145 results about "Transition probability matrix" patented technology

The invention discloses a medical data center network risk assessment method based on a random walk model. The method comprises the following steps of 1, obtaining topological structure information and a node set of a medical data center network; 2, constructing a weighted undirected graph according to the topological structure and the data traffic information between the nodes; 3, in the weighted undirected graph, performing simulation based on a random walk model to obtain a transition probability matrix between nodes; 4, calculating a node risk score of each node according to the transition probability matrix; 5, performing aggregation processing on the node risk scores of the nodes to obtain an overall network risk score; and step 6, generating an assessment report according to the overall network risk score. According to the method, the risk conduction probability between the nodes is quantified by constructing the combination of the weighted undirected graph and the random walk model, the limitation of a traditional static assessment method is broken through, and a cross-node risk propagation path caused by data flow can be identified.

The invention belongs to the technical field of network security, and particularly relates to a multi-source network risk information fusion and risk assessment method, which comprises the following steps of: extracting nodes and edges of a real-time event through an entity extraction tool, and constructing a multi-source heterogeneous threat knowledge graph based on the nodes and the edges; calculating a random walk transition probability matrix of the edge, and calculating a time-sensitive personalized random walk value based on the random walk transition probability matrix; a graph neural network model is adopted for training, graph neural network embedding and prediction are carried out in combination with the features of the nodes, and the classification probability is output; and carrying out three-dimensional sub-graph segmentation on the predicted multi-source heterogeneous threat knowledge graph, constructing a mapping index, carrying out aggregation calculation to obtain a propagation path and path popularity of a high-risk node, analyzing a sub-graph evolution trend and carrying out early warning. The method is suitable for carrying out dynamic analysis and prediction on multi-dimensional attack risks, and trend attack analysis of the industry changing along with time is achieved.

The invention relates to the technical field of air tightness detection, in particular to a diaphragm type energy accumulator air tightness detection method which is used for solving the problems that in the prior art, equipment operation characteristics cannot be accurately described, typical defect modes cannot be recognized in combination with spectral clustering, and a quantitative basis cannot be provided for equipment fault trend analysis and intelligent maintenance. The method comprises the following steps: constructing a low-dimensional state map to describe equipment operation characteristics, identifying typical defect modes in combination with spectral clustering, establishing a nonlinear correlation model to reveal a defect evolution relationship, optimizing classification model parameters by adopting an evolutionary algorithm, improving the identification accuracy, mining a most probable defect evolution path based on a transition probability matrix, and improving the identification efficiency. And a quantitative basis is provided for equipment fault trend analysis and intelligent maintenance.

The invention relates to the technical field of furniture production management, and discloses an intelligent management method and system for furniture production. The method comprises the following steps: collecting production data such as material consumption, equipment operation parameters and process completion time of each process node on a production line in real time; performing multi-dimensional feature extraction on the data to generate a comprehensive feature set containing time sequence, statistical and associated features; dividing continuous production stages according to a characteristic dynamic change rule and distributing identifiers; based on the identifier recombination data, calculating a transition probability matrix between adjacent stages; identifying a potential abnormal stage and generating a mark sequence by analyzing a state jump abnormal mode in the matrix; constructing a quality prediction model in combination with the abnormal mark and the real-time data, and outputting a quality prediction score of each process node; dynamically adjusting a procedure production parameter configuration scheme according to the deviation degree of the score and a preset threshold value; and performing similarity matching on the adjusted scheme and historical optimal configuration, and screening a to-be-verified configuration set to perform a simulation test.

The invention belongs to the technical field of medical data processing, and particularly relates to an electronic medical record intelligent quality control method and system based on a medical knowledge graph, and the method comprises the steps: obtaining target medical record data, and constructing an initial logic drive vector; calculating the dynamic propagation impedance of each edge in the general medical knowledge graph according to the deviation degree between the measured value of the physiological index in the target medical record data and the physiological constraint condition of the edge attribute in the general medical knowledge graph; constructing a transition probability matrix by using the dynamic propagation impedance, and obtaining a steady-state correlation distribution vector of each node through iterative calculation; and calculating a dynamic judgment threshold according to the steady-state correlation distribution vector, and performing exception verification on an actual disposal instruction in the target medical record data. According to the method, the real-time physiological status of the individual patient can be dynamically mapped into the constraint condition of map reasoning, potential taboo caused by abnormal physiological indexes is effectively identified, and personalized quality control of the electronic medical record is realized.

The invention relates to a sleep staging method based on Markov chain dynamic loss, and relates to the field of data processing. The method comprises the steps that electroencephalogram signals are preprocessed to obtain training samples, a sleep staging model is constructed and trained, sleep staging is conducted through the trained model, and training comprises the steps that basic classification loss is calculated; when the sleep stage of the training sample is transferred to different stages, obtaining a real physiological transition probability through a Markov transition probability matrix, and if the probability is smaller than a threshold value, calculating a loss weight factor of the training sample according to whether the model correctly predicts the sleep stage of the current training sample; calculating an average value of the sequence sensing loss according to the loss weight factor and the basic classification loss; and calculating the gradient of the average value to the parameters of the sleep staging model, and updating the model parameters of the sleep staging model. According to the method and the device, the physiological interpretability of understanding and prediction of the sleep staging model on the overall sleep structure is improved, and then the sleep staging accuracy is improved.

The invention provides a method, device and equipment for promoting retention through interaction scene prediction and a storage medium, and the method comprises the steps: obtaining a current multi-modal interaction data stream between users in an interaction scene, carrying out the preprocessing of noise filtering, framing processing, size normalization, time aggregation and the like, extracting emotion features, and generating a real-time multi-dimensional observation vector sequence. The sequence captures complementarity and time sequence dependence of multi-modal information, and overcomes limitation of single-modal static analysis. The observation vector sequence is input into a pre-trained hidden Markov model, and the model is trained based on historical sequences and defines a hidden state set, an initial probability, a transfer matrix and an emission probability; the optimal emotional state path at the current moment is obtained through Viterbi algorithm reasoning, and probability modeling and dynamic prediction of emotion transfer uncertainty are achieved. The current emotion state is extracted from the path, the mimicry representation of the virtual pet is driven to be displayed on the interactive interface, visual feedback is formed, emotion connection is enhanced, and the user retention rate is increased.

The embodiment of the invention discloses an explosion source flame detection method and system, and the method comprises the steps: collecting a multi-mode signal of an explosion source flame, and the multi-mode signal comprises an acoustic vibration type signal and an optical type signal; preprocessing the acoustic vibration type signal and the optical type signal to generate an acoustic vibration feature set and an optical feature set; performing topology analysis on the acoustic vibration feature set by using a persistent coherence algorithm to identify a dynamic behavior mode of the explosion source flame, and obtaining a dynamic behavior mode identification result of the explosion source flame; the dynamic behavior pattern recognition result and the optical feature set are converted into a symbol sequence through a symbolic dynamics algorithm, a cross-modal symbol pattern is analyzed through a transition probability matrix, and comprehensive feature representation of the explosion flame is generated; and based on the comprehensive feature representation, judging whether an explosion source flame exists through a classification algorithm, and outputting a detection result. According to the invention, high-precision explosion source flame detection suitable for a complex environment is realized.

The present application relates to a seamless positioning method, device and storage medium for an indoor and outdoor mobile robot, the method comprising: S1, constructing a GNSS / IMU-based combined positioning sub-model outdoors and a vision / IMU-based combined positioning sub-model indoors; S2, inputting interaction; S3, model filtering; S4, transition probability correction: correcting a transition probability matrix using a new information vector; S5, model probability updating: calculating a likelihood function of the combined positioning sub-model and updating model probabilities of the combined positioning sub-models; S6, outputting interaction: fusing current time state estimation values and covariances output by the combined positioning sub-models, weighting the model probabilities to obtain system final state estimation values and covariances, and finally inputting the updated model probabilities to step S2 for next iteration. Compared with the prior art, the present application realizes seamless positioning service in indoor and outdoor navigation.

The invention discloses an electric power system key line identification method based on a hypergraph model, and the method comprises the steps: 1, defining each power transmission path as a hyperedge through power flow tracking, defining each line as a node, and constructing a hypergraph model of an electric power system; 2, performing K-Shell decomposition on the hypergraph model, and calculating the structural importance of each power transmission line; 3, calculating the magnitude of the load flow of the hyperedge, and calculating the magnitude of the weight of the hyperedge by fusing the importance of the branch structure and the magnitude of the load flow of the hyperedge through an entropy weight method; 4, descending sorting is carried out according to hyperedge weights, and a key line sequence is output; step 5, constructing a node transition probability matrix based on hyperedge weight; step 6, screening unconnected high transition probability node pairs to generate a candidate branch set; and step 7, with maximization of the evaluation index load balancing entropy LBE as a target, adding branches in an accumulated manner until the LBE reaches a peak value, and completing structure optimization. According to the invention, the recognition precision of the key line can be improved, and balanced power flow distribution can be realized.

The present application provides a method, device and electronic device for determining the comprehensive benefits of a power distribution system. The method includes: obtaining multiple target power usage scenarios of the power distribution system after it is put into use, establishing an indicator set corresponding to each target power usage scenario of the power distribution system, and constructing a target evaluation vector corresponding to each target power usage scenario based on the indicator set; obtaining the initial scenario vector and the transition probability matrix of the power distribution system, calculating the product of the initial scenario vector and the transition probability matrix to obtain a probability vector; calculating the sum of the products of the probabilities of occurrence of multiple target power usage scenarios in the probability vector and the evaluation values ​​of the corresponding multiple target power usage scenarios in the target evaluation vector to obtain a comprehensive benefit score of the power distribution system. Through the present application, the problem that the evaluation index system of the power distribution system in a single scenario is difficult to fully reflect the comprehensive benefits of the system is solved, and the purpose of determining the comprehensive benefits of the power distribution system under multiple scenarios is achieved.

The invention provides a multi-target path planning method and a related device, and belongs to the technical field of path planning. The method comprises the following steps: dispersing a working space of a multi-target path planning model to obtain a real cell space; the multi-objective path planning model takes the minimum driving path and the maximum safety as objective functions; the constraint conditions of the multi-objective optimization model comprise kinematics constraint, boundary condition and obstacle constraint; generalized cell mapping is established on the real cell element space for each state variable and control variable of the constraint condition and the target function, and a cell mapping relation based on a transition probability matrix of all controllable cells in the real cell element space is obtained; the cell mapping relation is stored by adopting a hash table; based on the cell mapping relation, solving the multi-target path planning model to obtain an optimal solution set; and performing path planning on multiple targets according to the obtained optimal solution set to obtain a multi-target path planning result. According to the method, the problems of low accuracy and calculation efficiency of multi-target path planning are solved.

The invention relates to the technical field of logging evaluation in oil-gas exploration, and discloses a lithology identification method and device based on logging information and electronic equipment. The method comprises the following steps: determining sensitive logging information; preprocessing logging information of the source research area and the target research area; establishing a lithology prediction model Mo based on the convolutional neural network for the source research area; finely adjusting the lithology prediction model Mo based on a small number of samples in the target research area to obtain a lithology identification model Mt of the target research area; calculating a Markov prior probability and a transition probability matrix between different lithologies for the sample set of the target research area; predicting a lithology result of the test set sample of the target research area by adopting a lithology prediction model Mt; using Markov prior probability to correct the predicted lithology result to obtain a posterior probability; and determining a final lithology category. According to the technical scheme of the invention, effective lithology identification can be carried out on a complex lithology reservoir, and the lithology identification precision is high; petroleum geology research and dessert optimization are facilitated.

The application discloses a self-learning visual positioning method, comprising the following steps: S1, acquiring image data; S2, image enhancement; first, extracting brightness information from the picture, S3, dividing the acquired image information into grids based on the YOLOv5 algorithm, and extracting feature information and label information in the grid information, then adjusting the weight information of the feature information and the label information through the adaptive learning method of Bayes, then merging the feature information according to the adjusted weight information, and converting the merged image information into three-dimensional coordinate information; S4, adjusting the predicted speed value and the predicted three-dimensional coordinate information through the least square method, S5, predicting the posture data through the Markov data transition probability matrix, the predicted speed information, the predicted three-dimensional coordinate information and the target frame information of the current position in the image information in the S3 step, and then positioning and moving according to the predicted posture data, and the positioning data is good in reliability.

The application discloses a software abnormal behavior dynamic analysis method based on a non-uniform Markov chain, which comprises the following steps: S1, collecting and preprocessing software log data; S2, dynamically dividing states by using K-means clustering to generate a preliminary state set and a transition relationship; S3, constructing an improved non-uniform Markov chain model; S4, performing abnormal detection by the improved non-uniform Markov chain model and calculating an abnormal score; S5, constructing an objective function and optimizing by using a grey wolf optimization algorithm; S6, feeding back the optimized parameters to S3 to update a time-varying transition probability matrix and the abnormal score calculation; S7, judging the abnormality according to the updated abnormal score and a warning threshold, outputting a warning signal and feeding back. The application realizes real-time and accurate detection and self-adaptive optimization of software abnormal behaviors by constructing an improved non-uniform Markov chain and a software abnormal behavior dynamic analysis method based on a grey wolf optimization algorithm, thereby significantly improving system security and stability.

This invention relates to the fields of tunnel engineering and 3D geological modeling technology, specifically to an intelligent modeling method and system for 3D geological uncertainty in tunnels based on transition probability geostatistics and sparse boreholes. The method includes: S1, integrating multi-source data to construct a 3D geological conceptual model; S2, statistically characterizing a one-dimensional transition probability matrix, calculating and fitting a spatial continuity and 3D anisotropic variability function model; S3, calculating the prior spatial probabilities and transition adjustment factors for various lithologies, obtaining the posterior lithology distribution of nodes through Bayesian intelligent updating, and initially assigning lithology categories to nodes through random sampling; S4, assigning the most probable lithology category to each grid node, calculating the variance of lithology values ​​across all implementations, and measuring model uncertainty; S5, outputting the optimal 3D uncertainty model for the tunnel; and S6, verification and evaluation. This invention can effectively achieve 3D heterogeneous modeling and explicit quantification of uncertainty under strong geological constraints.

The present disclosure relates to systems, non-transitory computer readable media, and methods that provide graphical user interfaces comprising future transportation options with varying time windows at different transportation values and dynamically analyze the time windows to identify provider devices to fulfill transportation requests based on provider device efficiency metrics. For instance, the disclosed systems can delay selection of a provider device within a future time window utilizing a dynamic threshold provider device efficiency metric. For instance, the disclosed systems can analyze historical distributions of provider devices to generate a transition probability matrix that is utilized to analyze current provider devices and determine a threshold provider device efficiency metric that reflects the likelihood of identifying more efficient matches in the future. The disclosed systems can compare the determined threshold to anticipated efficiency metrics for individual provider devices to generate matches for digital transportation requests.

The invention discloses a hybrid off-road vehicle energy management method based on model reinforcement learning, and relates to the field of vehicle energy management control. The method comprises the following steps: acquiring historical information data of the hybrid off-road vehicle when the hybrid off-road vehicle is off-line; constructing a high-order Markov chain model at the current moment and training the intelligent agent by adopting a reinforcement learning method to obtain an energy management control strategy at the current moment; performing online updating on the high-order Markov chain model at the current moment by adopting a recursion method; based on the induction matrix norm, determining a matrix norm difference between the high-order Markov chain model at the current moment and the transition probability matrix corresponding to the updated high-order Markov chain model; and judging whether the norm difference of the matrix is greater than a set threshold value or not so as to update the model or control the driving condition of the hybrid off-road vehicle by adopting an energy management control strategy at the current moment. The invention aims to realize power distribution of the hybrid off-road vehicle.

The application provides a vehicle suspension state parameter estimation method and device, and relates to the technical field of intelligent control of vehicle suspension, and the method comprises the steps of: constructing an interactive state observer based on at least two physical characteristic parameters of a vehicle suspension; determining prior state prediction parameters corresponding to each sub-observer based on a previous model probability and a transition probability matrix corresponding to each sub-observer; updating the prior state prediction parameters of each sub-observer based on current observation data corresponding to each sub-observer to obtain posterior state estimation parameters of each sub-observer; adaptively determining a current forgetting factor corresponding to each sub-observer based on the current observation data, an observation matrix and the prior state prediction parameters of each sub-observer; and fusing the current model probability, the posterior state estimation parameters and the current forgetting factor corresponding to all sub-observers to obtain a current overall state parameter estimation value corresponding to the interactive state observer. The application can improve the estimation accuracy of the suspension state parameters.

The invention discloses a method for rapidly detecting the impurity content in the preparation process of organic silicon emulsion, and relates to the technical field of impurity detection.The method comprises the steps that technological parameters of all current steps in the preparation process of the organic silicon emulsion are collected in real time; calculating the probability of various impurities introduced in the step; when the impurity introduction probability of the single step exceeds a first preset threshold value, generating a high-risk impurity list and triggering online directional detection; if the impurity introduction probabilities of all the steps do not exceed a first preset threshold value, constructing a transition probability matrix for describing the state change of the impurities in the whole process; calculating and predicting steady-state probability distribution of various impurities in the finished product; and performing finished product terminal detection on the impurities with the prediction probability exceeding a second preset threshold. The method has the advantages that multi-level early warning of impurity risks is achieved, source abnormity is positioned through the probability model, a transfer matrix tracks a full-process impurity migration path, accurate directional detection is combined with a terminal focusing strategy, and the impurity detection efficiency is improved.

This invention discloses a few-shot learning named entity recognition method that integrates entity label encoding, comprising: step (1) obtaining character features and their sequences; step (2) obtaining word pair feature matrix, distance feature matrix, and region feature matrix, and concatenating them to obtain word pair relation feature matrix; step (3) obtaining the feature representation of each entity label; step (4) calculating the dot product similarity between each word pair relation feature and entity label feature to obtain the transition probability matrix; step (5) selecting the optimal model as the final pre-trained model; and step (6) loading the pre-trained model for few-shot learning, inputting sentences into the model, and outputting the word pair relation matrix. This method encodes word pairs and entity labels separately, and then matches the representations of word pairs and entity labels to obtain the entity relationships between words. It can effectively identify continuous entities, overlapping entities, and discontinuous entities under low resource conditions.

This invention provides a rapid depression detection method based on a weighted degree transfer network, comprising: acquiring and preprocessing the EEG signals of a subject to obtain a one-dimensional EEG time series; mapping the one-dimensional EEG time series into a complex network using a weighted horizontal visualization algorithm; extracting degree and intensity sequences; using the deduplicated set of degree values ​​as nodes of the new network; tracing the degree transfer path between adjacent time points; calculating the product of the degree difference and intensity difference between adjacent time points as the edge weights of the degree transfer path, thereby constructing a weighted degree transfer network; normalizing the network to obtain a transition probability matrix; calculating its Shannon entropy and MPR statistical complexity as a joint feature vector; and using a machine learning classifier to identify the depression state. This invention does not require presetting parameters such as embedding dimension, and by fusing the dynamic difference features of degree and intensity, it can quickly and accurately capture abnormal nonlinear dynamic patterns in the EEG signals of depression.

The application provides a watermark adding method, a watermark detection method, a watermark processing device and a computer storage medium. The watermark adding method comprises the following steps: inputting user interaction content and historical output text into a text generation model to obtain an original text probability distribution; obtaining a preset probability transition matrix; correcting the original text probability distribution by using the probability transition matrix to obtain a corrected text probability distribution; and taking a text corresponding to a maximum probability value in the corrected text probability distribution as current output text after watermark adding. By using the watermark adding method, the accuracy and usability of generated content are not damaged by directly setting a red-green set for a single character by using a hash function or other methods, and the output of the large model is corrected by using the probability transition matrix, so that the accuracy of watermark adding is improved.

The invention belongs to the technical field of medical data mining and rehabilitation evaluation, and provides a skeletal muscle spasm-to-contracture evolution rule quantitative analysis method and system and application, and the method comprises the steps: preprocessing spasm contracture time sequence data, and obtaining one-dimensional feature vector data; k-means clustering is carried out, state labeling of spasm and contracture is carried out, and clustering labels are obtained; determining the change rate of the torque characteristic and the adjacent angular velocity characteristic, and generating an eight-bit pseudo-sequential sequence representing the evolution rule from the spasm state to the contracture state in combination with a state transition threshold value determined by an ROC curve; constructing a Markov chain model to calculate a transition probability matrix from a spasm state to a contracture state, and identifying torque as a key driving feature of state evolution through grouping risk ratio; and constructing a dynamic correlation model based on the key driving features to complete quantitative analysis of the evolution rule from skeletal muscle spasm to contracture. According to the method, objective division from skeletal muscle spasm to contracture state can be realized, and the transition probability and key driving factors between the skeletal muscle spasm and the contracture state can be excavated.

The application relates to the technical field of automobile network security, in particular to a vehicle-mounted safety early warning method and device, a vehicle and a storage medium, the method comprising the following steps: mapping initial multi-modal data of a current vehicle to a unified space-time dimension to obtain calibrated multi-modal data, inputting the calibrated multi-modal data into a space-time Markov chain model, inferring the running state of the current vehicle by using a dynamic transition probability matrix, predicting the state evolution trend in future continuous time steps to obtain a prediction result, comparing the prediction result with a preset risk threshold, and determining a current risk level according to the prediction result in the case that the prediction result meets a preset early warning risk condition, and triggering a target alarm mechanism based on the current risk level. Therefore, the problems that a traditional VSOC depends on artificial rules, has a high false alarm rate and is difficult to identify complex attacks are solved, intelligent fusion of multi-modal data and dynamic risk prediction are realized, and the accuracy and real-time performance of threat identification are improved.

The application relates to the field of artificial intelligence and information retrieval technology, and discloses a temperature control acupuncture field knowledge retrieval method based on deep learning. In the field of temperature control acupuncture, acupoints, temperature and literature are abstracted as nodes and numbered according to a unified rule, a graph space is constructed, an adjacency matrix is filled according to the relationship among the three types of nodes, the node degree is calculated and normalized into a transition probability matrix; the acupoint number and temperature interval input by a user are received, a query node is obtained and an initial access vector is constructed, random walk is iterated on the graph according to the transition probability within a preset number of rounds, a converged access vector is obtained, and a literature node component is extracted therefrom as a retrieval score, the literature is sorted, and index or metadata information of the literature with a high score is output. Through the above steps, the acupoint and temperature conditions and the literature structure are uniformly modeled, the problems of dispersed constraints and low efficiency are alleviated, and the matching degree of the result and the temperature control context is improved.

This invention belongs to the technical field of logistics management, specifically relating to a method and system for predicting the arrival time of goods in transit based on real-time environmental data. It addresses the technical problems of existing methods, such as difficulty in quantifying uncertainty and neglecting real-time data quality and vehicle status. The prediction method includes: S1, calculating Shannon entropy based on the prior probability distribution of arrival time at the current moment; S2, adjusting the basic transition probability matrix to obtain a spatiotemporally correlated transition probability matrix; S3, generating evidence update weights by comprehensively considering the quantification index of prediction uncertainty and the credibility of observational evidence; and S4, obtaining the posterior probability distribution of arrival time at the current moment through Bayesian updating. This invention can improve the accuracy of arrival time prediction in complex traffic environments.

The invention discloses an advertisement click rate prediction method and system based on user behaviors, and relates to the field of click prediction, and the method comprises the steps: collecting user behavior data and spatio-temporal trajectory data, constructing a user movement graph structure, extracting a spatial embedding vector through a graph neural network, extracting a time embedding vector through the user behavior data, and obtaining a user movement graph structure; generating a space-time behavior vector according to the space embedding vector and the time embedding vector, inputting the space-time behavior vector into a generative adversarial network, separating a short-term interest vector and a long-term interest vector through a generator, and constructing a Bayesian network based on historical behavior data to obtain an interest state transition probability matrix; based on the short-term interest vector, the long-term interest vector and the interest transition probability matrix, utilizing a Bayesian-Markov model to predict a user behavior link probability, and generating a preliminary click rate prediction probability; according to the method, the user behavior link probability is predicted through the Bayesian-Markov model and the variational auto-encoder, and dynamic probability modeling is realized.

The invention provides a robot scene understanding method and system based on visual deep learning, and relates to the technical field of robot visual identification, and the method comprises the steps: building a target relation matrix through multi-layer pyramid feature decomposition and bidirectional feature transmission between local regions; and calculating a target transition probability matrix by using historical scene data to optimize the target attribute of the current scene. According to the method, dynamic modeling of the relationship between targets in the scene and historical scene knowledge migration are realized, and the accuracy and robustness of robot scene understanding in a complex environment are improved.

The invention discloses a probability model-driven smelting furnace life prediction method and system. The method comprises the following steps of collecting multi-source alarm and process data of smelting key components; feature engineering is carried out, degradation features are extracted, and discrete health states are divided; the current health state of the component is accurately recognized under the small sample condition by using the gray correlation degree theory; constructing a state transition probability matrix in combination with a Markov chain, and quantifying the randomness of the degradation process; and predicting the remaining service life of the component by solving an equation set taking failure as an absorption state. According to the method, the problem of life prediction caused by lack of historical failure data is effectively solved, dynamic, quantitative and interpretable evaluation of the residual life of key components such as an oxygen lance, a furnace lining and a flue is realized, and reliable technical support is provided for predictive maintenance decision-making of the smelting furnace.