Quality prediction system and method based on multi-dimensional feature coupling driving

Abstract

The invention relates to the technical field of data processing and intelligent prediction, in particular to a quality prediction system and method based on multi-dimensional feature coupling driving, and the system comprises a business logic solidification module which is used for extracting and solidifying business influence logic, and generating static risk factors; the real-time working condition comparison module is used for comparing the real-time operation data and generating a dynamic deviation file; the risk trend extrapolation module is used for processing the time sequence of the dynamic deviation file and outputting a prospective risk track; and the final risk grading module is used for integrating the static risk factor and the prospective risk track, grading and outputting a final quality risk assessment value. According to the method, the static inherent risk of the business process and the dynamic evolution trend of the real-time data are deeply coupled, so that the problem that multi-dimensional information is difficult to fuse in a traditional method is solved, the global association effect of the current deviation in the business process is quantified, and the accuracy and interpretability of quality risk prediction are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing and intelligent prediction technology, and specifically to a quality prediction system and method based on multi-dimensional feature coupling drive. Background Art

[0002] In modern industrial production, energy management, and environmental monitoring, analyzing massive amounts of operational data to predict future states and assess potential risks has become a core requirement for ensuring system security and improving operational efficiency. For example, in manufacturing, accurately predicting product quality risks can guide dynamic adjustments to the production process; in power grid dispatching, power quality risk assessment is key to maintaining grid stability.

[0003] To achieve this goal, existing technologies mostly rely on advanced data-driven paradigms, particularly deep learning-based models. These methods build complex neural network structures, such as graph neural networks to capture spatial associations between entities, or employ models such as long short-term memory networks and temporal convolutional networks to learn the temporal dependencies of historical data and directly predict future operational indicators.

[0004] However, existing technologies generally have problems such as difficulty in integrating business knowledge, lack of ideas for bias modeling, and rough integration of static and dynamic risks. These defects restrict the performance of predictive models in terms of accuracy, stability and interpretability.

[0005] To this end, a quality prediction system and method based on multi-dimensional feature coupling drive is proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide a quality prediction system and method based on multi-dimensional feature coupling drive, aiming to solve the problem that the quality prediction results of existing technologies are one-sided and lack foresight. To solve the problems existing in the existing technology, the present invention uses a business logic solidification module to refine and solidify the business impact logic to generate static risk factors; a real-time operating condition comparison module compares real-time operating data to generate dynamic deviation files; a risk trend extrapolation module processes the time series of dynamic deviation files and outputs forward-looking risk trajectories; and a final risk grading module integrates static risk factors and forward-looking risk trajectories to grade and output the final quality risk assessment value.

[0007] To achieve the above objectives, the present invention provides a quality prediction system based on multi-dimensional feature coupling drive, which specifically includes the following modules: Business logic solidification module: This module collects historical operational data of the monitored process and, based on the pre-set business knowledge ontology, extracts and solidifies the inherent influencing logic between business entities to form a structured business rule base. This structured business rule base is then processed to generate static risk factors. Real-time operating condition comparison module: used to compare the real-time operating data of the monitored process at each time step using the structured business rule library as a benchmark, determine the current state deviation between business entities, and generate a dynamic deviation file; Risk Trend Extrapolation Module: This module processes the time series of dynamic deviation archives, summarizes the deviation information of each time step through the correlation effect processor, and generates a stage-by-stage status summary. This stage-by-stage status summary is input into the time series extrapolator, which uses the preset evolutionary algorithm to extrapolate the future trend of the state deviation and output a forward-looking risk trajectory. Final risk rating module: used to integrate forward-looking risk trajectories and static risk factors to form a comprehensive input; apply the comprehensive input to the preset scoring logic, grade and output the final quality risk assessment value.

[0008] Preferably, the step of generating a static risk factor specifically includes: representing the business knowledge ontology as a process topology diagram containing nodes and directed edges, wherein the nodes represent business entities and the directed edges represent the impact transmission paths between entities; defining an inherent basic severity level for each node in the process topology diagram and defining an impact propagation coefficient for each directed edge as the core content of the structured business rule library through a knowledge engineering method; extracting the basic severity levels of all business entities in the structured business rule library to constitute the static risk factor.

[0009] Preferably, the step of generating a dynamic deviation archive specifically includes: calculating the monitoring indicators corresponding to each business entity from the historical health operation data, and storing the statistical benchmark distribution of the monitoring indicators, wherein the statistical benchmark distribution includes the expected mean and standard deviation; obtaining the real-time operation data value at each time step, and calculating the standardized discrete score according to the statistical benchmark distribution corresponding to the real-time operation data value; using the standardized discrete score as the quantified current state deviation, recording the standardized discrete scores of all business entities, and forming the dynamic deviation archive.

[0010] Preferably, the step of generating a stage status summary specifically includes: loading a process topology diagram; for each node in the process topology diagram, setting the current state deviation of the node in the dynamic deviation file as the initial deviation value of the node; according to the influence transfer path of the topology diagram, performing calculations based on the deviation values ​​of each node in the previous round of iterations, iteratively updating the deviation value of each node until the iteration is terminated after any condition is met: first, in two adjacent iterations, calculating the absolute value of the difference between the deviation value after the update and the deviation value before the update of all nodes, and the maximum value of the absolute value is less than a preset convergence threshold; second, reaching a preset maximum number of iterations; and using the final obtained deviation value of each node as the cumulative deviation vector representing the global correlation effect, i.e., the stage status summary.

[0011] Preferably, the step of outputting the forward-looking risk trajectory specifically includes: the time series extrapolator is a state space predictor; the state space predictor internally maintains a state vector representing the risk state, and the state vector includes the current level of risk and the rate of change of risk; when processing the time series of the stage state summary, the state space predictor alternately performs time update and measurement update at each time step, and through the time update step, predicts the risk state at the current moment based on the state vector at the previous moment to generate a prediction result; through the measurement update step, the prediction result is corrected in combination with the actual observation value at the current moment to obtain the optimal estimate of the risk state at the current moment; after the entire time series is processed, the state space predictor continues to apply the time update step, and performs multi-step extrapolation to the future based on the finally obtained state vector to generate a state vector sequence containing prediction values ​​for multiple future time steps as the forward-looking risk trajectory.

[0012] Preferably, the step of forming a comprehensive input specifically includes: for each business entity, extracting the basic severity level and risk prediction peak of the business entity from the static risk factors and the forward-looking risk trajectory respectively; applying the extracted basic severity level and the risk prediction peak as mutually related input variables to a predefined risk assessment model to output a quantitative comprehensive risk index; the comprehensive risk indices generated by all business entities are aggregated into a set to form the comprehensive input; the risk assessment model reflects the nonlinear interaction between the basic severity level and the risk prediction peak.

[0013] Preferably, the steps of grading and outputting the final quality risk assessment value specifically include: the preset scoring logic is a risk assessment matrix, the risk assessment matrix pre-defines multiple non-overlapping numerical intervals, and each interval is mapped to a unique and ordered quality risk level; the highest comprehensive risk index is matched with the numerical interval of the risk assessment matrix; and the quality risk level corresponding to the interval where the highest comprehensive risk index is located is output as the final quality risk assessment value.

[0014] Preferably, a quality prediction method based on multi-dimensional feature coupling drive includes: Collect historical operational data of the monitored process and, based on the preset business knowledge ontology, extract and solidify the inherent influence logic between business entities to form a structured business rule base; process the structured business rule base to generate static risk factors; At each time step, using the structured business rule base as a benchmark, comparing the real-time operational data of the monitored process, determining the current state deviations between business entities, and generating a dynamic deviation profile; Process the time series of dynamic deviation archives, summarize the deviation information of each time step through the correlation effect processor, and generate a stage-by-stage state summary; input the stage-by-stage state summary into the time series extrapolator, and according to the preset evolutionary algorithm, extrapolate the future trend of the state deviation and output a forward-looking risk trajectory; Integrate forward-looking risk trajectories and static risk factors to form a comprehensive input; apply the comprehensive input to the preset scoring logic to grade and output the final quality risk assessment value.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention first solidifies the inherent logic of business processes and expert knowledge into static risk factors, enabling the prediction model to be built on stable business rules rather than relying solely on fluctuating real-time data. This ensures that the final risk assessment conclusion is supported by clear business logic, overcoming the limitation of purely data-based models that make it difficult to explain the internal decision-making process, and making the prediction results more convincing.

[0016] 2. By using a correlation effect processor to simulate the propagation and accumulation of deviations within a business process network, this invention can quantify the cascading impact of a single local anomaly on the entire system. This transcends the limitations of traditional methods that analyze individual monitoring points in isolation, effectively identifying potential systemic risks resulting from the coupling of multiple small deviations and significantly improving the accuracy and completeness of risk identification.

[0017] 3. This invention uses a time-series extrapolator to predict the evolution of risk trends and generate future risk trajectories. This enables the system to transition from passive, responsive alerts to proactive, preventative warnings, buying valuable decision-making time for managers to take intervention measures, helping to prevent quality issues from occurring at the source and reduce potential losses.

[0018] 4. This invention integrates the assessment of static risks, reflecting the inherent importance of a business entity, with predictive risks, reflecting its future dynamics. This multi-dimensional, comprehensive assessment mechanism can more effectively distinguish between risks of varying natures, such as a minor anomaly in a core link versus a significant fluctuation in a non-critical link. This helps users prioritize resources to the most critical risk points, resulting in more efficient and targeted decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of the structure of a quality prediction system based on multi-dimensional feature coupling drive according to the present invention; Figure 2 This is an overall interaction sequence diagram of a quality prediction system and method based on multi-dimensional feature coupling drive according to the present invention; Figure 3 This is a flow chart of the multi-dimensional feature processing and risk fusion of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] See also Figures 1 to 3 The present invention provides a quality prediction system and method based on multi-dimensional feature coupling drive, and the technical solution is as follows: A quality prediction system based on multi-dimensional feature coupling drive, comprising: Business logic solidification module: This module collects historical operational data of the monitored process and, based on the pre-set business knowledge ontology, extracts and solidifies the inherent influencing logic between business entities to form a structured business rule base. This structured business rule base is then processed to generate static risk factors. Real-time operating condition comparison module: used to compare the real-time operating data of the monitored process at each time step using the structured business rule library as a benchmark, determine the current state deviation between business entities, and generate a dynamic deviation file; Risk Trend Extrapolation Module: This module processes the time series of dynamic deviation archives, summarizes the deviation information of each time step through the correlation effect processor, and generates a stage-by-stage status summary. This stage-by-stage status summary is input into the time series extrapolator, which uses the preset evolutionary algorithm to extrapolate the future trend of the state deviation and output a forward-looking risk trajectory. Final risk rating module: used to integrate forward-looking risk trajectories and static risk factors to form a comprehensive input; apply the comprehensive input to the preset scoring logic, grade and output the final quality risk assessment value.

[0022] Example 1 This embodiment provides a specific application of a quality prediction system and method based on multi-dimensional feature coupling drive, and takes a key thin film deposition process in semiconductor manufacturing as an example to illustrate the specific implementation of the present invention. Figure 2 、 3 This embodiment performs detection according to the process of a quality prediction method based on multi-dimensional feature coupling drive provided by the present invention and displays the data flow.

[0023] See Figure 1 , the system structure of the present invention is shown in the figure.

[0024] Furthermore, historical operational data of the monitored process is collected, and based on the preset business knowledge ontology, the inherent influence logic between business entities is refined and solidified to form a structured business rule base. The structured business rule base is processed to generate static risk factors. The specific steps are as follows: During the initial deployment phase, the system first implemented the business logic solidification module. Based on the domain knowledge of experienced process engineers, the system constructed a business knowledge ontology describing the deposition process. This knowledge ontology is represented as a process topology graph consisting of nodes and directed edges. Nodes represent business entities such as the gas flow controller, chamber temperature controller, and film thickness, the final quality indicator. Directed edges clearly depict the impact transmission paths between them. Using a knowledge engineering approach, the process topology graph was assigned quantified logical parameters. Each node representing a process parameter was assigned an inherent base severity level. The chamber temperature controller was assigned a level of 5 (maximum) due to its critical impact on final quality. Furthermore, each impact path was assigned an impact propagation coefficient. For example, the coefficient from chamber temperature to film thickness was set to 0.8 to indicate the strength of the impact transmission. These parameters collectively constitute a structured business rule base, from which the base severity level set extracted forms a static risk factor.

[0025] The knowledge engineering method further includes an offline calibration step for the impact propagation coefficient and the basic severity level. The system periodically extracts dynamic deviation files and corresponding final product quality data for a period from historical data records; uses a preset statistical association analysis method to calculate the actual correlation strength between the state deviation of each business entity and the final product quality during the period, and uses this as the data-driven correlation degree; compares the data-driven correlation degree with the impact propagation coefficient of the corresponding directed edge in the process topology diagram, and adjusts the impact propagation coefficient based on the difference between the two; and adjusts the basic severity level of a business entity based on the frequency and severity of quality problems caused by the deviation of a specific business entity during the period.

[0026] The offline calibration step combines the initial expert knowledge with long-term actual production data, enhancing the accuracy and reliability of the entire prediction system under long-term operation and avoiding the decline in prediction performance due to outdated knowledge base.

[0027] This method transforms domain knowledge into a structured process topology diagram that includes basic severity levels and impact propagation coefficients. This step provides clear, business-logic-based parameters for subsequent risk calculations, making the foundation of the assessment process traceable and explainable. This approach reduces the reliance on large numbers of annotated samples in purely data-driven models and aligns assessment results with actual process logic.

[0028] Furthermore, at each time step, the structured business rule library is used as a benchmark to compare the real-time operational data of the monitored process, determine the current state deviations between business entities, and generate a dynamic deviation file. The specific steps are as follows: After the system is put into operation, the real-time operating condition comparison module continues to work. The system pre-calculates the monitoring indicators corresponding to each business entity from the massive historical healthy production batch data and stores its statistical benchmark distribution. Define a data window with a preset fixed length; the data window is used to select a newest data subset from the historical healthy operation data; the calculation of the statistical benchmark distribution is performed only based on the data subset contained in the data window; as new healthy operation data is generated, the data window slides forward to include new data and exclude the oldest data, and the statistical benchmark distribution is periodically recalculated and updated for storage. For the monitoring indicator of chamber temperature, its statistical benchmark distribution is determined to be an expected mean of 200°C and a standard deviation of 0.5°C. At any production time step, when the module obtains the real-time operating data value of the chamber temperature of 201.2°C, it will immediately calculate its standardized discrete score of 2.4 based on the statistical benchmark distribution corresponding to the value. This score, together with the standardized discrete scores of other entities (such as gas flow controllers), constitutes the dynamic deviation profile at that moment, accurately quantifying the degree to which the current operating conditions deviate from the ideal state.

[0029] By limiting the calculation of the statistical benchmark distribution to the latest data subset, the present invention ensures that the benchmark can reflect the recent actual operating status of the monitored process, improves the accuracy of the subsequently calculated standardized discrete scores, and thus provides more timely and reliable input data for the entire risk prediction system.

[0030] This invention uses a method based on the statistical distribution of historical data to calculate the deviation of the current state. This method converts different types of monitoring data into standardized, dimensionless values, facilitating unified subsequent processing and comparison. It also takes into account the historical fluctuation range of the data, providing a relatively objective benchmark for determining current deviations and helping to reduce potential misjudgments caused by improperly set fixed thresholds.

[0031] Furthermore, the time series of the dynamic deviation archive is processed, and the deviation information of each time step is summarized through the correlation effect processor to generate a stage-by-stage state summary. The stage-by-stage state summary is input into the time series extrapolator. Based on the preset evolutionary algorithm, the future trend of the state deviation is extrapolated and the forward-looking risk trajectory is output. The specific steps are as follows: The risk trend extrapolation module then receives and processes the time series consisting of the dynamic deviation file. First, the associated effect processor loads the process topology and, for each node in the diagram, sets the current state deviation in the dynamic deviation file as the node's initial deviation value. Based on the impact transmission path in the topology, the processor iteratively updates the deviation value of each node.

[0032] When calculating the contribution of each upstream node to its downstream node, the deviation value of the upstream node is first processed by a preset nonlinear activation function; the nonlinear activation function has a saturation characteristic, and its function is that when the absolute value of the input deviation value is small, its output is approximately proportional to the input, and when the absolute value of the input deviation value exceeds a certain range, its output tends to a constant upper limit value; the output value after processing by the nonlinear activation function is then multiplied by the influence propagation coefficient as the effective influence input of the upstream node on the downstream node.

[0033] By performing nonlinear transformations on the deviation values ​​of upstream nodes, the system effectively limits the excessive impact of extreme outliers at a single node on the cumulative deviation assessment of the entire system. This step enhances the computational stability of the correlation effect processor when faced with extreme anomalies or noise in the input data, making the results of the staged status summary more robust and reasonable, and providing higher-quality input data for subsequent trend extrapolation.

[0034] Between two consecutive iterations, the processor calculates the absolute difference between the updated and pre-update deviations for all nodes and finds the maximum. The iteration terminates when this maximum value falls below a preset convergence threshold of 0.01. The resulting deviations for each node form a cumulative deviation vector representing the global correlation effect, effectively summarizing the state at that moment.

[0035] This method simulates the mutual impact between different business entities by performing iterative calculations on a process topology. This method aims to quantify the cumulative effect of local deviations propagating through the system, thereby generating a comprehensive indicator reflecting the overall state at the current moment. This helps identify potential correlation risks arising from the combined effects of multiple independent, smaller deviations.

[0036] Next, the state-space predictor, which acts as a time series extrapolator, begins to work. It internally maintains a state vector that contains the current level of risk and the rate of change. When processing a time series of periodic state summaries, the predictor alternates between time updates and measurement updates at each time step: first, through the time update step, the risk state at the current moment is predicted based on the state vector at the previous moment; then, through the measurement update step, the prediction result is corrected in combination with the actual observation value at the current moment to obtain the optimal estimate of the current risk state. When the entire historical time series is processed, the predictor continues to apply the time update step, performing multi-step extrapolation into the future based on the final state vector, generating a state vector sequence containing predicted values ​​for multiple future time steps as a forward-looking risk trajectory. For example, the trajectory may show that the comprehensive risk will reach a peak of 4.5 in the third time step in the future.

[0037] This invention uses a state-space predictor to process time series data representing the overall state. Based on the dynamic characteristics of the historical series, this predictor estimates its future trends and generates forward-looking data trajectories. This trajectory provides temporal input for subsequent risk assessments, allowing them to incorporate consideration of future states.

[0038] Furthermore, the forward-looking risk trajectory and static risk factors are integrated to form a comprehensive input. The comprehensive input is applied to the preset scoring logic to determine the level and output the final quality risk assessment value. The specific steps are as follows: Finally, the final risk assessment module integrates and evaluates the dynamic and static information. For the chamber temperature controller entity, the module extracts a baseline severity level of 5 and a predicted peak risk level of 4.5 from the static risk factors and forward-looking risk trajectory, respectively. These two values ​​are used as interrelated input variables in a predefined risk assessment model.

[0039] When extracting the risk prediction peak from the forward-looking risk trajectory, the future time step corresponding to the risk prediction peak is also extracted; the time decay weight is calculated based on the future time step and the preset time decay coefficient; the risk prediction peak is operated with the time decay weight to generate a time-corrected risk prediction peak; the time-corrected risk prediction value is subsequently used to replace the original risk prediction peak to calculate the comprehensive risk index.

[0040] The risk assessment model reflects the nonlinear interaction between the two. When the base severity level is high, the output comprehensive risk index may be amplified, even if the risk prediction peak is only moderate. In this example, the model outputs a quantitative comprehensive risk index of 28.0. The comprehensive risk indices generated by all business entities are aggregated into a single set to form the comprehensive input.

[0041] The risk prediction peak is subjected to time decay processing, so that the final risk assessment result not only reflects the severity of the risk, but also indirectly reflects the urgency of its occurrence, thereby improving the guiding value of early warning information for production intervention.

[0042] This method integrates static factors representing inherent importance with dynamic predictive values ​​representing future trends through a model that reflects nonlinear interactions. This approach attempts to capture both the inherent severity of an event and its potential for future development in a single value, resulting in a composite risk index that encompasses a richer range of information.

[0043] During the grading step, the system compares the highest comprehensive risk index (28.0) in the input set with a pre-defined risk assessment matrix. The risk assessment matrix pre-defines multiple, non-overlapping numerical ranges and corresponding unique quality risk levels. The matrix defines values ​​between 15 and 25 as "Level 3 (High Risk)" and values ​​greater than 25 as "Level 4 (Extreme Risk)." Since 28.0 falls within the "Level 4 (Extreme Risk)" range, the system outputs this final quality risk assessment value.

[0044] The present invention uses a risk assessment matrix to map the calculated comprehensive risk index to pre-set risk levels. This step converts continuous numerical results into discrete, ordered levels, making the assessment output easier to understand and use, and facilitating alignment with existing operating procedures or risk management standards.

[0045] This invention provides a quality risk prediction method. This method integrates static business logic with dynamic, real-time data and extrapolates future data trends to generate a quality risk assessment for the monitored process. This allows users to obtain a quantitative reference for potential risks before final product quality is achieved, providing data support for production process adjustments and helping to reduce potential losses caused by unknown risks.

[0046] Example 2 This example describes the specific workflow of the business logic curing module and the real-time operating condition comparison module in the present invention. The scenario is a plasma etching process in semiconductor manufacturing. The final etching uniformity quality of this process is affected by multiple process parameters such as RF power and reaction gas flow rate.

[0047] During the initial deployment phase, the system first executes the business logic solidification module. Based on the knowledge of process experts, the system represents the business knowledge ontology of the etching process as a process topology graph consisting of nodes and directed edges. Nodes represent business entities, such as RF power controllers and gas flow controllers, while directed edges represent the transmission paths of their impact on final product quality. Using knowledge engineering methods, this topology graph is assigned quantified logical parameters: the RF power controller node is assigned an inherent base severity level of 4 (maximum 5), and the gas flow controller node is assigned a base severity level of 3. Furthermore, the edge from the RF power controller to the final quality is assigned an impact propagation coefficient of 0.9, and the edge from the gas flow controller to the final quality is assigned an impact propagation coefficient of 0.7. These parameters together form a structured business rule base. The system then extracts the base severity levels of all business entities to form static risk factors.

[0048] During process operation, the real-time operating condition comparison module continues to work. The system has calculated the monitoring indicator of RF power from historical healthy production data and stored its statistical benchmark distribution, which has an expected mean of 500 watts and a standard deviation of 2 watts. At a certain production time step, the system obtained the real-time operating data value of RF power as 505 watts. Based on the statistical benchmark distribution corresponding to this value, the system calculated its standardized discrete score to be 2.5. By performing the same operation on all relevant business entities, the system records the standardized discrete scores of all entities at that moment, forming a dynamic deviation file for that time step. This dynamic deviation file, together with the files generated at other time steps, constitutes a time series, which serves as input data for the subsequent risk trend extrapolation module.

[0049] Example 3 This example describes the specific workflow of the risk trend extrapolation module and the final risk grading module in the present invention. In this example, the system has generated a time series of dynamic deviation profiles for the chemical mechanical polishing process through the preceding modules and is ready for in-depth trend analysis and risk grading.

[0050] The risk trend extrapolation module starts the analysis after receiving the time series of the dynamic deviation file. First, the correlation effect processor loads the process topology at each time step. For one of the time steps, it sets the current state deviation in the dynamic deviation file at that moment as the initial deviation value of the corresponding node in the process topology. According to the impact transfer path and the impact propagation coefficient of the topology, the processor updates the deviation value of each node through iterative calculation to simulate the propagation and accumulation of risks in the system. When the maximum absolute value of the difference between the deviation value of all nodes after the update and the deviation value before the update in two adjacent iterations is less than the preset convergence threshold, the iteration terminates. The deviation values ​​of each node obtained after the iteration constitute a cumulative deviation vector representing the global correlation effect, that is, the stage status summary at that moment.

[0051] Next, the state-space predictor, acting as a time-series extrapolator, processes a time series consisting of multiple staged state summaries. It maintains a state vector containing the current risk level and rate of change. While processing this time series, the predictor alternates between time update and measurement update steps at each time step, accurately tracking the dynamic evolution of the risk state. After processing the entire time series, the predictor applies the final state vector and continuously applies the time update step to perform multiple extrapolations into the future. This generates a sequence of state vectors containing predicted values ​​for multiple future time steps, known as the forward-looking risk trajectory. Assume that this trajectory indicates that the system's overall risk will reach a predicted peak risk value of 5.0 at some point in the future.

[0052] Finally, the final risk rating module begins its work. For the critical business entity, the RF power controller, it extracts a base severity level of 4 and a predicted risk peak of 5.0 from static risk factors and forward-looking risk trajectories, respectively. These two values ​​are used as interrelated input variables and applied to a predefined risk assessment model that reflects nonlinear interactions. After calculation, the model outputs a quantitative comprehensive risk index of 24. The system then aggregates the comprehensive risk indices generated by all business entities into a single set and identifies the highest index value, 24. Finally, the preset scoring logic, a risk assessment matrix, matches this highest comprehensive risk index with a preset numerical range in the matrix. The matrix defines the range of 15-25 as corresponding to "Level 3 (High Risk)," and the system outputs a final quality risk assessment value of "Level 3 (High Risk)."

[0053] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A quality prediction system based on multi-dimensional feature coupling drive, characterized in that: include: Business logic solidification module: This module collects historical operational data of the monitored process and, based on the pre-set business knowledge ontology, extracts and solidifies the inherent influencing logic between business entities to form a structured business rule base. This structured business rule base is then processed to generate static risk factors. Real-time operating condition comparison module: used to compare the real-time operating data of the monitored process at each time step using the structured business rule library as a benchmark, determine the current state deviation between business entities, and generate a dynamic deviation file; Risk Trend Extrapolation Module: This module processes the time series of dynamic deviation archives, summarizes the deviation information of each time step through the correlation effect processor, and generates a stage-by-stage status summary. This stage-by-stage status summary is input into the time series extrapolator, which uses the preset evolutionary algorithm to extrapolate the future trend of the state deviation and output a forward-looking risk trajectory. Final risk rating module: used to integrate forward-looking risk trajectories and static risk factors to form a comprehensive input; Apply the comprehensive input to the preset scoring logic to rank and output the final quality risk assessment value.

2. A quality prediction system based on multi-dimensional feature coupling drive according to claim 1, characterized in that: The step of generating a static risk factor specifically includes: The business knowledge ontology is represented as a process topology diagram containing nodes and directed edges, wherein the nodes represent business entities and the directed edges represent the impact transmission paths between entities; through the knowledge engineering method, an inherent basic severity level is defined for each node in the process topology diagram, and an impact propagation coefficient is defined for each directed edge, which serves as the core content of the structured business rule library; the basic severity levels of all business entities in the structured business rule library are extracted to form the static risk factor.

3. The quality prediction system based on multi-dimensional feature coupling drive according to claim 1, characterized in that: The steps of generating a dynamic deviation file specifically include: calculating the monitoring indicators corresponding to each business entity from the historical health operation data, and storing the statistical benchmark distribution of the monitoring indicators, wherein the statistical benchmark distribution includes the expected mean and standard deviation; obtaining the real-time operation data value at each time step, and calculating the standardized discrete score based on the statistical benchmark distribution corresponding to the real-time operation data value; using the standardized discrete score as a quantified current state deviation, recording the standardized discrete scores of all business entities, and forming the dynamic deviation file.

4. The quality prediction system based on multi-dimensional feature coupling drive according to claim 2, characterized in that: The step of generating a phased status summary specifically includes: Load the process topology diagram; for each node in the process topology diagram, set the current state deviation of the node in the dynamic deviation file as the initial deviation value of the node; calculate based on the deviation value of each node in the previous round of iteration according to the influence transfer path of the topology diagram, iteratively update the deviation value of each node until any condition is met and the iteration is terminated: first, in two adjacent iterations, calculate the absolute value of the difference between the deviation value after the update and the deviation value before the update of all nodes, and the maximum value of the absolute value is less than the preset convergence threshold; second, reach the preset maximum number of iterations; use the final deviation value of each node as the cumulative deviation vector representing the global correlation effect, that is, the stage state summary.

5. The quality prediction system based on multi-dimensional feature coupling drive according to claim 1, characterized in that: The step of outputting the forward-looking risk trajectory specifically includes: the time series extrapolator is a state space predictor; the state space predictor internally maintains a state vector representing the risk state, and the state vector includes the current level of risk and the rate of change of risk; when processing the time series of the stage state summary, the state space predictor alternately performs time update and measurement update at each time step, and through the time update step, predicts the risk state at the current moment based on the state vector at the previous moment to generate a prediction result; through the measurement update step, the prediction result is corrected in combination with the actual observation value at the current moment to obtain the optimal estimate of the risk state at the current moment; after the entire time series is processed, the state space predictor continues to apply the time update step, and performs multi-step extrapolation to the future based on the final state vector to generate a state vector sequence containing prediction values ​​for multiple future time steps as the forward-looking risk trajectory.

6. The quality prediction system based on multi-dimensional feature coupling drive according to claim 1, characterized in that: The steps of forming the comprehensive input specifically include: for each business entity, extracting the basic severity level and risk prediction peak of the business entity from the static risk factors and the forward-looking risk trajectory respectively; applying the extracted basic severity level and the risk prediction peak as mutually related input variables to a predefined risk assessment model to output a quantitative comprehensive risk index; the comprehensive risk indices generated by all business entities are aggregated into a set to form the comprehensive input; the risk assessment model reflects the nonlinear interaction between the basic severity level and the risk prediction peak.

7. The quality prediction system based on multi-dimensional feature coupling drive according to claim 1, characterized in that: The steps of grading and outputting the final quality risk assessment value specifically include: the preset scoring logic is a risk assessment matrix, the risk assessment matrix pre-defines multiple non-overlapping numerical intervals, and each interval is mapped to a unique and ordered quality risk level; the highest comprehensive risk index is matched with the numerical interval of the risk assessment matrix; and the quality risk level corresponding to the interval where the highest comprehensive risk index is located is output as the final quality risk assessment value.

8. A quality prediction method based on multi-dimensional feature coupling drive, characterized in that: include: Collect historical operational data of the monitored process and, based on the preset business knowledge ontology, extract and solidify the inherent influence logic between business entities to form a structured business rule library; Process the structured business rule base to generate static risk factors; At each time step, using the structured business rule base as a benchmark, comparing the real-time operational data of the monitored process, determining the current state deviations between business entities, and generating a dynamic deviation profile; Process the time series of dynamic deviation archives, summarize the deviation information of each time step through the correlation effect processor, and generate a stage-by-stage state summary; input the stage-by-stage state summary into the time series extrapolator, and according to the preset evolutionary algorithm, extrapolate the future trend of the state deviation and output a forward-looking risk trajectory; Integrate forward-looking risk trajectories and static risk factors to form a comprehensive input; apply the comprehensive input to the preset scoring logic to grade and output the final quality risk assessment value.

Images