51 results about "Diagnostic strategy" patented technology

The invention relates to the technical field of electric valves, in particular to an electric valve abnormal working condition monitoring method and system. Comprising the steps of setting a plurality of monitoring points based on equipment parameters of the electric valve; operation deviation values of all the monitoring points are generated according to a preset state evaluation model, and a first-level diagnosis strategy is set according to all the operation deviation values; generating an abnormal risk value according to the primary diagnosis strategy, and judging whether an early warning instruction is generated or not according to the abnormal risk value; a plurality of monitoring points are established based on the equipment parameters and the historical fault parameters of the electrically operated valve, and the abnormal operation state of each monitoring point is analyzed and pre-warned in time by setting the diagnosis sub-model of each monitoring point, so that the pre-warning efficiency of the abnormal working condition of the electrically operated valve is improved, the fault hidden danger is eliminated in time, and the safe operation of the electrically operated valve is ensured.

The invention discloses a mental stress assessment and intelligent dredging method and system, and belongs to the technical field of digital medical treatment and health informatics. Multi-modal physiological signals are continuously collected through the wearable device, medical diagnosis level pressure state evaluation is carried out based on the personalized physiological baseline, and personalized pressure indexes and levels are generated; secondly, when it is diagnosed that the pressure level exceeds the standard, the system serves as an intelligent decision support system, the environment and schedule information after privacy protection processing is fused, and an optimal grooming action is dynamically selected from a predefined intervention action library and executed; finally, the system serves as a continuous learning system, the intervention efficiency is evaluated in real time according to feedback data of the pressure index after execution, the decision model is updated, and collaborative self-optimization of the diagnosis strategy and the intervention strategy is achieved. According to the method, medical diagnosis, personalized treatment decision and adaptive learning are integrated, and the systematicness, accuracy and intelligent level of mental stress related health problem management are improved.

The invention relates to the technical field of pipeline structure health monitoring, and discloses a trend analysis and early warning method for pipeline expansion monitoring data, which comprises the following steps: acquiring actually measured total strain, temperature and pressure data of a pipeline, and calculating to obtain residual strain; segmenting the residual strain data into response segments according to the change of the working condition; according to the current diagnosis strategy, response morphological characteristics such as hysteretic loop area or creep curvature are extracted from the response fragment; diagnosing the state of the material as one of a healthy elastic state, an initial plastic state, a stable creep state and an accelerated creep state according to the characteristics; adaptively updating a diagnosis strategy based on the current diagnosis result; and when it is judged that state transition to a higher risk level occurs, graded early warning information is generated. By establishing a self-adaptive updating closed loop from a material state diagnosis result to a diagnosis strategy, the problem that early damage is difficult to accurately recognize by adopting a fixed threshold in a traditional method is solved, and the nonlinear degradation process of material performance can be dynamically tracked.

The invention discloses a fault diagnosis method and system for hydraulic element general equipment, and relates to the technical field of hydraulic element fault diagnos.The method comprises the steps that control channels of the general equipment are recognized based on hydraulic elements, transient signals are collected in all the control channels in real time, preset interference constraints are introduced, and fault diagnosis is conducted on the general equipment; generating an interference factor through a high-entropy random disturbance source, performing disturbance analysis on the amplitude, the frequency and the waveform state of the transient signal, and generating an interference identifier; a health state vector is formed in combination with the operation working state, the deviation degree of the health state vector and the health baseline characteristic spectrum is compared, if the deviation degree is smaller than a standard deviation threshold value, a state evaluation strategy is executed in the first diagnosis channel, and a health stability trend is recognized; if the deviation degree is greater than or equal to the standard deviation threshold value, executing a fault diagnosis strategy in the second diagnosis channel, and identifying a fault sign; according to the invention, efficient division of labor of health assessment and fault diagnosis is realized, and the effect of accurately identifying fault signs is achieved.

The invention discloses a multi-mode feature fusion power equipment fault diagnosis model and system, and the model comprises a gating network layer which is used for automatically selecting a mode suitable for participating in fusion according to the features of input data; the DNN and CNN parallel network layer is used for performing feature extraction and classification on different types of input data; the third-level decision-making layer is used for dynamically selecting a diagnosis strategy according to the integrity and the abnormal condition of the input data; the system comprises an acquisition module used for acquiring multi-modal monitoring data of power equipment; the FPGA control module is used for splitting, packaging and caching the collected data; the GPU processing module is used for performing verification, feature extraction and diagnosis on the received data; and the display control module is used for man-machine interaction and data and diagnosis result display. According to the model and the system, efficient, accurate and flexible diagnosis of power equipment faults is realized through multivariate feature fusion and multiple decision-making mechanisms, and the model and the system have important application value.

The invention discloses a regulation and control fault positioning method fusing a diagnosis feature library and power flow verification, and relates to the field of intelligent power grids, and the method comprises the following steps: extracting a power grid regulation and control abnormity related protocol message, carrying out the multi-dimensional protocol analysis, and generating a preliminary fault criterion; performing mode matching on the initial fault criterion and an intelligent diagnosis feature library to generate a mode matching result; constructing a regional power grid power flow distribution model, comparing an expected state with actual power flow distribution, and generating a physical verification result; fusing the preliminary fault criterion, the mode matching result and the physical verification result to obtain a final fault criterion, and outputting a fault positioning result; and performing reliability evaluation on a fault positioning result based on a confidence model, and optimizing a diagnosis strategy through reinforcement learning. According to the method, multi-dimensional diagnosis information and physical closed-loop verification are fused, and an adaptive learning mechanism is introduced, so that more systematic, accurate and reliable positioning of the regulation and control faults of the intelligent power grid is realized.

The application discloses a strategy for extracting fault features of electronic circuit systems, named as SSEST strategy, which is used for perceiving global information and paying attention to notable local information, and mining important local information means realizing expression of intermittent fault features of electronic circuit systems, specifically, first, S transformation is performed on a circuit output time sequence signal to acquire time-frequency domain features, then a squeeze and excitation network attention module is used to distribute channel weights, subsequently, input into a Swin Transformer framework, and pay attention to local information related to faults from global signals, and deep mining is performed on fault features, and two electronic circuits are taken as experimental circuits, the proposed diagnostic strategy realizes rapid and high-precision diagnosis, and shows that the proposed multiple attention mechanism is efficient for feature mining of intermittent faults of electronic circuit systems.

This application provides an AI-based diagnostic fee adjustment method and related apparatus. The method includes: acquiring basic diagnostic data of a target vehicle; performing preliminary AI diagnosis on the basic diagnostic data to obtain a first diagnostic result; the first diagnostic result includes a preliminary diagnostic report; determining a target diagnostic strategy based on the preliminary diagnostic report; the target diagnostic strategy includes any one of the following: AI-only diagnosis, AI-assisted and manual determination, or manual diagnosis only; determining a target diagnostic report based on the target diagnostic strategy; determining the AI ​​contribution based on the target diagnostic report; and determining the target diagnostic fee based on a preset basic fee standard and the AI ​​contribution. By constructing a differentiated human-machine collaboration process based on diagnostic confidence, quantifying the AI ​​contribution, and differentiating fees based on the AI ​​contribution, the accuracy of AI fault diagnosis applications and user engagement are improved.

The invention discloses an intelligent fault diagnosis method and device for a complex aerospace system based on correlation modeling, and relates to the technical field of aerospace system fault diagnosis, and the method comprises the steps: constructing a testability model of the aerospace system; the distinguishing capability of different faults is simplified, and test points are optimized; automatically generating a fault diagnosis strategy based on the optimized testability model; performing simulation evaluation on the fault diagnosis strategy, and counting a fault detection rate and a fault isolation rate; and performing iterative optimization by judging whether the fault detection rate and the fault isolation rate meet preset requirements or not. According to the method, the technical problems of low diagnosis efficiency and insufficient fault detection and isolation accuracy caused by dependence on artificial experience and incomplete static fault tree coverage of complex spaceflight system fault diagnosis in the prior art are solved, intelligentization and precision of complex spaceflight system fault diagnosis are realized, and the fault diagnosis efficiency is improved. And the fault detection rate and the isolation rate are improved.

The application provides application of serum sIL-2R as an ADAM10 inhibitor companion diagnostic marker and a corresponding kit, including two core application scenarios: before treatment, screening of a patient population with high activity of an ADAM10-sIL-2R pathway and most likely to benefit from ADAM10 inhibitor treatment by detecting the serum sIL-2R level of the patient; during and after treatment, evaluating the inhibition effect of the ADAM10 inhibitor on the target pathway and the treatment efficacy by dynamically monitoring the change of the serum sIL-2R level; the liquid companion diagnostic strategy will provide an indispensable supporting tool for the precise application of the ADAM10 inhibitor in pancreatic cancer, and significantly improve the precision and effectiveness of clinical medication.

A digital twin-driven electric servo mechanism health index system construction method belongs to the technical field of electric servo mechanism health management, and is characterized in that a unified functional unit digital twin fault model of an electric servo mechanism covering functional units such as a controller, a power amplifier, a motor, a transmission mechanism, a sensor and the like is constructed; various typical fault modes such as poor contact, power tube open circuit and magnetic steel demagnetization can be simulated, influence rules and characteristic expressions of various faults on dynamic performance of the system are revealed through a step response and sine response simulation system, and accurate twinborn simulation data can be provided for diagnosis strategy design without relying on physical entity fault tests. The complex process of fault influence mechanism analysis and diagnosis basic data acquisition is greatly simplified; the full-period health management requirement of the electric servo mechanism can be met, the universality is high, and the technical application efficiency can be greatly improved.

The invention relates to the technical field of automatic equipment fault diagnosis, in particular to an automatic equipment fault diagnosis method and system based on deep learning. According to the method, the health degree state of the execution mechanism is comprehensively evaluated by taking the life cycle stage and the performance state as the basis and combining with the evaluation model based on deep learning. In consideration of different health degree states of the execution mechanisms, in order to identify potential fault risks as much as possible to ensure smooth proceeding of a production plan, the fault diagnosis strategies for the execution mechanisms with different health degree states are also different, and the execution mechanisms with poor health degree states need to adopt diagnosis strategies with more strict standards; and more monitoring parameters need to be acquired, and the acquisition strategy also needs to be finer, so that potential fault risks can be checked as much as possible and timely maintenance can be carried out. The monitoring data acquired according to the health degree state self-adaptive acquisition strategy is input into the diagnosis model to identify potential faults as much as possible, so that the automation equipment is better managed.

This application relates to the field of computer technology and provides a method and related equipment for generating intelligent agents for diagnosis. The method generates a virtual object containing temporal pathological feature data based on a preset disease evolution template; controls an initial diagnostic agent to interact with the virtual object in multiple rounds to obtain a temporal diagnostic strategy output by the initial diagnostic agent; generates feedback results for the initial diagnostic agent based on the diagnostic strategy and the target diagnostic scheme corresponding to the virtual object; and optimizes the parameters of the initial diagnostic agent based on the feedback results to generate a target diagnostic agent. By constructing a disease evolution template, this application can generate virtual patient data with continuous temporal features at low cost and on a large scale, enabling the diagnostic agent to be exposed to the complete disease progression logic from the latent stage to the critical stage during the training phase, thereby improving the model's generalization ability.

The invention discloses a hand skin health condition detection system and method, and the system comprises a three-dimensional structure and multispectral image collection module which is used for synchronously obtaining the three-dimensional structure contour and multispectral image data of the hand of a user; the multi-modal data fusion and feature extraction module is used for performing spatial registration on the data and extracting features; the skin health state vector generation module is used for fusing the multi-source features into a standardized multi-dimensional health state vector; the method comprises the following steps: firstly, carrying out rapid low-resolution macroscopic pre-scanning, and identifying and positioning a potential risk area; the system dynamically plans a detailed investigation strategy, and only drives the acquisition module to carry out high-resolution multi-mode detailed investigation on the risk area; and carrying out comprehensive diagnosis by fusing macroscopic background and local accurate data. Through deep fusion of multi-modal data, accurate quantitative evaluation of the skin health condition is realized, and meanwhile, an innovative diagnosis strategy remarkably improves the detection efficiency and reduces the system overhead on the premise of ensuring the precision.

The application provides a leakage detection method of a fuel evaporation system and an electronic device, and is applied to a fuel evaporation system of a vehicle. The method comprises the following steps: when the parking time of the vehicle exceeds a first preset time length, and the enabled working condition of the fuel evaporation system is normal, the vacuum degree of a first space in which a first electric control valve is in a closed state is obtained; a first diagnosis strategy or a second diagnosis strategy is selected according to the vacuum degree, and leakage detection is performed on the first space and a second space to obtain a detection result, wherein the first diagnosis strategy is a diagnosis strategy corresponding to that the first space has no leakage when the absolute value of the vacuum degree is greater than a first preset threshold. In this way, an air pump does not need to be additionally arranged, so that the hardware cost of the system is reduced. In addition, different diagnosis strategies are beneficial to improving the diagnosis completion rate. Since the first space of the fuel evaporation system rarely leaks, the frequency of using the first diagnosis strategy is usually high, and the detection process of the first diagnosis strategy is less, so that the diagnosis efficiency is improved.

This invention discloses a multimodal state perception and fault diagnosis method based on equipment identification guidance. The method includes: acquiring multimodal information of the power equipment under test to obtain multimodal perception data; performing time synchronization, spatial alignment, and data normalization processing on the multimodal perception data; automatically identifying the power equipment based on visible light imaging information to obtain equipment type information and / or key structural location information; selecting a multimodal fusion and diagnosis strategy matching the corresponding power equipment according to the equipment type information, and jointly analyzing the multimodal perception data; and outputting the state assessment result and / or fault diagnosis result of the power equipment based on the joint analysis result. By using equipment identification results as a priori constraints to guide the multimodal information fusion and diagnosis process, this invention achieves differentiated state perception and fault diagnosis for different types of power equipment, improving detection accuracy and engineering applicability.

The application discloses a fault diagnosis strategy optimization method considering test uncertainty, and the specific steps are as follows: a simulation model of a measured object is established based on a Modelica language; the relationship between test points and faults is counted based on the simulation model, and a test-fault uncertainty matrix is established; on the basis of the test-fault uncertainty matrix, an error diagnosis cost caused by test uncertainty is introduced, an information entropy algorithm is improved, and a diagnosis strategy is obtained based on the improved information entropy algorithm. The application firstly proposes to establish a test-fault uncertainty matrix based on a simulation model; then, the error diagnosis cost is introduced into the information entropy algorithm, and the accuracy of the diagnosis strategy is improved. Finally, a backtracking step is added to the information entropy algorithm, and the algorithm is prevented from falling into local optimization. The research on the fault diagnosis strategy optimization based on the test uncertainty has important significance for improving the fault diagnosis capability of the measured object. The diagnosis strategy is fed back to the forward design of the system, and a reference is provided for the design of the measured object.

The invention provides a vehicle remote diagnosis method, platform and system based on a gray task, and belongs to the technical field of Internet of Vehicles, and the method comprises the steps: storing a vehicle relation table and a corresponding diagnosis strategy to the remote diagnosis platform; the remote diagnosis platform obtains vehicle information and sends a corresponding diagnosis strategy to the corresponding vehicle end to execute a diagnosis task; and the remote diagnosis platform generates a diagnosis report according to the diagnosis log sent by the vehicle end, and sends the diagnosis report to the vehicle end when an abnormality is found. Through an instant or periodic task triggering mechanism, regular and real-time remote diagnosis can be carried out on the vehicle, so that potential problems can be found in time before a fault occurs or deteriorates, post-remedy is changed into pre-prevention, and the time cost and the maintenance cost of a vehicle owner are effectively saved.

The invention discloses a testability diagnosis strategy generation method and system based on a correlation model, and relates to the technical field of testability engineering.The method comprises the steps that a multi-signal correlation model containing functions, faults and test signals is constructed, a correlation matrix is generated, test points are optimized through an information gain greedy algorithm, a test point optimization set is obtained, and the test point optimization set is obtained; and automatically generating a diagnosis strategy, then performing testability parameter simulation prediction, comparing a predicted value with a distribution index to generate a verification result, and when the verification result is not passed, triggering a closed-loop iterative optimization mechanism to perform optimization adjustment until the predicted value meets the distribution index, and generating a diagnosis optimization strategy. According to the method, the technical problems that index distribution and engineering practice are disjointed and diagnosis scheme design and verification links are separated in complex equipment testability design are solved, and the technical effects that complex equipment testability index distribution fits the practice, a diagnosis scheme and verification form collaboration and can be dynamically optimized, and later reworking is avoided are achieved.

The invention relates to a batch fault abnormity positioning method based on man-machine cooperation and comparative analysis, and the method comprises the following steps: 1, collecting multi-source original fault data, evaluating whether the data meets the feasible conditions of batch fault diagnosis or not, obtaining a final evaluation result, and generating a standardized table mapping suggestion; 2, through structure and integrity verification, semantic check and automatic repair and executable verification, obtaining tabular data which is subjected to structure and semantic verification and potential problem correction and is confirmed to be generated and can be directly executed; and step 3, obtaining historical experience most similar to the current scene through searching a historical scene library and matching, obtaining a recommended diagnosis strategy, performing lightweight trial operation on each candidate strategy, comparing to obtain a most suitable strategy, and further performing a final search process on complete data, discovering a key attribute mode and a complete anomaly positioning report. According to the invention, an end-to-end intelligent diagnosis process from scene identification, data verification to abnormal positioning can be realized.

The invention provides a test equipment abnormal state self-diagnosis report generation and secure transmission system, which adopts a Jenkins-based equipment abnormity automatic diagnosis system, and realizes automatic execution of equipment abnormity diagnosis monitoring and automatic generation and secure transmission of a fault report by utilizing timing triggering of Jenkins. The test efficiency is greatly improved, real-time monitoring of the test equipment is realized, and thus full-automatic periodic execution of diagnosis and monitoring of the test equipment is realized; according to the invention, an abnormal state diagnosis strategy based on principal component analysis and a random forest BP neural network (PCA-RF-BP) is used for improving the diagnosis efficiency of an equipment monitoring system; the method has the advantages of automatically diagnosing and monitoring the abnormity, automatically generating and pushing the fault report, improving the test efficiency and realizing the safety and reliability of information transmission.

The invention relates to the technical field of industrial automation, in particular to a production line anomaly diagnosis method and device, computer equipment and a medium, and the method comprises the steps: obtaining multi-source heterogeneous sensing data of a production line, and carrying out the standardization processing of the multi-source heterogeneous sensing data, and obtaining multi-source isomorphic sensing data; performing feature extraction processing on the multi-source isomorphic sensing data to obtain production line features of a production line; a target diagnosis graph corresponding to the production line features is determined, the target diagnosis graph comprises a directed acyclic graph composed of a plurality of diagnosis nodes and directed edges, each diagnosis node corresponds to a diagnosis strategy, and the directed edges between the diagnosis nodes represent the execution sequence of the diagnosis strategies; the production line features are input to the initial diagnosis node of the target diagnosis graph, diagnosis strategies corresponding to the diagnosis nodes are executed in sequence, and a production line abnormity diagnosis result is output, self-adaptive generation of the diagnosis path can be achieved, and the adaptability to diversified production line scenes is enhanced.

The invention relates to the technical field of antibodies, and discloses an antibody for specifically recognizing porcine pseudorabies virus gE (52-260aa) or an antigen binding fragment thereof and application. The antibody or the antigen binding fragment thereof comprises a heavy chain and a light chain; the antibody can realize specific recognition on gE protein and is used for detecting a target antigen in a sample. The antibody can be used as a core reagent in an immunodetection system, is suitable for laboratory detection and on-site rapid screening of porcine pseudorabies, is beneficial to improving the monitoring efficiency of basic-level and large-scale breeding links, and provides a technical guarantee for early discovery and accurate prevention and control of epidemic situations. Meanwhile, the antibody can also be used as a research tool to be applied to research on gE protein related biological functions and effects of the gE protein in a virus pathopoiesis process, and important support is provided for clarification of virus variation characteristics and optimization of subsequent prevention and control and diagnosis strategies.

The application discloses a kind of multimodal fusion medical information processing methods, comprising the following steps: collecting the multimodal medical image dataset of chest X-ray of patient and healthy individual;Image dataset and diagnosis report are preprocessed, and disease label is generated;Using multimodal convolutional neural network, the image dataset after preprocessing is encoded as visual feature, and fusion feature sequence is generated;With fusion feature sequence as initial input, graph convolutional neural network is constructed;Establish reinforcement learning model, with the comprehensive performance of diagnostic strategy as index, dynamically optimize the graph convolution process of convolutional neural network by defining reward function, output final neural network feature parameter;With neural network feature parameter as input, pooling, simplification is carried out, corresponding probability range is output by integration and transformation.The application is based on deep reinforcement learning technology, improves preliminary diagnosis efficiency, and improves the accuracy, coherence and integrity of diagnostic result words.

The invention provides a distributed micro-vibration monitoring system and method for a photovoltaic support, and belongs to the field of photovoltaic system structure monitoring, and the system comprises a sensor node, a wireless gateway, a cloud platform server and an operation and maintenance terminal. The sensor nodes are distributed at the joints of the photovoltaic supports and are composed of MEMS vibration sensors, low-power-consumption microcontrollers and wireless communication and power supply modules, and all the modules are connected according to functions. The wireless gateway covers a preset range, manages a plurality of sensor nodes, comprises a receiving module, a transmission module and a local storage module, and is used for collecting data uploaded by the nodes; the cloud platform server comprises a data storage module, a preprocessing module, a loosening diagnosis module and an early warning module, and vibration data storage, feature extraction, diagnosis and early warning are achieved. The operation and maintenance terminal comprises a PC and a mobile terminal, and can carry out remote parameter allocation and generate an operation and maintenance report. Through low-cost and large-scale layout of a distributed MEMS network, in combination with a low-power-consumption design and a diagnosis strategy, support vibration is accurately monitored, and early warning and quantitative diagnosis of bolt looseness are achieved.

This invention relates to the field of computer-aided skills training program analysis technology, and more particularly to a method and system for evaluating training programs for maintenance technicians. It includes the following steps: S1: Based on the leap characteristics of trainees' cognitive structures during the formation of maintenance skills, the training program is divided into N consecutive stages with capability qualitative change points as boundaries; S2: At the capability qualitative change boundaries of adjacent consecutive stages, based on the degree of influence of the baseline cognitive unit corresponding to the boundary on subsequent consecutive stages, data variation acting on a single observable parameter is constructed; S3: Based on the matching results of the constraint relationship between the single observable parameter involved in the data variation and the associated unvaried observable parameter, logical conflicts acting on trainees are generated. This invention quantifies cognitive stability through staged training and data variation, and transforms diagnostic strategies into behavioral biases by combining conflict experiments, significantly improving the accuracy of higher-order thinking assessment and attribution reliability in maintenance skills training.

The application discloses a kind of transformer area line loss abnormal diagnosis and loss reduction strategy intelligent decision method and system.The method of the present application comprises: collecting the meter data of each node in transformer area, and calculating the line loss rate of each time scale of transformer area;The 96-point time-sharing line loss data of transformer area in the past year is obtained by the time series anomaly detection algorithm based on adaptive k-nearest neighbor, and the line loss abnormal time series of the transformer area is obtained;Establish transformer area line loss abnormal reason diagnosis algorithm library, and construct targeted diagnostic algorithm for each type of abnormality that may occur;Loss reduction strategy knowledge base is constructed, loss reduction strategy is associated, and line loss abnormal event management strategy is obtained by using knowledge base matching;At the same time, each line loss abnormal event management strategy is divided.The present application can accurately detect line loss abnormal time series and calculate the amount of loss reduction power under the condition that the operation mode of transformer area changes greatly, and realizes the automation of transformer area line loss abnormal monitoring, reason diagnosis and strategy generation.

The invention provides a multi-phase motor IGBT open-circuit fault identification method and system and a storage medium, and relates to the technical field of motor control, and the method comprises the steps: obtaining the actual rotating speed of a multi-phase motor; when the actual rotating speed is larger than a set rotating speed threshold value, a first diagnosis strategy is started, and the first diagnosis strategy is used for conducting harmonic analysis based on all phases of currents of the multi-phase motor to locate a fault phase where an IGBT open-circuit fault occurs; when the actual rotating speed is smaller than a set rotating speed threshold value, a second diagnosis strategy is started, and the second diagnosis strategy is used for positioning a fault phase with an IGBT open-circuit fault based on the duty ratio and the current of each phase of the multi-phase motor; and generating an IGBT open-circuit fault identification result containing the fault phase. According to the method, different diagnosis strategies are dynamically selected according to the rotating speed of the motor, the fault positioning accuracy can be effectively improved, the misdiagnosis situation is avoided, and accurate identification of the IGBT open-circuit fault under all working conditions is achieved.

The invention discloses a multi-modal fusion medical information processing method. The method comprises the following steps: collecting a multi-modal medical image data set of chest X-rays of a patient and a healthy individual; preprocessing the image data set and the diagnosis report to generate a disease label; a multi-modal convolutional neural network is adopted to encode the preprocessed image data set as view table features, and a fusion feature sequence is generated; constructing a graph convolutional neural network by taking the fused feature sequence as initial input; establishing a reinforcement learning model, taking the comprehensive performance of the diagnosis strategy as an index, dynamically optimizing the graph convolution process of the convolutional neural network by defining a reward function, and outputting final neural network feature parameters; the neural network feature parameters are used as input, pooling and simplification are carried out, and a corresponding probability range is output through integration and transformation. Based on the deep reinforcement learning technology, the preliminary diagnosis efficiency is improved, and the accuracy, coherence and integrity of word use of the diagnosis result are improved.