69 results about "Statistical process control" patented technology

The invention provides a reliability detection method and system based on chip module packaging and a medium. The method comprises the following steps: querying a basic threshold value by obtaining chip module packaging process detection parameters and design data; combining the real-time influence data with a preset rule to obtain compensation data, and calculating a correction threshold value; obtaining a process capability state and threshold control data through historical data and statistical process control, further calculating a dynamic threshold, and comparing detection parameters to obtain a process state; after all procedures are qualified, evaluating edge states to obtain risk factors and summing the risk factors to obtain a comprehensive risk value, obtaining a final threshold by combining reliability test data and a reference threshold, and determining the reliability state of the module after comparison; therefore, through calculation processing and threshold comparison of the basic threshold, the corrected threshold, the dynamic threshold and the final threshold, the technology for detecting the packaging reliability of the chip module is realized.

The invention discloses a numerical control machining quality monitoring method and system based on statistical process control, and relates to the technical field of intelligent manufacturing and quality control. Comprising the following steps: 1, generating a multi-modal data collaborative perception and dynamic control chart; 2, federal learning optimization of knife compensation and quality association; 3, mixing enhanced quality control analysis; 4, causal reasoning driven abnormity traceability; and 5, performing digital twin closed-loop verification. Through multi-mode collaborative perception, federated learning optimization and digital twin closed-loop verification, the abnormal response speed of numerical control machining quality monitoring is increased, the compensation precision is improved, the new tool adaptive sample size is reduced, finally, the product percent of pass reaches 99.5% or above, and the comprehensive quality cost is reduced by 42%.

The invention provides a production control method and system based on SPC alarm linkage process locking, and relates to the technical field of quality management and control, and the method comprises the steps: deploying an SPC sensing node used for collecting technological parameters at a cutting edge position where a machining tool of production equipment is in contact with a workpiece, or at a contact point where a probe of online measurement equipment is in contact with the surface of the workpiece; the method comprises the following steps: acquiring process parameter data of a key production process in real time through an SPC sensing node deployed at a cutting edge position or a contact point; inputting the process parameter data as real-time data or small sample expansion data into an SPC judgment engine; and performing statistical process control analysis on the real-time data or the small sample expansion data through an SPC judgment engine so as to model data clusters which are densely distributed and have stable parameters in the data into a virtual reference curved surface representing a normal state domain. According to the invention, the accuracy of abnormity determination can be improved, and timely management and control of quality abnormity and minimization of production loss can be realized.

The invention discloses an enterprise digital operation real-time data analysis method based on big data, and particularly relates to the field of data analys.The enterprise digital operation real-time data analysis method comprises the steps that enterprise operation data are collected in real time through a multi-protocol adapter, abnormity is dynamically cleaned and detected, and data standardization is achieved in combination with statistical process control and BERT sentiment analysis; in the hierarchical stream processing architecture, edge nodes pre-aggregate data and filter anomalies, a center engine dynamically divides time windows to generate minute-level / second-level indexes, and a self-adaptive rolling window and dual-threshold session cutting support inventory monitoring and user behavior analysis; the incremental machine learning model updates prediction in real time based on a sliding window, screens a dynamic optimization decision tree through Hoeffering splitting conditions and feature importance, and triggers fine tuning or full training when data distribution drift is detected; and the cross-source data association engine fuses real-time and historical data, calculates a KPI anomaly score, and triggers hierarchical early warning.

The invention discloses a welding wire production control system and method based on visual feedback, and belongs to the technical field of precision machining. The system comprises a multi-mode intelligent sensing subsystem, a self-adaptive model prediction controller and a safety and exception handling subsystem, and a two-stage control architecture of a quality closed loop and an execution closed loop is constructed. According to the method, the microcosmic roughness, macroscopic geometry, defects and other multi-dimensional quality characteristics of the surface of the welding wire are obtained in real time through multi-modal sensing and fused into a comprehensive quality index; carrying out rolling optimization by adopting a model prediction control algorithm for updating parameters on line, and generating a control instruction; meanwhile, trend early warning based on statistical process control and sudden fault diagnosis based on model residual analysis are operated in parallel, and active safety protection is achieved. According to the method, the problems of unstable quality, control lag and passive safety caused by supplied material fluctuation and tool time variation in the welding wire flaking machining process are solved, and self-adaptive accurate control and intelligent safety protection in the machining process are achieved.

The invention provides a quality control system based on a multivariate exponential weighted moving average control chart, and belongs to the technical field of quality management and statistical process control. The system comprises a quality data acquisition module, a data preprocessing module, a statistical process control module, an anomaly judgment module and a quality evaluation and improvement module, and is used for monitoring and controlling the production process of a product with a plurality of quality characteristics in real time. The method comprises the following steps of: acquiring quality data of a plurality of quality characteristics in a production process by a system, preprocessing the quality data, and constructing a multivariate quality characteristic data set; estimating a mean vector and a covariance matrix in a controlled state based on the historical quality data, performing weighted updating on the multivariate quality data by adopting a multivariate exponential weighted moving average method, calculating a corresponding MEWMA statistic and generating an MEWMA control chart; and comparing the MEWMA statistical magnitude with a preset control limit to judge whether the production process is in an out-of-control state or not, and outputting early warning information and a corresponding quality evaluation result and improvement suggestion when abnormality is detected. The method can comprehensively analyze related information among multivariate quality characteristics, improves the sensitivity and accuracy of anomaly detection in the production process, and effectively improves the quality control level of the production process in the manufacturing industry.

The invention discloses a fault early warning method and system based on a robot, and the system comprises a signal processing module, a signal processed by the signal processing module comes from a sensor in a robot system and an operation data of a control system, and the signal processing module carries out the processing of the operation data of the sensor in the robot system and the operation data of the control system. The state detection module is used for testing and reporting the physical state or performance of a certain component or subsystem; the early warning evaluation module diagnoses and reports the health state of each assigned component or subsystem; the decision module re-plans the next task of the system according to the future health state of the robot system to generate a maintenance scheme of the system; compared with the prior art, the method has the advantages that a large amount of historical data are learned and analyzed, the potential faults of the robot are accurately predicted, and the different causes are found and solved in time in the operation process through statistical process control, so that fault prediction of the system is achieved.

The invention discloses an automatic fault detection and diagnosis system and method based on an artificial intelligence optimization operation system and a function library, and belongs to the technical field of intelligent building and industrial equipment operation and maintenance. The system comprises a data acquisition and preprocessing layer, a fault detection layer, a fault diagnosis layer, an intelligent recommendation layer and a visual display layer. The method comprises the following steps: collecting equipment data in real time and preprocessing; performing anomaly detection and performance trend analysis in combination with statistical process control and a machine learning algorithm; performing root cause diagnosis by using a Bayesian reasoning model; classifying faults according to an energy performance type, a controllability type and a thermal comfort correlation type, and evaluating priorities through a multi-dimensional weighted scoring model; and intelligently generating a solution and updating the knowledge base. According to the method, whole-process automation from early warning, diagnosis, classification to recommendation is achieved, the equipment reliability, the energy efficiency and the operation and maintenance intelligent level are improved, and the method is suitable for application scenes such as artificial intelligence middleware and computer visual and auditory software.

The invention provides a non-uniform ion implantation process monitoring method based on template matching image recognition, and the method comprises the steps: 1, carrying out the to-be-monitored non-uniform ion implantation of a wafer, carrying out the modeling according to the measurement data of the wafer, and obtaining a template used for monitoring the dose / resistance distribution of the non-uniform ion implantation; step 2, carrying out resistance measurement on the batch of wafers after the non-uniform ion implantation, and drawing a resistance distribution diagram according to the coordinates and the resistance values of the measurement point positions; and 3, carrying out image identification comparison operation on the template obtained in the step 1 and the resistance distribution diagram obtained in the step 2 to obtain a similarity score, and carrying out statistical process control and control on the similarity score. Abnormalities such as deviation and rotation of dose distribution patterns of the non-uniform ion implantation process are effectively captured, and monitoring of the non-uniform ion implantation process is achieved.

The invention relates to a vibration suppression method of a high-precision hardware part continuous punching machine in the field of intelligent manufacturing equipment industry in high-end equipment manufacturing industry, which comprises the following steps of: identifying a main vibration frequency and a secondary vibration frequency by adopting a peak detection algorithm according to vibration spectrum distribution, and determining a key frequency point set of multi-band vibration; through a pre-established vibration mode and damping parameter mapping model, for the dominant frequency vibration feature vector, a table look-up method is adopted to obtain a corresponding initial damping parameter combination, and a preliminary damping adjustment scheme is determined; the optimized damping parameter combination is adopted, the rigidity and the damping coefficient of a damping device are dynamically adjusted through a servo control system, multi-frequency-band vibration is restrained, and stable stamping process data are obtained; geometric dimension data of the suspender and the check claw are collected in real time through a laser measurement system, the dimensional deviation trend is analyzed through a statistical process control algorithm, and the stability of the stamping process is judged.

The invention discloses a fault early warning method for a chemical equipment explosion-proof inspection robot based on vibration analysis, and belongs to the field of robot fault early warning. Comprising the steps of multi-source vibration data acquisition and anti-explosion signal processing, vibration-environment multi-source data fusion diagnosis, statistical process control (SPC) vibration health assessment, anti-explosion environment self-adaptive early warning mechanism and graded early warning and maintenance decision. According to the chemical equipment anti-explosion inspection robot fault early warning method based on vibration analysis, the Db4 wavelet denoising technology is adopted, the fault detection rate is greatly improved compared with a traditional method, mechanical faults and electrical risks are accurately distinguished, the ATEX highest anti-explosion standard is achieved, the sensor self-healing capacity is achieved, the system reliability is improved, and the fault early warning effect is good. Through the daily increasing trend of the entropy value of the vibration envelope spectrum, faults are pre-judged 72 hours ahead of time, most of accidental shutdown losses are reduced, meanwhile, three-level response strategy synergy is achieved, similar vibration modes are combined through the K-means clustering algorithm in each quarter, and the annual diagnosis precision is improved.

The invention discloses an intelligent wearable watch case precision detection method based on visual inspection, and relates to the technical field of machine vision, and the method comprises the steps: collecting a watch surface point cloud and a watch case image, carrying out the point cloud denoising and image distortion correction, and obtaining registration multi-modal data; performing multi-layer pyramid convolution on the registered multi-modal data, and outputting geometric-texture features of the watchcase; curvature gradient calculation and spatial relation modeling are carried out on the geometric-texture features of the watchcase, and curvature distribution parameters and a spatial relation adjacency matrix are obtained; the curvature distribution parameters are used as nodes of the topological structure, the spatial relation adjacent matrix is used as edges of the topological structure, and a case geometrical relation graph is constructed. According to the method, the watch surface point cloud and the watchcase image are fused and subjected to multi-layer pyramid convolution, the comprehensiveness and positioning precision of watchcase defect recognition are improved, and meanwhile, the engineering practicability of the detection process is effectively improved by utilizing a graph attention defect prediction model and an SPC statistical process control algorithm.

The invention provides a self-starting control chart sequence rebound parameter updating method based on warning limit triggering, and relates to the field of statistical process control. Different from a traditional product quality characteristic distribution model based on normal distribution, the method mainly takes product quality monitoring in a high-quality manufacturing process as a target, modeling is performed on an event occurrence interval variable Xt in the process by using gamma distribution, a pivot vector # imgabs0 # is constructed based on the modeling, and mapping from the Xt to self-starting event occurrence interval data St is completed. Further constructing a monitoring statistic Qt of the self-starting control chart by using an exponential weighted moving average method, dynamically switching a parameter updating process based on a time sequence and a sequence bounce according to position information of the Qt in a safety domain, a warning domain and an out-of-control domain of the control chart, delaying the progress that an out-of-control sample flows into parameter estimation, and calculating the out-of-control sample according to the parameter updating process. Data pollution caused by out-of-control samples to parameter estimation is reduced, and the monitoring efficiency of the self-starting control chart is improved.

The invention provides an alumina carrier with a honeycomb-like structure and an additive manufacturing forming method thereof, and belongs to the field of advanced ceramic and structured reaction engineering. The alumina carrier is prepared by adopting a photoetching additive manufacturing technology, and high-precision consistency of intra-batch or inter-batch size and quality is realized by establishing a geometric manufacturable threshold system of a honeycomb-like structure, combining anisotropic shrinkage pre-compensation, local thickening correction and photoetching process in-situ metering closed-loop parameter adjustment and matching with a statistical process control acceptance release system. The unit characteristic scale L, the inner wall normal vector minimum thickness t and the fillet radius r of the honeycomb-like alumina carrier meet a specific constraint relationship, the total porosity is greater than 30%, the internal porosity is 20-30%, the dimensional deviation delta L / L is less than or equal to 10%, the mass deviation delta m / m is less than or equal to 10%, the process capability index Cpk is greater than or equal to 1.33, and the honeycomb-like alumina carrier is suitable for scenes such as medicine synthesis and fine chemical engineering which have strict requirements on carrier consistency.

The invention discloses a robust failure judgment method, system and equipment for a low-cost indoor environment sensor group and a medium, and the core of the method is to construct a serial five-step processing assembly line, utilize redundant information of the sensor group and combine robust statistics and statistical process control theories to judge the robustness failure of the indoor environment sensor group. According to the method, accurate decoupling of environmental events and sensor faults is achieved, and the problems that an existing method cannot distinguish environmental fluctuation and sensor degradation, and judgment lacks robustness and quantitative basis are solved. According to the method, the instantaneous deviation is judged, and misinformation caused by short-term fluctuation is avoided by setting the cumulant and the minimum duration time. The sensitivity to progressive degradation is improved.

The invention discloses a method and system for automatically processing AOI detection data in the semiconductor industry, and belongs to the technical field of intelligent manufacturing of semiconductor manufacturing enterprises, and the method comprises the steps: based on an equipment feature knowledge base, according to AOI equipment feature information, matching an equipment feature knowledge base, loading a corresponding analysis engine, and extracting measurement data from AOI equipment; analyzing the measurement data and cleaning abnormal values to generate data in a standardized format: transmitting the generated data in the standardized format to a statistical process control system through branches by a dynamic routing mechanism according to real-time requirements of detection parameters; the integrated statistical process control algorithm calculates the data in the standardized format, a multi-dimensional detection report is generated, and when the process capability index value is lower than a set threshold value, abnormal early warning is triggered. According to the method, the misjudgment rate of the system is reduced, the data processing efficiency and accuracy are improved, height overrun, yield fluctuation and the like of the solder balls are positioned in time, and the method has reliability and engineering applicability.

The invention discloses a real-time assembly error recognition and early warning method based on multi-dimensional dynamic threshold adjustment. The method comprises the following steps: extracting a multi-modal assembly state descriptor and a field assembly information sequence; field assembly information fragments are intercepted for similarity calculation; constructing a time sequence prediction module, and taking the field assembly information sequence as input to predict the assembly state of the next frame; constructing a weighted Euclidean distance matrix of the field fragment and the template fragment, and calculating a minimum cumulative distance to quantify the difference between the two sequences; a dynamic threshold model is established based on statistical process control, and whether errors exist in the assembly process or not is judged according to the model; based on an incremental dynamic time planning algorithm and a real-time data acquisition module, a real-time identification system is constructed to realize early warning and standard step pushing, and a real-time adaptive quality control closed loop is formed. According to the method, a real-time and self-adaptive quality control closed-loop system is formed through integration with multi-mode sensing and AR guide technologies, and an effective solution is provided for a complex industrial assembly scene.

Computer Device for Statistical Process Control Visualization

The invention provides a photomask center measurement method, a photomask center measurement device and a photoetching machine table detection method. The photomask centrality measurement method comprises the steps of obtaining measurement marks in a photomask centrality preset area, determining coordinates of all the measurement marks, calculating the distance between the coordinates of all the measurement marks and the boundary of a photomask, and obtaining the centrality of the photomask; the centrality of the photomask is applied to a conventional process control system for visual monitoring. According to the method for measuring the center of the photomask, about 20 minutes of measurement processing in a traditional mode are shortened to about 2 minutes, the measurement efficiency is greatly improved, meanwhile, the center data are uploaded to an SPC system to generate a visual graph, the center of a photoetching machine is monitored in real time through an SPC chart, center monitoring automation is achieved, the method is particularly suitable for all photomasks, and monitoring of each photomask is achieved.

The invention discloses a blowing and sucking integrated fan and a control system. The blowing and sucking integrated fan comprises a shell. The guide mechanism composed of the arc pieces and the fixing pipes is adopted, airflow is guided to form the spiral air duct, the air speed is remarkably increased, and the problem of insufficient wind power caused by unreasonable design of an existing fan air duct is solved; according to the invention, by arranging the efficiency analysis module, accurate modeling and dynamic management are carried out on the charging efficiency of the battery in a mode of charging while operating, and the optimal charging trigger threshold value can be comprehensively calculated based on multiple parameters such as real-time electric quantity, temperature and aging degree; the problem of low efficiency possibly caused by a traditional fixed low-electric-quantity charging strategy is effectively avoided, and the charging opportunity is intelligently judged on the premise that the charging efficiency is guaranteed, so that the single-time operation time of the battery is prolonged, and invalid charging circulation is reduced; the fault early warning module is arranged, and a deviation coefficient method and a time sequence trend analysis method based on statistical process control are adopted, so that dual health monitoring is carried out on core operation parts of the fan.

The application discloses a slab continuous casting machine crystallizer liquid level control method and an automatic casting method, and belongs to the technical field of metallurgy. In the liquid level control stage one, the crystallizer liquid level is raised from the initial liquid level value to the first liquid level set value; in the liquid level control stage two, the crystallizer liquid level is raised from the first liquid level set value to the second liquid level set value, and the second liquid level set value is the liquid level when the casting blank starts to be pulled out. The control method determines whether the parameters are abnormal through parameter collection and parameter analysis, adjusts the stopper opening degree for the crystallizer liquid level control according to the determination result, carries out single variable statistical process control and multivariable statistical process control through single variable analysis and multivariable analysis, and then adjusts the control stopper opening degree value, so that the crystallizer liquid level control in the automatic casting operation is realized, unnecessary shutdown and adjustment time are avoided, the efficiency of the continuous casting production line is maximally improved, the production capacity of the production equipment is fully utilized, and the utilization rate of the production line is greatly improved.

The present invention discloses an advertising placement optimization method and system based on traffic consumption, belonging to the technical field of advertising placement. The advertising placement optimization method and system based on traffic consumption include the following specific steps: Step 1: Real-time multi-dimensional data collection: The collection objects include: the number of requests, advertisement filling response, effective exposure volume, and previous display cost; Step 2: Sliding time window modeling: A dynamic sliding time window (default window length T = 24 hours) is adopted. The data within the window is divided into multiple sub-intervals according to time slices (each 5 minutes is a storage unit), and a ring buffer is used to store the window data. When new data enters, the oldest data block exceeding the window length is eliminated. By establishing a mathematical model between the network layer bandwidth cost and the business layer economic benefits, and at the same time proposing a dynamic regulation mechanism based on statistical process control, and designing a hybrid architecture combining SDN and big data processing, the present invention reduces the bandwidth cost.

The present application relates to the technical field of image detection, and discloses a card surface topography defect detection method and system based on a relief effect, which comprises the following steps: double-frequency structured light adaptive projection of a laminated card, relief three-dimensional height map reconstruction based on phase deflection, automatic segmentation of a relief area based on a variational level set, analysis of relief profile curvature continuity and defect detection, and relief depth statistical process control. The present application eliminates mirror reflection interference of a laminated layer through a double-frequency phase shift algorithm, and the measurement accuracy of the relief depth reaches ±5 microns, and the detection accuracy reaches 99.5%.

The invention provides a statistical process control method and system based on a large language model. The system comprises the large language model and a statistical process control model. Decoupling the statistical process control model and the large language model on the architecture; the large language model is used for receiving an analysis request input by a user; reasoning the analysis request, and determining an analysis intention and an input parameter of the user; according to the analysis intention, a predefined Schema parameter mode of a target analysis tool and an input parameter, generating a tool calling instruction for the target analysis tool in the statistical process control model; sending the tool calling instruction to a statistical process control model, and receiving returned analysis data determined based on a statistical algorithm library; and analyzing the analysis data, generating an analysis result in a natural language form, and feeding back the analysis result to the user. In this way, the problems that in the prior art, a statistical process control analysis tool is complex in operation process, high in requirement for professional knowledge of a user, low in use efficiency and unreliable in calculation result of a universal AI model can be solved.

The invention relates to the field of water treatment automation, and discloses an intelligent disinfection method, system, medium and equipment combining feedforward optimization and feedback correction, and the method comprises the following steps: in a feedforward stage, based on measurable parameters such as dosage, water inflow, water temperature, turbidity, pH and the like, predicting factory residual chlorine by utilizing a machine learning model, and optimizing and calculating basic dosage by combining a genetic algorithm; in the feedback stage, a short-time-delay chlorine consumption prediction model is established, and the deviation between the actual chlorine consumption and the theoretical chlorine consumption predicted after dosing is calculated; and monitoring the deviation in real time by using a statistical process control method, judging that unmonitored water quality disturbance exists when the deviation exceeds a statistical control limit, and calculating the compensation dosage according to the magnitude of the deviation. And superposing the optimal basic dosing amount and the compensation dosing amount to obtain a final execution dosing amount, and controlling a dosing pump to execute dosing.

This application discloses a coating system and a coating method, wherein the system includes: a statistical process control system, a manufacturing execution system, and a controller; the controller is configured to generate a roll change signal when it detects that the winding device of the coating machine performs a tape breakage operation and connects the tape to the target film roll; the statistical process control system is configured to, in response to the roll change signal, send a processing request for the film roll number to the manufacturing execution system, obtain a first film roll number of the target film roll sent by the manufacturing execution system based on the processing request, and create a first film roll order based on the first film roll number; wherein the first film roll order is used to record the production process information of the target film roll during the coating process; the manufacturing execution system is configured to, in response to the processing request, determine the first film roll number of the target film roll and send the first film roll number to the statistical process control system.

The invention discloses an SLCC converter valve state diagnosis method based on a multi-parameter correlation residual error. The method comprises the following steps: constructing a multi-parameter health correlation model HCM; collecting multi-parameter operation data in real time; calculating an expected healthy temperature under the current working condition; calculating a real-time residual error; and performing diagnostic decision-making based on residual error statistical process control. According to the method, early warning is realized, + 3 DEG C tiny temperature drift caused by aging of heat dissipation silicone grease can be stripped from strong working condition interference by detecting whether the real-time residual error breaks through the statistical threshold value or not, the sub-health hidden danger is found in advance, and the problem of diagnosis lag is solved. According to the method, the normal influence of the working condition is automatically stripped, the diagnosis basis is a'non-existing 'residual value, and false alarm and missing alarm under heavy load or light load of a traditional method are avoided.

The invention relates to a coking coal ash composition control method based on partial least squares, which comprises the following steps: collecting n coal ash samples, detecting the mass contents of SiO2, Al2O3, Fe2O3, CaO, MgO, Na2O, K2O, TiO2, P2O5 and MnO2 in the components of each coal ash sample, generating an n * 10 independent variable matrix X, detecting the CSR value of each coal ash sample, and generating an n * 1 dependent variable matrix Y; standardizing the independent variable matrix X and the dependent variable matrix Y, and marking the standardized matrixes as Xs and Ys; constructing a latent variable regression model; training the regression model, and evaluating the performance of the trained regression model; screening the first four important components according to the VIP values; defining a sample set of target range CSR values; and calculating the total amount of coal ash components with the top four VIPs of each target range CSR sample. The control range of the coal ash components is determined through VIP analysis and statistical process control.

The invention discloses a silk hydrophilic non-woven fabric production process online optimization method, and relates to the field of textile intelligent manufacturing, and the method comprises the steps: obtaining synchronization process data of a preset process node of a production line, the synchronization process data comprising a cloth cover image subjected to time-space alignment and a corresponding upstream process parameter; calculating and generating a microscopic uniformity characteristic time sequence based on the cloth cover image; performing correlation analysis on the characteristic time sequence and the corresponding process parameter time sequence, and identifying a process disturbance source; performing real-time monitoring on the microscopic uniformity characteristic time sequence by adopting a statistical process control method, and performing abnormal early warning according to trend prediction to generate an early warning signal; and according to the early warning signal and the process disturbance source, querying a preset compensation rule to generate a process parameter compensation instruction of the downstream process and executing the process parameter compensation instruction. The quantitative characteristics of the microscopic uniformity of the cloth cover are extracted in real time, a process disturbance source is traced, and prospective statistical early warning is combined, so that closed-loop optimization of the production process and improvement of the product quality are realized.

The present application relates to the technical field of enterprise operation, in particular to an enterprise operation early warning monitoring method and system based on data analysis, the method comprising: acquiring index data of multiple dimensions of enterprise operation; performing time series decomposition on the index of each dimension, constructing a dynamic baseline that dynamically changes with the business cycle, and calculating the deviation intensity of the actual value of each index based on the dynamic baseline; inputting the deviation intensity into a statistical process control model and an anomaly detection model respectively to obtain first and second outputs; nonlinearly fusing the first and second outputs, and calculating a fusion risk score and a corresponding confidence level according to the divergence between the first and second outputs; determining an early warning level according to the fusion risk score and the confidence level, and generating a structured disposal scheme; collecting feedback data on the disposal scheme, and dynamically updating the threshold coefficient of the dynamic baseline, the anomaly detection model, and the fusion weight of the nonlinear fusion according to the feedback data.