523 results about "False positive rate" patented technology

The invention provides a water conservancy project safety monitoring method and system based on data processing, and relates to the technical field of monitoring, and the method comprises the steps: obtaining and carrying out the multi-dimensional preprocessing of water conservancy project multi-source heterogeneous monitoring data through the deployment of a sensor network, and extracting multi-scale space-time fusion features from the data; performing structural state modeling, anomaly prediction, risk assessment and early warning by using a long-short-term memory neural network model integrated with a multi-head attention mechanism; and intelligent suggestions oriented to maintenance decisions are generated, so that comprehensive, accurate and prospective evaluation and early warning of the structural state of the water conservancy project are finally realized, the exception identification and risk prediction capabilities are effectively improved, the false alarm rate is reduced, refined and initiative intelligent maintenance decisions are provided, resource allocation is optimized, and the service life of the project is prolonged.

The invention belongs to the technical field of network security, and provides an abnormal traffic detection and attack recognition method and system based on deep learning, and the method comprises the steps: data preprocessing and feature extraction, cross-modal semantic alignment and knowledge graph construction, causal enhancement association reasoning, intelligent engine optimization, cloud edge collaborative resource scheduling, and result output. According to the method, statistical features and signature features are mapped to a unified semantic space through a cross-modal semantic alignment and knowledge graph construction module, a semantic barrier between heterogeneous features is broken through, time sequence causal discovery and transfer entropy calculation are introduced, a simple correlation and a reliable causal can be distinguished, and the method has a good application prospect. According to the method, the accuracy and credibility of attack chain reasoning are improved, the false alarm rate is reduced, online self-evolution of a detection model and dynamic optimal allocation of system resources are realized through intelligent engine optimization and cloud edge collaborative resource scheduling modules, and the overall adaptability, robustness and practicability of the system are enhanced.

The invention discloses a server fault diagnosis method and electronic equipment, and relates to the technical field of server operation and maintaining.The server fault diagnosis method comprises the steps that firstly, by collecting operation state data of a server and extracting structured events containing index types, context tags and attribute information, associated contexts and dynamic characteristics between indexes are captured; the method comprises the following steps of: converting dispersed operation data into a node and edge relationship with logic association based on a fact graph constructed by a structured event, and intuitively presenting a multi-dimensional dependency relationship related to a fault; the pre-trained graph neural network extracts causal embedding features from a graph, identifies causal association between nodes, captures a conduction path with abnormal indexes in a complex fault mode, and reduces missed report caused by neglected index linkage change; the causal embedding features and the original operation data are fused to form joint state representation, collaborative decisions are generated through a fault diagnosis model, abnormal changes of complex scenes are recognized, the false alarm rate is reduced, and reliable diagnosis results and disposal strategies are provided for operation and maintenance.

The invention provides an anti-cheating method and system based on game user behavior analysis, and the method comprises the steps: constructing a multi-dimensional behavior feature collection module, and collecting the operation behavior data of a game user in real time; based on the operation behavior data, constructing a user behavior basic model through a deep learning neural network; a multi-stage discrimination mechanism for behavior anomaly detection is established, and comprehensive credibility evaluation is obtained in combination with the basic model; implementing a self-adaptive risk grade division and dynamic response strategy, dividing users into different risk grades according to comprehensive credibility evaluation, and taking corresponding anti-cheating measures for different grades; a distributed collaborative verification network is constructed, encrypted transmission and cross verification are carried out on behavior feature vectors of suspicious users among a plurality of server nodes, the reliability of a verification result is ensured through a Byzantine fault-tolerant algorithm, and meanwhile user privacy data are protected. According to the method, the personalized behavior pattern of the user can be accurately captured, the accuracy of cheating detection is remarkably improved, and the misjudgment rate is reduced.

The invention relates to the technical field of side slope geological disaster monitoring, and discloses a side slope long-term monitoring early warning and threshold value dynamic adjustment method, which comprises the following steps: S1, deploying a plurality of sensors on a side slope; s2, preprocessing the collected original data; s3, setting an early warning threshold value for the key feature index; s4, dynamically evaluating and predicting the stable state of the slope in real time; s5, dynamically correcting and optimizing the initial early warning threshold value set in the step S3; s6, when the key characteristic index value exceeds the newest early warning threshold value, early warning is triggered; and S7, generating early warning information of different levels. The method solves the problems of tedious configuration of slope monitoring early warning threshold values and false alarm and missing alarm, achieves the adjustment and configuration of a unified threshold value through an efficient early warning trigger judgment disposal and threshold value dynamic adjustment mechanism, effectively improves the alarm success rate and early warning disposal efficiency, remarkably reduces the false alarm rate / missing alarm rate of slope early warning, and improves the safety of slope early warning. And the interference of the early warning system on residents is greatly reduced.

The invention discloses a power grid island detection and protection method and device based on an intelligent agent, equipment and a storage medium. The method is applied to an agent server side and comprises the steps that a plurality of edge agents are registered and managed, the edge agents comprise a monitoring agent, a decision agent, a memory agent and an execution agent, and the decision agent carries out strategy optimization based on historical log data stored by the memory agent; receiving and processing power grid parameter data reported by the edge agent, and performing time synchronization and anomaly detection; a global decision result is generated through a multi-agent collaborative judgment algorithm, the algorithm is combined with a voting mechanism and dynamic weight adjustment, and dynamic weights are comprehensively calculated based on agent historical false alarm rates, data credibility, geographic position importance and real-time abnormal degrees; and issuing a protection action instruction to the execution agent according to the global decision result. The island detection accuracy is remarkably improved, and the problems that a traditional method is high in misjudgment rate and inflexible in judgment threshold presetting are solved.

The invention relates to the technical field of industrial carbon emission monitoring, and particularly discloses an industrial carbon emission visual monitoring method and system based on digital twinning, and the method comprises the following steps: S1, carrying out the collaborative collection and preprocessing of carbon emission data; s2, dynamically updating the hierarchical digital twin model; s3, intelligent early warning and accurate tracing are carried out; and S4, virtual-real linkage visual interaction is carried out. According to the invention, by establishing a multi-factor driven hierarchical iteration mechanism, hierarchical updating and closed-loop verification of parameters-structures are realized, by establishing a space-time correlation-process coupling-mode matching three-dimensional fusion early warning rule, the false alarm rate and the missing report rate are reduced, a visual interaction system of virtual-real linkage is built, the decision-making efficiency is improved, and the decision-making efficiency is improved. And an accurate and efficient carbon emission control technical support is provided for multi-process industrial scenes such as metallurgy and chemical engineering.

The invention belongs to the technical field of motor fault diagnosis, and particularly relates to a method for constructing a motor fault diagnosis model, which comprises the following steps: deploying a multi-modal sensor and carrying out data acquisition; preprocessing the collected data at the edge end; carrying out multi-modal data fusion and carrying out credibility estimation on the multi-modal fusion data; learning and identifying the depth representation of the fault mode according to the self-supervised time sequence comparison representation; incremental recognition and disastrous forgetting suppression are adopted; performing influence estimation and linear correction according to environmental causality; a non-linear compensation and edge-cloud cooperative triggering strategy of uncertainty gating is constructed; according to the method, the fault identification accuracy and the recall rate can be remarkably improved under the complex working condition, the false alarm rate and the missing report rate are remarkably reduced, the robustness to environment covariant and short-time interference is enhanced, and meanwhile, an end-cloud cooperative fault diagnosis system which supports online incremental updating, privacy protection, real-time alarm of a limited bandwidth field station and cloud continuous optimization is supported.

The invention discloses a software source code security vulnerability detection method based on an artificial intelligence large model, and belongs to the technical field of information security and source code detection, and the method comprises the following steps: carrying out user registration and login, uploading a file, judging whether static analysis is carried out or not, judging whether dynamic analysis is carried out or not, introducing the AI large model, and generating a detection report; according to the method, through multi-level technology integration and a self-evolution mechanism, normal form innovation of vulnerability detection from a single view angle to global perspective is realized. According to the system, on the basis of unified intermediate representation, a closed-loop architecture of static rule screening, dynamic behavior verification and AI semantic reasoning is constructed, and the core contradiction of high false alarm rate, insufficient path coverage and semantic understanding deficiency of a traditional method is effectively solved.

The invention discloses a strip steel outlet detection method and related equipment, and relates to the technical field of industrial automatic detection, and the method comprises the steps: obtaining image data of a target detection point in a strip steel outlet area; target detection is conducted on the image data based on a strip steel state detection model, the abnormal state of the strip steel is determined, and the abnormal state comprises the steel stacking state, the steel clamping state or the strip steel arching state; performing motion artifact compensation verification on the abnormal state based on the roller way vibration amplitude data to obtain a verification result; and generating a control instruction based on the verification result and the abnormal state to adjust the operation action of the production line. According to the method, manual monitoring is replaced by automatic detection, the omission ratio and the misjudgment rate are reduced, response and interlocking control of the abnormal state are achieved, non-planned shutdown of a production line is effectively avoided, and continuity and high efficiency of the production process are guaranteed.

The invention relates to the technical field of component collision detection, in particular to a building component-level collision detection method and system fusing multi-model semantics. The method comprises the steps of obtaining building component model data; component attribute extraction and semantic recognition are carried out according to the obtained building component model data; carrying out space conflict detection on the extracted information based on a bounding box and grid Boolean operation; judging conflict types and levels by utilizing a component attribute semantic library; and outputting a judgment result. By introducing a component attribute semantic database and combining semantic discrimination rules to match conflict types, 'acceptable conflicts' and'substantive conflicts' can be accurately distinguished, non-key conflict information can be automatically filtered, the false alarm rate is greatly reduced, the screening workload of designers is reduced, and collaborative design verification focuses on the core problem affecting construction safety and function implementation.

The invention discloses a quick data traceability method and system based on multi-source log analysis, and relates to a data processing system or method specially suitable for administrative, commercial, financial, management, supervision or prediction.The method comprises the steps that firstly, the system responds to a service instruction to generate a transaction receipt containing a unique serial number and service attribute parameters; and establishing an association relationship between the two records and generating a to-be-confirmed record and a timestamp. And sending a service message containing the service attribute parameters to a receiving end. When an ownership right reverse search request of a receiving end is received, the system calculates an effective time interval based on current time and a preset time delay threshold value, and screens out a target to-be-confirmed record which is matched with a service attribute parameter and generates a timestamp in the interval. And finally, extracting the unique serial number from the target record, generating a confirmation receipt and sending the confirmation receipt to a receiving end system. The method is used for reducing the calculation overhead and false alarm rate of accurately positioning specific abnormal transactions in massive logs by a system, and improving the confidence and timeliness of an automatic traceability result.

The invention discloses an eddy current testing reliability evaluation method and system based on signal space distribution characteristics, and belongs to the technical field of nondestructive testing. According to the method, eddy current signal space distribution characteristics and the detection probability (POD) and the false detection rate (PFA) in nondestructive testing reliability evaluation indexes are fused, and the method is used for comprehensively evaluating the defect detection capacity and the reliability of a detection system. By introducing the Gini coefficient to quantitatively detect the imbalance of signal space distribution, the method effectively characterizes the distinguishability of signals in a defect state and a defect-free state, and replaces traditional POD and PFA analysis methods based on single-point amplitude or phase characteristics. Compared with a traditional method, the Gini coefficient can more comprehensively represent the overall distribution characteristics of the signals, the influence of noise is small, and the method is particularly suitable for the eddy current detection environment under the condition of the low signal-to-noise ratio. Besides, a POD and PFA relation curve is constructed based on Gini coefficients, ROC analysis is performed, the defect detection rate can be effectively increased, the false detection rate can be reduced, and the reliability of the nondestructive testing system can be evaluated more efficiently and accurately.

The invention discloses a security vulnerability detection method and device, a computer program product and a storage medium, and belongs to the field of code detection.A structured graph of a to-be-detected code can be constructed according to structured information of the to-be-detected code, and then the to-be-detected code is coded to obtain a text coding vector; and coding the structured graph of the to-be-detected code to obtain a structured coding vector, and finally inputting the comprehensive vector into a pre-trained detection model so as to obtain a security vulnerability detection result. The problem of how to reduce the omission ratio and the false alarm rate of security vulnerability detection of codes is solved, and the beneficial effects that security vulnerability detection is performed from two dimensions of semantics of code texts and semantics of structured information, and the omission ratio and the false alarm rate of security vulnerability detection of codes are reduced are achieved.

The invention provides an aircraft system key parameter anomaly detection method and system based on an auto-encoder, and belongs to the field of aviation fault analysis. The method comprises the following steps: S1, preprocessing key parameters of an aircraft system, and dividing the key parameters into a training set, a verification set and a test set; s2, constructing abnormal data for simulating different forms of key parameter anomalies of the aircraft system; s3, implanting a plurality of anomalies; s4, constructing a composite loss function and training the auto-encoder; and S5, determining an exception threshold and performing exception detection, adaptively determining an exception judgment threshold based on a reconstruction residual statistical result of the verification set sample, and performing exception detection on the test set sample according to the exception judgment threshold. According to the method, the accuracy and robustness of unknown anomaly detection can be remarkably improved, the false alarm rate caused by noise and slight disturbance is reduced, and the method is suitable for high-dimensional aircraft system key parameter monitoring and anomaly early warning of aircraft environment control, flight control, hydraulic and other key systems.

The invention discloses an intelligent detection method and system for a flexible circuit board based on machine vision, and relates to the field of circuit board detection.The method comprises the steps that firstly, through a global coarse registration step, rigid transformation such as translation, rotation and zooming of the whole to-be-detected FPC image is rapidly corrected, and preliminary alignment with a standard template is achieved; then, local fine registration is introduced for nonlinear deformation such as local stretching and wrinkles caused by the characteristics of the FPC material, so that the fine and non-uniform local dislocation is accurately sensed and compensated, and nearly pixel-level accurate matching between the to-be-detected image and the template is realized. On the basis of high-precision alignment, differential attention deficit segmentation is carried out, so that false differences caused by inaccurate deformation registration can be effectively eliminated, real defects can be accurately identified, the false alarm rate is remarkably reduced, and the reliability and practicability of a detection scheme are greatly improved.

An intrusion detection method based on a traceability graph comprises the following steps that S1, an original log data set not containing hostile attacks serves as input, logs are analyzed, and an initial full-system traceability graph is constructed; s2, deleting redundant nodes and edges on the initial full-system traceability graph; s3, specific domain knowledge is fused in the node feature construction and graph representation process, and initial embedding is generated for nodes on the traceability graph; s4, constructing a GraphSAGE-based neural network model, performing node classification on the traceability graphs after training is finished, storing sub-models obtained by training each time, and jointly using the sub-models for abnormal node detection; s5, the log data containing attack traces serve as input, a traceability graph is constructed in real time, and the traceability graph constructed in real time is detected; writing a single abnormal node on the identification traceability graph into an alarm file; and S6, carrying out attack scene reconstruction by taking the abnormal node alarm as an entry event of attack investigation. The false alarm rate can be reduced, and the detection performance of the intrusion detection method can be improved.

The invention belongs to the technical field of numerical control machine tool machining, and particularly relates to a numerical control boring and milling tool cutting force intelligent monitoring method and system, and the method comprises the steps: obtaining a real-time main force, a real-time transverse force, a real-time longitudinal force sequence, a real-time main shaft rotating speed, a signal sampling frequency and a basic statistical threshold value in a machining task of numerical control boring and milling equipment; obtaining characteristic force; calculating a historical mean value at the current moment, and recording the historical mean value as a historical mean value; calculating a current historical deviation, and determining the width of a sliding window; calculating an instantaneous deviation; obtaining the change amplitude of the characteristic force at the current moment, the severity of the change trend deviating from the historical normal state, and the maximum value of the change amplitude and the severity to obtain a risk index; and comparing whether the risk index is abnormal or not according to the basic statistical threshold, and generating a processing process digital file. The technical problems of high false alarm rate and poor reliability caused by incapability of adapting to normal wear of the cutter in the prior art are solved.

The invention discloses an unmanned aerial vehicle subsystem health assessment method and system, and relates to the technical field of unmanned aerial vehicle health management, and the method comprises the steps: obtaining the data of a multi-source sensor of an unmanned aerial vehicle, screening the features strongly correlated with a fault state through a Spearman rank correlation coefficient, and obtaining key monitoring parameters; according to the self-adaptive mutation mechanism, the virtual evolution strategy and the double-objective optimization strategy, an improved multi-objective genetic algorithm optimization model is constructed, then key monitoring parameters are input, and a visual report of a health state assessment result and confidence rating is obtained. According to the method, the problems that traditional single-target optimization neglects the misjudgment rate, the parameter search efficiency is low and the evaluation credibility is insufficient are solved, accurate evaluation is achieved, the evaluation accuracy is improved to 84.86%, the key misjudgment rate is reduced to 0.009%, the convergence speed is improved by 40% compared with a traditional method, the safety and real-time performance of unmanned aerial vehicle health state recognition are remarkably optimized, and the method is suitable for popularization and application. The method is suitable for embedded system deployment.

The invention discloses a power load event detection algorithm based on an adaptive window model, and belongs to the field of power detection, and the algorithm comprises the steps: collecting power load total power time sequence data, carrying out the preprocessing, carrying out the rough detection of the preprocessed data, positioning all suspected event points, constructing a composite feature vector, and screening the suspected event points. And eliminating misinformation and repeated event points, and outputting a final power load event detection result. According to the method, the missing detection rate and the false detection rate are greatly reduced, the high recall rate is achieved through the dynamic threshold value and the improved CUSUM algorithm in the coarse detection stage, the load event detection accuracy is improved through composite feature vector and mahalanobis distance screening in the fine detection stage, the false detection rate is reduced, and the method can adapt to different types of loads and different sampling frequency data.

The embodiment of the invention relates to the technical field of information, and discloses an interface verification and test method and device based on a large model, a medium and a product, and the method comprises the steps: obtaining a to-be-tested interface document, analyzing the interface document through a first large model, constructing a test request parameter based on a request parameter definition, acquiring a real-time response message by calling a real interface; performing semantic comparison on a response example in the document structure information and the real-time response message by using the second large model to obtain a corrected interface document; comparing the parameter constraint condition in the corrected interface document with the verification logic in the business logic code snippet by using a third model to generate a logic verification passing result; and on the basis of the corrected interface document and a logic verification passing result, a Mock rule and an automatic test case script are generated by utilizing the fourth model, so that the maintenance cost and the false alarm rate of the automatic test script are reduced, and the automatic test efficiency is improved.

The invention discloses an abnormal Transform model-based electricity larceny detection method and device and a storage medium, and relates to the field of electric power system electricity larceny detection, and the abnormal Transform model-based electricity larceny detection method comprises the following steps: capturing complex correlation among different time points in sequence data by using a self-attention mechanism of a Transform model; the characteristics of the normal time point and the abnormal time point are effectively distinguished by learning the relevance difference of each time point, the accuracy of electricity stealing user identification is improved, and the misjudgment rate is reduced. The method comprises the following steps: S1, obtaining user multi-dimensional power consumption time series data, preprocessing the data, and constructing a sliding window sample; s2, constructing an electricity larceny detection model based on an Angle-Attention mechanism, wherein the electricity larceny detection model comprises a prior correlation branch and a sequence correlation branch; s3, quantifying the time dependence deviation of the power consumption behavior by taking the relevance difference as an anomaly judgment standard; s4, training the model by using a minimax optimization strategy, and amplifying the relevance difference between normal electricity utilization and electricity stealing electricity utilization; and S5, performing electricity larceny risk prediction and model performance evaluation through the joint score of the relevance difference and the reconstruction error.

The invention discloses a power grid data management intelligent checking method and system based on four-dimensional cross validation. The method comprises the following steps: acquiring data from each service system of a power grid and preprocessing to generate a to-be-verified data set; performing value, rule, relation and attribute dimension verification on the to-be-verified data set in sequence; summarizing four-dimensional verification results of value-rule-relationship-attribute and dividing anomaly levels to generate an anomaly diagnosis report; and based on feedback data of an abnormal diagnosis result and an actual rectification result in the report, mining an association rule to expand a rule base, dynamically adjusting the weight of the rule in the rule base, and detecting and eliminating a failure rule. According to the method, full-dimension anomaly recognition can be achieved, complex scenes can be covered, the false alarm rate is effectively reduced, design rules are self-optimized, a rule base can be autonomously evolved, manual intervention does not need to be completely relied on, the rule updating period is greatly shortened, and dynamic changes of power grid services and data can be rapidly responded.

The invention discloses a PCB defect detection method and system based on adaptive multiple submodels, and the method comprises the steps: collecting an image of the surface of a PCB through a high-resolution camera, and carrying out the preprocessing of the image; performing dynamic priority decision through reinforcement learning, and reasonably selecting the priority of the sub-model according to the equipment state and the load of the computing resource; yOLOv8 is adopted as a backbone network, a plurality of sub-models work in parallel, and the sub-models are trained in combination with gradient masks; realizing multi-stage cascade detection by adopting a confidence threshold screening and variance analysis mode, and reversely optimizing parameters of the model according to a batch statistical result; and according to the recall rate and the false detection rate of each detection, optimizing the sub-model through an incremental training mechanism. According to the method, the adaptive multiple sub-models are provided, and through a dynamic priority control and feature sharing mechanism, the high accuracy is ensured, the detection strategy is dynamically adjusted, the calculation burden is reduced, and efficient and accurate PCB defect detection is realized.

The invention provides a user interest migration point detection method and device, electronic equipment and a storage medium, the method comprises the following steps: slicing a behavior vector sequence by adopting a sliding window mode, and aggregating windows of the sliced behavior vector sequence to obtain an interest state representation; determining an interest migration point based on the difference between the current interest state representation of the current window and the future interest state representation of the next window, and in the process, sensitively capturing the interest migration point by introducing an interest state modeling mechanism based on a sliding window and combining the difference between the interest state representations of the front window and the back window; the limitation that a traditional migration point identification method based on a static portrait or time weighted behavior aggregation method is insensitive is broken through; the interest migration points are further recognized through the interest state judgment network, the interest state mutation points are obtained, complex scenes such as frequent fluctuation of user behaviors can be effectively dealt with, the misjudgment rate and the missed judgment rate are remarkably reduced, and the adaptability to multiple behavior modes is enhanced.

The invention discloses a cross-border e-commerce logistics risk assessment method and system based on artificial intelligence, and belongs to the technical field of e-commerce logistics intelligent management, and the method comprises the steps of multi-source data integration, risk factor modeling, dynamic risk assessment and risk visualization. According to the method, intelligent fault detection is carried out by adopting a multi-mode self-adaptive fault detection method, link operation data, equipment operation records and network structure information are integrated, potential abnormal nodes are dynamically identified, the false alarm rate is reduced, faults possibly influencing network operation are found in advance, quantifiable abnormal credibility scores are provided, and the reliability of the network is improved. The subsequent dynamic risk assessment is effectively supported; dynamic risk assessment is carried out by adopting a risk assessment method combined with a deep Q network, continuous prediction and evolution tracking can be carried out on the risk state of each node, and the influence of different logistics operation strategies on the risk is simulated, so that quantifiable and comparable comprehensive risk scores are generated, and dynamic and visual risk information is provided for logistics management personnel.

The invention relates to the field of cutter detection, in particular to a numerical control machine tool cutter wear identification method and system, and the method comprises the steps: obtaining a cutter image, and extracting a main body surface region of a cutter; calculating the micro texture entropy energy of the main body surface area, wherein the micro texture entropy energy is used for representing the texture complexity of the tool surface; calculating a macroscopic defect index of the surface area of the main body, wherein the macroscopic defect index is used for representing the geometric deviation degree of the edge contour of the cutter; calculating an environment interference coefficient based on the highlight area proportion and gray level distribution characteristics of the main body surface area; determining an adaptive fusion weight of the microscopic features based on the environmental interference coefficient, and performing weighted fusion on the microscopic texture entropy energy and the macroscopic defect index according to the adaptive fusion weight to obtain a wear state value of the tool; and identifying whether the cutter is worn or not by using the wear state value. According to the invention, the false alarm rate in the tool wear identification process is reduced, and the accuracy of the identification result is improved.

The invention discloses a pump station engineering fault diagnosis and prediction system and method based on a fault tree, and relates to the field of intelligent water conservancy. The system comprises a fault tree database, a data acquisition module, a data analysis module, a fault diagnosis module and a fault prediction module. The data analysis module transmits the abnormal data according to the following preset rules: when the associated parameter data exceeds a top event preset standard, the abnormal data is transmitted to the fault diagnosis module; when the associated parameter data exceed a sub-event preset standard, abnormal data are transmitted to a fault prediction module; when the trend of the associated parameter data deteriorates but does not exceed any preset standard, the preset standard in the fault tree database is dynamically adjusted, and then processing is carried out according to a preset rule; the fault diagnosis module is used for finding out a fault reason corresponding to the top event; and the fault prediction module is used for predicting whether the top event can occur or not. The invention aims to solve the problems that the fault reason positioning efficiency is low, the misjudgment rate is high, and missed judgment is easy to occur when a static threshold value is used for early warning.

The invention relates to the technical field of power generation equipment fault diagnosis, and discloses a fan gear box fracture early warning method and device based on an acoustic emission positioning technology, and the method comprises the steps: obtaining the internal structure parameters of a fan gear box, and constructing a sound wave propagation model based on the internal structure parameters of the fan gear box; performing sound wave propagation simulation at the positions of the plurality of sound emission sources by using the sound wave propagation model to obtain sound wave propagation simulation data; acquiring an acoustic emission real-time signal, and performing path separation on the acoustic emission real-time signal by using the graph convolutional network model to obtain classification probability data of a multipath signal path; and based on the sound wave propagation simulation data and the classification probability data of the multipath signal paths, fault signal detection is carried out by using a reinforcement learning agent model, and fan gearbox rupture early warning information is obtained. According to the invention, the false alarm rate is effectively reduced, the accuracy and reliability of the early warning system are greatly improved, and a more reliable guarantee is provided for the early warning of the rupture of the fan gearbox.