14511results about How to "Good effect" patented technology

The invention provides an unstructured text data calculation method and system, and relates to the technical field of natural language processing, and the method comprises the steps: carrying out the multi-level processing of input unstructured text data, dynamically adjusting the word segmentation granularity according to the word frequency distribution, constructing a part-of-speech co-occurrence matrix in combination with context semantic information, and extracting entity information. And fusing the part-of-speech co-occurrence matrix and the entity information to generate a hierarchical semantic tag sequence. Feature extraction units with different convolution kernel sizes are adopted to extract feature representation, and cosine similarity between different semantic hierarchies is calculated to establish an association weight matrix. A semantic enhancement vector is constructed based on entity information, and adversarial training is carried out to obtain a multi-modal semantic feature matrix. And calculating the semantic similarity between the fusion feature vectors for clustering, setting difficulty weight sorting according to the complexity, consistency and ambiguity of class clusters, then inputting the class clusters into a classifier, and iteratively optimizing a classification result to obtain a calculation result.

The invention relates to the field of geological disaster monitoring and early warning, in particular to a slope catastrophe early warning method and system. The method comprises the following steps: establishing a space-time database, storing and managing preprocessed slope data from multiple monitoring sources, and sorting to obtain a slope catastrophe data set; performing feature extraction on the slope catastrophe data set to obtain a key information set reflecting slope catastrophe features; analyzing the key information set to obtain a slope catastrophe value; calculating an early warning threshold value and a real-time displacement prediction value of the side slope based on the side slope data in the space-time database, and fusing the real-time displacement prediction value and the side slope catastrophe value to obtain a fusion value; and evaluating the real-time stable state of each monitored section of the slope according to the obtained fusion value and an early warning threshold value, and judging whether to send out an early warning signal and an alarm signal or not according to an evaluation result. According to the invention, accurate identification of the slope catastrophe value and the potential risk area is realized, and the accuracy and reliability of the early warning system are improved.

The invention relates to the technical field of data processing, and discloses a BIM-assisted steel structure construction safety management early warning method, device and equipment. The method comprises the following steps: carrying out multi-source heterogeneous data acquisition and space-time alignment processing on a steel structure construction site to obtain an engineering data stream; carrying out dynamic noise reduction and feature extraction on the engineering data flow to obtain a standardized multi-modal feature vector, and carrying out time sequence correlation analysis and mode mining to obtain a potential risk index set; performing multi-level threshold calculation and context sensitivity evaluation on the potential risk index set to obtain an adaptive early warning trigger matrix; performing multi-dimensional cross validation and causal chain analysis on the self-adaptive early warning trigger matrix to obtain a graded and classified risk assessment report; and carrying out scene analysis and response scheme generation on the graded and classified risk assessment report to obtain an executable multi-step emergency plan. According to the invention, the efficiency and accuracy of BIM-assisted steel structure construction safety management early warning are improved.

The invention discloses an intelligent management method and system for IT equipment asset maintenance, and relates to the technical field of intelligent management. During system operation, multi-dimensional parameters of IT equipment are collected in real time and transmitted to a data processing module, the collected data are preprocessed, the current state of the equipment is comprehensively analyzed, and the health state coefficient Hstate of the equipment is calculated; the method comprises the following steps: acquiring a current equipment operation state, performing baseline establishment on historical operation data of the equipment, comparing the current state with the historical data, analyzing the performance trend and change of the equipment, calculating to obtain a trend analysis coefficient Ttend, and identifying potential problems and performance reduction trend of the equipment through long-term data accumulation. And calculating a final fault prediction probability index Pfault, comparing the final fault prediction probability index Pfault with a set threshold value, obtaining a grade prediction scheme, generating early warning information, prompting a system administrator to sort and distribute according to the urgency and priority of tasks, automatically generating a comprehensive report by the system, and providing decision support for the administrator.

The invention discloses an elevator control system based on artificial intelligence, which belongs to the technical field of elevator control, and comprises a plurality of layers of architectures such as data acquisition, transmission, processing, intelligent decision, execution control and the like. The data acquisition layer collects multi-source data through a car inside and outside and operation state sensor group; the data processing layer provides support for intelligent decision-making through preprocessing, analysis mining and model training optimization. An elevator dispatching module of the intelligent decision-making layer carries out peak dynamic zoning and flat peak energy-saving dispatching according to information such as passenger flow; the safety protection module monitors and processes hidden dangers in real time; the personalized service module provides customized services based on passenger behaviors and demands. And driving of the control layer, a door machine and a car environment control system are executed, and stable lifting of the car, safe operation of a door and intelligent environment adjustment are achieved correspondingly. Experiments show that the system can remarkably improve the elevator operation efficiency, safety and passenger satisfaction and reduce energy consumption, and has important application value.

The invention discloses a forest defense method and system based on digital twin and intelligent sensors, and the method comprises the steps: collecting and monitoring environment data, and carrying out the integration processing through data fusion; a big data analysis technology is adopted to analyze a potential risk area where a fire occurs, a high-risk area is identified by evaluating a risk level, and a fire spreading trend is predicted; constructing a three-dimensional virtual model of the forest region by adopting a digital twinborn technology; the spreading conditions of the fire at different time and spatial positions are simulated; according to the predicted fire spreading trend, when a potential fire risk is detected, an early warning mechanism is started; automatic generation and screening of the emergency plans are realized, and the emergency plans are generated. The invention provides an emergency protection system integrating the functions of multi-source sensing data fusion, intelligent early warning analysis, dynamic emergency plan generation and the like. In addition, the problem that a traditional emergency plan is static and cannot be dynamically adjusted according to actual conditions is solved.

The invention provides an ocean wave disaster defensive area demarcation optimization method based on ship AIS data, and belongs to the technical field of natural disaster prevention and control, and the ocean wave disaster defensive area demarcation optimization method based on ship AIS data comprises the steps: dividing a target sea area into regular grids, and building coordinate mapping; cleaning the data, removing abnormal values and segmenting the data; a distributed track segment pre-clustering method is adopted, and clustering is carried out through dimension reduction, feature extraction and an Euclidean distance algorithm; establishing a multi-dimensional feature tensor according to ship length classification; constructing a composite sea wave disaster danger index model, and comprehensively evaluating the risk; carrying out statistics on ship track space-time characteristics in each grid, and constructing a passing frequency tensor; fusing the trajectory frequency and the disaster risk model, and constructing a dynamic risk assessment hypergraph; based on multi-criterion threshold clustering, defensive areas are identified, boundaries are optimized, and key defensive areas are delimited; according to the invention, sea area risk conditions can be provided for marine mobile disaster-bearing bodies such as sailing ships and the like.

The invention relates to the technical field of power grid fault repair, in particular to a power grid fault emergency rescue method. The method comprises the following steps: acquiring topological structure data, key equipment parameters and real-time operation state data of a power grid; constructing a power grid digital twin model according to the power grid topological structure data, the key equipment parameters and the real-time operation state data, and performing real-time state evaluation to obtain fault risk early warning data and real-time fault condition data; when the real-time fault condition data shows that a fault occurs, performing fault point positioning and fault type determination based on deep learning according to the fault risk early warning data and the real-time operation state data to obtain fault diagnosis data; and constructing a fault influence range model according to the fault diagnosis data and a preset FACTS and HVDC characteristic model. According to the method, deep learning, a graph theory and a multi-objective optimization technology are fused, and intelligent fault identification, influence range modeling and accurate emergency disposal scheme optimization are realized.

The invention discloses a real-time relative pose estimation method in an unmanned aerial vehicle-unmanned vehicle cooperative system, and the method comprises the steps: 1, achieving the real-time pose estimation of each platform through visual feature point tracking and IMU pre-integration through a monocular camera and IMU data, and constructing a sparse feature point cloud map; step 2, combining laser radar point cloud data with an IMU, and applying a Fast-LIO algorithm and an IESKF to complete dense three-dimensional point cloud map construction of each platform; and step 3, fusing the data in the step 1 and the step 2, constructing a factor graph optimization problem through point-surface matching residual error, visual reprojection residual error and IMU pre-integration residual error factors, and solving the relative pose in real time. According to the method, vision and laser radar information are integrated, the pose estimation precision and robustness are improved, the method is suitable for an unstructured environment, and the collaborative operation efficiency is enhanced.

The invention discloses a dual-prevention intelligent interaction system based on a large language model, and belongs to the technical field of artificial intelligence. The system comprises a big language base model, a knowledge base management module, an agent module, a multi-modal big model and an evaluation module, and the big language base model understands user intentions and processes natural language tasks and provides token processing capacity for other modules; the knowledge base management module is used for classified storage and management of professional knowledge, establishing indexes according to semantics and finally performing localized storage; the intelligent agent module is used for analyzing input of a user, and different workflows are packaged in different intelligent agents so as to provide personalized services for the user; the multi-modal large model module receives voice input by a user, converts the voice into characters, transmits description of information to the Agent, generates answers by analyzing semantics, converts the characters into the voice and sends the voice to the user; the evaluation module establishes an evaluation index system from the four aspects of accuracy, answer correlation, context correlation and context recall rate.

The invention relates to the field of hard disk data protection, in particular to a hard disk data protection method and device based on artificial intelligence. The method comprises the following steps: acquiring a hard disk history monitoring log; performing multi-period hard disk load analysis on the historical hard disk monitoring log, and performing load peak fluctuation fitting so as to construct a hard disk load peak fluctuation curve; carrying out peak value trend evolution analysis on the hard disk load peak value fluctuation curve, and carrying out abnormal load peak value detection to obtain a time period where an abnormal peak value point is located; carrying out current time period redundant data calculation on the hard disk according to the time period of the abnormal peak point, and carrying out dynamic global data backup decision, so as to generate an abnormal peak global backup strategy; and performing abnormal write behavior detection and attack node tracing positioning on the hard disk historical monitoring log, and marking a hostile attack tracing positioning point. According to the invention, safe and reliable hard disk data protection is realized.

The invention relates to the technical field of unmanned aerial vehicle clusters, in particular to an unmanned aerial vehicle cluster formation method. Comprising the following steps: system initialization and configuration, cluster communication and data exchange, collaborative algorithm calculation and control input, positioning and navigation assistance, global and dynamic task planning setting, energy management and optimization strategy implementation, and biological enlightenment formation control application. Compared with a traditional formation operation method, the unmanned aerial vehicle cluster formation method is higher in autonomy, and the unmanned aerial vehicle cluster formation is more accurate through an advanced autonomous flight control system and a sensor technology. Information sharing and cooperative actions among the unmanned aerial vehicles in the cluster formation are realized through an efficient communication technology, and even if part of the unmanned aerial vehicles fail, other unmanned aerial vehicles can still continue to execute tasks.

The invention discloses an insulator defect detection method based on multi-scale features and channel perception, and belongs to the technical field of computer-aided algorithms. The method comprises the following steps: firstly, constructing a multi-scale feature fusion module, and fully extracting context information by combining multi-scale feature information and through a series of parallel deep convolution; secondly, a self-adaptive channel sensing down-sampling module is embedded in the backbone network, so that small-scale feature information is prevented from being lost during down-sampling of a feature map, and the sensing capability of the model to important channel information is improved; and finally, on the basis of adding a small-scale target detection layer, designing a lightweight shared phantom convolution detection head so as to improve the multi-scale target detection capability of the model under the condition of reducing the model parameter quantity. According to the method, a good effect is achieved on insulator defects of different types and multiple scales, the average detection precision is 91.4% and is improved by 5.2% compared with a reference model, and the overall performance is superior to that of other insulator defect detection algorithms.

The invention discloses a multi-type energy complementary island micro-grid coordination control method and system, and relates to the technical field of energy coordination control, and the method comprises the steps: collecting an island micro-grid structure and a predetermined prediction time zone, and carrying out the historical data analysis; performing energy coordination control on the island microgrid based on the energy coordination control scheme; and performing simulation evaluation through a simulation tool. According to the island micro-grid coordination control method for multi-type energy complementation, the island micro-grid structure and the preset prediction time zone are collected for historical data analysis, the accuracy and consistency of data analysis are improved, different power utilization levels are set according to the actual power utilization condition and management requirements of the island, data are arranged according to the time sequence, and the power utilization rate of the island is improved. The power supply reliability and efficiency are improved, power generation prediction is carried out based on the prediction model, the perspectiveness and accuracy of energy management are improved, and better effects are achieved in the aspects of the power supply reliability, the power generation prediction precision and the power generation prediction efficiency.

The invention discloses a remote sensing surveying and mapping image enhancement processing method for an unmanned aerial vehicle, and relates to the technical field of image enhancement processing. During operation of a system, flight attitude and position information are recorded through GNSS and IMU equipment, the flight path of the unmanned aerial vehicle is controlled, and collected original images are primarily processed; calculating and obtaining a flight stability coefficient Fs, an image acquisition quality coefficient Iq and a topographic feature retention coefficient Tf, removing random noise in a remote sensing image by using an improved non-local mean NLM algorithm, optimizing and processing an edge region, improving edge definition and texture details of a topographic image through a Laplacian pyramid algorithm, strengthening landform features of mountains and hills, and obtaining a remote sensing image; according to the method, color distortion caused by complex mountain illumination conditions is corrected, image colors are enabled to truly restore topographic features, important landform information is prevented from being lost due to over-brightness or over-darkness, topographic feature extraction is optimized, a final index Zfinal is generated, and quality evaluation is performed on an enhanced and corrected image.

The invention discloses an environment-friendly calcium-zinc composite heat stabilizer and a preparation method thereof, and relates to the technical field of heat stabilizers. The environment-friendly calcium-zinc composite heat stabilizer is prepared from the following raw materials in parts by mass: 10-20 parts of zinc stearate, 20-30 parts of calcium stearate, 10-15 parts of an auxiliary stabilizer, 5-8 parts of a plasticizer, 1-5 parts of an antioxidant and 2-5 parts of filler, wherein the auxiliary stabilizer is a mixture of beta-diketone, modified organic acid intercalated hydrotalcite and modified silicon dioxide; the modified organic acid intercalated hydrotalcite is obtained by splitting expanded vermiculite, adding zinc and lanthanum compounds, coprecipitating with organic acid, and adding a modifier; the modified silicon dioxide is obtained by adding a coupling agent into a solid substance obtained by splitting expanded vermiculite for modification and carrying out a series of chemical reactions with maleic anhydride and pentaerythritol. PVC plastic prepared by adding the calcium-zinc composite stabilizer into PVC paste resin and resin powder has good thermal stability.

The invention relates to a tree neighbor context-based big language model RAG optimization method, which comprises the following steps of: reading a directory structure and content of a document, generating a corresponding directory tree, converting the directory tree into a Markdown format containing a title and a level, and automatically extracting according to a document template if no title exists; segmenting the document into knowledge blocks according to the priority; after the knowledge blocks are recalled, the context is expanded by using a tree adjacent context method, and a new knowledge block list is formed; grouping and sorting the expanded knowledge blocks according to document identifiers, splicing reference data based on different fields and confidence coefficients, and generating a character string set from the original document; and splicing prompt words by using the constructed reference data and the new knowledge block list, and sending the spliced prompt words to the large model for processing. According to the scheme, the performance and practicability of the large model in the information retrieval and answer generation process are comprehensively enhanced by optimizing context information, emphasizing key content, adapting to various document formats, improving user experience and the like.

The invention discloses a personalized medical record analysis recommendation method and system based on a patient portrait. The method comprises the following steps: S1, generating a preprocessed health data set; s2, according to the preprocessed health data set, detecting a health event of the patient through a predefined health event triggering rule and an event classification model; s3, based on the priority of the health event, setting a trigger condition to start an updating process of the patient portrait; s4, the preprocessed health data set is combined with health event information, and a dynamic patient portrait is generated through a multi-modal data fusion technology and a time sequence modeling technology; s5, calculating a similarity index and selecting an optimally matched similar medical record; s6, according to the intelligent medical record analysis result and the medical knowledge graph, a diagnosis and treatment suggestion for the current patient is generated in combination with the rule base. The intelligent level and the precision level of the personalized medical record analysis recommendation system are remarkably improved.

The invention discloses a novel electric power system source network load storage coordinated optimization planning method, and relates to the technical field of electric power system optimization, and the method comprises the steps: S1, building a load prediction model based on various user side load dynamic adjustment data, and improving the prediction precision, S2, determining parameters and constraint conditions of each link of source network load storage of an electric power system, s3, carrying out source network load storage collaborative optimization scheduling by utilizing an optimization algorithm; S4, evaluating and verifying an optimization result; according to the novel power system source-network-load-storage coordinated optimization planning method, the precision of load prediction is greatly improved, so that the problem that supply and demand matching is difficult is effectively solved, rapid fluctuation or sudden change of user loads can be rapidly responded, the adaptability and flexibility of a scheduling strategy are remarkably improved, and the scheduling efficiency is improved. The problems of supply and demand matching and scheduling flexibility caused by low load prediction precision under the condition of dynamic adjustment of various user side loads in the prior art are solved.

The invention belongs to the technical field of artificial intelligence and big data analysis, and particularly relates to a scenic spot passenger flow trend prediction analysis method based on an AI big language model. According to the method, big data and artificial intelligence technologies are utilized, LBS data, scenic spot data, holiday and festival information, third-party maps, weather data and other multi-source data are comprehensively applied, a comprehensive multi-dimensional knowledge base is constructed, and meanwhile, the knowledge search and generation capability is improved by utilizing an RAG technology; through model pre-training and fine tuning, the universal model capability is developed, and a multi-agent system is further constructed; based on the natural language processing capability of the AI large language model, the LLM can intelligently analyze the text input of the scenic region administrator, drive the cooperative work among multiple agents, and achieve the intelligent prediction and analysis of the scenic region passenger flow trend. The method can more accurately predict the passenger flow trend of the scenic area.

The invention belongs to the technical field of data management, and discloses a decision management method and system based on machine learning. The method comprises the following steps: constructing an engineering construction decision analysis model and an engineering construction decision optimization model based on a cloud data center; acquiring real-time engineering construction decision data, and performing engineering construction decision analysis by using the engineering construction decision analysis model; performing engineering construction decision optimization on the real-time engineering construction decision data by using an engineering construction decision optimization model according to a real-time engineering construction decision analysis result; according to the optimized real-time engineering construction decision data, an engineering construction decision analysis model is used to carry out analysis after an engineering construction decision; and performing uplink storage on the optimized real-time engineering construction decision data and the optimized real-time engineering construction decision analysis result by using a preset decision storage block chain system. According to the method, the problems of low data security, low analysis accuracy and low practicability in the prior art are solved.

The invention provides an advertisement recommendation method and system based on deep reinforcement learning. The method comprises the following steps: receiving a real-time interaction signal of a user, analyzing time sequence characteristics in the real-time interaction signal to extract behavior characteristics of the user, and identifying a contextual model of the user; modeling the interest preference of the user to generate a personalized interest dynamic graph; performing intelligent exploration and decision processing on the optimal advertisement content and the corresponding pushing opportunity to obtain an advertisement recommendation strategy; calculating the optimal pushing opportunity corresponding to the optimal advertisement content from the optimal advertisement content and the corresponding pushing opportunity, and pushing the optimal advertisement content to the user based on the optimal pushing opportunity; establishing a real-time feedback mechanism to obtain an adjusted advertisement recommendation strategy; and performing comprehensive evaluation processing on the advertisement effect of the adjusted advertisement recommendation strategy to generate an advertisement effect evaluation result. According to the technical scheme provided by the invention, the flexibility and efficiency of advertisement recommendation can be improved.

The invention relates to the technical field of petroleum pipelines, in particular to a petroleum pipeline leakage detection method and system, and the method comprises the following steps: S1, data collection and preprocessing: collecting the detection data of a petroleum pipeline in real time through a plurality of sensors and geographic information collection equipment; s2, feature extraction: performing feature extraction on the preprocessed detection data by adopting a multi-scale feature extraction algorithm; s3, model construction: constructing a petroleum pipeline state recognition model; s4, real-time detection: inputting detection data acquired in real time into the petroleum pipeline state recognition model; s5, leakage path prediction: predicting the diffusion path and range after petroleum leakage in real time; and S6, leakage alarm: when leakage of the petroleum pipeline is detected, an alarm mechanism is triggered, and related personnel are notified to take emergency measures. According to the method, the diffusion path and range after leakage are accurately predicted, and a reliable basis is provided for emergency response.

The invention relates to a nano antibody targeting cadherin 17 and application thereof, and relates to the technical field of immunology and molecular biology, the amino acid sequence of a complementary determining region of the nano antibody targeting cadherin 17 comprises CDR1 shown as any one of SEQ ID NO: 21-32, CDR2 shown as any one of SEQ ID NO: 34-45, and CDR3 shown as any one of SEQ ID NO: 47-58. According to the invention, an alpaca immune phage display library is constructed, a series of nano antibody molecules targeting cadherin 17 (CDH17) are screened from the alpaca immune phage display library, and functions and applications of the nano antibody are verified at various levels of CDH17 antigen protein, CDH17 target cells, CDH17 CAR-T cells and the like.

The invention discloses a drainage pipe network water quality monitoring, early warning and traceability system and method, and the system comprises the steps: arranging pipe network water quality monitoring stations, building pipe network water quality early warning monitoring stations, carrying out water quality early warning, building a traceability business platform, and building a water quality early warning traceability model. Constructing a water quality pollution traceability mechanism flow; the pipe network early-warning traceability supervision mechanism can quickly study and judge the water quality of the inlet of the pipe network, can track pollution sources of secret discharge, leakage discharge and excessive discharge, provides early warning for inlet water of a sewage treatment plant, reserves emergency disposal time, realizes early-warning traceability of the sewage pipe network, quickly and efficiently searches for the pollution sources, improves supervision efficiency, and ensures that the water quality of the inlet of the sewage treatment plant is ensured. Stable operation of a sewage treatment plant is guaranteed, and impact is effectively reduced; the development of the project provided by the invention plays a demonstration role in an environmental protection system, plays a leading role in improving the water quality of a sewage pipe network and surface water, and provides a scientific basis for water environment management.

The invention relates to the technical field of remote monitoring, in particular to a mobile terminal remote control monitoring system and method, and the system comprises a video dynamic coding module, a content importance analysis module, a cognitive load adjustment module and an audio optimization and alarm module. According to the method, the change of the network bandwidth can be responded in real time by dynamically adjusting the video coding parameters, the stability and adaptability in data transmission are remarkably enhanced, the dynamic content and color change are accurately captured by analyzing the video frames, and key events are immediately positioned and extracted, so that the emergency response speed and precision are greatly improved, and the data transmission efficiency is improved. Cognitive resources are dynamically distributed by utilizing video stream complexity evaluation, resource configuration is adjusted according to actual requirements, monitoring quality is guaranteed, resource use efficiency is optimized, audio processing is optimized, real-time alarm strategies are adjusted, clear transmission of sound information and alarm timeliness are guaranteed, and the overall response capability and effect are improved.

The invention discloses a point cloud segmentation method and system based on two channels, and relates to the technical field of computer vision and artificial intelligence, and the method comprises the steps: collecting point cloud data; creating a feature channel according to the three-dimensional point cloud data; and performing feature fusion and classification according to the feature channel. According to the point cloud segmentation method based on the two channels, different types of features are processed in different channels, high calculation efficiency is kept while global feature dependence is captured, and the method is suitable for processing and segmentation tasks of large-scale point cloud data. Two different network architectures are used, the structure and parameters of the model are flexibly adjusted to adapt to different application requirements, and a more efficient and reliable solution is provided for a point cloud segmentation task in a complex three-dimensional scene. The method has better effects in the aspects of segmentation precision, segmentation result accuracy and universality.

The invention provides a lake ecological water level calculation method, device and system and a storage medium, and relates to the technical field of water resource regulation and control. The method comprises the following steps: obtaining the water turbidity of a target lake and the biomass of all submerged plants existing in the target lake; the water depth required by growth is calculated; dividing the target lake into habitat units; calculating the actual water depth of each habitat unit; calculating the plant coverage by using the water depth required for growth and the actual water depth; and obtaining an ecological water level range and an annual water level regulation and control range based on the plant coverage and taking the ranges as lake ecological water level results. By constructing the growth-illumination model and the underwater illumination prediction model of the submerged plants, the monthly water level accurate regulation and control based on biological requirements are realized, and by subdividing habitat units and analyzing ecological requirements, the growth conditions of the plants can be optimized, the water quality and biological diversity of a lake ecosystem can be improved, and the water quality and the biological diversity of the lake ecosystem can be improved. And the scientificity and systematicness of lake management are enhanced.

The invention discloses a self-supervised learning and multi-modal biomedical signal processing method, system and terminal, and belongs to the technical field of biomedical engineering and information, and the method comprises the steps: employing an unsupervised learning auto-encoder to extract the feature representation of multi-modal data; and taking the feature representation of the source domain data as input, taking the feature representation of the target domain data as output, and training the large language model based on the Transform architecture based on maximum likelihood estimation. According to the method, relevance and common features among multi-modal data are learned through an auto-encoder, and feature representation extracted by the auto-encoder is applied to pre-training of a large language model, so that the large language model learns more abstract and advanced feature representation, and the processing capacity and generalization performance of the multi-modal data are improved; meanwhile, due to the fact that an unsupervised learning method is adopted, a large amount of unmarked data can be used for pre-training the large language model, and therefore data resources can be better utilized.

The utility model relates to the technical field of waste water collection, and discloses waste water collection equipment convenient to clean, which comprises a tank body, an anti-blocking structure is arranged in the tank body, and the anti-blocking structure comprises a screening disc, a rotating shaft, a sleeve, a scraper blade, a fixing frame, a sliding chute, a reciprocating screw rod, a connecting cylinder, an L-shaped rod and a cleaning plate, impurities in waste water are separated from water liquid through a screening disc, at the moment, separated water continues to move downwards to be subjected to follow-up filtering work again, then part of fixed impurities are left on the surface of the screening disc, and when a scraper rotates to the position of a collecting assembly, an output shaft of a waterproof motor rotates to drive a reciprocating lead screw to rotate; and the connecting cylinder moves to drive an L-shaped rod on the side wall of the connecting cylinder and a cleaning plate to move so as to push and clean impurities pushed by the scraper, the impurities accumulated on the surface of the scraper are pushed into a collecting assembly to be collected through movement of the cleaning plate, and the using effect of the device is further improved.