for 1-h ahead predictions and 0.0151 for 35-h ahead predictions. Under static conditions, the RMSE for 1-h ahead is 0.0012, and for 35-h ahead prediction, it is 0.0097. Comparative analysis with hybrid models using the same dataset reveals that this novel approach demonstrates superior accuracy and efficiency.

Towards Non-Region Specific Large-Scale Inundation Modelling with Machine Learning Methods

Traditional flood inundation modelling methods are computationally expensive and not suitable for near-real time inundation prediction. In this study we explore a data-driven machine learning method to complement and, in some cases, replace existing methods. Given sufficient training data and model capacity, our design enables a single neural network instance to approximate the flow characteristics of any input region, opening the possibility of applying the model to regions without available training data. To demonstrate the method we apply it to a very large >8000 km2 region of the Fitzroy river basin in Western Australia with a spatial resolution of 30 m × 30 m, placing an emphasis on efficiency and scalability. In this work we identify and address a range of practical limitations, e.g., we develop a novel water height regression method and cost function to address extreme class imbalances and by carefully constructing the input data, we introduce some natural physical constraints. Furthermore, a compact neural network design and training method was developed to enable the training problem to fit within GPU memory constraints and a novel dataset was constructed from the output of a calibrated two-dimensional hydrodynamic model. A good correlation between the predicted and groundtruth water heights was observed.

A weak leak location method based on signal attenuation mechanism for energy transportation system

Traditional leak location methods based on signal saltation points make it hard to locate the position of the weak leak event in the energy transportation system. To address it, a data-driven method based on signal attenuation mechanism is proposed in this paper. First, a wavelet adaptive threshold denoising procedure is proposed to obtain the pure signal from the collected data, and the decimal-coded genetic mechanism is designed to calculate the optimal threshold of wavelet decomposition for different layers. Second, the two-stage leak location method is proposed to calculate the position of the leak point through the combination analysis of pressure data change of both ends of the pipeline. Through the presented pressure signal attenuation mechanism, the leak location result is obtained from the data changes of both ends of the pipeline, not the time difference of arrived signal. With the above signal denoising and analysis way, the proposed method has the 1% location deviation of the leak point, which is superior to leak location methods based on signal saltation points such as wavelet transform. Finally, different experiment results are studied, and the analysis results show the effectiveness of the proposed method for the leak location problem.

Distribution of Relaxation Times Analysis of PEMFC through Loewner Framework: A Direct and Regularization Free Approach

The analysis and interpretation of electrochemical impedance spectroscopy (EIS) data are often complex and challenging, despite its popularity within the field of electrochemistry. Hence, recently, we have introduced novel data-driven approach for extracting the distribution of relaxation times (DRT) from impedance data in order to improve further analysis [1, 2]. This approach is based on the Loewner framework, which is a versatile and straightforward data-driven modeling method that computes reduced-order models in a state-space representation directly from frequency response data. Thus, it allows estimation of DRT without need of regularization procedures or iterative optimization algorithms.The effectiveness of this approach is successfully demonstrated by analysing the experimental EIS data of a polymer electrolyte membrane fuel cell (PEMFC). The results show that this method allows clear identification of the individual polarization processes based on their characteristic time constants (see Figure). Furthermore, the peaks observed in the DRT plot are correctly associated with different processes by examining spectra under various operating conditions. This is further validated by employing a physics-based model [3]. In this way, unambiguous fault identification and monitoring of the degradation state of the cell is possible and it can be the foundation for an on-board diagnostic tool.

A Long Short-Term Memory Neural Network Algorithm for Data-Driven Spatial Load Forecasting

Based on the LSTM neural network, the author proposes a new data-driven spatial load prediction method to analyze the time series within the neural network, avoid data depression, and determine the correlation of the training data space. Establish prediction models based on different neurons, reduce the dimensionality of collected data through data preprocessing, and ensure data integrity. At the same time, provide data management base, control model input and output, unify process model, ensure training sequence model, combine LSTM neural network model, select prediction method, and finish data-driven related transportation. Experimental results show that the data driven global load forecasting model based on LSTM neural network proposed in this paper takes 1.23 seconds to complete 8000 training data, when traditional data drive spatial load forecasting method based on CNN neural network takes 3.56 seconds to finish 8000 training data. It can be seen that the prediction method proposed in this article has a good prediction accuracy.

When houses wear faces: Reverse correlation applied to architectural design

Reverse correlation (RC) is a data-driven method from social psychology that has been effectively shown to visualize the mental representations that humans hold regarding facial attributes. The method helps to understand what features are relevant in terms of the evaluation of faces, such as dominance or submissiveness. To the best of our knowledge, RC has solely been applied to faces within the area of psychology until this point. However, there are many other areas where it is of interest to understand how humans evaluate and visualize content, one of them being the evaluation of house facades. With this work, we extended the application of RC to architectural design, specifically focusing on the evaluation of house facades with respect to the psychological attributes of facelikeness, invitingness, and likeability. Furthermore, we propose a novel approach to create the base image, by utilizing a generative adversarial network. In an online study with a between-subject design, 121 participants completed the RC task, with 40 to 41 participants assigned to each of the three attributes. The resulting classification images (CIs) from the RC task unveil face-related features for the attribute facelikeness, signifying the potential extension of the RC methodology beyond the established domain of facial analysis to other domains, such as architectural design.

B-Spline signature responses in structural change detection: method development

Damage detection is perhaps the most critical objective of structural health monitoring. Proper documentation of the progression of structural damage from onset to failure enables performance-based maintenance that mitigates risk and contains socioeconomic impact. Recently, damage detection methods that focus on change detection through analyses of the structural global response in the frequency or time domain have garnered attention due to their efficiency and accuracy. This work proposes a novel time-domain, nonparametric, data-driven change detection method. The time-domain B-Spline transfer function is defined as a signature time history response of the structure, referred to in this work as the B-Spline signature response (BSR). BSR may be computed directly through a physics-based simulation, or it be extracted in a discrete form from measurements of displacement, velocity, or acceleration time histories of structural response to any external excitation. The BSR is independent of loading, is a characteristic of the system, and captures the condition of the structure at the time of acquisition without the need for identifying structural properties. In this work, the fundamentals of the BSR concept are introduced first along with the process of extracting a BSR from structural responses to arbitrary excitations. Verification and validation studies are conducted based on computer simulations and laboratory testing and are discussed in detail. Subsequently, an approach is proposed to detect change through correlation of BSR signals acquired at different times during the monitoring period. Any changes in the intrinsic characteristics affecting the response cause BSR changes and correlation loss. The change detection methodology is demonstrated through computer simulations and experimental investigations. Initial parametric studies are conducted to investigate the suitability of the proposed approach to detect small structural changes. This article demonstrates the validity of the proposed method and establishes the framework for further development and case-specific applications.

Hybrid CMOD-Diffusion Algorithm Applied to Sentinel-1 for More Robust and Precise Wind Retrieval

Synthetic Aperture Radar (SAR) imagery presents significant advantages for observing ocean surface winds owing to its high spatial resolution and low sensitivity to extreme weather conditions. Nevertheless, signal noise poses a challenge, hindering precise wind retrieval from SAR imagery. Moreover, traditional geophysical model functions (GMFs) often falter, particularly in accurately estimating high wind speeds, notably during extreme weather phenomena like tropical cyclones (TCs). To address these limitations, this study proposes a novel hybrid model, CMOD-Diffusion, which integrates the strengths of GMFs with data-driven deep learning methods, thereby achieving enhanced accuracy and robustness in wind retrieval. Based on the coarse estimation of wind speed by the traditional GMF CMOD5.N, we introduce the recently developed data-driven method Denoising Diffusion Probabilistic Model (DDPM). It transforms an image from one domain to another domain by gradually adding Gaussian noise, thus achieving denoising and image synthesis. By introducing the DDPM, the noise from the observed normalized radar cross-section (NRCS) and the residual of the GMF methods can be largely compensated. Specifically, for wind speeds within the low-to-medium range, a DDPM is employed before proceeding to another CMOD iteration to recalibrate the observed NRCS. Conversely, a posterior-placed DDPM is applied after CMOD to reconstruct high-wind-speed regions or TC-affected areas, with the prior information from regions characterized by low wind speeds and recalibrated NRCS values. The efficacy of the proposed model is evaluated by using Sentinel-1 SAR imagery in vertical–vertical (VV) polarization, collocated with data from the European Centre for Medium-Range Weather Forecasts (ECMWF). Experimental results based on validation sets demonstrate significant improvements over CMOD5.N, particularly in low-to-medium wind speed regions, with the Structural Similarity Index (SSIM) increasing from 0.76 to 0.98 and the Root Mean Square Error (RMSE) decreasing from 1.98 to 0.63. Across the entire wind field, including regions with high wind speeds, the validation data obtained through the proposed method exhibit an RMSE of 2.39 m/s, with a correlation coefficient of 0.979.

Sensitivity analysis and robust optimization to high-dimensional uncertainties of compressors with active subspace method

The performance of compressors often deviates significantly from the design value due to the substantial geometric deviations arising from manufacturing errors and in-service degradation. Robust design optimization (RDO) plays a crucial role in ensuring the high performance and reliability of compressors. The study proposes an active subspace-based dimensionality-reduction RDO method for high-dimensional uncertainties of compressors. By employing the active subspace (AS) method, a dimensionality-reduction artificial neural network (ANN) is raised to solve the "dimension disaster" problem for high-dimensional uncertainty quantification (UQ) analysis. Additionally, to decrease the computation cost of the original AS method, a data-driven method is utilized to calculate the gradient of the quantity of interest (QoI) within the AS dimensionality-reduction framework. Furthermore, Shapley method is applied to study the sensitivity of geometric variables and their variation direction on the mean and variance of the performance, and the robustness improvement method by only modifying key geometric variables based on Shapley results is discussed. The RDO of high dimensional geometrical uncertainties of a transonic compressor stage is studied by using this method. The results demonstrate a reduction from the original 24-dimensional uncertainties to a 4-dimensional space, achieved through this method. Consequently, the required sample size is reduced by 75 % while maintaining nearly unchanged model accuracy. Furthermore, the findings reveal that after RDO, the mean efficiency increases by 1.04 %, with a 42 % reduction in variance. Compared to RDO, which requires simultaneous optimization of all 24 geometric variables, the robustness improvement based on Shapley only necessitates modifying 4 key geometric variables, resulting in a substantial increase in both the mean and variance of efficiency.

Intelligent measurement method of transition resistance of insulation fasteners in urban rail transit system based on SSA-optimized algorithm

Rail-to-earth transition resistance characterizes the amplitude of stray current leakage in urban rail transit system using electric traction, which is critical for evaluating failure risk of metal infrastructures caused by electrochemical corrosion. However, detection method commonly used in the field exhibit measurement errors resulting from the limitation of measuring principle and measuring distance, which will seriously affect the accuracy of electrical insulation performance evaluation of the whole system. In view of this, this paper proposes a data-driven method for accurately measuring rail-to-earth transition resistance in urban rail transit system based on CDEGS-based simulation and SSA-optimized algorithm. CDEGS model was built to construct database for machine learning of SSA-RF network structure. SSA algorithm was used in the measurement model to optimize the topology of the random forest (RF) structure. Proposed ensemble in this paper was proved to conduct transition resistance regression with a mean relative accuracy rate of 98.85 %. Results of sensitivity analysis indicate that proposed measurement method shows acceptable robustness, which allows the algorithm parameters to fluctuate within a certain range when applied in the field. Besides, in the case of uniform and non-uniform insulation degradation, the proposed method shows good performance in assessing insulation performance. This provides a feasible foundation for future applications of the proposed method in failure risk evaluation of buried infrastructures.

Heat transfer optimization for MH reactor using combined taguchi design and data-driven optimization method

Adding the heat transfer enhancement components into the MH reactor can improve the hydrogen absorption/desorption rate, but the vital indicators as the gravimetric and volumetric hydrogen storage density decrease at the same time, which is hardly considered in the published researches. In this study, a comprehensive hydrogen storage performance (CHSP) indicator for the MH reactor has been proposed considering the above contradiction, which can be applied to evaluate the comprehensive performance of the MH reactor and compare the reactor performances with different heat transfer configurations. Then, with the constraint of the CHSP, the heat transfer structure of the classical finned-tube MH reactor is optimized using a novel sieving optimization strategy which combines the Taguchi design and the data-driven optimization (ANN model + GA) to increase the optimization efficiency and the reliability of optimization solutions. The optimized results indicate that the optimal structural parameters of the finned-tube MH reactor are the diameter of heat exchange tube d of 5.13 mm, the fin thickness h of 0.28 mm, the fin radius r of 21.73 mm and the fin number N of 11, and the optimal CHSP is 0.547 mg/s.