Artificial Intelligence Meets Flexible Sensors:Emerging Smart Flexible Sensing Systems Driven by Machine Learning and Artificial Synapses

The recent wave of the artificial intelligence(AI)revolution has aroused unprecedented interest in the intelligentialize of human society.As an essential component that bridges the physical world and digital signals,flexible sensors are evolving from a single sensing element to a smarter system,which is capable of highly efficient acquisition,analysis,and even perception of vast,multifaceted data.While challenging from a manual perspective,the development of intelligent flexible sensing has been remarkably facilitated owing to the rapid advances of brain-inspired AI innovations from both the algorithm(machine learning)and the framework(artificial synapses)level.This review presents the recent progress of the emerging AI-driven,intelligent flexible sensing systems.The basic concept of machine learning and artificial synapses are introduced.The new enabling features induced by the fusion of AI and flexible sensing are comprehensively reviewed,which significantly advances the applications such as flexible sensory systems,soft/humanoid robotics,and human activity monitoring.As two of the most profound innovations in the twenty-first century,the deep incorporation of flexible sensing and AI technology holds tremendous potential for creating a smarter world for human beings.

Stress-assisted corrosion mechanism of 3Ni steel by using gradient boosting decision tree machining learning method

Traditional 3Ni weathering steel cannot completely meet the requirements for offshore engineering development,resulting in the design of novel 3Ni steel with the addition of microalloy elements such as Mn or Nb for strength enhancement becoming a trend.The stress-assisted corrosion behavior of a novel designed high-strength 3Ni steel was investigated in the current study using the corrosion big data method.The information on the corrosion process was recorded using the galvanic corrosion current monitoring method.The gradi-ent boosting decision tree(GBDT)machine learning method was used to mine the corrosion mechanism,and the importance of the struc-ture factor was investigated.Field exposure tests were conducted to verify the calculated results using the GBDT method.Results indic-ated that the GBDT method can be effectively used to study the influence of structural factors on the corrosion process of 3Ni steel.Dif-ferent mechanisms for the addition of Mn and Cu to the stress-assisted corrosion of 3Ni steel suggested that Mn and Cu have no obvious effect on the corrosion rate of non-stressed 3Ni steel during the early stage of corrosion.When the corrosion reached a stable state,the in-crease in Mn element content increased the corrosion rate of 3Ni steel,while Cu reduced this rate.In the presence of stress,the increase in Mn element content and Cu addition can inhibit the corrosion process.The corrosion law of outdoor-exposed 3Ni steel is consistent with the law based on corrosion big data technology,verifying the reliability of the big data evaluation method and data prediction model selection.

Enhancing Deep Learning Soil Moisture Forecasting Models by Integrating Physics-based Models

Accurate soil moisture(SM)prediction is critical for understanding hydrological processes.Physics-based(PB)models exhibit large uncertainties in SM predictions arising from uncertain parameterizations and insufficient representation of land-surface processes.In addition to PB models,deep learning(DL)models have been widely used in SM predictions recently.However,few pure DL models have notably high success rates due to lacking physical information.Thus,we developed hybrid models to effectively integrate the outputs of PB models into DL models to improve SM predictions.To this end,we first developed a hybrid model based on the attention mechanism to take advantage of PB models at each forecast time scale(attention model).We further built an ensemble model that combined the advantages of different hybrid schemes(ensemble model).We utilized SM forecasts from the Global Forecast System to enhance the convolutional long short-term memory(ConvLSTM)model for 1–16 days of SM predictions.The performances of the proposed hybrid models were investigated and compared with two existing hybrid models.The results showed that the attention model could leverage benefits of PB models and achieved the best predictability of drought events among the different hybrid models.Moreover,the ensemble model performed best among all hybrid models at all forecast time scales and different soil conditions.It is highlighted that the ensemble model outperformed the pure DL model over 79.5%of in situ stations for 16-day predictions.These findings suggest that our proposed hybrid models can adequately exploit the benefits of PB model outputs to aid DL models in making SM predictions.

Assessment of compressive strength of jet grouting by machine learning

Jet grouting is one of the most popular soil improvement techniques,but its design usually involves great uncertainties that can lead to economic cost overruns in construction projects.The high dispersion in the properties of the improved material leads to designers assuming a conservative,arbitrary and unjustified strength,which is even sometimes subjected to the results of the test fields.The present paper presents an approach for prediction of the uniaxial compressive strength(UCS)of jet grouting columns based on the analysis of several machine learning algorithms on a database of 854 results mainly collected from different research papers.The selected machine learning model(extremely randomized trees)relates the soil type and various parameters of the technique to the value of the compressive strength.Despite the complex mechanism that surrounds the jet grouting process,evidenced by the high dispersion and low correlation of the variables studied,the trained model allows to optimally predict the values of compressive strength with a significant improvement with respect to the existing works.Consequently,this work proposes for the first time a reliable and easily applicable approach for estimation of the compressive strength of jet grouting columns.

Prediction of corrosion rate for friction stir processed WE43 alloy by combining PSO-based virtual sample generation and machine learning

The corrosion rate is a crucial factor that impacts the longevity of materials in different applications.After undergoing friction stir processing(FSP),the refined grain structure leads to a notable decrease in corrosion rate.However,a better understanding of the correlation between the FSP process parameters and the corrosion rate is still lacking.The current study used machine learning to establish the relationship between the corrosion rate and FSP process parameters(rotational speed,traverse speed,and shoulder diameter)for WE43 alloy.The Taguchi L27 design of experiments was used for the experimental analysis.In addition,synthetic data was generated using particle swarm optimization for virtual sample generation(VSG).The application of VSG has led to an increase in the prediction accuracy of machine learning models.A sensitivity analysis was performed using Shapley Additive Explanations to determine the key factors affecting the corrosion rate.The shoulder diameter had a significant impact in comparison to the traverse speed.A graphical user interface(GUI)has been created to predict the corrosion rate using the identified factors.This study focuses on the WE43 alloy,but its findings can also be used to predict the corrosion rate of other magnesium alloys.

Anti-Byzantine Attacks Enabled Vehicle Selection for Asynchronous Federated Learning in Vehicular Edge Computing

In vehicle edge computing(VEC),asynchronous federated learning(AFL)is used,where the edge receives a local model and updates the global model,effectively reducing the global aggregation latency.Due to different amounts of local data,computing capabilities and locations of the vehicles,renewing the global model with same weight is inappropriate.The above factors will affect the local calculation time and upload time of the local model,and the vehicle may also be affected by Byzantine attacks,leading to the deterioration of the vehicle data.However,based on deep reinforcement learning(DRL),we can consider these factors comprehensively to eliminate vehicles with poor performance as much as possible and exclude vehicles that have suffered Byzantine attacks before AFL.At the same time,when aggregating AFL,we can focus on those vehicles with better performance to improve the accuracy and safety of the system.In this paper,we proposed a vehicle selection scheme based on DRL in VEC.In this scheme,vehicle’s mobility,channel conditions with temporal variations,computational resources with temporal variations,different data amount,transmission channel status of vehicles as well as Byzantine attacks were taken into account.Simulation results show that the proposed scheme effectively improves the safety and accuracy of the global model.

Accurate and efficient remaining useful life prediction of batteries enabled by physics-informed machine learning

The safe and reliable operation of lithium-ion batteries necessitates the accurate prediction of remaining useful life(RUL).However,this task is challenging due to the diverse ageing mechanisms,various operating conditions,and limited measured signals.Although data-driven methods are perceived as a promising solution,they ignore intrinsic battery physics,leading to compromised accuracy,low efficiency,and low interpretability.In response,this study integrates domain knowledge into deep learning to enhance the RUL prediction performance.We demonstrate accurate RUL prediction using only a single charging curve.First,a generalisable physics-based model is developed to extract ageing-correlated parameters that can describe and explain battery degradation from battery charging data.The parameters inform a deep neural network(DNN)to predict RUL with high accuracy and efficiency.The trained model is validated under 3 types of batteries working under 7 conditions,considering fully charged and partially charged cases.Using data from one cycle only,the proposed method achieves a root mean squared error(RMSE)of 11.42 cycles and a mean absolute relative error(MARE)of 3.19%on average,which are over45%and 44%lower compared to the two state-of-the-art data-driven methods,respectively.Besides its accuracy,the proposed method also outperforms existing methods in terms of efficiency,input burden,and robustness.The inherent relationship between the model parameters and the battery degradation mechanism is further revealed,substantiating the intrinsic superiority of the proposed method.

Identification of Software Bugs by Analyzing Natural Language-Based Requirements Using Optimized Deep Learning Features

Software project outcomes heavily depend on natural language requirements,often causing diverse interpretations and issues like ambiguities and incomplete or faulty requirements.Researchers are exploring machine learning to predict software bugs,but a more precise and general approach is needed.Accurate bug prediction is crucial for software evolution and user training,prompting an investigation into deep and ensemble learning methods.However,these studies are not generalized and efficient when extended to other datasets.Therefore,this paper proposed a hybrid approach combining multiple techniques to explore their effectiveness on bug identification problems.The methods involved feature selection,which is used to reduce the dimensionality and redundancy of features and select only the relevant ones;transfer learning is used to train and test the model on different datasets to analyze how much of the learning is passed to other datasets,and ensemble method is utilized to explore the increase in performance upon combining multiple classifiers in a model.Four National Aeronautics and Space Administration(NASA)and four Promise datasets are used in the study,showing an increase in the model’s performance by providing better Area Under the Receiver Operating Characteristic Curve(AUC-ROC)values when different classifiers were combined.It reveals that using an amalgam of techniques such as those used in this study,feature selection,transfer learning,and ensemble methods prove helpful in optimizing the software bug prediction models and providing high-performing,useful end mode.

Well-defined high entropy-metal nanoparticles:Detection of the multi-element particles by deep learning

Characterizing and control the chemical compositions of multi-element particles as single metal nanoparticles(mNPs) on the surfaces of catalytic metal oxide supports is challenging.This can be attributed to the heterogeneity and large size at the nanoscale,the poorly defined catalyst nanostructure,and thermodynamic immiscibility of the strongly repelling metallic elements.To address these challenges,an ultrasonic-assisted coincident electro-oxidation-reduction-precipitation(U-SEO-P) is presented to fabricate ultra-stable PtRuAgCoCuP NPs,which produces numerous active intermediates and induces strong metal-support interactions.To sort the active high-entropy mNPs,individual NPs are described on the support surface and the role of deep learning in understanding/predicting the features of PtRuAgCoCu@TiO_(x) catalysts is explained.Notably,this deep learning approach required minimal to no human input.The as-prepared PtRuAgCoCu@TiO_(x) catalysts can be used to catalyze various important chemical reactions,such as a high reduction conversion(100% in 30 s),with no loss of catalytic activity even after 20 cycles of nitroarene and ketone/aldehyde,which is several times higher than commercial Pt@TiO_(x) owing to individual PtRuAgCoCuP NPs on TiO_(x) surface.In this study,we present the "Totally Defined Catalysis" concept,which has enormous potential for the advancement of high-activity catalysts in the reduction of organic compounds.

GP‐FMLNet:A feature matrix learning network enhanced by glyph and phonetic information for Chinese sentiment analysis

Sentiment analysis is a fine‐grained analysis task that aims to identify the sentiment polarity of a specified sentence.Existing methods in Chinese sentiment analysis tasks only consider sentiment features from a single pole and scale and thus cannot fully exploit and utilise sentiment feature information,making their performance less than ideal.To resolve the problem,the authors propose a new method,GP‐FMLNet,that integrates both glyph and phonetic information and design a novel feature matrix learning process for phonetic features with which to model words that have the same pinyin information but different glyph information.Our method solves the problem of misspelling words influencing sentiment polarity prediction results.Specifically,the authors iteratively mine character,glyph,and pinyin features from the input comments sentences.Then,the authors use soft attention and matrix compound modules to model the phonetic features,which empowers their model to keep on zeroing in on the dynamic‐setting words in various positions and to dispense with the impacts of the deceptive‐setting ones.Ex-periments on six public datasets prove that the proposed model fully utilises the glyph and phonetic information and improves on the performance of existing Chinese senti-ment analysis algorithms.

Study on Quantitative Precipitation Estimation by Polarimetric Radar Using Deep Learning

Accurate radar quantitative precipitation estimation(QPE)plays an essential role in disaster prevention and mitigation.In this paper,two deep learning-based QPE networks including a single-parameter network and a multi-parameter network are designed.Meanwhile,a self-defined loss function(SLF)is proposed during modeling.The dataset includes Shijiazhuang S-band dual polarimetric radar(CINRAD/SAD)data and rain gauge data within the radar’s 100-km detection range during the flood season of 2021 in North China.Considering that the specific propagation phase shift(KDP)has a roughly linear relationship with the precipitation intensity,KDP is set to 0.5°km^(-1 )as a threshold value to divide all the rain data(AR)into a heavy rain(HR)and light rain(LR)dataset.Subsequently,12 deep learning-based QPE models are trained according to the input radar parameters,the precipitation datasets,and whether an SLF was adopted,respectively.The results suggest that the effects of QPE after distinguishing rainfall intensity are better than those without distinguishing,and the effects of using SLF are better than those that used MSE as a loss function.A Z-R relationship and a ZH-KDP-R synthesis method are compared with deep learning-based QPE.The mean relative errors(MRE)of AR models using SLF are improved by 61.90%,51.21%,and 56.34%compared with the Z-R relational method,and by 38.63%,42.55%,and 47.49%compared with the synthesis method.Finally,the models are further evaluated in three precipitation processes,which manifest that the deep learning-based models have significant advantages over the traditional empirical formula methods.

Intelligent Design of High Strength and High Conductivity Copper Alloys Using Machine Learning Assisted by Genetic Algor

Metallic alloys for a given application are usually designed to achieve the desired properties by devising experimentsbased on experience, thermodynamic and kinetic principles, and various modeling and simulation exercises.However, the influence of process parameters and material properties is often non-linear and non-colligative. Inrecent years, machine learning (ML) has emerged as a promising tool to dealwith the complex interrelation betweencomposition, properties, and process parameters to facilitate accelerated discovery and development of new alloysand functionalities. In this study, we adopt an ML-based approach, coupled with genetic algorithm (GA) principles,to design novel copper alloys for achieving seemingly contradictory targets of high strength and high electricalconductivity. Initially, we establish a correlation between the alloy composition (binary to multi-component) andthe target properties, namely, electrical conductivity and mechanical strength. Catboost, an ML model coupledwith GA, was used for this task. The accuracy of the model was above 93.5%. Next, for obtaining the optimizedcompositions the outputs fromthe initial model were refined by combining the concepts of data augmentation andPareto front. Finally, the ultimate objective of predicting the target composition that would deliver the desired rangeof properties was achieved by developing an advancedMLmodel through data segregation and data augmentation.To examine the reliability of this model, results were rigorously compared and verified using several independentdata reported in the literature. This comparison substantiates that the results predicted by our model regarding thevariation of conductivity and evolution ofmicrostructure and mechanical properties with composition are in goodagreement with the reports published in the literature.

A Modified Iterative Learning Control Approach for the Active Suppression of Rotor Vibration Induced by Coupled Unbalance and Misalignment

This paper proposes a modified iterative learning control(MILC)periodical feedback-feedforward algorithm to reduce the vibration of a rotor caused by coupled unbalance and parallel misalignment.The control of the vibration of the rotor is provided by an active magnetic actuator(AMA).The iterative gain of the MILC algorithm here presented has a self-adjustment based on the magnitude of the vibration.Notch filters are adopted to extract the synchronous(1×Ω)and twice rotational frequency(2×Ω)components of the rotor vibration.Both the notch frequency of the filter and the size of feedforward storage used during the experiment have a real-time adaptation to the rotational speed.The method proposed in this work can provide effective suppression of the vibration of the rotor in case of sudden changes or fluctuations of the rotor speed.Simulations and experiments using the MILC algorithm proposed here are carried out and give evidence to the feasibility and robustness of the technique proposed.

Properties of radiation defects and threshold energy of displacement in zirconium hydride obtained by new deep-learning potential

Zirconium hydride(ZrH_(2)) is an ideal neutron moderator material. However, radiation effect significantly changes its properties, which affect its behavior and the lifespan of the reactor. The threshold energy of displacement is an important quantity of the number of radiation defects produced, which helps us to predict the evolution of radiation defects in ZrH_(2).Molecular dynamics(MD) and ab initio molecular dynamics(AIMD) are two main methods of calculating the threshold energy of displacement. The MD simulations with empirical potentials often cannot accurately depict the transitional states that lattice atoms must surpass to reach an interstitial state. Additionally, the AIMD method is unable to perform largescale calculation, which poses a computational challenge beyond the simulation range of density functional theory. Machine learning potentials are renowned for their high accuracy and efficiency, making them an increasingly preferred choice for molecular dynamics simulations. In this work, we develop an accurate potential energy model for the ZrH_(2) system by using the deep-potential(DP) method. The DP model has a high degree of agreement with first-principles calculations for the typical defect energy and mechanical properties of the ZrH_(2) system, including the basic bulk properties, formation energy of point defects, as well as diffusion behavior of hydrogen and zirconium. By integrating the DP model with Ziegler–Biersack–Littmark(ZBL) potential, we can predict the threshold energy of displacement of zirconium and hydrogen in ε-ZrH_(2).

Recent advances in protein conformation sampling by combining machine learning with molecular simulation

The rapid advancement and broad application of machine learning(ML)have driven a groundbreaking revolution in computational biology.One of the most cutting-edge and important applications of ML is its integration with molecular simulations to improve the sampling efficiency of the vast conformational space of large biomolecules.This review focuses on recent studies that utilize ML-based techniques in the exploration of protein conformational landscape.We first highlight the recent development of ML-aided enhanced sampling methods,including heuristic algorithms and neural networks that are designed to refine the selection of reaction coordinates for the construction of bias potential,or facilitate the exploration of the unsampled region of the energy landscape.Further,we review the development of autoencoder based methods that combine molecular simulations and deep learning to expand the search for protein conformations.Lastly,we discuss the cutting-edge methodologies for the one-shot generation of protein conformations with precise Boltzmann weights.Collectively,this review demonstrates the promising potential of machine learning in revolutionizing our insight into the complex conformational ensembles of proteins.

Reliable calculations of nuclear binding energies by the Gaussian process of machine learning

Reliable calculations of nuclear binding energies are crucial for advancing the research of nuclear physics. Machine learning provides an innovative approach to exploring complex physical problems. In this study, the nuclear binding energies are modeled directly using a machine-learning method called the Gaussian process. First, the binding energies for 2238 nuclei with Z > 20 and N > 20 are calculated using the Gaussian process in a physically motivated feature space, yielding an average deviation of 0.046 MeV and a standard deviation of 0.066 MeV. The results show the good learning ability of the Gaussian process in the studies of binding energies. Then, the predictive power of the Gaussian process is studied by calculating the binding energies for 108 nuclei newly included in AME2020. The theoretical results are in good agreement with the experimental data, reflecting the good predictive power of the Gaussian process. Moreover, the α-decay energies for 1169 nuclei with 50 ≤ Z ≤ 110 are derived from the theoretical binding energies calculated using the Gaussian process. The average deviation and the standard deviation are, respectively, 0.047 MeV and 0.070 MeV. Noticeably, the calculated α-decay energies for the two new isotopes ^ (204 )Ac(Huang et al. Phys Lett B 834, 137484(2022)) and ^ (207) Th(Yang et al. Phys Rev C 105, L051302(2022)) agree well with the latest experimental data. These results demonstrate that the Gaussian process is reliable for the calculations of nuclear binding energies. Finally, the α-decay properties of some unknown actinide nuclei are predicted using the Gaussian process. The predicted results can be useful guides for future research on binding energies and α-decay properties.

Evolution of pore systems in low-maturity oil shales during thermal upgrading--Quantified by dynamic SEM and machine learning

In-situ upgrading by heating is feasible for low-maturity shale oil,where the pore space dynamically evolves.We characterize this response for a heated substrate concurrently imaged by SEM.We systematically follow the evolution of pore quantity,size(length,width and cross-sectional area),orientation,shape(aspect ratio,roundness and solidity)and their anisotropy—interpreted by machine learning.Results indicate that heating generates new pores in both organic matter and inorganic minerals.However,the newly formed pores are smaller than the original pores and thus reduce average lengths and widths of the bedding-parallel pore system.Conversely,the average pore lengths and widths are increased in the bedding-perpendicular direction.Besides,heating increases the cross-sectional area of pores in low-maturity oil shales,where this growth tendency fluctuates at<300℃ but becomes steady at>300℃.In addition,the orientation and shape of the newly-formed heating-induced pores follow the habit of the original pores and follow the initial probability distributions of pore orientation and shape.Herein,limited anisotropy is detected in pore direction and shape,indicating similar modes of evolution both bedding-parallel and bedding-normal.We propose a straightforward but robust model to describe evolution of pore system in low-maturity oil shales during heating.

Machine learning for parameters diagnosis of spark discharge by electro-acoustic signal

Discharge plasma parameter measurement is a key focus in low-temperature plasma research.Traditional diagnostics often require costly equipment,whereas electro-acoustic signals provide a rich,non-invasive,and less complex source of discharge information.This study harnesses machine learning to decode these signals.It establishes links between electro-acoustic signals and gas discharge parameters,such as power and distance,thus streamlining the prediction process.By building a spark discharge platform to collect electro-acoustic signals and implementing a series of acoustic signal processing techniques,the Mel-Frequency Cepstral Coefficients(MFCCs)of the acoustic signals are extracted to construct the predictors.Three machine learning models(Linear Regression,k-Nearest Neighbors,and Random Forest)are introduced and applied to the predictors to achieve real-time rapid diagnostic measurement of typical spark discharge power and discharge distance.All models display impressive performance in prediction precision and fitting abilities.Among them,the k-Nearest Neighbors model shows the best performance on discharge power prediction with the lowest mean square error(MSE=0.00571)and the highest R-squared value(R^(2)=0.93877).The experimental results show that the relationship between the electro-acoustic signal and the gas discharge power and distance can be effectively constructed based on the machine learning algorithm,which provides a new idea and basis for the online monitoring and real-time diagnosis of plasma parameters.

Estimating wheat spike-leaf composite indicator(SLI)dynamics by coupling spectral indices and machine learning

The contribution of spike photosynthesis to grain yield(GY)has been overlooked in the accurate spectral prediction of yield.Thus,it’s essential to construct and estimate a yield-related phenotypic trait considering spike photosynthesis.Based on field and spectral reflectance data from 19 wheat cultivars under two nitrogen fertilization conditions in two years,our objectives were to(i)construct a yield-related phenotypic trait(spike–leaf composite indicator,SLI)accounting for the contribution of the spike to photosynthesis,(ii)develop a novel spectral index(enhanced triangle vegetation index,ETVI3)sensitive to SLI,and(iii)establish and evaluate SLI estimation models by integrating spectral indices and machine learning algorithms.The results showed that SLI was sensitive to nitrogen fertilizer and wheat cultivar variation as well as a better predictor of yield than the leaf area index.ETVI3 maintained a strong correlation with SLI throughout the growth stage,whereas the correlations of other spectral indices with SLI were poor after spike emergence.Integrating spectral indices and machine learning algorithms improved the estimation accuracy of SLI,with the most accurate estimates of SLI showing coefficient of determination,root mean square error(RMSE),and relative RMSE values of 0.71,0.047,and 26.93%,respectively.These results provide new insights into the role of fruiting organs for the accurate spectral prediction of GY.This high-throughput SLI estimation approach can be applied for wheat yield prediction at whole growth stages and may be assisted with agronomical practices and variety selection.

Adaptable and Dynamic Access Control Decision-Enforcement Approach Based on Multilayer Hybrid Deep Learning Techniques in BYOD Environment

Organizations are adopting the Bring Your Own Device(BYOD)concept to enhance productivity and reduce expenses.However,this trend introduces security challenges,such as unauthorized access.Traditional access control systems,such as Attribute-Based Access Control(ABAC)and Role-Based Access Control(RBAC),are limited in their ability to enforce access decisions due to the variability and dynamism of attributes related to users and resources.This paper proposes a method for enforcing access decisions that is adaptable and dynamic,based on multilayer hybrid deep learning techniques,particularly the Tabular Deep Neural Network Tabular DNN method.This technique transforms all input attributes in an access request into a binary classification(allow or deny)using multiple layers,ensuring accurate and efficient access decision-making.The proposed solution was evaluated using the Kaggle Amazon access control policy dataset and demonstrated its effectiveness by achieving a 94%accuracy rate.Additionally,the proposed solution enhances the implementation of access decisions based on a variety of resource and user attributes while ensuring privacy through indirect communication with the Policy Administration Point(PAP).This solution significantly improves the flexibility of access control systems,making themmore dynamic and adaptable to the evolving needs ofmodern organizations.Furthermore,it offers a scalable approach to manage the complexities associated with the BYOD environment,providing a robust framework for secure and efficient access management.