50results about How to "Guarantee data quality" patented technology

The invention relates to the technical field of cell culture, and discloses a stem cell differentiation degree identification method based on image features. The method comprises the following steps: continuously capturing multi-temporal image data in a stem cell culture process through a high-resolution microscopic imaging system, and generating an image and environment synchronous data set; processing the data set, extracting dynamic morphological features and texture change modes of the stem cells, and constructing a fusion feature set; outputting a differentiation process index based on the fusion feature set; a differentiation degree grade division rule is set, and a differentiation process index and differentiation grade mapping table is established; in a real-time application stage, acquiring real-time imaging data and environment readings, and inputting the real-time imaging data and the environment readings into an evaluation network to calculate real-time differentiation process indexes; querying the mapping table according to the real-time index to obtain a differentiation grade judgment result; comparing the judgment result with a preset target range, and if the judgment result falls into the target range, automatically executing differentiation state marking and culture parameter adjusting operation.

The application discloses a lining cloth dyeing control method and system based on artificial intelligence, and is used for the control field, and the method comprises the following steps: real-time monitoring of multi-source data in the lining cloth dye vat, capturing the video stream inside the lining cloth dye vat; using a Gaussian mixture variational autoencoder model to extract features from the multi-source data, obtaining multi-source feature data; obtaining the changing moment of the fabric state in the video stream, and extracting the fabric state change feature data according to the observed fabric color change and the heterogeneity of the dye distribution in the video stream; fusion of multi-source feature data and fabric state change feature data; introducing a model predictive control layer, predicting the result to adjust the dyeing parameters in real time; establishing an intelligent feedback mechanism. The Gaussian mixture variational autoencoder model is trained and optimized through the small-batch stochastic gradient descent and back propagation algorithm, important features are learned and captured from the data.

This invention belongs to the field of spectral data processing technology, specifically a spectral data processing system and method based on in-situ data sensing and intelligent analysis. This invention employs in-situ optical imaging, absorption spectroscopy, fluorescence spectroscopy, and PL mapping to acquire real-time three-dimensional spectral data of time, wavelength, and intensity for real-time monitoring of material crystallization, defects, and luminescence properties. It uses an automatic invalid value detection and cleaning algorithm to address data quality issues; it extracts peak position and half-peak width (WHM) characteristic parameters from the spectral data using peak position extraction and WHM calculation methods; it uses a nonlinear optimization algorithm to perform multi-peak Gaussian fitting on the spectral data; and it employs a signal-slot mechanism to achieve real-time synchronization of data presentation and fitting analysis, forming a complete closed loop of in-situ testing, data sensing, processing, analysis, synchronization, and export. This invention achieves real-time acquisition, efficient processing, accurate analysis, and automatic synchronization of spectral data, providing technical support for standardized processing and traceable analysis of spectral data.

The invention discloses a mathematical model-based chemical process simulation data management system and method, and relates to the technical field of data management. The system comprises a data feature system construction module, a dynamic mapping rule generation module, a cross-scale collaborative data processing module and a matching relation adaptive optimization module. The method comprises the following steps: constructing a multi-scale data quantization feature system and generating feature vectors; establishing a dynamic mapping mechanism and an association matching rule set, wherein the association matching rule set comprises a cross-scale collaborative weight and a working condition dynamic threshold factor; checking data in multiple dimensions, complementing missing attributes and pushing the data as required; the feature vectors and mapping rules may be iteratively optimized through deviation analysis. The problem of data and model hidden mismatching is solved, working condition fluctuation is adapted, cross-scale simulation accuracy and robustness are improved, and safe and efficient circulation of data is achieved.

This invention relates to the field of perioral muscle function detection technology, and discloses a flexible sensing integrated tongue muscle function detection system and method. The system includes: a brace-type flexible thin-film sensor, a wireless transmission module, and a control module. The brace-type flexible thin-film sensor is configured as a wearable brace structure, and internally contains a sensor array composed of multiple sensing units distributed in the maxillary and lateral tooth regions corresponding to tongue muscle activity. The wireless transmission module is located in the lateral region of the upper molars and is connected to the brace-type flexible thin-film sensor via a flexible printed circuit. The control module is connected to the wireless transmission module and includes a calibration module, an initial value determination module, a correction module, and a data storage and communication module. This invention achieves precise, non-invasive, and real-time monitoring of tongue muscle activity.

The application relates to the technical field of Internet of Things and edge computing, and discloses an Internet of Things data access and processing system based on edge computing, which comprises a data acquisition module, an environment sensing module, a data preprocessing module, a transmission strategy analysis module, an edge distributed analysis module, a multi-source data fusion module, a deep intelligent analysis module, a resource scheduling analysis module and a global resource optimization module; through the arrangement of the nine functional modules, the whole-process processing of Internet of Things data from access, preprocessing, transmission, edge analysis, multi-source fusion, deep intelligent analysis to resource scheduling optimization is realized; compared with the prior art, the data processing delay is significantly reduced, the resource utilization rate and the data processing accuracy are improved, the application is suitable for various Internet of Things application scenarios, and has wide popularization value.

The application belongs to the technical field of industrial internet data communication, and relates to a method and system for intercommunication and adaptation of industrial data in a multi-protocol environment, comprising: collecting original data carrying protocol identifiers from a device layer, an application layer and a platform layer based on a pre-set three-layer interconnected bus; performing analysis and processing on the original data based on a pre-set protocol stack, and obtaining intermediate data in a unified format; performing semantic adaptation on the intermediate data based on a pre-set information exchange semantic model, and obtaining semantic specification data; performing data preprocessing on the semantic specification data, and obtaining standardized semantic specification data; and performing layered transmission of target data in the standardized semantic specification data through a corresponding interconnected bus according to a specified interaction mode of a target layer. The application can improve the protocol coverage rate of an industrial scene, realize semantic unification and data intercommunication, improve the cross-system collaborative analysis efficiency, and reduce the deployment time and configuration cost.

The present application relates to the field of wind power generation technology, and discloses a wind turbine fault diagnosis and early warning method, device, equipment and medium, the method comprises the following steps: obtaining various types of faults corresponding to the physical characteristics generated by the diagnosis object and the working condition information associated with the physical characteristics; the obtained physical characteristics and working condition information are identified and classified, and the principal component characteristics, auxiliary characteristics, and target working condition information associated with the principal component characteristics and auxiliary characteristics are obtained; the principal component characteristics and auxiliary characteristics, target working condition information and single feature threshold value set by the user are fused and analyzed; the alarm conclusion of the diagnosis object obtained after the analysis and processing is output and displayed. The wind turbine fault diagnosis and early warning method can adapt to the habits of users, automatically analyze and diagnose the fault of the diagnosis object for different users, has multi-feature fusion analysis capability, improves the fault analysis efficiency and fault diagnosis and early warning accuracy.

The application discloses a kind of dynamic regulation and grid-connected response control methods of new energy station, through real-time data acquisition, multi-parameter fusion analysis and dynamic regulation strategy, realize the stable grid-connected operation of new energy station and power grid, first, data acquisition and pretreatment are carried out, respectively collect power generation parameters, power grid parameters and environmental parameters, and through data cleaning, outlier processing and normalization, ensure data quality, based on preprocessed data, three kinds of regulation models are constructed, power regulation, voltage regulation and frequency regulation, based on regulation model, the key data of new energy station before grid connection is regulated, finally, after grid connection evaluation and feedback, if current control strategy is unqualified, control strategy is re-adjusted, through the combination of data driving and model prediction, the transformation of new energy station from "passive compliance" to "active support" power grid is realized, and the grid-connected reliability and power grid consumption capacity are significantly improved.

The present application relates to the technical field of simulation sandbox, and particularly relates to a simulation sandbox data base construction method and device, wherein the method constructs compatible domain object models in the localization BIM forward design stage; according to unstructured drawings, localization BIM models and other data types, SHP data, geometric feature data and structured topology data are extracted or generated; simulation element data is obtained through field-level direct mapping, SHP data is converted twice and coordinates are unified, and then geometric feature completion, line segment merging and topology reconstruction are performed, object construction and labeling are completed based on the simulation element data, and simulation sandbox data base is obtained. The present application solves the problems of low construction efficiency, poor compatibility and insufficient geometric accuracy, realizes standardized data processing and efficient construction, can be directly loaded into a simulation sandbox, is suitable for airport and other scenes, and improves simulation accuracy and reliability.

This invention provides a lithium carbonate price prediction method based on multimodal data collaborative driving, comprising the following steps: multimodal data source identification and data acquisition; multimodal data classification and standardization; personalized noise filtering and outlier correction of multimodal data; extraction of trend and correlation features from structured data; extraction of semantic and visual / speech features from unstructured data; extraction of key information and feature quantification from semi-structured data; weighted collaborative fusion and feature optimization of multimodal features; construction and training of a two-branch collaborative prediction model; model prediction and preliminary verification of prediction results; optimization and dynamic correction of prediction results; interpretability analysis and visualization of prediction results; and method effectiveness verification and continuous improvement. This method, based on multimodal data collaborative driving, addresses the pain points of existing lithium carbonate price prediction technologies, forming a comprehensive, high-precision, and highly practical prediction solution.

The invention discloses a dual-mode electrophysiology acquisition system and method for a multi-mode fingertip monitor, and belongs to the technical field of medical electronics and biological signal measurement. The system is integrated in fingerstall type wearable equipment supporting double working modes of high-fidelity recording and continuous monitoring, and is characterized by comprising at least two electrodes in contact with the skin; the high-precision bioelectricity analog front-end chip conforms to the medical standard; a low power consumption bioelectricity monitoring unit; and the channel switching circuit is controlled by the main controller. The high-precision analog front-end chip is configured to switch or perform time division multiplexing on at least one signal input channel between a first configuration for collecting electrocardiogram signals and a second configuration for collecting skin electrical activity signals by dynamically reconfiguring an internal register of the high-precision analog front-end chip; therefore, hardware-level synchronous acquisition of two kinds of bio-electricity signals in a high-fidelity recording mode is realized, and sampling of the two kinds of bio-electricity signals is based on the same main clock source. Through the architecture of dual-core isomerism, mode driving and hardware dynamic multiplexing, the contradiction between medical-level precision electrocardio / skin electricity synchronous acquisition and ultra-long endurance continuous monitoring which are simultaneously realized in micro fingerstall equipment is solved.

This invention relates to the field of artificial intelligence technology, specifically to a method, apparatus, and system for training large-scale industry models based on efficient fine-tuning, solving the technical problem of poor training data quality in existing technologies. The method includes: collecting multi-source heterogeneous data from various power plants; the multi-source heterogeneous data includes meteorological time-series data, power generation time-series data, a set of operation and maintenance records, and geographical environmental parameters; each operation and maintenance record in the set records the type and time of faults occurring at the power plant; performing correlation analysis on the multi-source heterogeneous data from various power plants to obtain the impact sequence corresponding to each operation and maintenance record; performing similarity analysis on the impact sequence corresponding to each operation and maintenance record to determine the importance index corresponding to each operation and maintenance record; processing the impact sequence based on the importance index corresponding to each operation and maintenance record, and training a large-scale industry model using the training dataset obtained from the data processing.

The application discloses an indoor situation generation method and device combining thin-plate spline interpolation and natural neighborhood interpolation and a medium, and relates to the field of electric digital data processing.The method comprises the following steps: acquiring original sampling data; performing grid division, merging and extracting data falling into the same grid to obtain a first sample data set; generating preliminary interpolation data by using thin-plate spline interpolation; extracting supplementary sampling point data from the preliminary interpolation data in an area not covered by the original sampling data to obtain a supplementary sample data set; performing grid division and merging to obtain a second sample data set; combining the supplementary sample data set and the second sample data set, filtering the supplementary data, and obtaining a target sample data set; and generating final situation data by using natural neighborhood interpolation.The application can effectively improve the situation generation quality under the condition of uneven sample distribution by filling in sparse area samples through thin-plate spline interpolation and finely processing the samples through natural neighborhood interpolation.

The invention discloses a pressure instrument detection data processing method and system based on cloud computing. The method comprises the following steps: collecting a feature data set through an edge computing node; the feature data set is uploaded to a cloud platform, an XGBoost and GNN graph neural network are combined to construct a multi-modal anomaly detection model, a cross-modal attention mechanism is utilized to fuse numerical features and spatial features to obtain fused features, the fused features are predicted, and a diagnosis report is generated and at least comprises fault recognition, health degree evaluation and residual life prediction; and the cloud platform generates an adjustment instruction according to the diagnosis report, and issues the adjustment instruction to a corresponding edge computing node to guide an edge side to update a preprocessing strategy. And the fault identification accuracy is greatly improved.

The invention discloses an on-line monitoring method for an ignition symptom of an easy-spontaneous-combustion multi-coal-seam goaf, and belongs to the technical field of coal mine safety monitoring. According to the method, multi-coal-seam stratified gas and temperature data are synchronously collected through photoacoustic spectrometry and infrared temperature measurement, and data transmission is awakened in a time-sharing mode through an underground low-power-consumption ad hoc network; a solid-gas-heat multi-field coupling seepage model is combined with a Lagrange particle tracking and physical information neural network, a gas migration path is reduced, and extrinsic abnormal data of interlayer gas blow-by and ventilation interference is eliminated; fusing oxidation kinetic parameters through a space-time attention Transformer network, and introducing dynamic causal discovery and time sequence comparative learning to suppress pseudo-correlation interference, so as to realize advanced identification and positioning of an ignition symptom; and uploading the result to a ground terminal. The method can reduce the false alarm rate, and improves the early-stage monitoring accuracy and reliability of spontaneous combustion of the multi-coal-seam goaf.

The application discloses a cross-medium laser radar detection system and method. The system comprises a transmitting module, a receiving module and a control processing module. The transmitting module generates a large pulse laser through the same laser, which is divided into a detection beam and an acoustic beam by a beam splitter, and respectively enters a water body to be detected to form the detection beam and the acoustic beam; the receiving module receives reflected light signals and reflected sound signals through first and second receiving units, and converts them into corresponding first and second electric signals; the control processing module fuses the two types of electric signals to obtain a detection result. The method combines the advantages of optical short-range high-resolution and acoustic long-range wide coverage through photoacoustic synchronous detection, solves the problem that a single detection method cannot balance distance and resolution, realizes cross-medium detection of underwater targets in full range and high precision, is suitable for underwater search and rescue, resource exploration and other scenes, and improves the reliability and efficiency of detection.

The application belongs to the technical field of detection, and particularly relates to a karst cave three-dimensional model automatic construction method based on multi-source point cloud fusion, which comprises the following steps: step S100, multi-hole multi-source sensor cooperative arrangement and original detection data acquisition; step S200, single-hole detection data preprocessing and karst cave space contour point cloud data extraction; step S300, multi-hole point cloud constrained by control points is registered and globally fused, the karst cave space contour point cloud data of all detection drill holes are registered and fused, and global karst cave point cloud and fusion quality signals are output; and step S400, a karst cave three-dimensional grid model or a fusion quality insufficient prompt information is output. The present application executes a three-level registration fusion strategy, and has high multi-hole registration precision. The present application effectively solves the error accumulation problem in multi-hole data fusion, ensures uniform distribution of overall fusion precision, and is suitable for large-scale chain-bead string-shaped karst cave detection scenes with many caves and many drill holes.

The invention discloses a geological operation-oriented operator mental fatigue real-time monitoring method and device. The geological operation-oriented operator mental fatigue real-time monitoring method comprises the following steps: S1, determining the layout of a measured target and a millimeter wave radar; s2, millimeter wave radar intermediate frequency echoes are obtained, after a measured target is positioned through a distance domain, complex phases of corresponding distance information are extracted, phase unwrapping and steady state correction are carried out, and a continuous phase sequence representing chest micrometric displacement is obtained; s3, IIR digital band-pass filtering is applied to the continuous phase sequence, and the continuous phase sequence is decoupled into a breathing channel and a heartbeat channel; s4, extracting time domain and frequency domain features of heart rate variability (HRV) and respiratory rhythm from the respiratory channel and the heartbeat channel in the sliding time window to form a multi-dimensional vital sign feature vector set; s5, taking the multi-dimensional vital sign feature vector set as input, performing incremental learning by using OS-ELM, and judging and outputting a mental fatigue level in real time; s6, implementing hierarchical risk management, and triggering early warning and intervention measures; and S7, locally storing and visually displaying the fatigue level and the event log, and synchronizing the fatigue level and the event log to a remote management platform.

The invention relates to the technical field of silicon steel production, and discloses a silicon steel hot rolling convexity prediction and head convexity control method, device and equipment, and the method comprises the steps: obtaining silicon steel components, process parameters and historical data of corresponding hot rolling convexity, and constructing a training set and a test set after preprocessing; evaluating the model precision of the prediction model by using the test set, and if the model precision is not within a preset precision range, performing iterative tuning on the prediction model by using hyper-parameter grid search and feature variable optimization to obtain an optimized prediction model; and silicon steel component data and process parameter data in actual production are obtained, the convexity of the hot rolling target serves as an optimization target, and finish rolling adjustment and optimization process parameters are determined through the optimization prediction model. The prediction precision of the silicon steel hot rolling convexity is improved through the method of combining model parameter adjustment and optimization with characteristic variable adjustment and optimization, silicon steel head finish rolling process parameter optimization is achieved through the silicon steel hot rolling convexity prediction model, and the control precision of the silicon steel head hot rolling convexity is improved.

The application discloses an artificial intelligence psychological evaluation method and system based on multi-modal input, and belongs to the technical field of artificial intelligence, and solves the problem that the existing system does not consider individual differences of the tested person to generate a customized evaluation strategy, resulting in insufficient evaluation targeting of the tested person, the method comprises the following steps: generating a personalized evaluation psychological evaluation strategy of the tested person based on basic information of the tested person, collecting multi-modal evaluation information of the tested person in real time, and identifying and analyzing a comprehensive state characteristic vector based on a pre-constructed psychological evaluation model; the application utilizes basic information to construct an initial feature portrait, uses test appeal as a constraint, adopts a collaborative attention network to generate strategy demand information, and individually adjusts item priorities based on an adjusted feature portrait, so that an evaluation strategy that adapts to individual cognitive preferences and emotional characteristics is generated, different individual differences can be adapted to, and implicit and defensive psychological problems of the tested person can be accurately captured, and the accuracy and efficiency of psychological evaluation are improved.

The application relates to the technical field of plant maturity intelligent monitoring, and discloses a tree flower bud maturity intelligent monitoring method and system, which comprises the following steps: according to a pre-acquired tree flower image, analyzing flower bud distribution density, collecting multi-angle images in a region with a flower bud distribution density greater than a preset density threshold value to obtain a first image set; extracting a flower bud shape parameter, identifying a flower bud contour, and calculating a flower bud volume; extracting a morphological feature of an inner region of the flower bud contour, combining the flower bud volume, calculating a probability value of the flower bud belonging to a preset maturity grade through feature fusion analysis, and obtaining the maturity of the flower bud; combining the maturity of the flower bud and historical growth data, simulating a growth process of the flower bud, predicting a flower bud maturity time, and intelligently monitoring the maturity of the tree flower bud; the application can improve the prediction accuracy of the flower bud maturity time, realizes accurate harvesting and yield increase of the tree flower.

The invention discloses a device and a method for testing dissolved oxygen and ozone isotopes in water, and belongs to the field of gas stable isotope analysis and testing. Comprising a constant-pressure constant-current automatic sampling device and a gas sampling system, a certain volume of water sample in a sealed sample bottle is removed by using the constant-pressure constant-flow automatic sampling device for dissolved oxygen in water, the water sample is placed in the constant-temperature balance oscillation device for balancing, back flushing of a gas circuit is realized by using the gas sampling system, the influence of residual gas is eliminated, and then the sample is collected and purified. Purified gas passes through a multi-channel gas path switching system, separated oxygen is loaded into a gas stable isotope mass spectrometer, oxygen isotopes of dissolved oxygen are detected, and therefore automatic pretreatment of the dissolved oxygen in water and accurate testing of the three oxygen isotopes of the dissolved oxygen are achieved. According to the method, full-automatic operation is realized, man-made interference and air pollution are effectively avoided, and the detection precision and efficiency are remarkably improved.

The invention discloses an intelligent voltage regulating transformer fault diagnosis method and system based on an improved CNN-BiLSTM model. The method comprises the following steps: firstly, collecting historical data of a voltage regulating transformer for preprocessing; then training is carried out based on historical data of the voltage regulating transformer to obtain a transformer fault diagnosis model, the transformer fault diagnosis model is an improved CNN-BiLSTM model, and an ECA-TAM attention mechanism is introduced into the CNN-BiLSTM model; and finally, inputting the current operation data of the voltage regulating transformer into the transformer fault diagnosis model to obtain a fault diagnosis result of the voltage regulating transformer. According to the method, an ECA-TAM attention mechanism is introduced into the CNN-BiLSTM network model for voltage regulating transformer fault diagnosis, purer and more representative feature input can be provided, interference of redundant information on time sequence modeling is reduced, the key dynamic learning ability is improved, efficient and accurate diagnosis of voltage regulating transformer fault types is achieved, and the fault diagnosis efficiency is improved. The accuracy and timeliness of transformer fault prediction are improved, and stable and safe operation of an electric power system is guaranteed.

This invention discloses a method and system for joint prediction of meteorological, wind, solar, and energy storage power based on ultra-long context, belonging to the field of power system prediction technology. The method constructs a multi-agent collaborative architecture for data cleaning, time alignment, context generation, and joint prediction, achieving automatic processing and efficient organization of historical data on wind power, solar power, and energy storage. By introducing an ultra-long context input mechanism spanning seasons and years, it overcomes the limitations of traditional model context length, achieving unified modeling and collaborative representation of multi-source data from different frequencies. Based on this, it utilizes a large time-series model to conduct zero-sample joint prediction, and combines physical constraints such as wind power curves, solar irradiance, and energy storage state of charge to construct a wind-solar-storage coupling consistency correction mechanism to optimize the prediction results. This invention effectively improves the multi-source covariate fusion capability and prediction accuracy, possessing good generalization performance and engineering adaptability.

The present application relates to the technical field of industrial water treatment and intelligent control, in particular to a metallurgical wastewater advanced treatment system, comprising a multi-source sensing module, which obtains real-time water quality and hydraulic parameters through a sensor array, and calculates heat load prediction data in combination with production plans; a digital twin engine module, which inputs data into a water quality-thermal energy coupling model, simulates the evolution of the physicochemical properties of water flow, and generates a full-network scaling risk thermodynamic map; an adaptive control module, which analyzes electromagnetic wave modulation instructions based on the thermodynamic map, and generates cascade waterway scheduling instructions; and a field execution module, which dynamically adjusts electronic scale-inhibiting equipment parameters in response to the instructions, and controls valves to reconstruct water supply paths; the present application realizes dynamic matching of water quality and thermal energy through digital twin technology, converts passive chemical cleaning into active scale inhibition by physical fields, and effectively reduces the dependence on chemical reagents.

This invention discloses an intelligent real-time monitoring method and system for water conservancy projects, relating to the field of water conservancy project safety monitoring technology. The method includes: constructing the physical response relationship between various physical quantities and environmental quantities, a normal operating condition coupling characteristic reference benchmark, and anomaly mode judgment criteria; generating monitoring residuals; verifying the validity of the monitoring residuals based on the spatial consistency of the monitoring residuals at various measuring points of the same type of physical quantity, generating a set of valid monitoring residuals; performing time-delay compensation on the monitoring residuals of each physical quantity, generating a multi-physical quantity monitoring residual sequence; constructing a coupling feature vector, comparing it with the normal operating condition coupling characteristic reference benchmark, obtaining the deviation amount and deviation trend; matching the deviation amount and deviation trend with the anomaly mode judgment criteria, generating anomaly level identification results and early warning instructions. This invention achieves intelligent real-time identification and early warning of structural anomalies in water conservancy projects through multi-physical quantity coupling feature analysis and dynamic anomaly level judgment.

The invention provides an elevator door lock short circuit detection method, electronic equipment and a medium. The method comprises the steps that visual time sequence data, vibration time sequence data, spectrum time sequence data and environment time sequence data of an elevator door lock indicator light are obtained; extracting a first visual feature parameter and a second visual feature parameter; extracting a first vibration characteristic parameter and a second vibration characteristic parameter; extracting a first spectral characteristic parameter and a second spectral characteristic parameter; according to the first judgment condition, the initial state of the elevator door lock is determined; the optimization state of the elevator door lock is determined through a second judgment condition; and an improved Bayesian network is used for fusion decision making, and an elevator door lock short circuit detection result is obtained. According to the method, the spectral information of the indicator lamp is introduced, vision, vibration and environmental data are combined, deep physical characteristics such as wavelength drift and bandwidth change are extracted, an improved Bayesian network is constructed to realize dynamic fusion decision, an existing detection blind area is effectively broken through, and the detection accuracy and robustness in a complex environment are remarkably improved.

The application discloses a construction safety hidden danger detection method and system based on a multi-modal large language model, which comprises the following steps: adopting a streaming multi-modal data acquisition strategy to synchronously collect text logs, image streams and sensor data of a construction site in real time; using an edge computing device to perform parallel preprocessing on real-time multi-modal construction safety data streams to eliminate inter-modal time inconsistency; constructing a multi-modal data feature correlation graph to identify and repair modal data anomalies caused by environmental interference in real time; performing real-time feature extraction and fusion on the preprocessed data streams, and completing rapid coding and identification of cross-modal hidden danger features through a sliding window mechanism and attention weight dynamic adjustment; based on the real-time feature fusion result, performing instant analysis through an online hidden danger detection method; and according to the real-time detection result, completing real-time visualization and tracking of the hidden danger, thereby achieving timely early warning and tracking management of the hidden danger.

The invention relates to the field of data processing, and provides a data access and cleaning method and system based on entity snapshots. The method comprises the following steps: monitoring a trigger request in real time, triggering a conflict coordination mechanism when detecting that an uncompleted processing request exists in the same entity, generating an entity snapshot and storing the entity snapshot in a snapshot storage unit; a latest-version snapshot of the entity is called from the snapshot storage unit, the current access data and the latest snapshot are associated and bound through the entity ID, a data set only containing increment fields is obtained, hash verification is conducted on each increment field, and whether a data cleaning link is allowed to enter or not is judged; after entering a data cleaning link, entity snapshot data in a historical set time period is called from a snapshot storage unit, abnormity identification and trend prediction are performed by combining static rule verification and a machine learning model, and the machine learning model is updated adaptively, so that the data transmission and storage cost is reduced, and the implicit noise is accurately identified.