Adaptive optimization method for servo driver based on multi-source heterogeneous data fusion

Abstract

The invention belongs to the technical field of servo drive control, and provides a multi-source heterogeneous data fusion servo driver adaptive optimization method, which comprises the following steps: dividing working condition groups by clustering multi-dimensional working condition features, determining a reference interval of each working condition group, positioning a current working condition, identifying unqualified parameters, and determining a reference interval of each working condition group; key parameters are screened through the fault influence indexes, mismatch indexes are calculated according to the adaptive optimization parameter library and the proportion of unqualified parameters in the key parameters, whether the mismatch phenomenon exists between the parameter library and the current working condition or not is judged, and if yes, the mismatch index is calculated; the optimization priority of the unqualified parameters is determined according to the fault influence index and the deviation between the adaptive interval and the reference interval, the unqualified parameters are adjusted in stages according to the priority, dynamic optimization of the parameter library is finally achieved, the problems that in traditional adaptive optimization, the parameter interval is too wide and the like are solved, and the optimization efficiency is improved. And the self-adaptive capability and the operation reliability of the servo driver under complex working conditions are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of servo drive control, and particularly relates to a servo driver adaptive optimization method based on multi-source heterogeneous data fusion. BACKGROUND

[0002] As a core component of industrial automation equipment, the control precision and stability of a servo driver directly affect production efficiency and product quality. With the complication of industrial scenes, the servo driver needs to adapt to dynamic environments such as multiple working conditions and variable loads, and the traditional fixed parameter control or simple adaptive optimization method has been difficult to meet the demand.

[0003] The existing adaptive optimization method has the following deficiencies: the influence of working condition difference on parameters is not fully considered, parameters under different working conditions are mixed together, the optimization interval is too wide or too narrow, the adaptability is poor, the fusion and utilization of multi-source heterogeneous data (such as different types and dimensions of data such as current, speed and temperature) are insufficient, the parameter adjustment lacks comprehensive data support, the parameter interval rationality judgment relies on experience, unqualified parameters are not identified through quantitative analysis, the optimization efficiency is low, the importance of parameters is not distinguished, there is no priority during adjustment, which easily leads to the lag of key parameter optimization and causes system stability risk, and in the long run, it may cause faults, seriously affecting the normal work and service life of the servo driver.

[0004] Therefore, the application provides a servo driver adaptive optimization method based on multi-source heterogeneous data fusion. SUMMARY

[0005] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background.

[0006] The technical scheme adopted by the application to solve the technical problems is: a servo driver adaptive optimization method based on multi-source heterogeneous data fusion, comprising: Step 1: using the K-Means algorithm to cluster multi-dimensional working condition characteristics to obtain multiple working condition groups, and determining the servo driver parameter reference interval of each working condition group; Step 2: performing similarity analysis on the current working condition characteristics and the working condition characteristics of each working condition group, identifying the working condition group to which the current working condition belongs and the corresponding servo driver parameter reference interval, and performing deviation analysis on the adaptive interval and the reference interval of each servo driver parameter in the adaptive optimization parameter library to identify unqualified parameters in the adaptive optimization parameter library; Step 3: calculating the fault influence index of each servo driver parameter, identifying the key parameters in the servo driver parameters, and performing frequency analysis on the unqualified parameters and the key parameters to determine whether there is a mismatch between the adaptive optimization parameter library and the current working condition; Step four: if there is, according to the deviation of the adaptive interval of the unqualified parameter from the reference interval and the fault influence index, determine the optimization priority of the unqualified parameter; Step five: adjust the adaptive interval of the unqualified parameter according to the optimization priority, and update the adjusted parameter interval to the adaptive optimization parameter library.

[0007] Further, the process of clustering the multi-dimensional working condition characteristics is: Calculate the Pearson coefficient of each working condition variable and the servo driver parameter, and compare it with the preset correlation coefficient to select the working condition variable with a Pearson correlation coefficient greater than the preset correlation coefficient, which is standardized as a multi-dimensional working condition characteristic; Calculate the average contour coefficient of different K values using the contour coefficient method, and select the K with the largest average contour coefficient as the clustering number; Randomly select K cluster centers, calculate the Euclidean distance of each sample to the K centers, and assign the sample to the cluster with the nearest Euclidean distance; After all samples are assigned to clusters, recalculate the average value of all sample feature values in each cluster as the new cluster center; Repeat the sample assignment and cluster center update, record the change value of each cluster center, and compare it with the preset threshold value, when the change value is less than the preset threshold value, the clustering is completed, and each cluster is a working condition group.

[0008] Further, the acquisition method of the servo driver parameter reference interval of each working condition group is: Collect the historical data of the servo driver parameters of the servo driver under normal operation in each working condition group as the normal value of the servo driver parameter; Based on any one working condition group; For any one servo driver parameter; Calculate the mean μ and standard deviation σ of the normal value of the servo driver parameter, and use the "±3σ" principle to take [μ-3σ, μ+3σ] as the reference interval of each servo driver parameter in the working condition group.

[0009] Further, the way to determine the working condition group to which the current working condition belongs and the corresponding servo driver parameter reference interval is: Obtain the key features of the current working condition through the sensor, and calculate the Euclidean distance between the current working condition features and the features of each working condition group after standardization; Compare the Euclidean distance with the threshold value, if the Euclidean distance is less than the threshold value, mark the working condition group as a similar working condition group; If there are multiple similar working condition groups, select the working condition group with the smallest Euclidean distance as the working condition group to which the current working condition belongs; The servo driver parameter reference interval corresponding to the current working condition is the servo driver parameter reference interval of the working condition group to which the current working condition belongs.

[0010] Further, the process of identifying unqualified parameters in the adaptive optimization parameter library is: Mark the interval of each servo driver parameter in the adaptive optimization parameter library as the adaptive interval of the servo driver parameter; Calculate the width of the servo driver parameter reference interval and the adaptive interval respectively, and compare them. If the adaptive interval width of the servo driver parameter is greater than the reference interval width, calculate the interval width deviation of the servo driver parameter; Compare the interval width deviation with the preset deviation. If the interval width deviation is greater than the preset deviation, mark the servo driver parameter as an unqualified parameter.

[0011] Further, the way of identifying key parameters in the servo driver parameter is: For any one servo driver parameter; Get the parameter historical fault frequency and the average repair time of the fault caused by the servo driver parameter anomaly, and calculate the fault trigger frequency and repair time proportion; Add the fault trigger frequency and the repair time proportion to obtain the fault impact index; Compare the fault impact index with the preset impact index. If the fault impact index is greater than the preset impact index, mark the servo driver parameter as a key parameter.

[0012] Further, the calculation method of the fault trigger frequency and the repair time proportion is: Calculate the frequency of the servo driver parameter anomaly in the historical fault, and calculate the proportion with the total historical fault times to obtain the fault trigger frequency; Get the average repair time of each servo driver parameter anomaly causing fault, and sum it up to obtain the total repair time; Calculate the average repair time of the servo driver parameter and the total repair time to obtain the repair time proportion.

[0013] Further, the judgment method of whether the adaptive optimization parameter library and the current working condition exist mismatch phenomenon is: Calculate the number of unqualified parameters in the adaptive optimization parameter library, and calculate the proportion with the total number of parameters in the adaptive optimization parameter library to obtain the unqualified parameter proportion; Calculate the proportion of unqualified parameters in key parameters to obtain the key unqualified parameter proportion; Add the unqualified parameter proportion and the key unqualified parameter proportion to obtain the mismatch index; The mismatch index is compared with a preset threshold value, and if the mismatch index is greater than the preset threshold value, the adaptive optimization parameter library is mismatched with the current working condition.

[0014] Further, the determination manner of the adjustment priority of the unqualified parameter is: For any one unqualified parameter; The priority index is obtained by adding the interval width deviation and the fault influence index, and the unqualified parameters are arranged in descending order of the priority index to obtain the optimization priority of the unqualified parameters.

[0015] Further, the process of adjusting the adaptive interval of the unqualified parameter according to the optimization priority is: For any one unqualified parameter; The adjustment period and the adjustment ratio are set, and in each adjustment period, the upper limit and the lower limit of the unqualified parameter adaptive interval are adjusted at the same time according to the adjustment ratio; After each adjustment period ends, the interval width deviation of the unqualified parameter adaptive interval after adjustment and the reference parameter interval is calculated and compared with a preset deviation; If the interval width deviation is less than the preset deviation, the adjustment is completed.

[0016] The beneficial effects of the present application are as follows: by K-Means clustering, the working conditions with similar running states are grouped into a group, avoiding parameter analysis distortion caused by working condition difference, the reference interval is constructed based on the "±3σ" principle, providing a reliable working condition reference for optimization, the current working condition and the corresponding reference interval are located through similarity analysis, the unqualified parameters are identified through interval width deviation, avoiding indiscriminate adjustment, concentrating resources on key problems, quantifying parameter interval rationality, improving optimization targeting, distinguishing parameter importance, screening key parameters through fault influence index, judging parameter library adaptability combined with mismatch index, ensuring key parameter optimization priority, reducing fault probability caused by core parameter abnormality, dynamically adjusting to ensure system stability, adjusting parameter interval in cycles and according to priority, avoiding parameter mutation impact on the system, ensuring dynamic stability during the adjustment process, adapting to real-time changes under complex working conditions, solving the problem of excessively wide parameter interval in traditional adaptive optimization, and improving the adaptive ability and running reliability of the servo driver under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0017] The present application will be further described below with reference to the accompanying drawings.

[0018] Figure 1 is a step flow chart of a multi-source heterogeneous data fusion servo driver adaptive optimization method according to embodiment 1 of the present application; Figure 2The logical judgment diagram of the multi-source heterogeneous data fusion servo driver adaptive optimization method is shown in Figure 1. DETAILED DESCRIPTION

[0019] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application will be further described below in conjunction with specific embodiments.

[0020] Embodiment 1: Please refer to Figure 1 Figure 1 The multi-source heterogeneous data fusion servo driver adaptive optimization method comprises the following steps: Step 1: using the K-Means algorithm to cluster the multi-dimensional working condition characteristics to obtain multiple working condition groups, and determining the servo driver parameter reference interval of each working condition group; In step 1, the multi-dimensional working condition characteristics refer to working condition variables that have strong correlation with servo driver parameters, for example: Mechanical processing equipment: load torque (N) ), spindle speed (r / min), and environmental temperature (℃); Conveyer belt system: conveying speed (m / s), material weight (kg), and belt tension (N); Calculate the Pearson coefficient of each working condition variable and servo driver parameter, and compare it with the preset correlation coefficient to screen the working condition variables with Pearson correlation coefficient greater than the preset correlation coefficient, and then standardize them as multi-dimensional working condition characteristics; Calculate the average silhouette coefficient of different K values using the silhouette coefficient method, and select the K with the largest average silhouette coefficient as the clustering number; Randomly select K cluster centers, calculate the Euclidean distance of each sample to the K centers, and assign the sample to the cluster with the nearest Euclidean distance; After all samples are assigned to clusters, recalculate the average value of all sample characteristic values of each cluster as the new cluster center; Repeat the sample assignment and cluster center update, record the change value of each cluster center, and compare it with the preset threshold value, when the change value is less than the preset threshold value, the clustering is completed; Each cluster center is a working condition group; For example, assume that the physical quantities in the mechanical processing equipment that have strong correlation with servo driver parameters (such as PID parameters and current loop parameters) are selected, including load torque (N) ), spindle speed (r / min), and environmental temperature (℃), calculate the Pearson correlation coefficient of each physical quantity and parameter, screen the characteristics with coefficient > 0.6 (load torque: 0.82; spindle speed: 0.75; environmental temperature: 0.63), and use them as multi-dimensional working condition characteristics; The screened feature data is standardized to eliminate dimensional differences, and the sliding window mean method is used to filter noise to ensure data validity. The average profile coefficient when K=2~6 is calculated by the profile coefficient method, and the coefficient is maximum (0.78) when K=3, and the cluster number is determined as 3; Three cluster centers are randomly selected initially, the Euclidean distance of each sample to the center is calculated, and the sample is classified into the nearest cluster; The cluster center is iteratively updated (the average value of the feature of each sample is calculated), until the center change value is less than 0.01 (preset threshold), the clustering is completed, and three working condition groups are obtained: Group 1 (light load-low speed): load torque [5, 15] , main shaft speed [500, 1500] r / min, and environment temperature [20, 25] ℃; Group 2 (medium load-medium speed): load torque [15, 30] , main shaft speed [1500, 3000] r / min, and environment temperature [25, 30] ℃; Group 3 (heavy load-high speed): load torque [30, 50] , main shaft speed [3000, 5000] r / min, and environment temperature [30, 35] ℃; In step one, the determination process of the servo driver parameter reference interval of each working condition group includes: Collect the target parameter historical data of the servo driver under normal operation in each working condition group as the normal value of the servo driver parameter; Based on any one working condition group; For any one servo driver parameter; Calculate the mean μ and standard deviation σ of the normal value of the servo driver parameter, and use the “±3σ” principle to take [μ-3σ, μ+3σ] as the reference interval of each servo driver parameter in this working condition group; It should be noted that the principle of determining the servo driver parameter reference parameter interval by the “±3σ” principle is to combine statistical rules with engineering practice requirements, and the core logic is to frame a parameter interval that can cover most normal working conditions through the distribution characteristics of historical normal data, which avoids excessive strictness leading to misjudgment and prevents the interval from being too wide and burying risks; The role of clustering to divide the working condition groups and determine the servo driver parameter reference interval of each working condition group is: Action 1: Cluster the working conditions with similar running characteristics into groups to ensure that the core conditions such as the running environment and load characteristics of the servo driver in the same working condition group are highly consistent, provide a unified working condition background for parameter analysis, avoid misjudgment of normal fluctuations of parameters in different working conditions as abnormal, or meaningless widening of the parameter reference interval due to mixed working conditions, and fundamentally ensure the pertinence and accuracy of parameter analysis; Action 2: Calculate the reference interval based on the historical normal data of the same working condition group through the "±3σ" principle, so that the reasonable range of each parameter is strongly bound to the specific working condition; Action 3: Set the parameter reference by working condition to avoid over-optimization, response lag or system oscillation caused by mismatch between the parameter interval and the actual working condition; Step two: Similarity analysis of the current working condition characteristics and the working condition characteristics of each working condition group to determine the working condition group to which the current working condition belongs and the corresponding servo driver parameter reference interval, and deviation analysis of the adaptive interval and the reference interval of each servo driver parameter in the adaptive optimization parameter library to identify unqualified parameters in the adaptive optimization parameter library; In step two, the process of similarity analysis of the current working condition characteristics and the working condition characteristics of each working condition group includes: Obtain the key characteristics of the current working condition through sensors, and calculate the Euclidean distance between the current working condition characteristics and the characteristics of each working condition group after standardization; Compare the Euclidean distance with the threshold value, if the Euclidean distance is less than the threshold value, mark the working condition group as a similar working condition group; If there are multiple similar working condition groups, select the working condition group with the smallest Euclidean distance as the working condition group to which the current working condition belongs; In step two, the servo driver parameter reference interval corresponding to the current working condition is the servo driver parameter reference interval of the working condition group to which the current working condition belongs; In step two, the identification process of unqualified parameters includes: Mark the interval of each parameter in the adaptive optimization parameter library as the adaptive interval of the parameter; Calculate the width of the parameter reference interval and the adaptive interval respectively, and compare them, if the adaptive interval width of the parameter is greater than the reference interval width, calculate the interval width deviation of the parameter, the formula is ; It can be understood that the physical meaning of the interval width deviation is that the difference between the adaptive interval width and the reference interval width accounts for the degree of deviation of the adaptive interval from the reasonable range, the greater the value, the more unreasonable the adaptive interval; Compare the interval width deviation of the parameter with the preset deviation, if the interval width deviation is greater than the preset deviation, mark the parameter as an unqualified parameter; The role of identifying unqualified parameters is: Effect 1: By identifying unqualified parameters, the optimization resources are concentrated on the parameters with excessively wide intervals, avoiding indiscriminate adjustment and improving optimization efficiency; Effect 2: Timely finding parameters with excessively wide adaptive intervals (which may lead to over-optimization or failure), and reducing system failures caused by parameter abnormalities from the source; Step three: Calculate the fault impact index of each servo drive parameter, identify the key parameters in the servo drive parameters, and perform frequency analysis on unqualified parameters and key parameters to determine whether there is a mismatch between the adaptive optimization parameter library and the current working condition; Please refer to Figure 2 In step three, the identification process of the key parameters includes: For any servo drive parameter: Calculate the frequency of abnormality of the parameter in historical faults, and calculate the fault trigger frequency by proportion with the total number of historical faults; Get the average repair time of each parameter abnormality leading to failure, and sum up to get the total repair time; Calculate the average repair time of the parameter and the total repair time by proportion to get the repair time proportion; Add the fault trigger frequency and the repair time proportion to get the fault impact index; It can be understood that the physical meaning of the fault impact index is: comprehensively quantifying the destructive degree of a single parameter abnormality to the normal operation of the servo drive system, and the numerical value directly reflects the key degree of the parameter in the system. The higher the value, the greater the threat of parameter abnormalities to the system, which needs to be monitored and controlled first; Compare the fault impact index with the preset impact index. If the fault impact index is greater than the preset impact index, mark the parameter as a key parameter; In step three, the process of determining whether there is a mismatch between the adaptive optimization parameter library and the current working condition includes: Calculate the number of unqualified parameters in the adaptive optimization parameter library, and calculate the unqualified parameter proportion by proportion with the total number of parameters in the adaptive optimization parameter library; Calculate the proportion of unqualified parameters in key parameters to get the key unqualified parameter proportion; Add the unqualified parameter proportion and the key unqualified parameter proportion to get the mismatch index; It can be understood that the physical meaning of the mismatch index is: quantitatively evaluating the overall mismatch degree of the adaptive parameter library and the current working condition, reflecting the rationality and effectiveness of the parameter range setting in the parameter library, and the numerical value directly reflects the support ability of the parameter library to the normal operation of the equipment. The lower the value, the more matched the parameter library is to the actual working condition, the higher the system stability, and vice versa, which needs to optimize the parameter library first; The proportion of unqualified parameters reflects the basic quality of the parameter library. The higher the proportion of unqualified parameters, the worse the overall quality of the parameter library, and the parameter library needs to be calibrated comprehensively. The proportion of key unqualified parameters reflects the support capability of the parameter library for the core function. The higher the proportion of key unqualified parameters, the worse the adaptability of the parameter library to the key function, and the greater the potential risk. The mismatch index is compared with the preset threshold. If the mismatch index is greater than the preset threshold, the adaptive optimization parameter library is mismatched with the current working condition. The function of determining whether the adaptive optimization parameter library is mismatched with the current working condition is to: The overall quality of the adaptive optimization parameter library is quantitatively evaluated, the potential risk of the system is identified, and a decision basis is provided for whether to start the parameter library optimization. Step four: If there is, the optimization priority of the unqualified parameter is determined according to the deviation of the adaptive interval of the unqualified parameter from the reference interval and the fault influence index. In step four, the determination process of the optimization priority of the unqualified parameter includes: For any unqualified parameter: The interval width deviation and the fault influence index are added to obtain a priority index. The unqualified parameters are arranged in descending order of the priority index to obtain the optimization priority of the unqualified parameters. It can be understood that the physical meaning of the priority index is that the priority index is a core index for comprehensively quantifying the optimization urgency and necessity of a single parameter under the current working condition, and the numerical value directly reflects the urgency of the parameter needing to be adjusted first. The function of determining the parameter optimization priority is: On the premise that it has been identified that the adaptive optimization parameter library is mismatched with the current working condition, the most urgent object to be adjusted is selected from a large number of parameters to be optimized, a clear sequence guidance is provided for subsequent optimization actions, resource waste is avoided, and the potential risk is minimized. Step five: The adaptive interval of the unqualified parameter is adjusted according to the optimization priority, and the adjusted parameter interval is updated to the adaptive optimization parameter library. In step five, the process of adjusting the adaptive interval of the unqualified parameter according to the optimization priority includes: An adjustment period and an adjustment ratio are set. In each adjustment period, the upper limit and the lower limit of the adaptive interval of the unqualified parameter are adjusted simultaneously according to the adjustment ratio. After each adjustment period ends, the interval width deviation of the adjusted adaptive interval from the reference interval is calculated and compared with the preset deviation. If the interval width deviation is less than the preset deviation, the adjustment of the unqualified parameter is stopped, and the finally obtained parameter interval is updated to the adaptive optimization parameter library. When the adaptive intervals of all unqualified parameters in the adaptive optimization parameter library are adjusted and updated, the adjustment is completed, and the adjustment is ended; It should be noted that the core goal of adjusting the adaptive parameter interval in different periods and stages is to gradually reduce the deviation of the parameter interval from the benchmark under the premise of ensuring system stability. The role of adjusting the adaptive interval of unqualified parameters is: Effect 1: By adjusting the adaptive interval of unqualified parameters to the benchmark interval, the problem of parameter deviation from the standard is directly solved, avoiding system failure, performance degradation or incorrect decision caused by abnormal parameters, and providing basic guarantee for stable operation of the system; Effect 2: Avoiding sudden changes in parameter interval, system usually has an adaptive threshold for parameter changes, sudden changes may trigger a chain reaction, phased adjustment combined with deviation check can make parameter interval gradually approach the benchmark, reduce the impact on the current running state of the system, and maintain dynamic stability during the adjustment process; Effect 3: Adjusting according to the optimization priority order ensures that resources (such as computing power and adjustment opportunity) are tilted to more critical parameters, and parameters that have a greater impact on the core function of the system are processed first, which can reduce system risk in the shortest time, improve overall adjustment efficiency, and avoid resource waste caused by indiscriminate processing.

[0021] The technical scheme and advantages of the embodiments of the present application are as follows: the K-Means algorithm is used to cluster multi-dimensional working condition characteristics to obtain multiple working condition groups, and the servo driver parameter reference interval of each working condition group is determined, similarity analysis is performed on the current working condition characteristics and the working condition characteristics of each working condition group, the working condition group to which the current working condition belongs and the corresponding servo driver parameter reference interval are determined, deviation analysis is performed on the adaptive interval and the reference interval of each servo driver parameter in the adaptive optimization parameter library, unqualified parameters are identified, the fault influence index of each servo driver parameter is calculated, key parameters are identified, frequency analysis is performed on the unqualified parameters and the key parameters, and it is judged whether the adaptive optimization parameter library and the current working condition exist mismatch phenomenon, if so, the optimization priority of the unqualified parameters is determined according to the deviation of the adaptive interval and the reference interval of the unqualified parameters and the fault influence index, the adaptive interval of the unqualified parameters is adjusted according to the optimization priority, and the adjusted parameter interval is updated to the adaptive optimization parameter library. The present application divides the working condition groups by clustering the multi-dimensional working condition characteristics and determines the reference interval of each working condition group, locates the current working condition, identifies unqualified parameters, selects key parameters through the fault influence index, and calculates the mismatch index according to the proportion of unqualified parameters in the adaptive optimization parameter library and the key parameters, judges whether the parameter library and the current working condition exist mismatch phenomenon, if so, determines the optimization priority of the unqualified parameters according to the fault influence index and the deviation of the adaptive interval and the reference interval, and adjusts the unqualified parameters in stages according to the priority, finally realizes the dynamic optimization of the parameter library, solves the problem of too wide parameter interval in the traditional adaptive optimization, and improves the adaptive ability and operation reliability of the servo driver under complex working conditions.

[0022] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A servo drive adaptive optimization method based on multi-source heterogeneous data fusion, characterized in that: include: Step 1: Use the K-Means algorithm to cluster the multi-dimensional operating condition features to obtain multiple operating condition groups, and determine the baseline range of servo drive parameters for each operating condition group. Step 2: Perform a similarity analysis between the current operating condition characteristics and the operating condition characteristics of each operating condition group, determine the operating condition group to which the current operating condition belongs and the corresponding servo drive parameter reference range, and perform a deviation analysis between the adaptive range and the reference range of each servo drive parameter in the adaptive optimization parameter library, and identify unqualified parameters in the adaptive optimization parameter library. Step 3: Calculate the fault impact index of each servo drive parameter, identify key parameters in the servo drive parameters, and perform frequency analysis on unqualified parameters and key parameters to determine whether there is a mismatch between the adaptive optimization parameter library and the current operating conditions. Step 4: If present, determine the optimization priority of the non-conforming parameters based on the deviation between the adaptive range and the baseline range of the non-conforming parameters and the fault impact index. Step 5: Adjust the adaptive range of unqualified parameters according to the optimization priority, and update the adjusted parameter range to the adaptive optimization parameter library.

2. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The process of clustering multi-dimensional working condition features is as follows: Calculate the Pearson coefficients of each operating condition variable and servo drive parameters, compare them with preset correlation coefficients, and select operating condition variables with Pearson correlation coefficients greater than preset correlation coefficients. After standardization, these variables are used as multi-dimensional operating condition features. The average silhouette coefficient for different K values ​​is calculated using the silhouette coefficient method, and the K with the largest average silhouette coefficient is selected as the number of clusters. Initially, K cluster centers are randomly selected. The Euclidean distance from each sample to the K centers is calculated, and the sample is assigned to the cluster with the closest Euclidean distance. After all samples are assigned to clusters, the average value of all sample feature values ​​in each cluster is recalculated and used as the new cluster center. The distribution of duplicate samples and the updating of cluster centers are recorded. The change value of each cluster center is recorded and compared with a preset threshold. When the change value is less than the preset threshold, the clustering is completed, and each cluster is a working condition group.

3. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The method for obtaining the reference range of servo driver parameters for each operating condition group is as follows: Collect historical data of servo drive parameters during normal operation of the servo drive under each working condition group, and use this data as the normal values ​​of the servo drive parameters; Based on any set of operating conditions; For any given servo driver parameter; The mean μ and standard deviation σ of the normal values ​​of the servo drive parameters are calculated. The "±3σ" principle is adopted, and [μ-3σ,μ+3σ] is used as the reference range of each servo drive parameter under the working condition group.

4. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The method for clearly defining the current operating condition group and the corresponding servo driver parameter reference range is as follows: The key features of the current working condition are obtained by sensors, and the Euclidean distance between the features of the current working condition and the features of each working condition group is calculated after standardization. Compare the Euclidean distance with a threshold. If the Euclidean distance is less than the threshold, mark the working condition group as a similar working condition group. If there are multiple similar working condition groups, select the working condition group with the smallest Euclidean distance, which is the working condition group to which the current working condition belongs. The reference range of servo driver parameters corresponding to the current operating condition is the reference range of servo driver parameters for the operating condition group to which the current operating condition belongs.

5. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The process of identifying unqualified parameters in the adaptive optimization parameter library is as follows: Mark the range of each servo driver parameter in the adaptive optimization parameter library as the adaptive range of the servo driver parameter; Calculate the baseline interval width and the adaptive interval width of the servo driver parameters respectively, and compare them. If the adaptive interval width of the servo driver parameters is greater than the baseline interval width, calculate the interval width deviation of the servo driver parameters. The interval width deviation is compared with the preset deviation. If the interval width deviation is greater than the preset deviation, the servo drive parameter is marked as an unqualified parameter.

6. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The method for identifying key parameters in the servo driver parameters is as follows: For any given servo driver parameter; The historical failure frequency of parameters and the average repair time of failures caused by abnormal servo drive parameters are obtained, and the failure trigger frequency and repair time ratio are calculated. The fault impact index is obtained by adding the fault trigger frequency to the proportion of repair time. The fault impact index is compared with the preset impact index. If the fault impact index is greater than the preset impact index, the servo drive parameter is marked as a critical parameter.

7. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 6, characterized in that: The calculation method for the ratio of fault trigger frequency to repair time is as follows: The frequency of abnormal servo drive parameters in historical faults is statistically analyzed and calculated as a proportion of the total number of historical faults to obtain the fault trigger frequency. Obtain the average repair time for each servo drive parameter abnormality caused by the fault, and sum them to get the total repair time; The repair time percentage is calculated by proportionally dividing the average repair time of the servo drive parameters by the total repair time.

8. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The method for determining whether there is a mismatch between the adaptive optimization parameter library and the current operating conditions is as follows: The number of unqualified parameters in the adaptive optimization parameter library is counted, and the ratio is calculated with the total number of parameters in the adaptive optimization parameter library to obtain the proportion of unqualified parameters. The proportion of non-compliant parameters among the key parameters is statistically analyzed to obtain the percentage of key non-compliant parameters. The mismatch index is obtained by adding the percentage of non-conforming parameters to the percentage of critical non-conforming parameters. The mismatch index is compared with a preset threshold. If the mismatch index is greater than the preset threshold, then there is a mismatch between the adaptive optimization parameter library and the current working condition.

9. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The optimization priority of the non-compliant parameters is determined as follows: For any one of the non-compliant parameters; The priority index is obtained by adding the interval width deviation to the fault impact index. The parameters are then sorted in descending order of the priority index to obtain the optimization priority of the unqualified parameters.

10. The adaptive optimization method for servo drives based on multi-source heterogeneous data fusion according to claim 1, characterized in that: The process of adjusting the adaptive range of unqualified parameters according to optimization priority is as follows: For any one of the non-compliant parameters; Set the adjustment period and adjustment ratio. Within each adjustment period, adjust the upper and lower limits of the adaptive range of the non-compliant parameters simultaneously according to the adjustment ratio. After each adjustment cycle, the deviation of the adaptive range of unqualified parameters from the baseline range is calculated and compared with the preset deviation. If the deviation in the interval width is less than the preset deviation, the adjustment is complete.

Images