Fatigue damage analysis method and system for industrial speed reducer based on finite element analysis

Abstract

The invention provides an industrial speed reducer fatigue damage analysis method and system based on finite element analysis, and relates to the technical field of finite element analysis, and the method comprises the steps: for each expert, generating an analysis strategy currently asked by the expert based on collaborative interaction behavior data analysis, calculating a weight coefficient of the analysis strategy in combination with preset expert feature data and the correlation between the analysis strategy and the collaborative task target; aggregating analysis strategies and weight coefficients of all experts, and constructing a multi-dimensional decision matrix; performing strategy fusion processing based on the multi-dimensional decision matrix to generate a target simulation scheme; the finite element analysis model is controlled to execute fatigue damage analysis of the target simulation scheme, and the execution rhythm is dynamically adjusted in the execution process so as to adapt to the real-time rhythm of collaborative discussion and decision making of experts. By quantifying expert opinions and optimizing the simulation scheme through strategy fusion, the real-time performance of expert cooperation and the effectiveness of decision making are enhanced, and the analysis result is closer to the actual demand.

Description

Technical Field

[0001] The present invention relates to the technical field of finite element analysis, and in particular to a fatigue damage analysis method and system for an industrial reducer based on finite element analysis. Background Art

[0002] With the rapid development of industrial technology, fatigue damage to industrial reducers, core components of mechanical equipment, is a growing concern. Traditional fatigue damage analysis methods rely heavily on the empirical judgment of a single expert, lacking real-time support and multi-party collaboration, resulting in insufficient scientific and applicable analysis results. In recent years, finite element analysis (FEA) technology has been widely used in reducer fatigue damage analysis. However, effectively integrating the collaborative wisdom of multiple experts with FEA technology to create a dynamic and efficient analysis process remains a pressing technical challenge. Summary of the Invention

[0003] One of the purposes of the present invention is to provide a method and system for analyzing fatigue damage of an industrial reducer based on finite element analysis, so as to solve the problems pointed out in the background technology.

[0004] In a first aspect, an embodiment of the present invention provides an industrial reducer fatigue damage analysis method based on finite element analysis, comprising: 101. On the expert collaboration platform, real-time collection of collaborative interaction behavior data generated by experts for the finite element analysis model of industrial reducers; 102. For each expert, generate the analysis strategy currently advocated by the expert based on the collaborative interaction behavior data analysis, and calculate the weight coefficient of the analysis strategy by combining the preset expert feature data and the correlation between the analysis strategy and the collaborative task goal; 103. Aggregate all experts’ analysis strategies and their weight coefficients to construct a multi-dimensional decision matrix; 104. Perform strategy fusion processing based on the multi-dimensional decision matrix to generate the target simulation plan for the current round of finite element simulation; 105. Control the finite element analysis model to execute fatigue damage analysis of the target simulation scheme, and dynamically adjust the execution rhythm during the execution process to adapt to the real-time rhythm of collaborative discussion and decision-making among experts; 106. After the current round of finite element simulation is completed, return to step 101 to perform the next round of operations until the analysis termination condition is met.

[0005] Optionally, in step 102, generating the analysis strategy currently advocated by the expert based on the analysis of the collaborative interaction behavior data includes: extracting features of target data related to the expert in the collaborative interaction behavior data to obtain a plurality of first feature values; Constructing a first feature description vector based on each first eigenvalue; Based on the preset analysis strategy library, the analysis strategy currently advocated by the expert is matched according to the first feature description vector.

[0006] Optionally, in step 102, the weight coefficient of the analysis strategy is calculated by combining the preset expert feature data and the correlation between the analysis strategy and the collaborative task goal, including: Based on the preset quantitative system, the authority of the analysis strategy represented by the expert feature data and the degree of relevance are quantified respectively; Based on the preset weights, the authority and relevance are weighted to obtain the weight coefficient of the analysis strategy.

[0007] Optionally, the step 104 of performing strategy fusion processing based on the multidimensional decision matrix to generate a target simulation plan for the current round of finite element simulation includes: Based on the local matrix interception constraint, multiple local matrices are intercepted from the multidimensional decision matrix; For each local matrix, feature extraction is performed on the local matrix to obtain multiple second eigenvalues. Based on each second eigenvalue, a second feature description vector is constructed. Based on the preset strategy collaborative execution value library, the value degree of collaborative execution of each analysis strategy in the local matrix is ​​matched according to the second feature description vector; Aggregate the analysis strategies in the local matrix with the largest value to obtain the target simulation solution for the current round of finite element simulation; The local matrix truncation constraints include: The proportion of analysis strategies whose weight coefficients in the local matrix exceed a preset coefficient threshold exceeds a preset proportion threshold; The analysis purposes of each analysis strategy in the same local matrix are different; the overall difference between the analysis strategies in the two local matrices exceeds the preset difference threshold.

[0008] Optionally, in step 105 , the execution rhythm is dynamically adjusted during the execution process to adapt to the real-time rhythm of collaborative discussion and decision-making among the experts, including: Obtain the remaining fatigue damage analysis tasks of the finite element analysis model; Predict future trends in the pace of collaborative discussion and decision-making among experts; Planning experts collaboratively discuss and decide on the best next rhythm based on the current remaining fatigue damage analysis tasks, future rhythm change trends, and real-time rhythm; Split and serialize the remaining fatigue damage analysis tasks into task units to obtain a subtask unit sequence; Select the first N subtask units from the subtask unit sequence; where N is a positive integer, and the value of N is selected so that when the finite element analysis model sequentially executes the selected subtask units, the expected actual rhythm of guiding the experts to discuss and make decisions together is closest to the optimal next rhythm; Dynamically adjusting the execution rhythm includes: controlling the finite element analysis model to sequentially execute the selected first N subtask units within the next unit rhythm time.

[0009] Optionally, the predicted trend of future changes in the pace of collaborative discussion and decision-making among experts includes: Obtain the historical rhythm of collaborative discussions and decision-making among experts; Based on the real-time rhythm and historical rhythm, predict the future rhythm change trend of experts' collaborative discussion and decision-making.

[0010] Optionally, the optimal next cadence, discussed and decided by planning experts in a collaborative manner based on the currently remaining fatigue damage analysis tasks, future cadence change trends, and the real-time cadence, may include: Based on the current remaining fatigue damage analysis tasks, future rhythm change trends and real-time rhythm, a third feature description vector is constructed; Based on the optimal rhythm library, the optimal next rhythm for expert collaborative discussion and decision-making is matched according to the third feature description vector.

[0011] Optionally, the step of acquiring the real-time rhythm of collaborative discussion and decision-making among the experts includes: determining the real-time rhythm of collaborative discussion and decision-making among the experts based on updating the collaborative interaction behavior data collected in real time.

[0012] Optionally, the analysis termination condition includes: The simulation results of multiple rounds of historical finite element simulations are consistent with the collaborative mission goals.

[0013] In the second aspect, an embodiment of the present invention provides an industrial reducer fatigue damage analysis system based on finite element analysis, comprising a processor and a memory, wherein a computer program is stored on the memory, and when the computer program is executed by the processor, the steps of any one of the above-mentioned industrial reducer fatigue damage analysis methods based on finite element analysis are implemented.

[0014] The present invention has achieved the following beneficial effects: The expert collaboration platform collects real-time data on the collaborative interaction behaviors of experts, analyzes and generates their analytical strategies, and calculates weight coefficients based on the correlation between expert characteristic data and the strategies and task objectives. This then constructs a multidimensional decision matrix for strategy integration and generates target simulation plans. The execution rhythm is dynamically adjusted during execution to accommodate the experts' real-time discussions and decision-making. Ultimately, through multiple rounds of iteration, high-precision and efficient fatigue damage analysis is achieved. This significantly improves the scientific nature of the analysis, optimizes the simulation plan through the integration of quantitative expert opinions and strategies, enhances the real-time nature of expert collaboration and the effectiveness of decision-making, and makes the analysis results more aligned with actual needs.

[0015] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0016] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Flowchart of a fatigue damage analysis method for an industrial reducer based on finite element analysis in an embodiment of the present invention; Figure 2 is another flow chart of a fatigue damage analysis method for an industrial reducer based on finite element analysis in an embodiment of the present invention; Figure 3 Another flow chart of the fatigue damage analysis method for industrial reducers based on finite element analysis in an embodiment of the present invention; Figure 4 This is another flow chart of the fatigue damage analysis method for industrial reducers based on finite element analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0019] The research and development idea of ​​this application is to build a dynamic, iterative analysis process for the fatigue damage problem of industrial reducers by combining expert collaboration with finite element analysis technology. The core is to use the real-time interactive data of experts to generate analysis strategies, form an optimized simulation plan through weight quantification and strategy fusion, and dynamically adjust the rhythm during execution, and finally achieve high-precision and high-efficiency fatigue damage analysis through multiple rounds of iteration. This method not only improves the scientific nature of the analysis, but also enhances the real-time nature of expert collaboration and the effectiveness of decision-making.

[0020] Figure 1 The present invention provides a flowchart of a fatigue damage analysis method for an industrial reducer based on finite element analysis, as shown in FIG. Figure 1 As shown, the method includes: 101. On the expert collaboration platform, the collaborative interaction behavior data generated by experts for the finite element analysis model of industrial reducers are collected in real time.

[0021] In this step, the expert collaboration platform collects real-time data on the collaborative interactions between experts working on the finite element analysis model of an industrial reducer. This data reflects the experts' actions, decisions, and communication during the analysis process, and serves as the foundation for subsequent analysis of expert strategies. The real-time nature of the data ensures that the analysis remains current with the experts' latest thinking, while the collaborative nature of the data reflects the collective wisdom of multiple experts.

[0022] Here's an implementation example: In an industrial reducer fatigue analysis project, three experts operated the same finite element model through a collaborative platform. Expert A adjusted the mesh density of the gear area from 50 units to 100 units, ran a stress analysis, and recorded the maximum stress value. Expert B modified the material properties of the bearing part, adjusting the elastic modulus from 200 GPa to 210 GPa, and submitted the thermal stress simulation results. Expert C proposed in the discussion that attention should be paid to the fatigue crack propagation at the gear meshing point and marked the relevant areas. These behavioral data (including parameter adjustments, simulation runs, and communication discussions) are collected in real time and stored in the database of the expert collaborative platform as collaborative interactive behavioral data. Each data item is timestamped and identified by the expert.

[0023] 102. For each expert, the analysis strategy currently advocated by the expert is generated based on the analysis of the collaborative interaction behavior data. The weight coefficient of the analysis strategy is calculated by combining the preset expert feature data and the correlation between the analysis strategy and the collaborative task goal.

[0024] In this step, for each expert, their current analytical strategy is analyzed based on their collaborative interaction data. The weight coefficient for that strategy is calculated by combining expert characteristic data (such as experience and domain expertise) and the correlation between the analytical strategy and the task objective. The weight coefficient reflects the importance of the analytical strategy in the overall decision-making process and serves as a key basis for subsequent integration.

[0025] like Figure 2 As shown, step 102 specifically includes the following sub-steps: 201. Perform feature extraction on target data related to the expert in the collaborative interaction behavior data to obtain multiple first feature values.

[0026] In this step, a plurality of first characteristic values ​​are extracted from the target data related to the experts in the collaborative interaction behavior data. These first characteristic values ​​include quantitative indicators such as operation frequency, parameter adjustment range, and simulation type.

[0027] Continuing with the previous example, let's analyze the collaborative interaction behavior data of Expert A and determine the following first eigenvalues: a grid density adjustment frequency of 2 times per hour, a simulation type of stress analysis, and a parameter adjustment amplitude of a 100% increase in the number of grid cells. These first eigenvalues ​​quantify Expert A's behavior pattern.

[0028] 202. Construct a first feature description vector based on each first eigenvalue.

[0029] In this step, a first feature description vector is constructed based on the first eigenvalue as a structured representation of the expert behavior.

[0030] Continuing with the above implementation example: the feature vector of expert A is [2, stress analysis, 100%], corresponding to frequency, simulation type, and adjustment amplitude, respectively.

[0031] 203. Based on the preset analysis strategy library, match the analysis strategy currently advocated by the expert according to the first feature description vector.

[0032] Based on a library of pre-configured analysis strategies, the first characteristic description vector is used to match the analysis strategy currently advocated by the expert. The strategy library contains analysis strategies corresponding to different first characteristic description vectors, such as high-density mesh stress analysis and material property optimization analysis.

[0033] Continuing with the above implementation example: Expert A's eigenvector is [2, stress analysis, 100%], which matches the high-density mesh stress analysis in the analysis strategy library, indicating that the analysis strategy currently advocated by the expert is a refined analysis focusing on stress concentration areas.

[0034] 204. Based on a preset quantitative system, the authority of the analysis strategy represented by the expert feature data and the relevance of the relevance are quantified respectively.

[0035] In this step, based on a pre-configured quantification system, the authority and relevance of the analysis strategy represented by the expert's characteristic data (e.g., historical industrial reducer fatigue analysis experience) are quantified. When quantifying authority, the proportion of relevant experience of the analysis strategy within the expert's historical industrial reducer fatigue analysis experience is used as the authority level. When quantifying relevance, a mapping relationship is pre-set between different relevances and the relevance levels they represent. During quantification, this mapping relationship is directly queried based on the relevance.

[0036] Continuing with the previous implementation example, Expert A's analysis strategy focuses on detailed analysis of stress concentration areas. This strategy accounts for 90% of their historical experience analyzing fatigue for industrial reducers, resulting in an authority level of 0.9. For the collaborative task objective of gear fatigue damage analysis, Expert A's analysis strategy is highly correlated, with a quantitative correlation level of 0.95.

[0037] 205. Based on the preset weights, perform weighted calculation on the authority and relevance to obtain the weight coefficient of the analysis strategy.

[0038] In this step, based on the preset weights (e.g., authority accounts for 60% and relevance accounts for 40%), the authority and relevance are weighted to obtain the weight coefficient of the analysis strategy. The weight coefficient represents the importance of the analysis strategy.

[0039] Continuing with the above implementation example: the weight coefficient of expert A's analysis strategy = 0.9×0.6+0.95×0.4=0.54+0.38=0.92, indicating that the analysis strategy is of great importance.

[0040] 103. Aggregate the analysis strategies and weight coefficients of all experts to construct a multi-dimensional decision matrix.

[0041] In this step, all experts' analysis strategies and their weight coefficients are aggregated to construct a multidimensional decision matrix. This matrix integrates the scattered expert opinions into a structured framework, providing a data foundation for strategy integration.

[0042] Here's an example implementation: Expert A's analysis strategy is high-density mesh stress analysis with a weighting factor of 0.92. Expert B's analysis strategy is material property optimization analysis with a weighting factor of 0.85. Expert C's analysis strategy is crack growth analysis with a weighting factor of 0.78. The resulting multidimensional decision matrix is ​​shown in Table 1 below: Table 1: Example of a multidimensional decision matrix

[0043] 104. Perform strategy fusion processing based on the multi-dimensional decision matrix to generate the target simulation plan for the current round of finite element simulation.

[0044] In this step, based on a multidimensional decision matrix, the analysis strategies proposed by multiple experts are integrated through local matrix extraction and collaborative value assessment to generate a target simulation solution for the current round of finite element simulation. The multidimensional decision matrix is ​​a structured data framework that aggregates all expert analysis strategies and their corresponding weight coefficients. Each row represents an expert's strategy and its importance assessment results. By performing strategy fusion processing on the matrix, the system can select the optimal strategy combination from the numerous expert opinions, ensuring that the simulation solution is both comprehensive and efficient. This process is divided into three specific sub-steps: local matrix extraction, feature extraction and value assessment, and target solution generation.

[0045] like Figure 3 As shown, step 104 specifically includes the following sub-steps: 401. Based on the local matrix truncation constraint, extract multiple local matrices from the multidimensional decision matrix. The local matrix truncation constraint includes: Constraint 1: The proportion of analysis strategies whose weight coefficients in the local matrix exceed a preset coefficient threshold exceeds a preset proportion threshold.

[0046] This constraint ensures that the strategies in the local matrix have sufficient importance. A preset coefficient threshold (such as 0.8) is used to filter out high-weight strategies, while a preset proportion threshold (such as 0.5) requires that most strategies in the local matrix meet this importance standard.

[0047] Constraint 2: The analysis objectives of each analysis strategy in the same local matrix are different.

[0048] This constraint ensures the diversity of strategies within the local matrix. The analysis objective refers to the execution goal of the strategy (such as stress analysis and material optimization), and each strategy must be unique in its objectives to avoid redundancy.

[0049] Constraint 3: The overall difference between each analysis strategy in the pairwise local matrix exceeds the preset difference threshold.

[0050] This constraint ensures that the strategy combinations between different local matrices are sufficiently distinct. The overall difference can be quantified by the difference in strategy type or weight coefficient, and a preset difference threshold is used to measure the degree of difference.

[0051] Through these three constraints, the extracted local matrix not only contains highly important strategies, but also has internal diversity and external differences, providing high-quality candidate combinations for subsequent fusion.

[0052] Continuing with the above implementation example: The multidimensional decision matrix contains the following data: Expert D: load spectrum optimization analysis, weight coefficient 0.88; Expert E: Thermal stress coupling analysis, weight coefficient 0.90.

[0053] Set the preset coefficient threshold to 0.8, the preset proportion threshold to 0.5, and the preset difference threshold to 0.6. The process of intercepting the local matrix is ​​as follows: Local matrix 1: Includes expert A, expert B, expert C and their respective analysis strategies and weight coefficients; The proportion of analysis strategies with weights exceeding 0.8 (analysis strategies of experts A and B) is 2 / 3≈0.67>0.5, satisfying constraint 1; Analysis objectives: stress analysis, material optimization, and crack growth, which are different from each other and satisfy constraint 2; Local matrix 2: Includes expert D, expert E and their respective analysis strategies and weight coefficients; The proportion of strategies with weights exceeding 0.8 (both the analysis strategies of Expert D and Expert E) is 2 / 2 = 1 > 0.5, which satisfies Constraint 1. Analysis objectives: load optimization and thermal stress analysis, which are different from each other and satisfy constraint 2; In addition, the strategies of local matrix 1 and local matrix 2 are completely different (non-overlapping types), and the difference is 1>0.6, which satisfies constraint three.

[0054] 402. For each local matrix, feature extraction is performed on the local matrix to obtain multiple second eigenvalues. Based on each second eigenvalue, a second feature description vector is constructed. Based on the preset strategy collaborative execution value library, the value degree of the collaborative execution of each analysis strategy in the local matrix is ​​matched according to the second feature description vector.

[0055] In this step, feature extraction is performed on each local matrix to obtain multiple second eigenvalues, which are used to quantify the characteristics of the strategy combination. Based on these eigenvalues, a second feature description vector is constructed, and the value level of the strategy collaborative execution in the local matrix is ​​evaluated through the preset strategy collaborative execution value library. The second eigenvalues ​​include: analysis strategy type (such as stress analysis, crack extension analysis), average weight coefficient (reflecting overall importance), and the correlation between pairwise analysis strategies (such as whether there is execution dependency, for example, stress analysis provides data for crack extension). The second feature description vector is a structured representation of these eigenvalues, which is used to characterize the strategy combination characteristics of the local matrix. The preset strategy collaborative execution value library stores the value levels corresponding to different eigenvectors based on historical experimental data. The value level measures the positive effect of strategy collaborative execution on fatigue analysis of industrial reducers, such as improving the comprehensiveness of analysis or the efficiency of expert collaboration.

[0056] Continuing the above implementation example, process local matrix 1 and local matrix 2: Local matrix 1 (experts A, B, C): Second eigenvalue: Analysis strategy types: high-density mesh stress analysis, material property optimization analysis, crack growth analysis; Average weight coefficient: (0.92 + 0.85 + 0.78) / 3 ≈ 0.85; Correlation: High-density mesh stress analysis provides stress data for crack growth analysis, and there is a dependency relationship; Second feature description vector: [Type diversity = 3, Average weight = 0.85, Dependency = Yes] Value assessment: Based on the strategy-coordinated execution of the value library, the value degree of the second feature description vector is 0.9 (high value, due to enhanced comprehensiveness of analysis due to diversity and dependencies).

[0057] Local matrix 2 (experts D, E): Second eigenvalue: Analysis strategy types: load spectrum optimization analysis, thermal stress coupling analysis; Average weight coefficient: (0.88 + 0.90) / 2 = 0.89; Relationship: No direct dependency Second feature description vector: [Type diversity = 2, Average weight = 0.89, Dependency = No] Value assessment: According to the strategy collaborative execution value library, the value degree of the second feature description vector is 0.7 (medium value, due to the lack of dependency, the synergy effect is low).

[0058] 403. Aggregate the analysis strategies in the local matrix with the largest value to obtain the target simulation solution for the current round of finite element simulation.

[0059] In this step, the most valuable local matrix is ​​selected from all local matrices and the analysis strategies within it are aggregated to form the target simulation plan for the current round of finite element simulation. This plan represents the optimal combination of expert strategies and maximizes the value of collaborative execution.

[0060] Continuing with the above implementation example, select Local Matrix 1 and aggregate its strategies: high-density mesh stress analysis, material property optimization analysis, and crack growth analysis as the target simulation scheme for the current round.

[0061] 105. Control the finite element analysis model to execute fatigue damage analysis for the target simulation solution, and dynamically adjust the execution rhythm during execution to adapt to the real-time rhythm of collaborative discussion and decision-making by the experts. The step of obtaining the real-time rhythm of collaborative discussion and decision-making by the experts includes: determining the real-time rhythm of collaborative discussion and decision-making by the experts based on updating the collaborative interaction behavior data collected in real time.

[0062] In this step, the finite element analysis model is controlled to perform fatigue damage analysis of an industrial reducer according to the target simulation solution generated in step 104. Simultaneously, the simulation execution rhythm is dynamically adjusted using real-time collaborative interaction data to align with the experts' collaborative discussions and decision-making. This dynamic adjustment ensures that simulation results promptly reflect the experts' latest opinions, improving the real-time nature of the analysis and collaborative efficiency.

[0063] Here's an implementation example: the target simulation scenario includes high-density mesh stress analysis, material property optimization analysis, and crack propagation analysis. During execution, the system monitors the frequency of expert discussions. For example, if expert C frequently mentions crack propagation at a frequency of 5 times per minute, the system will speed up the execution of the crack propagation analysis (e.g., from 10 iterations per second to 15). If the expert discussion slows down (e.g., the frequency drops to 1 time per minute), the system will slow down the execution of the stress analysis and wait for more input. The real-time rhythm is determined by collaborative interaction behavior data (e.g., discussion frequency, number of parameter adjustments), ensuring seamless integration of simulation and expert collaboration.

[0064] 106. After the current round of finite element simulation is completed, return to step 101 to perform the next round of operations until the analysis termination condition is met. The analysis termination condition includes: the simulation results of multiple rounds of finite element simulation in the past meet the collaborative task objectives.

[0065] After completing the current round of simulation, the system returns to step 101, recollects expert interaction data, generates a new simulation plan, and enters the next iteration. This process continues until the simulation results meet the analysis termination criteria, that is, the multiple simulation results are consistent with the collaborative task objectives (e.g., fatigue life prediction error less than 5%).

[0066] Here's an example implementation: After the first round of simulation, the results show an 8% error in the gear fatigue life prediction, which falls short of the target. The system returns to step 101, collects new expert recommendations for crack propagation areas (such as increasing mesh density), generates a new solution, and performs a second round of simulation. After several iterations, the error is reduced to 4%, meeting the termination criteria, and the analysis concludes.

[0067] Overall, the method collects real-time data on the collaborative interaction behaviors of experts through an expert collaboration platform, analyzes and generates their analytical strategies, and then calculates weight coefficients based on the correlation between expert characteristic data and the strategies and task objectives. This method then constructs a multidimensional decision matrix for strategy integration and generates a target simulation plan. The execution rhythm is dynamically adjusted during execution to accommodate the real-time discussions and decision-making of the experts. Ultimately, through multiple rounds of iteration, high-precision and efficient fatigue damage analysis is achieved. This significantly improves the scientific nature of the analysis, optimizes the simulation plan by integrating quantitative expert opinions and strategies, and enhances the real-time nature of expert collaboration and the effectiveness of decision-making, making the analysis results more aligned with practical needs.

[0068] Specifically, step 102 collects real-time data on the experts' interactive behaviors on the collaborative platform, analyzes and generates each expert's analysis strategy, and calculates a weight coefficient for each strategy based on the expert's characteristic data (such as experience and expertise) and the relevance of the strategy to the task objective. This transforms the expert's subjective opinions into quantitative, structured analysis strategies and weights, achieving a scientific expression of expert knowledge. This not only provides an objective and reliable data foundation for subsequent strategy integration, but also reflects the importance of each expert in decision-making through weight calculation, ensuring the authority and pertinence of the analysis plan, thereby significantly improving the scientific nature of fatigue damage analysis for industrial reducers.

[0069] Step 104, based on the multidimensional decision matrix, integrates the analysis strategies of multiple experts through local matrix extraction and collaborative value assessment to generate a target simulation plan for the current round of finite element simulation. The technical benefit lies in selecting highly important and complementary strategy combinations by setting constraints (such as weight thresholds, diversity, and differentiation), ensuring that the integrated simulation plan is both comprehensive and efficient. This approach effectively integrates the collective wisdom of experts, optimizes the design of simulation plans, and significantly improves the scientificity and accuracy of fatigue damage analysis for industrial reducers. It also provides a high-quality basis for the execution of finite element simulations.

[0070] After completing the current round of finite element simulation in step 106, the system returns to step 101, recollects the experts' collaborative interaction data, generates a new simulation plan, and enters the next round of iteration until the simulation results meet the preset analysis termination criteria (such as the fatigue life prediction error being less than a certain threshold). Through multiple rounds of dynamic iteration, the analysis process is continuously optimized, and the simulation results gradually approach the collaborative task objectives. This cyclical mechanism ensures that the analysis can adapt to the real-time feedback and adjustments of experts, significantly improving the accuracy and efficiency of fatigue damage analysis for industrial reducers and achieving continuous improvement and optimization of analysis.

[0071] In some embodiments, in the above 105, the execution rhythm is dynamically adjusted during the execution process to adapt to the real-time rhythm of the experts' collaborative discussion and decision-making, such as Figure 4 As shown, including: 501. Obtain the currently remaining fatigue damage analysis tasks of the finite element analysis model.

[0072] The purpose of this step is to identify the unfinished work in the finite element analysis model during the current round, providing a foundation for subsequent pacing adjustments. The remaining tasks are the unexecuted parts of the target simulation solution generated in step 104, reflecting the progress of the analysis. Examples include stress distribution calculation and crack growth simulation.

[0073] Here is an example implementation: The target simulation solution generated in step 104 includes the following three main tasks: High-density mesh stress analysis (for gear meshing areas); Material property optimization analysis (for bearing materials); Crack growth analysis (for gear surfaces).

[0074] During the execution process, the high-density mesh stress analysis has been completed by 50%, and the remaining two tasks have not yet started. The system obtains the current remaining tasks by querying the execution log of the finite element analysis model as follows: High-density mesh stress analysis of the remaining part (calculation of stress distribution in the remaining mesh area); Material property optimization analysis (complete task, not yet executed); Crack Growth Analysis (Full Task, not yet performed).

[0075] This task information is stored as a task list, with each task accompanied by an estimate of the computational effort (such as the number of mesh elements and the number of iterations) and the estimated time required (for example, the remaining part of the stress analysis will take 30 minutes).

[0076] 502. Predict future trends in the pace of collaborative discussion and decision-making among experts.

[0077] This step analyzes the historical and current cadence of expert discussions to predict future trends in the cadence of expert collaboration. This cadence reflects potential changes in the frequency, depth, and focus of expert discussions, providing a key basis for dynamically adjusting the execution cadence.

[0078] Step 502 specifically includes the following sub-steps: Capture the historical rhythm of collaborative discussions and decision-making among experts.

[0079] Historical cadence data is a quantitative metric extracted from interaction records on the expert collaboration platform. It includes discussion frequency (e.g., number of speeches per minute), decision speed (e.g., interval between parameter adjustments), and duration of focus on a topic. This data reflects the behavioral patterns of experts in past analyses.

[0080] Here's an example implementation: During the past two hours of analysis, the system records that: Expert A made a mesh adjustment suggestion every 10 minutes (a frequency of 0.1 times / minute); Expert B adjusted material parameters every 15 minutes (a frequency of 0.067 times / minute); and Expert C mentioned crack growth issues every 5 minutes (a frequency of 0.2 times / minute). This historical rhythm data is stored as a time series, annotated with the timestamp and content of each interaction.

[0081] Based on the real-time rhythm and historical rhythm, predict the future rhythm change trend of experts' collaborative discussion and decision-making.

[0082] Combining real-time and historical cadences, statistical analysis or machine learning models (collecting extensive data on the changing cadences of discussions and decisions made by experts during collaborative fatigue damage analysis using finite element analysis models for industrial reducers, and using this data for machine learning training) can be used to infer future cadences. For example, if the real-time cadence shows an increase in discussion frequency, historical trends can be used to predict whether this increase will continue.

[0083] Continuing with the previous implementation example, the current real-time cadence shows that Expert C's discussion frequency has increased to 0.3 times / minute (once every 3 minutes) over the past 10 minutes, indicating a concern about crack expansion. Experts A and B's frequency remains stable. Combining historical data (Expert C's highest frequency was 0.2 times / minute), the system uses a machine learning model to predict that Expert C's discussion frequency is likely to rise to 0.35 times / minute and stabilize within the next 30 minutes, while Experts A and B's frequency remains low. This prediction indicates that crack expansion will become a hot topic of discussion in the short term.

[0084] 503. Based on the current remaining fatigue damage analysis tasks, future rhythm change trends and real-time rhythm, planning experts collaboratively discuss and decide the best next rhythm.

[0085] This step plans a cadence target that is optimal for the next phase of expert collaboration, based on the complexity of the remaining tasks, the predicted cadence trend, and the current discussion cadence. The optimal next cadence aims to balance computational efficiency with expert decision-making requirements, ensuring timely simulation results to support discussions while remaining neither too fast nor too slow, resulting in wasted resources or disconnected collaboration.

[0086] Step 503 specifically includes the following sub-steps: Based on the current remaining fatigue damage analysis tasks, future rhythm change trends and real-time rhythm, a third feature description vector is constructed.

[0087] The third feature description vector is a multidimensional data structure that integrates task characteristics (such as computational effort, priority), prediction rhythm (such as discussion frequency trend), and real-time rhythm (such as current speaking frequency) to characterize the comprehensive state of the current analysis scenario.

[0088] Continuing with the above implementation example, the remaining tasks include high-density mesh stress analysis (medium computational complexity), material property optimization analysis (high computational complexity), and crack growth analysis (medium computational complexity). The predicted cadence shows that Expert C's discussion frequency has increased to 0.35 times / minute (focusing on crack growth). The real-time cadence is 0.3 times / minute for Expert C, while Experts A and B have lower frequencies. The third feature description vector is constructed as: [Task Complexity = Medium to High, Predicted Frequency = 0.35, Real-Time Frequency = 0.3, Focus = Crack Growth].

[0089] Based on the optimal rhythm library, the optimal next rhythm for expert collaborative discussion and decision-making is matched according to the third feature description vector.

[0090] The optimal cadence library is a pre-trained database that stores the mapping between different feature vectors and recommended cadences. A cadence includes discussion intervals (e.g., every 5 minutes) and focus allocation (e.g., 50% of the time discussing crack expansion). By matching these vectors, the optimal cadence is found.

[0091] Continuing with the previous implementation example, the third feature description vector is input into the optimal cadence library. The matching result indicates that the optimal next cadence is a discussion every 4 minutes (a frequency of 0.25 times / minute), with 60% of the time focused on crack growth, 30% on stress analysis, and 10% on material optimization. This cadence is slightly lower than the predicted frequency (0.35) to avoid expert fatigue while prioritizing responses to crack growth.

[0092] 504. Perform task unit splitting and serialization processing on the currently remaining fatigue damage analysis task to obtain a subtask unit sequence.

[0093] This step breaks down the remaining fatigue damage analysis tasks into smaller subtask units and arranges them in a sequential order, forming a sequenced subtask list. The granularity of task decomposition should be sufficiently fine-grained to support flexible pacing adjustments while maintaining the consistency of the computational logic. Sequencing ensures that subtasks are executed in an orderly manner based on priority and dependencies.

[0094] Continuing with the above implementation example, the remaining fatigue damage analysis tasks are decomposed as follows: High-density mesh stress analysis: Subtask Unit 1: Refine the mesh of the gear meshing area (1000 units); Subtask Unit 2: Calculate the stress distribution in the area; Subtask Unit 3: Generate stress visualization diagram.

[0095] Material property optimization analysis: Subtask Unit 4: Adjust the elastic modulus of the bearing material (210GPa); Subtask Unit 5: Run thermal stress simulation; Subtask Unit 6: Evaluate material fatigue life.

[0096] Crack growth analysis: Subtask unit 7: Set the initial crack parameters (length 0.1mm); Subtask Unit 8: Simulate crack growth for 10 iterations; Subtask unit 9: Output crack propagation path.

[0097] The serialization result is a subtask unit sequence: [subtask unit 1, subtask unit 2, subtask unit 3, subtask unit 4, subtask unit 5, subtask unit 6, subtask unit 7, subtask unit 8, subtask unit 9], where each subtask unit is arranged in the logical order of stress analysis, material optimization, and crack propagation.

[0098] 505. Select the first N subtask units from the subtask unit sequence, where N is a positive integer, and the value of N is selected so that when the finite element analysis model sequentially executes the selected subtask units, the expected actual rhythm of guiding the experts to collaboratively discuss and make decisions is closest to the optimal next rhythm.

[0099] Select the first N subtasks (N is a positive integer) from the subtask sequence so that the finite element analysis model's execution of these subtasks leads the experts' actual discussion rhythm to approach the optimal next rhythm planned in step 503. N is selected based on the matching of the subtask duration and the discussion focus, ensuring that the simulation progress is synchronized with the experts' needs.

[0100] Continuing with the above implementation example: The optimal next-step cadence is a discussion every 4 minutes, with 60% focus on crack propagation. The estimated subtask unit times are: Subtask Unit 1 (5 minutes), Subtask Unit 2 (10 minutes), Subtask Unit 7 (3 minutes), and Subtask Unit 8 (8 minutes). To match the cadence, select N = 3, and the subtask units are [Subtask Unit 1, Subtask Unit 2, Subtask Unit 7]: Subtask Units 1 and 2 provide stress data (15 minutes) and support 30% stress discussion; Subtask unit 7 initiates crack growth (3 minutes), accounting for 60% focus.

[0101] The total time was 18 minutes, with results output every 6 minutes on average, close to a 4-minute rhythm. After adjustment, experts were guided to discuss every 5 minutes.

[0102] 506. Dynamically adjust the execution rhythm, including: controlling the finite element analysis model to execute the selected first N subtask units in sequence within the next unit rhythm time.

[0103] Control the finite element analysis model to sequentially execute the selected N subtasks within the next unit of time (e.g., the next hour). By adjusting the execution speed and output interval, the experts' discussion rhythm is dynamically influenced to align with the optimal next rhythm. This step is the execution phase of rhythm adjustment.

[0104] Continuing the above implementation example: Select [Subtask Unit 1, Subtask Unit 2, Subtask Unit 7], and the next unit rhythm time is 30 minutes. System Settings: Subtask unit 1: Complete in 5 minutes and output grid data; Subtask Unit 2: Complete in 10 minutes and output stress diagram; Subtask unit 7: Completed in 3 minutes, output crack parameters.

[0105] The execution cadence was adjusted to produce results every 6 minutes (slightly slower than the 4-minute pace to accommodate expert response times), resulting in a total of 18 minutes for completion, with the remaining time for buffer discussion. After the experts received the results, they discussed them every 5 minutes, approaching the optimal cadence.

[0106] Overall, this method obtains the remaining fatigue damage analysis tasks, predicts the future cadence of expert collaborative discussions and decision-making, plans the optimal next cadence, and then decomposes and serializes the tasks into units. It selects the top N subtask units to guide the experts' actual discussion cadence closer to the optimal cadence, ultimately controlling the finite element analysis model to execute the selected subtasks within the next unit of time. This ensures that the simulation process can promptly respond to the experts' latest discussions and decisions, improving the real-time nature of the analysis and collaborative efficiency. Furthermore, refined cadence adjustments enhance the synchronization of simulation results with expert opinions.

[0107] Specifically, step 505 selects the first N subtask units (N is a positive integer) from the subtask unit sequence. By controlling the pacing of the finite element analysis model as it executes these subtasks, the experts' actual discussion and decision-making rhythm is guided to approach the pre-planned optimal next rhythm. This meticulous task segmentation and dynamic pacing ensures that the simulation process remains synchronized with the experts' real-time collaborative needs. This approach not only ensures that simulation results promptly respond to the experts' discussion focus and decision-making changes, but also enhances the real-time nature and collaborative efficiency of the analysis process, thereby improving the overall effectiveness and synergy of fatigue damage analysis for industrial reducers.

[0108] An embodiment of the present application provides an industrial reducer fatigue damage analysis system based on finite element analysis, including a processor and a memory, wherein a computer program is stored on the memory, and when the computer program is executed by the processor, the steps of any one of the above-mentioned industrial reducer fatigue damage analysis methods based on finite element analysis are implemented.

[0109] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. The fatigue damage analysis method of industrial reducer based on finite element analysis is characterized by: include:

101. On the expert collaboration platform, real-time collection of collaborative interaction behavior data generated by experts for the finite element analysis model of industrial reducers; 102. For each expert, generate the analysis strategy currently advocated by the expert based on the collaborative interaction behavior data analysis, and calculate the weight coefficient of the analysis strategy by combining the preset expert feature data and the correlation between the analysis strategy and the collaborative task goal; 103. Aggregate all experts’ analysis strategies and their weight coefficients to construct a multi-dimensional decision matrix; 104. Perform strategy fusion processing based on the multi-dimensional decision matrix to generate the target simulation plan for the current round of finite element simulation; 105. Control the finite element analysis model to execute fatigue damage analysis of the target simulation scheme, and dynamically adjust the execution rhythm during the execution process to adapt to the real-time rhythm of collaborative discussion and decision-making among experts; 106. After the current round of finite element simulation is completed, return to step 101 to perform the next round of operations until the analysis termination condition is met.

2. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: In the step 102, generating the analysis strategy currently advocated by the expert based on the analysis of the collaborative interaction behavior data includes: extracting features of target data related to the expert in the collaborative interaction behavior data to obtain a plurality of first feature values; Constructing a first feature description vector based on each first eigenvalue; Based on the preset analysis strategy library, the analysis strategy currently advocated by the expert is matched according to the first feature description vector.

3. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: In step 102, the weight coefficient of the analysis strategy is calculated by combining the preset expert feature data and the correlation between the analysis strategy and the collaborative task goal, including: Based on the preset quantitative system, the authority of the analysis strategy represented by the expert feature data and the degree of relevance are quantified respectively; Based on the preset weights, the authority and relevance are weighted to obtain the weight coefficient of the analysis strategy.

4. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: The step 104 of performing strategy fusion processing based on the multi-dimensional decision matrix to generate a target simulation plan for the current round of finite element simulation includes: Based on the local matrix interception constraint, multiple local matrices are intercepted from the multidimensional decision matrix; For each local matrix, feature extraction is performed on the local matrix to obtain multiple second eigenvalues. Based on each second eigenvalue, a second feature description vector is constructed. Based on the preset strategy collaborative execution value library, the value degree of collaborative execution of each analysis strategy in the local matrix is ​​matched according to the second feature description vector; Aggregate the analysis strategies in the local matrix with the largest value to obtain the target simulation solution for the current round of finite element simulation; The local matrix truncation constraints include: The proportion of analysis strategies whose weight coefficients in the local matrix exceed a preset coefficient threshold exceeds a preset proportion threshold; The analysis purposes of each analysis strategy in the same local matrix are different; the overall difference between the analysis strategies in the two local matrices exceeds the preset difference threshold.

5. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: In the above 105, the execution rhythm is dynamically adjusted during the execution process to adapt to the real-time rhythm of the experts' collaborative discussion and decision-making, including: Obtain the remaining fatigue damage analysis tasks of the finite element analysis model; Predict future trends in the pace of collaborative discussion and decision-making among experts; Planning experts collaboratively discuss and decide on the best next rhythm based on the current remaining fatigue damage analysis tasks, future rhythm change trends, and real-time rhythm; Split and serialize the remaining fatigue damage analysis tasks into task units to obtain a subtask unit sequence; Select the first N subtask units from the subtask unit sequence; where N is a positive integer, and the value of N is selected so that when the finite element analysis model sequentially executes the selected subtask units, the expected actual rhythm of guiding the experts to discuss and make decisions together is closest to the optimal next rhythm; Dynamically adjusting the execution rhythm includes: controlling the finite element analysis model to sequentially execute the selected first N subtask units within the next unit rhythm time.

6. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 5, characterized in that: The predicted future rhythm of changes in the experts' collaborative discussions and decisions includes: Obtain the historical rhythm of collaborative discussions and decision-making among experts; Based on the real-time rhythm and historical rhythm, predict the future rhythm change trend of experts' collaborative discussion and decision-making.

7. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 5, characterized in that: The optimal next rhythm is discussed and decided by planning experts in a collaborative manner based on the remaining fatigue damage analysis tasks, future rhythm change trends, and the real-time rhythm, including: Based on the current remaining fatigue damage analysis tasks, future rhythm change trends and real-time rhythm, a third feature description vector is constructed; Based on the optimal rhythm library, the optimal next rhythm for expert collaborative discussion and decision-making is matched according to the third feature description vector.

8. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: The step of acquiring the real-time rhythm of collaborative discussion and decision-making of the experts includes: determining the real-time rhythm of collaborative discussion and decision-making of the experts based on the updated collaborative interaction behavior data collected in real time.

9. The fatigue damage analysis method for industrial reducers based on finite element analysis according to claim 1, characterized in that: The analysis termination conditions include: The simulation results of multiple rounds of historical finite element simulations are consistent with the collaborative mission goals.

10. The fatigue damage analysis system for industrial reducers based on finite element analysis is characterized by: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the fatigue damage analysis method of an industrial reducer based on finite element analysis are implemented as described in any one of claims 1 to 9.

Images