101 results about "Spectral entropy" patented technology

The invention discloses an electrical fire monitoring method based on inherent residual current automatic compensation, and relates to the technical field of electrical safety monitoring, and the method comprises the following steps: S100, collecting a residual current signal, executing multi-component time-frequency deconstruction processing, extracting an energy distribution characteristic through wavelet packet transformation, and extracting a mutability index in combination with short-time spectrum entropy analysis, and constructing a preliminary distribution map of higher harmonic interference, and determining boundary features of interference signals in a time domain and a frequency domain. According to the method, high-frequency interference signals are accurately positioned through time-frequency deconstruction, wavelet packet analysis and spectral entropy indexes, non-fault harmonic components are effectively eliminated in combination with amplitude-frequency coupling recognition and a recursive rejection strategy, closed-loop control is constructed by introducing an adaptive compensation and stability backtracking mechanism, accurate recognition and dynamic correction of real electric leakage risks are achieved, and the method is suitable for large-scale popularization and application. The identification precision and the safety reliability of the monitoring system in a complex industrial environment are obviously improved, and the method has good engineering adaptability.

The invention provides a wind turbine generator voiceprint fault recognition method, and relates to the technical field of wind turbine generator state monitoring and fault diagnosis, and the method comprises the steps: carrying out the noise reduction of an original audio signal through variational mode decomposition, screening a target mode of which the frequency, energy and kurtosis accord with features, and reconstructing the signal; extracting a Mel frequency cepstrum coefficient and a sensing noise robust coefficient, and generating multi-dimensional voiceprint data in combination with statistical characteristics such as a frequency spectrum gravity center, a spectrum entropy, energy, kurtosis and a zero-crossing rate; constructing a support set based on the prototype network, realizing small sample fault classification by calculating the Euclidean distance between the feature vector and the prototype vector, and outputting a preliminary result; judging whether the voiceprint is abnormal according to a preset threshold value, if so, storing the voiceprint into a dynamic abnormal voiceprint knowledge base; frequently occurring abnormal samples are manually labeled and added into a support set, the prototype network is retrained to update the model, and continuous optimization of the fault recognition capability is achieved.

The invention discloses a wellhead data acquisition modeling system. The wellhead data acquisition modeling system comprises a data acquisition module, a hydraulic characteristic processing module, a well layer coupling modeling module and an interaction control module, the invention belongs to the technical field of well water monitoring management, based on an adaptive time-frequency analysis method, a dynamic window function enhanced by local frequency spectrum entropy is introduced, feature extraction and abnormal early warning are performed on transient hydraulic fluctuation signals, and the time-frequency resolution and change recognition capability of the hydraulic signals are remarkably improved; a dynamic coupling model between a water pumping system and an aquifer is constructed by combining a multi-scale parameter coordination mechanism, and accuracy and dynamic adaptability of well layer response modeling are realized by introducing a gap compression lag function and a joint parameter calibration method; and the interaction control module implements model prediction control according to the hydraulic feature vector, the hydraulic anomaly early warning signal and the well layer coupling prediction parameter combination, and outputs wellhead data acquisition decision support data.

The invention discloses an automatic marking system for an abnormal wave band of an epilepsy electroencephalogram image, and relates to the technical field of epilepsy electroencephalograms, the system carries out frequency domain analysis and time domain feature extraction on an electroencephalogram signal through Fourier transform and convolution operation, abnormal waveforms such as sharp waves, ratchet waves and slow waves can be accurately identified, and the accuracy of the abnormal wave band marking is improved. The sensitivity of the system to fine abnormal wavebands is enhanced, highly similar abnormal wavebands can be combined into a complete abnormal event, redundant annotation and repeated event annotation are avoided, further refinement is performed through calculation of the spectrum entropy and the spectrum entropy difference value, setting of a reasonable threshold value and combination of the ratio difference value, and the abnormal event annotation accuracy is improved. The system can identify and label different epilepsy stages more accurately, the system optimizes labeling through multi-dimensional feature differences such as frequency spectrum entropy difference and ratio difference values of different wavebands, seizure, precursor and normality of epilepsy in different stages can be distinguished more accurately, and the multi-level labeling method effectively improves the epilepsy prediction capacity.

The invention belongs to the technical field of radar signal processing, and discloses a sea surface small target detection method based on optimized characteristic mode decomposition, which comprises the following steps: S1, acquiring to-be-detected signal data; s2, decomposing an original signal into a plurality of modal components by using FMD, and selecting an envelope spectrum entropy as a fitness function; s3, performing global optimization on the fitness function in the FMD by using an SOS algorithm; s4, introducing a PSO algorithm to carry out local optimization on key parameters of the FMD; s5, components with low envelope spectrum entropy values and correlation coefficients larger than a threshold value are reserved; s6, extracting an envelope spectrum entropy and frequency band energy ratio feature from the screened modal components, introducing a Gini coefficient as a weighting factor, and constructing a GSEBE joint feature; and S7, inputting the entropy value of the envelope spectrum into a DELM classifier with a controllable false alarm, and realizing target detection based on comparison between a predicted value and a judgment threshold. According to the invention, the capability of distinguishing sea clutters and target echoes is enhanced, and more accurate classification detection is realized.

The invention provides a high-frequency welded pipe roller wear state monitoring method and device, and relates to the technical field of artificial intelligence and data processing, a holographic sensing system of working condition parameters and physical signals is constructed through synchronous monitoring of a multi-source sensor, and noise spectrum offset is tracked in real time based on time-frequency expression of data in different modes, so that the real-time monitoring of the wear state of a high-frequency welded pipe roller is realized. And an anti-error filtering protection channel is constructed, so that the inherent characteristics of the damage can be effectively captured. In addition, multi-physical field damage characterization information is fused from multiple complementary dimensions of transient impact (instantaneous frequency vector), spectrum complexity (spectral entropy) and amplitude distribution deflection (skewness), more comprehensive and richer state description can be provided, accurate quantitative evaluation of the roller wear level is achieved, and the capacity of real-time monitoring and accurate recognition of the roller wear state is remarkably improved.

The invention discloses a logic board comprehensive performance testing device and testing method, and relates to the technical field of electronic equipment testing. A time sequence signal, a power consumption curve and infrared temperature data are synchronously acquired through a logic board interface, a time-space aligned fusion data set is generated by using a dynamic time warping algorithm, a power consumption coupling analysis model is input, a dynamic incidence matrix of a signal integrity parameter and power consumption fluctuation is acquired, and key coupling characteristics are extracted through singular value decomposition. And constructing an adaptive neighborhood density clustering model according to the correlation mode parameters and the temperature gradient distribution, generating an abnormal region probability graph, and marking electrothermal coupling abnormal coordinates and confidence. And inputting the fusion data, the correlation parameters and the abnormal probability graph into a graph neural network, taking a time sequence signal spectrum entropy, a local temperature mean value and a power consumption fluctuation variance as node features, taking an electrothermal coupling coefficient as an edge weight, iteratively updating a node state through a graph attention mechanism, and outputting a defect type and a three-dimensional coordinate. And the defect detection accuracy is effectively improved.

The invention discloses a non-cooperative underwater sound pulse signal detection method based on power-w and fractional order spectral entropy. The method comprises the following steps: performing fractional order Fourier transform on a received non-cooperative underwater sound pulse signal; the optimal order p of fractional Fourier transform is found through a traversal method; performing v-th power processing on the energy of the frequency domain signal after the p-order fractional order Fourier transform to obtain an improved fractional order power spectrum; performing normalization processing on the improved fractional order power spectrum E, and calculating an entropy value T corresponding to the data; calculating an entropy value of the background noise, taking the entropy value as a fixed threshold value, comparing the fixed threshold value with the entropy value obtained by calculation, and if the fixed threshold value is greater than the entropy value obtained by calculation, judging that there is a signal; otherwise, no signal exists. According to the method, the detection effect can still keep excellent under the condition of low signal-to-noise ratio, and the robustness is higher.

The invention relates to the technical field of artificial intelligence and data processing, and discloses an energy storage equipment operation state monitoring method based on Internet of Things acquisition, which comprises the following steps: constructing a vibration data sample; performing state labeling on the vibration data sample to form a labeling database; performing sliding mean filtering and linear interpolation processing on the vibration data sample; carrying out adaptive spectrum noise suppression by adopting a dual-threshold wavelet packet noise reduction function; carrying out frequency band division by adopting a weighted multi-scale spectral entropy feature enhancement method; constructing a deep neural network model for classifying the running state of the energy storage equipment, and training the deep neural network model by using a labeling database; and inputting preprocessed vibration data acquired in real time into the trained deep neural network model, and performing state identification and fault alarm according to a category probability prediction vector output by the model. The running state of the energy storage equipment can be accurately monitored in real time, potential faults can be found in time, and measures can be taken.

The invention relates to the field of online analysis of chemical components of materials, and discloses an online monitoring process for components of a lead frame electroplating solution, which comprises the following steps: conventionally monitoring the spectrum of the electroplating solution to calculate the concentration; calculating a spectral entropy index based on the deviation degree of the real-time spectrum and a pre-established statistical baseline model, and comparing the spectral entropy index with a preset failure trigger threshold; a dynamic calibration step is started only when the spectral entropy index exceeds the preset failure trigger threshold, the dynamic calibration step compares a physical true difference spectrum with a model virtual difference spectrum by injecting a standard pulse to generate a model error vector, and the model error vector is utilized to correct the chemometrics model, so that the stoichiometry accuracy is improved. According to the method, on-demand calibration is realized, and the consumption of the standard substance is reduced while the real-time accuracy of the model is guaranteed.

The invention discloses a Lidar point cloud registration optimization method based on geometric feature reliability, and the method comprises the steps: carrying out the preprocessing and initial alignment of a laser radar scanning point cloud, extracting the features of a maximum principal curvature and a minimum principal curvature based on the local curvature of the point cloud, dividing a point region into two types of geometric features of angular points and plane points according to the threshold value of the maximum principal curvature, and carrying out the registration of the angular points and the plane points. On the basis, fitting quality factors including fitting errors, local curvatures and spectral entropies of the linear features and the plane features are calculated respectively, then the three factors are fused into a unified reliability weight based on a Bayesian probability model, and finally the feature reliability weight is introduced into an optimization objective function of iterative nearest point registration to execute weighted ICP registration. And outputting positioning and attitude determination results. According to the method, the reliability of geometric features is quantitatively evaluated, and a weighted optimization framework is constructed, so that the point cloud registration precision and robustness are remarkably improved, and the problem that a traditional ICP algorithm is sensitive to unreliable features in feature degradation or high-dynamic scenes is effectively solved.

The invention relates to the technical field of power monitoring, in particular to an online monitoring method and system for a power control cabinet. The method comprises the following steps: acquiring a time domain signal, tracking a fundamental wave frequency, adjusting a sampling window, and performing Fourier transform to obtain a harmonic amplitude and a phase; harmonic waves are divided into harmonic wave correlation groups according to association rules, harmonic wave phase consistency coefficients are calculated to weight the total energy in the groups, and weighted harmonic wave energy is obtained; total inter-harmonic energy is calculated, and an inter-harmonic structure factor is obtained based on the spectral entropy. And acquiring a real-time load rate and a change rate, and calculating a synchronous change coefficient and a load factor. And summing the weighted harmonic energy and the corrected inter-harmonic energy to obtain total distortion energy, and obtaining a comprehensive distortion index according to the total distortion energy. According to the scheme, the structured inter-harmonics possibly caused by the early failure of the equipment can be found earlier, and the requirements of early warning and predictive maintenance are met.

The present invention discloses an automatic labeling system for abnormal bands in epileptic EEG images, which relates to the technical field of epileptic EEG. The system performs frequency domain analysis and time domain feature extraction on EEG signals through Fourier transform and convolution operations, and can accurately identify abnormal waveforms such as sharp waves, spike waves and slow waves, thereby enhancing the system's sensitivity to subtle abnormal bands and merging highly similar abnormal bands into a complete abnormal event, avoiding redundant labeling and repeated event labeling. By calculating spectral entropy and spectral entropy difference, setting a reasonable threshold, and further refining it in combination with ratio difference, the system can more accurately identify and label different epileptic stages. The system optimizes labeling through multi-dimensional feature differences such as spectral entropy difference and ratio difference of different bands, and can more accurately distinguish between onset, precursors and normal in different stages of epilepsy. This multi-level labeling method effectively improves the predictive ability of epilepsy.

The invention discloses a pulmonary nodule detection method based on multi-kernel representation learning, and belongs to the field of medical image analysis. The method comprises the following steps: S1, obtaining suspected region image block data of which the form is similar to that of a nodule, and carrying out standardization processing; s2, performing feature extraction through principal component analysis according to the normalized image blocks; s3, according to the extracted feature representation, calculating corresponding kernel matrixes by using multiple kernel functions; s4, according to the obtained multiple kernel matrixes, based on spectral entropy weighted fusion, obtaining a mixed kernel matrix; s5, training a single-class support vector machine model according to the mixed kernel matrix, and establishing a discrimination boundary of a normal sample; and S6, in a test stage, repeatedly executing the steps S1 to S4 on suspected nodule region image blocks acquired from the CT image of the patient, extracting features, constructing a mixed kernel matrix, inputting the trained single-class support vector machine model for judgment, and if the suspected nodule region image blocks are abnormal, outputting a nodule region until judgment of all candidate regions is completed. The problems that an existing single-core support vector machine model is poor in adaptability and sensitive to data are solved, the deep learning method depends on a large number of labeled samples, and robust pulmonary nodule detection is achieved under the condition that pulmonary nodule samples are scarce by means of the unsupervised characteristic of anomaly detection.

The invention provides a regional electromagnetic intelligent evaluation method and system based on spectral entropy weighting, and the method comprises the following steps: collecting electromagnetic spectrum data and the motion state of a patrol vehicle, and constructing a virtual aperture data matrix through the motion information; space covariance analysis and subspace projection are carried out on the matrix, multipath scattering interference is separated and eliminated, and a pure field intensity sequence is obtained through reconstruction. By calculating the normalized spectrum entropy of the Doppler power spectrum of the pure field intensity sequence, an inverse proportion weight coefficient representing the signal confidence is generated; and weighting the weight to a pure field intensity sequence, and mapping the pure field intensity sequence to a geographic space. And dynamically adjusting the density of the evaluation grid according to the field intensity gradient, and finally carrying out weighted centroid calculation based on inverse proportion weight to obtain a regional electromagnetic environment evaluation result. According to the method, the technical problem that a regional evaluation result is not accurate due to the fact that multipath components cannot be separated and single-point data quality cannot be evaluated in traditional mobile patrol is solved.

The invention discloses a multi-modal data processing method and system based on an attention mechanism, and relates to the technical field of deep learning, and the method comprises the steps: collecting a multi-modal data set, carrying out the multi-scale time sequence calibration through dynamic time warping, and obtaining a time sequence alignment data stream; performing cross-modal semantic association on the time sequence alignment data stream to form a multi-modal feature vector; performing sparse processing on the multi-modal feature vector by using a multi-head self-attention mechanism to generate potential sparse representation; and carrying out coarse graining analysis and fluctuation mode capture on the potential sparse representation, generating a feature sequence length and a variance descriptor, and carrying out spectral entropy calculation to obtain a data complexity score. According to the method, cross-modal semantic association is performed by using canonical correlation analysis, and meanwhile, differential processing is performed on samples with different complexities through the hierarchical adaptive processing model, so that dynamic matching of computing resources is realized, and the resource utilization rate of multi-modal data processing is remarkably improved.

The invention relates to the technical field of power cable fault detection and positioning, and provides a power cable fault discharge sound recognition method and system based on multi-characteristic fusion, and the method comprises the following steps: obtaining a sound signal of a power cable, and carrying out the adaptive framing windowing processing according to a period peak value; calculating an envelope variance of the sound signal and a local maximum attenuation coefficient; extracting a high-frequency modal energy ratio of the sound signal and calculating a normalized spectrum entropy; inputting the local maximum attenuation coefficient and the envelope variance data into a first classifier; inputting the high-frequency modal energy ratio and the normalized spectral entropy into a trained second classifier; and fusing the two classification results. Based on a self-correlation analysis self-adaptive windowing method and a multi-dimensional feature fusion discharge sound recognition strategy, a dual-classifier structure is constructed, and time-frequency information is fused, so that the accuracy and robustness of cable fault discharge sound recognition are effectively improved.

The invention discloses an LED packaging tiny defect detection system based on machine vision, and relates to the field of LED packaging defect detection, the system comprises a data acquisition module, a feature preprocessing module, an edge chain construction module, a tensor construction module and an analysis decision module, the data acquisition module obtains a multispectral image and constructs an image matrix; the feature preprocessing module calculates a spectral gradient tensor and a spectral entropy value, and a multispectral enhanced image matrix is obtained after filtering; the edge chain construction module fuses multiple features to construct a joint edge response function, and generates a dynamic edge chain; the tensor construction module constructs a three-layer tensor and decomposes the three-layer tensor to obtain a defect sensitive parameter vector; the analysis decision module constructs a multi-level early warning mechanism based on parameter vectors, defect judgment and disposal are achieved in combination with a case library, the system achieves accurate detection of LED packaging tiny defects, and the detection efficiency and adaptability are improved.

The invention provides a processing chatter signal feature sample extraction method and a chatter detection method and device, and belongs to the technical field of industrial signal processing, and the method comprises the steps: carrying out the optimization variational mode decomposition of an original processing chatter signal, and obtaining a plurality of IMF signal components; performing fusion index evaluation on the IMF signal components to obtain a quality score of each IMF signal component, and performing screening reconstruction on the IMF signal components according to the quality scores to obtain reconstructed signals; recursive hierarchical decomposition is carried out on the reconstructed signal based on a preset matrix operator to obtain a plurality of hierarchical node components, and singular spectrum entropy feature sample extraction is carried out on the hierarchical node components to obtain a feature sample data set; according to the method, weakening of short-time abrupt change features in signal segments is avoided through recursive hierarchical decomposition, local abrupt change features in signals are captured through singular spectrum entropy, sensitivity to short-time abrupt change or transient events is improved through combination of hierarchical decomposition and hierarchical singular spectrum entropy of singular spectrum entropy, and accurate feature samples are obtained.

The invention relates to a node entanglement perception and heterogeneous agent collaborative reinforcement learning defense method oriented to an electric power information physical system, and belongs to the technical field of electric power information physical systems. According to the method, firstly, a simulation model in which a power system layer and a communication system layer are tightly coupled is constructed, a node entanglement degree is introduced to serve as a core index for quantifying network toughness, and key fragile nodes are accurately identified by analyzing the sensitivity of network spectrum entropy to node disturbance. On the basis, a heterogeneous multi-agent reinforcement learning framework composed of a communication agent and an electric power agent is designed and is responsible for route optimization and network reconstruction respectively. By constructing a collaborative objective function fusing a voltage recovery reward and a node entanglement perception promotion reward, the intelligent agent is guided to actively enhance the overall robustness of the system network while recovering the power supply quality. And a hierarchical cooperative training mechanism is adopted, so that a heterogeneous agent learns a cooperative decision strategy, and an information physical cooperative defense mechanism under a cross-domain fault is realized.

A system and method are provided for identifying breath variability events from patient respiratory data obtained during the operation of a mechanical ventilator. The ventilator monitors and saves patient respiratory data during operation. The method includes extracting treatment parameters from a ventilation prescription, and using the treatment parameters to determine an Epoch size and a frequency band for analysis. An input signal (flow or pressure) is extracted from the patient data and from the input signal at least one parameter is extracted and analyzed. A Spectral Energy (Es) of the input signal is determined based on the signal parameter, an Epoch size and a frequency band and a Spectral Entropy (SE) is determined based on the Spectral Energy (Es). The Spectral Entropy may then be displayed in graphical form to identify breath variability events over the Epoch size. The method may further classify the type and degree of the events.

The invention provides an insurance accident scene video evidence chain generation and credible evidence storage method and system, and the system comprises an edge multi-mode connector suite, a unified data bus, a multi-task neural network engine and a target system adaptation layer. Real-time videos are collected through the industrial vision module; the data bus provides event-driven time sequence-document-streaming hybrid transmission; the neural network engine introduces a spectral entropy modulation activation function of a production line load rate and semantic confidence, performs multi-modal coding on equipment data and visual features, and outputs semantic mapping, load prediction and control instructions; and the adaptation layer pushes the instruction to an MES platform, a WMS platform and a quality detection SaaS platform to form a millisecond-level closed loop. According to the technical scheme, private protocol zero code access, semantic unification, self-adaptive scheduling and defect detection are achieved, the data compression rate and the defect recall rate are greatly increased, and traceable safety is achieved.

The invention relates to the technical field of power equipment state monitoring and fault diagnosis, and discloses a fusion type high-voltage switch cabinet insulation state quantitative monitoring method and system, and the method comprises the steps: firstly obtaining a partial discharge ultrahigh frequency and sound wave signal, and calculating a generalized Renyi spectrum entropy and an entropy gradient index; carrying out nonlinear correction on the medium sound wave propagation velocity by using the entropy gradient index, establishing an equivalent sound velocity model in a medium degradation state, and obtaining a standardized equivalent time difference; then, constructing a two-dimensional phase plane state space taking the standardized equivalent time difference and the generalized Renyi spectral entropy as dimensions, and analyzing a differential evolution trajectory of a state vector on a time sequence; and finally, constructing a comprehensive state loss function, mapping to generate an insulation health degree index, and carrying out life prediction. According to the invention, through medium degradation wave velocity constitutive mapping and phase plane conjoint analysis, monitoring errors caused by medium aging are solved, and dynamic accurate quantitative evaluation of the insulation state is realized.

The present application relates to the field of on-line analysis of material chemical composition, and discloses an on-line monitoring process for the composition of a lead frame electroplating solution, comprising: calculating the concentration of the electroplating solution by conventional monitoring of the spectrum; calculating a spectrum entropy index based on the degree of deviation of the real-time spectrum from a pre-established statistical baseline model, and comparing the spectrum entropy index with a pre-set failure trigger threshold; starting a dynamic calibration step only when the spectrum entropy index exceeds the pre-set failure trigger threshold, the dynamic calibration step generating a model error vector by comparing the physically real differential spectrum with the model virtual differential spectrum through injection of a standard pulse, and correcting the chemometrics model using the model error vector, so that the present application realizes on-demand calibration, ensures real-time accuracy of the model, and reduces the consumption of the standard.

A method for automatically detecting pregnancy based on pulse signals at the Cun, Guan, and Chi positions, characterized in that it includes the following steps: (1) reading three pulse wave signals synchronously detected from the Cun, Guan, and Chi positions of a woman, and performing the same preprocessing on them to remove baseline drift and high-frequency noise therein; (2) performing the same Fourier transform on the three preprocessed pulse wave signals to produce their respective amplitude spectra or power spectra, and calculating their respective comprehensive spectral entropy indicators based on the above-mentioned Fourier transforms, and selects the signal with the smallest comprehensive spectral entropy indicator among the three signals, that is, the signal with the best and cleanest quality; (3) for the selected signal with the smallest comprehensive spectral entropy indicator, a frequency point is selected on its amplitude spectrum or power spectrum, and the ratio of the sum of the spectral line amplitudes greater than the frequency point to the sum of the spectral line amplitudes less than the frequency point is calculated. If the ratio is greater than a certain threshold, the woman can be determined to be pregnant; if the ratio is less than a certain threshold, the woman can be determined to be non-pregnant.

The invention discloses an enterprise energy consumption management method based on data fluctuation load analysis. The method comprises the following steps: collecting multi-dimensional real-time operation data of a power enterprise production system; in a sliding time window, extracting fluctuation characteristics such as standard deviation, kurtosis and spectral entropy from the time sequence data of the key parameters, and converting the fluctuation characteristics into fluctuation characteristic vectors; constructing a data fluctuation load model, and learning a non-linear relationship between fluctuation feature vectors and system energy efficiency and carbon emission indexes; on the basis of a model prediction result, a fluctuation load index (FLBI) is calculated and used for quantitatively evaluating the health bearing capacity of the system for operation fluctuation; and finally, according to the FLBI value and the change trend thereof, proactive sub-health early warning, root cause diagnosis and accurate optimization control suggestions are generated. According to the invention, the information is used as core information for analyzing the dynamic health degree of the system, and the transformation from post-event alarm to pre-event early warning is realized.

This invention relates to the fields of artificial intelligence and data processing technology, and discloses a method for monitoring the operating status of energy storage devices based on IoT data acquisition. The method includes: constructing vibration data samples; labeling the vibration data samples to form a labeled database; performing moving average filtering and linear interpolation on the vibration data samples; using a dual-threshold wavelet packet denoising function for adaptive spectral noise suppression; using a weighted multi-scale spectral entropy feature enhancement method for frequency band division; constructing a deep neural network model for classifying the operating status of energy storage devices, and training the deep neural network model using the labeled database; inputting the preprocessed vibration data acquired in real time into the trained deep neural network model, and performing status identification and fault alarm based on the class probability prediction vector output by the model. This invention can monitor the operating status of energy storage devices in real time and accurately, promptly detect potential faults, and take appropriate measures.

The present invention relates to the technical field of power cable fault detection and location, and proposes a method and system for identifying power cable fault discharge sounds based on multi-feature fusion, comprising the following steps: acquiring the sound signal of the power cable, and performing adaptive framing and windowing processing based on the periodic peak value; calculating the envelope variance and local maximum attenuation coefficient of the sound signal; extracting the high-frequency modal energy ratio of the sound signal and calculating the normalized spectral entropy; inputting the local maximum attenuation coefficient and envelope variance data into a first classifier; inputting the high-frequency modal energy ratio and normalized spectral entropy into a trained second classifier; and fusing the two classification results. Based on the adaptive windowing method of autocorrelation analysis and the discharge sound recognition strategy of multi-dimensional feature fusion, a dual classifier structure is constructed and time-frequency information is integrated, effectively improving the accuracy and robustness of cable fault discharge sound recognition.

The invention relates to the technical field of non-contact vital sign monitoring, and discloses a millimeter wave radar vital sign monitoring method and system based on VMD, and the method comprises the steps: obtaining sample data containing vital signs; performing multi-feature fusion analysis to obtain a frequency band energy ratio, a frequency spectrum entropy, signal-to-noise ratio estimation and a peak value concentration ratio of the vital signs; judging whether the frequency band energy ratio, the frequency spectrum entropy, the signal-to-noise ratio estimation and the peak value concentration ratio of the vital signs meet preset conditions or not so as to match corresponding penalty factors and decomposition layers in a grading manner; decomposing the micro-motion signal by adopting a VMD algorithm to obtain a plurality of sample components; performing multi-feature weighted scoring on the sample components; screening the sample component by adopting a bidirectional interference suppression strategy to obtain a target signal; and based on the target signal, reconstructing to obtain a vital sign signal, and outputting a monitoring result. The method has the effect of improving the monitoring precision and timeliness of the vital signs in the real-time processing scene.

The invention discloses a target orientation estimation method, system and equipment based on a single-vector hydrophone, and belongs to the technical field of underwater acoustic signal processing. The method comprises the following steps: acquiring a sound pressure signal in a sound pressure channel of the single-vector hydrophone and a particle vibration velocity signal in a three-dimensional vibration velocity channel; constructing a matrix based on the sound pressure signal and the mass point vibration velocity signal; calculating weights, wherein the weights comprise sound energy flow weights and spectral entropy weights; introducing an adaptive coefficient to fuse the sound energy flow weight and the spectral entropy weight to obtain a joint weight; weighting the covariance matrix through the joint weight to obtain a feature R, and performing decomposition and dimension reduction on the feature R to obtain a dimension-reduced covariance matrix; and constructing an MVDR beam former based on the dimensionality-reduced covariance matrix, and carrying out azimuth spectrum scanning so as to carry out target azimuth estimation. According to the method, the energy information and the spectrum structure information of the vector hydrophone are fused, and the noise is suppressed in combination with the dimension reduction technology, so that high-precision target orientation estimation in a complex environment is realized.