191 results about "Neighbor algorithm" patented technology

The invention discloses a dynamic self-adaptive recommendation strategy optimization method for a business handling failure scene, and relates to the technical field of business handling recommendation and intelligent decision making, and the method comprises the steps: firstly collecting various types of data of a whole business handling process, and guaranteeing the integrity and real-time performance at a frequency of 100 milliseconds per time; a decision tree and Bayesian network fusion algorithm is used for attribution, and direct and indirect reasons are clarified; integrating data to construct a user portrait, and mining potential and subsequent demands; generating a recommendation scheme set based on attribution and portraits, and adjusting priorities and forms in combination with scene features; feedback data is introduced, a strategy weight is optimized by using a gradient descent algorithm, and the scheme is updated regularly; a multi-dimensional index weighted evaluation effect is set, and emergency optimization is carried out if the evaluation result does not reach the standard; and establishing a distributed strategy library, and reusing the optimal strategy of the similar scene by using a K-nearest neighbor algorithm. According to the method, failure reason accurate positioning and personalized recommendation are realized, the recommendation effect is continuously optimized along with data accumulation, and the method is adaptive to multiple service types and user groups.

The invention provides a geological disaster prediction method and device fusing space-time sequence analysis and causal reasoning, and belongs to the technical field of geological disaster monitoring and early warning. Aiming at the problems of non-uniform data space-time reference, lack of causal logic, poor real-time performance and weak scene adaptability of a model in the prior art, the method comprises the following steps: performing standardization processing on acquired multi-source data, processing missing values by adopting an improved K nearest neighbor algorithm in combination with stratum characteristics, and processing abnormal values through a 3 sigma criterion and geological verification; based on an information theory and an improved SURD algorithm, three types of causal entropies among variables are calculated, a time attenuation coefficient is introduced, a core causal chain is constructed, and a dynamic causal graph is constructed; a core causal variable is used as input, a multi-feature attention-multi-relation space-time diagram recursive network model is constructed, a hour-level predicted value is output through space-time diagram convolution, residual training and a geological physical constraint layer, and'causal-space-time 'fusion is realized through a causal weight adjustment model; the method can be widely applied to early warning of geological disasters such as landslide and debris flow.

The invention belongs to the technical field of three-dimensional point cloud processing, and particularly discloses an automatic tooth segmentation method and system for oral cavity scanning point clouds, and the method comprises the following steps: collecting and preprocessing dentition three-dimensional point clouds, extracting point cloud features containing coordinates and normal directions, and constructing a neighborhood structure; setting a first-stage multi-branch network, and performing tooth and gingiva coarse segmentation and instance initial clustering on the point cloud; setting a second-stage semantic refinement network, and carrying out local cutting on the original point cloud; setting a boundary detection and boundary offset network to enhance a tooth contact area, and refining a boundary instance by clustering to obtain a boundary enhanced instance tag; and performing label fusion, and propagating the fused label to the original high-resolution point cloud through a nearest neighbor algorithm to obtain a final semantic label and a final instance label. By adopting the technical scheme, high-precision instance-level segmentation of the complete dentition point cloud is realized through semantic segmentation, boundary detection and clustering cooperative work based on a biased field.

The invention discloses a big data measurement asset supply and demand matching and inventory optimization method, and relates to the technical field of big data processing and supply chain management. The method comprises the following steps: segmenting supply chain asset data through a hash function according to multi-dimensional attributes to generate a data fragment set, and constructing a global index tree according to the data fragment set; and monitoring the load state of the leaf nodes of the global index tree in real time, and migrating and verifying data during unbalance. And querying the global index tree, positioning fragment data associated with the query request from the uniformly distributed leaf nodes, generating a preliminary matching asset set, calculating the matching degree of each asset and the query request, and generating a resource allocation result. And updating the global index tree, and generating the latest data view representation. And adjusting the attribute weight coefficient of each dimension in the hash function, and generating optimized storage layout configuration. Irrelevant data are filtered through a neighbor algorithm, and a final matching result set is generated. Balanced storage, efficient query and accurate matching of asset data are realized, and the overall response speed and the resource utilization rate are improved.

An automated system attempts to characterize code as safe or unsafe. For intermediate code samples not placed with sufficient confidence in either category, human-readable analysis is automatically generated to assist a human reviewer in reaching a final disposition. For example, a random forest over human-interpretable features may be created and used to identify suspicious features in a manner that is understandable to, and actionable by, a human reviewer. Similarly, a k-nearest neighbor algorithm may be used to identify similar samples of known safe and unsafe code based on a model for, e.g., a file path, a URL, an executable, and so forth. Similar code may then be displayed (with other information) to a user for evaluation in a user interface. This comparative information can improve the speed and accuracy of human interventions by providing richer context for human review of potential threats.

The invention discloses a mining area exploration system based on artificial intelligence. The mining area exploration system comprises a multi-source data acquisition module, a data optimization module, a mining area candidate area potential identification module and a mining area intelligent exploration module. The invention relates to the technical field of artificial intelligence data analysis and computer vision, in particular to a mining area exploration system based on artificial intelligence, according to the scheme, a mining area candidate area potential identification module and a mining area intelligent exploration module are innovatively combined, and the accuracy of mining area identification and the pertinence of exploration are improved; a weighted graph structure is constructed by adopting a K nearest neighbor algorithm, and an improved graph convolutional neural network with a parallel updating mechanism is introduced, so that high-precision automatic potential classification of a large-range candidate mining area is realized; and a composite chaotic mapping initialization method and a sine index inertia weight improved particle swarm optimization algorithm are introduced, so that the accuracy and stability of yield prediction are improved, and high-precision yield prediction of each sub-region of the mining area is realized.

The invention relates to the technical field of mechanical fault diagnosis, and provides a fault diagnosis and classification method for a bearing of an aluminum alloy impeller die-casting ladling machine, which comprises the following steps: acquiring a vibration signal of the bearing of the ladling machine, carrying out continuous wavelet transform on the vibration signal, generating a time-frequency diagram, and carrying out adaptive grid segmentation on the time-frequency diagram. Grid granularity is adjusted according to the local change rate of the time-frequency graph, multi-scale nodes are generated, edge connection is generated for the multi-scale nodes based on a K-nearest neighbor algorithm, and a multi-scale graph structure is constructed; performing unsupervised feature extraction on the multi-scale image structure, including: performing data enhancement on the multi-scale image structure through edge deletion and feature mask to generate an enhanced view; a graph attention network encoder is used for encoding the enhanced view, graph-level embedding is generated, and graph-level embedding is optimized by comparing a loss function; and based on the optimized graph-level embedding, performing fault classification by using a classifier constructed by a graph attention network and a multi-layer perceptron, and outputting a fault category.

The invention provides a mechanical equipment structure key pressure-bearing component real-time simulation method based on digital twinning, which takes a cubic press hinge beam as an example, and combines finite element analysis, Latin hypercube sampling, a K nearest neighbor algorithm, Gaussian interpolation and an RBF (Radial Basis Function) proxy model to realize stress-strain rapid prediction and three-dimensional visualization. According to the method, an efficient prediction model is established through structure database construction, dimension reduction processing, neighbor search and interpolation calculation, a simulation result is presented in real time by utilizing Python and Unity interaction, the design efficiency and accuracy are improved, and the method is suitable for structure optimization analysis under complex working conditions.

Devices and techniques are generally described for natural language processing interfaces. In various examples, first natural language data may be received from an input device. First embedding data representing the first natural language data may be generated. A nearest neighbor algorithm may determine first data representing similarity between the first embedding data and second embedding data, the second embedding data representing second natural language data associated with a first skill. The nearest neighbor algorithm may determine second data representing similarity between the first embedding data and third embedding data, the third embedding data representing third natural language data associated with a second skill. First output data that indicates that the first skill and the second skill are candidates for processing the first natural language data may be generated.

The invention relates to a gearbox fault diagnosis model construction method based on metric guide graph comparative learning, and belongs to the field of gearbox fault diagnosis model construction. The method comprises the following four core stages: firstly, carrying out frequency domain conversion and normalization on vibration signals of the gearbox to generate a node characteristic matrix; secondly, a cosine distance and an Euclidean distance are fused to construct a mixed distance matrix, and a fault diagnosis graph is generated based on a K-nearest neighbor algorithm; then unsupervised graph comparison pre-training is realized through graph data enhancement and a dynamic graph attention network (DGAT); and finally, weak supervision fine tuning is carried out by using a small number of marked samples to complete construction of the gearbox fault diagnosis model. According to the method, the construction of the high-precision gearbox fault diagnosis model can be realized in a scene with extremely few marked samples (1-10 samples per class), and the method is suitable for planetary gearbox health monitoring in the fields of wind turbines, helicopters, hybrid electric vehicles and the like.

The invention relates to the technical field of computers, and particularly discloses a security product self-evaluation report intelligent auditing method based on artificial intelligence assistance, and the method comprises the steps: analyzing a self-evaluation report to generate structured detection item data; carrying out multi-modal feature extraction and fusion on the text and image proof materials; semantic conflicts, configuration compliance, evidence credibility, image-text consistency and historical risk matching degree are analyzed in parallel through an artificial neural network, a support vector machine, a random forest, logistic regression and a weighted neighbor algorithm; according to a preset weight, dynamically fusing results of all dimensions to determine a compliance probability; similar historical cases are retrieved in combination with a security product compliance analysis knowledge graph, and auditing instructions and improvement suggestions with violation positioning bases are generated; and finally, outputting a structured auditing report and supporting continuous optimization of a manual reexamination feedback driving model. According to the method, high-precision, full-dimension and automatic security product self-evaluation report auditing can be realized, and the auditing efficiency, objectivity and large-scale processing capability are remarkably improved.

The invention discloses a multi-antenna RSSI human body indoor positioning method and system based on a CNN-LSTM model, and the method comprises the steps: arranging a plurality of groups of RFID antennas in an indoor positioning region, and constructing an RSSI fingerprint database; collecting real-time RSSI values respectively detected by four groups of RFID antennas at the position of the to-be-positioned human body; collecting an RSSI value of the to-be-positioned human body containing interference person mobile interference information at the moment; processing the collected real-time RSSI value of the to-be-positioned human body and the RSSI value containing the mobile interference information of the interference person through a deep learning model, eliminating dynamic interference, and outputting a corrected RSSI value; and through a weighted dynamic K-nearest neighbor algorithm based on a difference value, matching the corrected RSSI value with the RSSI average value, and calculating to obtain a three-dimensional coordinate of the to-be-positioned human body. In order to improve the accuracy of an indoor positioning algorithm and the adaptive capacity in a complex environment, the RFID technology and deep learning are fused, the multi-antenna RFID human body indoor positioning method is constructed, and the practicability is very high.

This invention relates to the field of marine multi-parameter information extraction technology, specifically to a method for extracting marine multi-parameter information based on terminology semantics. It introduces advanced clustering and classification algorithms to perform deep learning and optimization of marine terms; utilizes terminology expressive power and Chinese phrase structure rules to optimize the K-nearest neighbor algorithm to identify marine entities and their attributes in the text; performs entity and attribute extraction and standardizes the extracted entities and attributes; designs a maximum correlation algorithm to assess the importance of different parameters in specific marine events or processes; and matches the extracted parameter information with marine structured data through multi-dimensional correlation analysis to achieve multi-dimensional information association. Based on the analysis results, it outputs the optimized extracted content. By combining the ecosystem characteristics and seasonal changes of the sea area, it selects the parameters with the greatest impact on water quality, generating intuitive reports and recommendations to provide decision-making basis for managers.

The application discloses a kind of dynamic self-adapting recommendation strategy optimization methods for business handling failure scene, it is related to business handling recommendation and intelligent decision-making technical field, first, it is collected to handle the whole process of business multi-class data, with 100 millisecond / second frequency guarantee complete real-time;With decision tree and bayesian network fusion algorithm attribution, direct and indirect reasons are clear;Integrate data to build user portrait, mine potential and subsequent demand;Based on attribution and portrait generation recommendation scheme set, priority and form are adjusted in combination with scene characteristics;Feedback data is introduced, and the strategy weight is optimized using gradient descent algorithm, and the scheme is updated regularly;Set multi-dimensional index weighted evaluation effect, not up to standard then emergency optimization;Establish distributed strategy library, and reuse similar scene optimal strategy using K nearest neighbor algorithm.The application realizes accurate positioning and personalized recommendation of failure causes, and the recommendation effect is continuously optimized with data accumulation, and is suitable for multiple business types and user groups.

The invention discloses a BIM model number inputting and auditing method based on GH and Revit. A two-dimensional drawing is cleaned and number data is extracted through Rhino, the centroid of a BIM model component is extracted by utilizing Grasshopper and is projected to generate a point cloud, the number of the two-dimensional drawing and the BIM model component are matched based on a nearest neighbor algorithm, and batch inputting and dynamic synchronous updating of the number are realized. And reading an instance parameter value of the BIM model component through Grasshopper, and automatically comparing the instance parameter value with the two-dimensional drawing number data set to generate an Excel auditing report containing a consistency mark. The invention relates to the technical field of building modeling, and can solve the technical problems in the prior art.

The invention discloses a tennis ball collecting robot and a control method, system and equipment thereof, and the method comprises the steps: obtaining a coordinate data set of scattered tennis balls in a target area in response to a received ball collecting instruction; according to the coordinate data set, coordinates of a starting tennis ball are obtained, and the starting tennis ball is configured to be a tennis ball located at the edge of the scattered tennis ball; obtaining an initial path based on an improved nearest neighbor algorithm according to the coordinates of the initial tennis ball and the coordinate data set; and according to the initial path, eliminating path intersection by a local optimization algorithm to obtain an optimal path. According to the scheme, through an innovative mixed path planning algorithm, scattered tennis balls are collected within the shortest time, the collection efficiency is remarkably improved, and the energy consumption of the robot is reduced.

The invention relates to a particle-ball-based quantum HVS graph KNN method, and belongs to the field of quantum calculation and machine learning. Aiming at the technical problems of low calculation efficiency and large resource consumption when a classical nearest neighbor algorithm is used for processing high-dimensional big data, the invention provides a hierarchical search scheme fusing granular ball reduction and quantum parallel calculation. The hierarchical search scheme comprises the following steps: generating a low-dimensional feature unit through a granular ball compression original data set; constructing a hierarchical Voronoi diagram structure to realize a coarse-to-fine search path; encoding data by adopting a quantum random access memory, and accelerating similarity calculation by utilizing a quantum exchange test circuit; and dynamically screening nearest neighbor nodes in combination with the priority queue. The method significantly improves the classification efficiency, effectively reduces the occupation of quantum bit resources, guarantees the classification precision, and provides technical support for the landing of quantum machine learning in industrial detection and other scenes.

According to the artificial intelligence ammeter data visualization-based power distribution system fault position identification method provided by the invention, three basic learners, namely a random forest, a K-proximity algorithm and an artificial neural network, are combined, and an integration method of combining a plurality of different machine learning models into a prediction model can be adapted to different scenes, so that the accuracy of the prediction model is improved. For example, noise and data loss, fault resistance change, load or power distribution feeder structure change and the like can be realized. An integrated voting classifier is developed, and a random forest, kNN and an artificial neural network model are utilized to classify fault types and identify a limited number of data points of FL only using voltage measurement. According to the method, the key problem of insufficient multiple features is overcome, the fault position can be quickly identified, the method is not influenced by the initial angle of the fault and the resistance on the line, the anti-interference capability is high, the classification accuracy of multiple fault types is high, the detection precision is high, and the cost is not too high.

The application relates to a river closure date prediction method based on historical similar year performance driving, which comprises the following steps: using a plurality of feature selection algorithms and leave-one-out cross-validation to screen target predictor sets respectively matched with selected machine learning models and statistical models; based on parameter sensitivity analysis and a Bayesian optimization algorithm, sensitive hyperparameters of the machine learning models are optimized to obtain optimized machine learning models; the statistical models and the optimized machine learning models are configured as candidate river closure date prediction models; a K-neighbor algorithm is used to search a set of similar historical years in a historical observation data set according to current observation data, and a target river closure date prediction model is dynamically optimized according to the comprehensive prediction error of the candidate river closure date prediction models on the set of similar historical years, and a prediction result is output, so that the advantages of multiple models are effectively fused, the generalization limitation of a single model in a complex non-stationary environment is avoided, and the accuracy and robustness of the prediction result are significantly improved.

The invention discloses a tree point cloud branch and leaf separation method based on dynamic hypergraph guided Mama, and the method comprises the following steps: S1, carrying out the farthest point sampling of FPS, obtaining N sampling points, taking the N sampling points as central points, taking each central point as a center, obtaining K-1 neighbors through a K-nearest neighbor algorithm, and forming N neighborhoods, each neighborhood containing K points; s2, each neighborhood point cloud is coded into a feature vector through a point cloud coding layer, and N feature vectors are obtained through the N neighborhood point clouds in total; s3, three hypergraphs are constructed for the N feature vectors based on the distance relation, the representation relation and the similarity relation; the problems that in the branch and leaf separation task of an existing Transform frame, the patch size is limited, so that the receptive field is reduced, and the precision is insufficient are solved; the method aims at solving the problems that when an existing model processes a complex high-similarity local area of forest point cloud, it is difficult to effectively extract features with the identification degree, and a point cloud serialization method is insufficient in expression ability.

This invention discloses a crowdsourcing testing platform system and method based on intelligent matching. The method includes: receiving test task requests containing task description text; using a hierarchical attention network model to perform semantic understanding of the task description text, generating structured skill tags, and dynamically updating the tag confidence of testers using a Bayesian probabilistic framework; calculating the initial matching degree using a cosine similarity algorithm that incorporates a recent task completion quality correction factor, and then filtering and constructing a candidate pool; for each candidate, constructing a dynamic feature matrix, and using an improved K-nearest neighbor algorithm based on Mahalanobis distance for ranking and recommendation, where the K value is dynamically determined based on the number of candidates in the pool and the task dwell time. This invention significantly improves the efficiency and accuracy of task allocation through a capability assessment mechanism, achieving optimized allocation of testing resources.

This invention discloses a method for diverse image synthesis based on retrieval-enhanced diffusion, comprising: constructing a dynamic retrieval pool containing real images and generated images; extracting feature representations semantically aligned with intermediate states during the generation process using a pre-trained feature extractor; and, at a specific denoising time step, introducing Top-K reference samples retrieved through an approximate nearest neighbor algorithm into an anti-attention module to actively guide the generated features away from the existing data distribution, thereby continuously expanding the feature space coverage of the synthetic dataset. Using this invention, while maintaining the visual quality of the synthesized image, diversity data augmentation can be performed based on existing data, enhancing the diversity and practical value of the synthesized data.

The invention provides a large model entity alignment prompting method and system. The method comprises the following steps of: obtaining semantic embedding representation by utilizing joint embedding and fine tuning of a knowledge graph and a text corpus; retrieving a text block set related to the semantics of the query entity in a text corpus through a k-nearest neighbor algorithm retrieval technology to obtain a retrieval result; searching k neighbor entity pairs of the query entity pair from an expert labeled sample pool based on the key information, and taking the k neighbor entity pairs as positive samples; searching k neighbor entity pairs of the query entity pair from a public data set based on key information as supplementary examples; constructing the positive samples and the supplementary samples into a demonstration sample set; converting the demonstration sample set into a natural language text to obtain a sample text representation; and inputting the retrieval result and the sample text representation into the large model as an entity alignment prompt. According to the large model entity alignment prompting method and system, the reasoning ability of the large model in the entity alignment task is remarkably improved.

The application discloses a steel structure welding method, and relates to the technical field of 3D visual welding, and comprises the following steps: step S1, scanning point cloud data of a steel structure; step S2, based on the scanned data in step S1, a random sample consensus method is used to find a plane; step S3, according to a visual coordinate, a normal line of the plane obtained in step S2 is unified; step S4, a boundary between planes is calculated according to the plane obtained in step S2; step S5, a nearest neighbor algorithm is used to search for the nearest neighbor points of each two groups of boundaries in step S4, and a common boundary is formed; step S6, straight line fitting is performed on the common boundary obtained in step S5, and end points are extracted as welds; and step S7, path planning is performed based on a distance method. The welding method provided by the application automatically solves the plate plate welding problem in a steel structure, uses a 3D visual system to identify welds, automatically plans a welding path, saves labor cost, improves welding precision stability, and shortens a welding cycle.

The application provides a kind of information push method and system based on big data matching and GPS positioning, wherein, the application is processed by receiving position data, position data is processed using Kalman filtering algorithm, combined with time stamp and velocity vector information, trajectory prediction algorithm is applied to predict movement trend, and potential interest area list is generated;Using spatial clustering analysis algorithm, the historical position data of the user is analyzed, the behavior pattern and the preferred place type of the user are identified, and the user interest model is constructed according to it;The dynamic adaptive geofencing system is configured by using the user interest model;Weighted K nearest neighbor algorithm is used to sort service providing points, and multi-modal fusion algorithm is used to combine geographic location information in service recommendation list with other sensor data to generate customized information package;The technical scheme provided by the application improves the accuracy of position information, enhances the accuracy and timeliness of prediction, and enhances personalized service recommendation.

The embodiment of the present disclosure discloses a wiring software test method and device, electronic equipment and storage medium, wherein the method comprises: inputting test cases and at least one image taking angle in the reference environment and the test environment respectively, obtaining at least one set of reference images and test images, wherein a set of reference images and test images correspond to one wiring rule; in response to the existence of differences in the at least one set of reference images and test images, determining the difference pixel points of the at least one set of reference images and test images; classifying the difference pixel points according to the nearest neighbor algorithm to obtain at least one difference area; and visually displaying the at least one difference area. The embodiment of the present disclosure can realize the automation of wiring software testing based on image processing technology, and solve the problem of extremely high test cost caused by manual testing and verification of each wiring rule.

The invention discloses a three-dimensional point cloud registration method based on a Transform structure driven by a Format principle, and the method comprises the steps: S01, obtaining point cloud data, dividing the point cloud data into a training set, a verification set and a test set, carrying out the preprocessing, and generating a numpy file; s02, obtaining an initial corresponding relation between a source point cloud and a target point cloud according to the point cloud data through a nearest neighbor algorithm, and forming a matching pair; s03, constructing a neural network model, training the neural network model based on the training set, the verification set and the test set until the model converges, and obtaining a point cloud registration network model; s04, sending to-be-registered point cloud data with partially overlapped scenes into the trained point cloud registration network model to obtain a final point cloud registration result, and according to the scheme, the local-to-global structural features in the matching pairs can be effectively obtained, the distinguishing accuracy of the inner points and the outer points is improved, and the accuracy of the point cloud registration is improved. Furthermore, a better transformation matrix is solved in an attitude estimation task, and the method has relatively high precision performance and application value.

The invention provides an intelligent teaching method, system and device based on AI intelligence and a medium, and relates to the technical field of intelligent teaching, and the method comprises the steps: collecting learning track data, teaching feedback data and image data of students in a teaching process; constructing a data set based on the learning track data and the teaching feedback data, converting the image data into feature points, and encoding the feature points to obtain feature encoding data; performing collaborative analysis on the data set and the feature coding data by adopting a K-nearest neighbor algorithm to divide sample clusters in different learning states, and extracting typical features from the sample clusters; performing time sequence analysis on the typical characteristics by adopting a long-short-term memory network to obtain evaluation results of the teaching matching degree and the learning efficiency; and finally, a teaching resource pushing strategy of the platform is optimized based on an evaluation result. According to the method, accurate portraying and dynamic evaluation of the learning state of the student can be realized, so that adaptive optimization of a teaching resource pushing strategy is driven.

The application belongs to the technical field of genomic breeding, and provides a machine learning genomic selection system and method, which includes the whole process of data import, model selection, hyperparameter tuning, model training, evaluation and prediction. The platform integrates various machine learning algorithms, including kernel ridge regression, support vector regression, random forest, k-neighbor algorithm, and provides automatic hyperparameter optimization, model selection and result visualization functions. The platform also provides rich data visualization functions to help users better understand data and model results. The machine learning genomic selection platform of the application has the characteristics of high stability, simple operation and high integration, significantly reduces the technical threshold of genomic selection, enables breeding researchers without programming background to efficiently perform genomic data analysis, and promotes the wide application of machine learning technology in animal breeding.

The invention provides an anomaly detection method and device based on machine working audio and a storage medium. The method comprises the following steps: acquiring a machine audio sample; extracting Mel filter bank energy characteristics of the machine audio sample; deep features of all the Mel filter bank energy features are extracted through a transformer encoder, and a plurality of deep general features and a plurality of depth to-be-measured features are obtained; constructing a training feature set according to all the deep general features, and determining an abnormal threshold value from the training feature set by adopting a K-nearest neighbor algorithm; obtaining a distance metric value between the depth to-be-measured feature and each deep general feature in the training feature set one by one, and generating an abnormal score of the depth to-be-measured feature based on the distance metric value; and comparing the abnormal score with an abnormal threshold value, and judging whether the to-be-detected machine has a fault or not according to a comparison result. According to the method, the detection threshold is determined only through the audio sample when the machine works normally, and fault detection can be realized without participation of an abnormal sample in the judgment process.