74results about How to "Resolve dependencies" patented technology

The invention discloses an electrical switch cabinet detection method based on the Internet of Things, and particularly relates to the technical field of electrical switch cabinets. A sensor is installed inside or outside a switch cabinet, temperature, humidity, current, voltage and gas concentration are collected, data are transmitted to an edge computing node for preprocessing by using a wireless communication technology, the preprocessed data are uploaded to a cloud platform, storage and trend analysis are performed at a cloud end, and the data are analyzed based on big data analysis and an artificial intelligence algorithm. According to the invention, the system can predict potential faults, a user can check the operation state and the prediction result of the equipment in real time through the cloud platform, and carries out the emergency shutdown or equipment scheduling operation remotely according to the early warning information, thereby effectively solving a reliability problem caused by communication interruption or data delay in an existing Internet of Things system, and improving the safety and intelligent level of the monitoring of the switch cabinet.

The invention relates to a server application architecture migration method and device and a storage medium, and the method comprises the steps that an automatic detection module analyzes and checks a to-be-migrated application under an X86 architecture, and the to-be-migrated application is subjected to static code analysis, dynamic performance analysis and system compatibility check; according to analysis and check results of the automatic detection module, the automatic adjustment module replaces an X86 specific instruction and a library function, performs memory allocation optimization and memory mapping adjustment, and generates an automatic adjustment report; and the automatic optimization module executes a performance benchmark test, and executes memory access optimization, algorithm optimization, task scheduling optimization and multi-core parallelization adjustment according to a test result. According to the server application architecture migration method and device and the storage medium, the architecture adaptation process can be simplified, performance optimization support and automatic dependency relationship management can be provided, the migration efficiency can be improved, and the cost can be reduced.

According to the welding process parameter recommendation method using the knowledge graph to retrieve the enhanced large model, the welding process knowledge graph is constructed, multi-source heterogeneous welding knowledge is stored in the graph, and effective management and reuse of the welding knowledge are achieved; meanwhile, through user feedback and verification, the welding process knowledge graph can be continuously updated and perfected, and continuous accumulation and iterative optimization of knowledge are achieved; the welding process knowledge graph is used as an external knowledge base to retrieve and enhance the large model, and accurate and comprehensive welding process related knowledge can be provided for the large model to serve as a reasoning basis, so that the illusion problem of the large model is effectively relieved; according to the method, intelligent recommendation of the welding process parameters is realized, personalized analysis can be carried out according to specific requirements of the user, welding process parameter suggestions meeting user expectations are quickly generated, the efficiency of welding process parameter recommendation is effectively improved, and the problems of strong expert experience dependence and relatively high cost in process parameter recommendation are solved.

The invention provides a power distribution network and traffic network cooperative operation method considering multi-subject benefit coupling, and belongs to the technical field of network system control, and the method comprises the following steps: S1, constructing a three-layer power traffic coupling network cooperative operation model, comprising a lower-layer traffic network user balance model, a middle-layer charging network operator optimal pricing model and an upper-layer power distribution network second-order cone relaxation optimal power flow model; s2, converting the lower-layer traffic network user equilibrium model into a group of constraints by applying a KKT condition, thereby converting the three-layer electric power traffic coupling network collaborative operation model into a double-layer model; and S3, solving the double-layer model by using a CCG algorithm. According to the power distribution network and traffic network cooperative operation method considering multi-agent benefit coupling, the system operation efficiency is optimized, the cooperative interaction among the operation agents is ensured, and the overall cooperative effect of the power distribution network and the traffic network is improved.

The invention discloses a thermal generator set load prediction system based on machine learning, and the system overcomes the defects of a conventional load prediction method through an end-to-end solution composed of a comprehensive data collection module, a multi-level feature engineering module, a multi-model integration module, a self-adaptive optimization module and the like. The system collects multi-dimensional data from a thermal generator set control system, an environment monitoring system, a fuel detection system and a historical database, wherein the multi-dimensional data comprises load data, environment data, fuel characteristics and equipment operation data. According to the multi-dimensional data integration mode, the model not only depends on single historical load data, but combines multiple influence factors such as environment, fuel and equipment states, the problem that a traditional method depends on a single data source is solved, the load prediction precision is effectively improved, and the load prediction efficiency is improved. And the adaptive capacity of the prediction model to different environment conditions and equipment states is stronger.

The invention discloses a compact model parameter extraction method based on a DQN algorithm, and the method comprises the steps: firstly collecting a current-voltage characteristic curve and a capacitance-voltage characteristic curve of a device, and determining a BSIM model parameter needing to be extracted; secondly, simulating the initial BSIM model by using Hspice, regarding parameter extraction of the BSIM model as a Markov decision problem, combining with a deep Q network, constructing and training a NoisyNet-DQN model, updating network parameters through a gradient descent method, and improving the precision of the NoisyNet-DQN model through continuous iteration; and finally, verifying the effectiveness of the model through a parameter result after Hspice simulation training is completed. According to the compact model parameter extraction method, the accuracy of automatic parameter extraction is effectively improved, and the problems of high dependence on experience, low efficiency and the like in the traditional manual extraction process are solved.

The invention relates to the technical field of transportation or storage devices, in particular to a carrying robot for stacking warehousing, which comprises a machine main body and a lifting mechanism arranged on the machine main body and used for lifting goods to a preset height, the lifting mechanism is provided with two supporting parts, each supporting part comprises a supporting shaft group and a belt arranged outside the supporting shaft group in a sleeving manner, each supporting shaft set is provided with a supporting end and a compensation end perpendicular to the supporting end, the supporting ends of the two supporting shaft sets are arranged in the same supporting plane and arranged in the supporting plane in a telescopic mode in the first direction, and the compensation ends are constructed to be capable of synchronously stretching out and drawing back in the second direction when the supporting ends stretch out and draw back. Therefore, the overall length of the supporting shaft group is kept unchanged. By arranging the supporting shaft set structure with the telescopic supporting end and the compensation end, the stable lifting and placing functions of goods under the condition that no tray exists are achieved, the problem that a traditional carrying robot depends on the tray is solved, and the storage cost is reduced.

The invention discloses a time sequence generation method and system based on uncertainty quantization and frequency domain constraint, and belongs to the technical field of artificial intelligence and data mining. Outputting a mean value of the completion value and a variance representing the uncertainty of the completion result; an entropy weight adaptive mask generation mechanism is constructed, a soft mask matrix of continuous values is generated according to variance and local observation density, and high uncertainty corresponds to low mask weight; constructing a mixed loss function containing time domain weighted loss and frequency domain consistency loss; and training a diffusion model based on the soft mask matrix and a mixed loss function to generate a regularized high-fidelity time sequence. According to the method, the Bayesian uncertainty quantization and soft mask mechanism is introduced, so that the problems that a binary mask mechanism in the prior art is rigid and cannot sense completion errors are solved, and meanwhile, the periodic characteristics of generated data are ensured in combination with frequency domain constraints.

The invention discloses a GNSS (Global Navigation Satellite System) and wind cloud satellite assisted meteorological and hydrological coupled runoff forecasting method. The method comprises the following steps: 1, GNSS two-dimensional water vapor inversion; 2, performing water vapor inversion on the Fengyun No.4; 3, performing three-dimensional variation assimilation on the precipitable water volume inverted by the GNSS and the precipitable water volume inverted by the FY-4 star A based on a WRF meteorological mode; and 4, runoff forecasting is carried out based on the WRF-Hydro hydrological model. The method is simple in step and reasonable in design, three-dimensional variational assimilation is performed on the precipitable water volume inversed by the GNSS and the precipitable water volume inversed by the FY-4 star A based on the WRF meteorological mode, accurate meteorological driving data is provided for a WRF-Hydro hydrological model, the dependence of WRF meteorological mode data assimilation on conventional meteorological observation data in traditional meteorological and hydrological coupling forecast is avoided, and the method is suitable for popularization and application. And the applicability of the WRF and WRF-Hydro model coupling mode in the meso-and small-scale watershed in the conventional data scarcity region is further improved.

The invention relates to a neural network acceleration method and device based on algorithm and hardware cooperation and an accelerator, and the method comprises the steps: processing a value-level index of a quantization tensor based on a sparsity threshold value, so as to generate layer metadata, and dynamically adjusting a sparsity mode; identifying the positions and the number of non-zero bits according to symbol conversion and a bit lookup table, and putting the processed data into bit groups to eliminate non-contribution bits; processing the bit groups, distributing the related input and tensor of the self-adaptive indexing mechanism extracted from the layer metadata to an elastic calculation unit, and executing shift addition calculation according to the number of non-zero bits; and on the basis of the column type calculation array, instructing all the arrays to complete multiplication operation in a preset period, and storing partial products and data streams corresponding to respective bit widths. According to the method, the accelerator controller can efficiently skip the zero positions in the calculation process, and meanwhile, the performance and applicability of the bit-level accelerator are improved.

The invention provides an electric aviation fuel oil and lubricating oil integrated pump device, which relates to the technical field of oil pumps, and comprises a shell, a motor, a fuel oil pump, a lubricating oil pump, a fuel oil pump outlet filter, a lubricating oil pump inlet filter and a lubricating oil pump outlet filter, the motor, the fuel pump and the lubricating oil pump are all installed on the shell, the output end of the motor is connected with a driving gear, and the fuel pump is connected with a fuel transmission gear; the lubricating oil pump is connected with a lubricating oil transmission gear; the fuel oil transmission gear and the lubricating oil transmission gear are respectively meshed with the driving gear; the fuel oil pump outlet filter, the lubricating oil pump inlet filter and the lubricating oil pump outlet filter are respectively arranged in the shell; a fuel oil inlet plug, a fuel oil outlet plug, a lubricating oil inlet plug and a lubricating oil outlet plug are respectively arranged in the shell; and a plurality of internal flow channels are also arranged in the shell. The device has the advantages that the overall structure is simplified, the occupied space and cost are saved, meanwhile, fault points of the system are reduced, and installation and maintenance are convenient.

The invention relates to the technical field of image recognition, and particularly discloses an image recognition optimization method based on a deep learning algorithm, and the method comprises the steps: 1, selecting a training data set for image recognition optimization, and carrying out the transformation of the training data, and enabling a model to be able to learn better robust features; 2, constructing a model of an image recognition task; according to the method, a deep learning model generally has relatively strong dependence on a large amount of annotated data by utilizing the matching effect of a plurality of steps, so that the problem of data dependence is solved, which means that the model can more effectively utilize limited data to learn and predict, the adaptability of the model in different data sets and scenes is improved, and the method is suitable for large-scale popularization and application. Moreover, the generalization ability is one of important indexes for measuring the performance of the deep learning model, and the improvement of the generalization ability of the model means that the model can better adapt to new data and scenes, thereby improving the accuracy and robustness of image recognition, and improving the application of the image recognition technology in wider scenes.

The invention discloses a building edge segmentation method and system based on adaptive prompt of a segmentation large model, and belongs to the technical field of remote sensing image processing. The method comprises the following steps: firstly, inputting a to-be-segmented image into an SAM large model for full segmentation, and identifying semantic information of segmented patches by a classifier; performing instance object identification and spatial recombination on the segmented plaques, and automatically generating a positive prompt point set and a negative prompt point set through spatial clustering and similarity calculation; and re-inputting the positive and negative prompt point sets of the image to be segmented and the instance object into the SAM model to carry out building edge refinement re-segmentation, and finally obtaining a complete segmentation result of the image to be segmented. Compared with the existing SAM fine tuning method and prompt learning training method, the method provided by the invention greatly reduces the task learning difficulty, and can effectively realize accurate positioning and fine segmentation of the building edge.

The invention discloses a real-time topic tweet classification method and device based on small samples, and the method comprises the steps: S1, obtaining original tweet sample data, and marking the data; s2, dividing the original tweet data marked with the data into a training set and a verification set, and performing preprocessing; s3, establishing a word embedding model, inputting the preprocessed training set data into the model for training, and converting the labeled data into word vector representation; s4, establishing a classifier model, transmitting the training set data into the classifier model, and performing pre-training by taking a mask language modeling task as a target task; s5, obtaining tweet data containing each target topic specific keyword on each large social network site through a web crawler, and performing incremental training on the model; s6, inputting the verification set data into the text classifier to obtain a final topic tweet classification result; s7, repeatedly executing the incremental training process in the step S5 in real time by taking day as a unit; the problem that a traditional text classification method depends on large-scale marked data is solved.

The invention provides a soft photosensitive rapid convergence method for a low-power-consumption camera. The method comprises the following steps: S1, starting up a main processor; s2, initializing and operating a coprocessor, and parallelly operating the starting of the system in S3 and the convergence processing of the coprocessor in S5; s3, the main processor initializes the system to start other necessary resources; s4, the main processor sends a stop command to the coprocessor, and the step S6 is executed after the stop command is completed; s5, convergence processing of the coprocessor: S5.1, the coprocessor starts to operate; s5.2, configuring a sensor, and using a fixed exposure parameter; s5.3, initializing an ISP (Internet Service Provider); s5.4, obtaining statistical information of a brightness histogram; s5.5, looking up a table to configure Sensor parameters and sending the Sensor parameters to the main processor, and ending the coprocessor flow after receiving a stop command; s6, synchronizing a stop state, and synchronizing a table look-up result to the ISP; and S7, entering a normal process. Defects of the photoresistor can be solved, dependence on the photoresistor is avoided, and the first picture effect can be achieved without increasing hardware cost. The problem that a high-frame-rate small image depends on the Sensor can be solved, and common Sensors can use the method.

The invention discloses a tungsten electrode rotary welding gun for electric arc additive manufacturing, and belongs to the technical field of additive manufacturing. The base is used for being connected with additive manufacturing equipment; the DD motor is fixed at the lower end of the base; the sliding sleeve assembly is connected to the lower end of the DD motor. The sliding sleeve assembly comprises a rotating sleeve driven by a DD motor. The upper end of the wire guide tube is fixed to the base, the lower end of the wire guide tube penetrates through the DD motor and the axis of the sliding sleeve, and the wire guide nozzle is fixed to the lower end of the wire guide tube. The tungsten electrode module is fixed at the lower end of the rotating sleeve; the tungsten electrode module comprises a tungsten needle matched with the wire guide nozzle; when the DD motor rotates, the tungsten needle rotates with the wire guide nozzle as the axis to form an annular heat source. The tungsten needle rotating with the wire guide nozzle as the axis provides an approximately annular heat source for the welding wire, coaxial feeding of the welding wire and the central axis of the welding gun avoids the problem that non-consumable electrode paraxial wire feeding arc additive depends on the wire feeding direction, is suitable for manufacturing multi-layer and complex-shape large components, and has the advantages of being flexible, efficient, high in precision and the like.

The present application relates to a method for extracting an acetabular facies lunata model, in particular to a method for extracting an acetabular facies lunata model based on mixed surface fitting, which solves the technical problem of low precision and robustness of the existing method for extracting an acetabular facies lunata model. The present application realizes rough extraction of the acetabular facies lunata model through spherical surface fitting, and then realizes accurate extraction of the acetabular facies lunata model in combination with ellipsoidal surface fitting, which combines the stability of spherical surface fitting and the anatomical fidelity of ellipsoidal surface fitting. In particular, in the case of complex pathologies such as hip dysplasia, the present application can still maintain high extraction accuracy, and improve the accuracy and reliability of clinical diagnosis and surgical planning. Meanwhile, in the iteration process of ellipsoidal surface fitting, an adaptive fault tolerance threshold is introduced, which effectively solves the problems of sensitivity to noise and strong parameter dependence of traditional methods, and has better robustness in processing complex acetabular morphology.

The utility model relates to the field of robot trolleys, and particularly provides a composite robot trolley. The trolley comprises a carriage, the top of the carriage is provided with the robot and a tray, the trolley is provided with an access cover, an access cover lock catch, pulleys, a console, a controller mounting frame, foot cups, supporting feet and pulleys, and the sliding access cover is opened and closed through one-hand operation of a magnetic lock catch. A closed cable management channel is formed by a cable fixing buckle on the inner side of the cover plate and the hollow cart grip, the equipment troubleshooting difficulty is lowered, an operator can complete state monitoring without changing the position when pushing the grip, meanwhile, through the matching structure of pulleys arranged at the four corners of the bottom and lifting foot cups, the safety of the operator is improved, and the safety of the operator is improved. Manual free pushing is achieved to adapt to a narrow space, stable parking can be achieved through rigid connection of the supporting feet and the anti-skid rubber pads after positioning, and the problem of track dependence of a traditional AGV is solved.

The utility model discloses a massage device of replaceable battery, include: massage device main part, the massage device main part one side embedded type installation changes the electricity box, sets up the battery in the electricity box, and the battery is detachable design, the utility model discloses a detachable battery design, supports double battery alternation and charges, realizes " one uses one fills " mode, improves the endurance time of massage device, thoroughly solves the pain point of equipment downtime during charging, also is convenient for the convenience of the repair problem of the maintenance problem of massage device battery problem, and the user can complete opening and closing only by pressing the box cover, and the auxiliary fixing of screw is combined, and the dismounting efficiency and structural stability are taken into account, avoid the tool dependence problem caused by traditional screw full fixation, the conductive plug and the electric socket adopt card type connection, cooperate the hole limit design, ensure the insertion accuracy, the contact resistance reduces, avoid the interface loosening caused by vibration simultaneously, and the power supply stability is obviously superior to the welding or ordinary insertion scheme.

The invention relates to an interrupt response time detection method and device, equipment, a medium and a program product. The interrupt response time detection method comprises the steps that in response to monitoring that a timer used for interrupt response time detection in an operating system is overtime, a detection program reads a first timestamp from character equipment in a kernel mode of the operating system and triggers the character equipment to send an event notification signal to the detection program, the detection program is a program used for detecting the interrupt response time in a user mode of the operating system, and the first timestamp is a timestamp when the operating system generates the current interrupt; in response to the received event notification signal, the detection program calls a callback interface corresponding to the event notification signal in a user mode in the operating system, and obtains a second timestamp from the callback interface, the second timestamp being a timestamp that the operating system starts to process the current interruption by using the callback interface; based on the first timestamp and the second timestamp, the detection program determines an interrupt response time for the current interrupt.

The invention provides a seismic velocity signal and acceleration signal combined high-resolution processing method. The seismic velocity signal and acceleration signal combined high-resolution processing method comprises the following steps: step 1, carrying out combined processing on a seismic velocity signal and an acceleration signal; 2, signal decomposition based on independent component analysis is carried out; step 3, seismic signal joint high-resolution processing based on Kalman optimal estimation is carried out; and 4, inputting various signals to obtain a combined high-resolution processing result. According to the seismic velocity signal and acceleration signal combined high-resolution processing method, the problem that dependence on selected signals is large due to the fact that only one signal is used for high-resolution processing at present is solved, and the resolution and accuracy of seismic signals are improved through combined processing of velocity signals and acceleration signals.

This application provides a real-time communication management method, apparatus, and device for industrial control networks, relating to the field of industrial control communication technology. A real-time communication management method for industrial control networks, applied to a management node in the industrial control network, includes: parsing preset configuration information and network structure information in the industrial control network to obtain a communication cycle; periodically sending real-time control messages to multiple slave nodes in the industrial control network based on the communication cycle; calculating the amount of non-real-time messages sent by each slave node according to the real-time control messages and the amount of non-real-time messages sent by the management node to each slave node; and sending and receiving non-real-time messages in the industrial control network based on the calculation results. According to embodiments of this application, the response speed of real-time messages and the transmission and processing speed of non-real-time messages are improved, thereby increasing network bandwidth utilization.

This invention discloses an offline navigation method and system based on plain text and local AI. The method parses the user-input text start and end point information using a local offline AI model, combines it with a pre-stored local text-based geographic knowledge base to perform path reasoning, and generates plain text walking steps. The system does not rely on GPS, BeiDou, or other satellite positioning signals or external networks; all calculations and data storage are completed locally on the terminal, and it can generate simulated GPS coordinates and visually assisted navigation. This invention is applicable to scenarios such as signal-free spaces, indoor environments, and visually impaired assistance, and features high privacy, strong stability, and a wide range of applications.

This utility model discloses an automatic inventory device and system for livestock RFID tags. The device includes a power supply unit, a main control chip, sensor devices, and a communication device. The power supply unit is connected to the main control chip and supplies power to the automatic inventory device. The main control chip is connected to the sensor devices, which identify livestock and transmit the identification results to the main control chip. The communication device is connected to the main control chip and includes an RFID read / write module and a 4G wireless terminal module. The RFID read / write module reads RFID ear tags from livestock and transmits the read RFID tag data to the main control chip. The 4G wireless terminal module transmits the data to a server. This utility model significantly reduces server load and network requirements, solving the problems of excessive reliance on servers and high network demands in existing technologies.

The invention provides a generator and transformer protection setting value automatic generation system and method, and relates to the technical field of electrical protection parameter setting. By extracting voltage and current signals before a fault and corresponding setting values, a historical data chain is constructed; the real-time data acquisition module provides the real-time sensing capability of the system for the current operation state, a setting data link can be dynamically generated by acquiring voltage and current signals of a generator or a transformer, the current operation state of the system can be reflected in real time, and the signal processing module performs feature extraction on the voltage and current signals to obtain the current operation state of the system. A historical data link closest to the current state can be quickly matched and recognized, a data link matching module finds the historical data link most matched with the current running state in a historical knowledge base by calculating the error degree, and a setting value generation module achieves automatic generation of a protection setting value by referring to the protection setting value of the optimal historical data link.

The application discloses a bronze harmful rust risk assessment method based on hyperspectral imaging and belongs to the technical field of hyperspectral imaging. The method comprises the following steps: S1, hyperspectral data acquisition; S2, data preprocessing; S3, characteristic wave band extraction; S4, sub-index calculation and corresponding risk judgment; S5, comprehensive risk fusion; S6, environmental factor calculation and corrosion dynamic judgment; S7, risk grade judgment; and S8, prediction and evaluation. The method provides a quantitative, continuous, comparable risk score and discrete operation grade for cultural relic protection personnel, solves subjective dependence and experience difference problems, and realizes a leap from 'qualitative distribution' to 'quantitative risk assessment' and an upgrade from 'picture speaking' to 'digital decision-making'.

This application relates to a method, system, electronic device, and storage medium for early warning of acute mountain sickness. The method includes: collecting physiological data from a step test conducted on a subject at a target plain altitude; generating at least one physiological principal component based on the step test data; performing cluster analysis on the at least one physiological principal component to obtain risk classification clusters; obtaining the actual high-altitude Lake Louise score of the subject in each risk classification cluster and calculating the score interval proportion to determine the risk classification label; and providing an early warning of acute mountain sickness to the user based on the risk classification label. This solves the problems in related technologies, such as high dependence on specialized testing equipment and laboratories, time-consuming testing processes, inability to screen individuals before they enter high altitudes, difficulty in achieving non-invasive and simple screening, and inability to effectively determine the acute mountain sickness tolerance of new subjects, thus hindering personalized early warning and prevention.

The invention discloses an automatic network security risk quantitative evaluation method for a power distribution network information physical system. The method comprises the following steps: carrying out vulnerability scanning on a target network by utilizing a Nessus vulnerability scanner; the Mu lVAL tool generates an attack graph by using the vulnerability scanning result and the input network topology and security policy; searching corresponding vulnerability description in a vulnerability library according to a vulnerability scanning result, and outputting CVSS measurement classification of vulnerabilities through a CVSS measurement classification prediction model; and carrying out system risk quantitative evaluation according to the attack graph, the CVSS measurement and an input physical result, and calculating a risk value of each node in the target network. According to the method, the problem of automatic generation of the attack graph is solved by using an open source tool, and the problem of CVSS information dependence is solved through a CVSS measurement classification prediction model; by designing a plurality of classifiers sharing a multi-layer network, synchronous classification prediction and classification feature sharing of each CVSS metric are realized.

The present invention discloses a bearing fault detection method based on the Gershgorin circle theorem, which relates to the technical field of fault diagnosis. The technical solution identifies abnormal eigenvalues ​​T at an early stage and issues early warnings by establishing an autocorrelation matrix and calculating the eigenvalue T distribution. Secondly, the technical solution overcomes the dependence of traditional methods on high sampling rates, and uses fast Fourier transform and noise filtering technology to simplify data processing and improve system efficiency. Finally, by calculating and evaluating the health index Rhzs, the dependence on experience is reduced, making the detection process more automated and data-driven. The evaluation results are based on historical data and health threshold classification, which improves the accuracy and reliability of fault detection. Overall, the solution optimizes the effect of bearing fault detection through precise data processing and automated analysis.

The application discloses a knowledge-based temporal convolution neural network (KB-TGCN) blast furnace fault diagnosis method. The method can solve the space-time dependence and sample imbalance problem of the blast furnace ironmaking process. First, the calculated variables that can further reflect the process state are calculated by using the variables directly detected by the sensor. Then, the knowledge-based graph structure is constructed according to the spatial position and calculation relationship of the variables. Subsequently, a one-dimensional time information extraction module is embedded in the graph convolutional neural network. Therefore, the TGCN can capture time information while maintaining the original spatial relationship. By using the knowledge-based graph structure, the KB-TGCN can fully mine the spatial information of different blast furnace positions. In addition, the method uses a focal loss function with an adaptive balance factor instead of a traditional cross-entropy loss function to overcome the sample imbalance problem.