Sliding data dilution method and system based on SSA optimization SVMD noise reduction and multi-principle fusion

Abstract

The invention discloses a sliding data dilution method and system based on SSA optimization SVMD noise reduction and multi-principle fusion. The method comprises the steps that vibration signals and current, voltage and rotating speed parameter signals are collected; optimizing a parameter maxAlpha of SVMD based on SSA by combining a composite index fused by mutual information entropy and energy difference as a fitness function, decomposing an original vibration signal by using an optimized SVMD algorithm to obtain a plurality of modal components IMF, rejecting IMF components containing noise, and reconstructing the signal to obtain a denoised signal; calculating characteristic values of the denoised signal Y: a frequency pass speed effective value and a vibration impact value; performing transient working condition and steady working condition identification on each IMF component and characteristic value data, completely retaining transient working condition data with changed working conditions, and diluting steady working condition data; and diluting and retaining the whole data except the retained data according to a time rule. The data storage and processing burden is relieved, and the data processing and analysis efficiency and accuracy are ensured.

Description

Technical Field

[0001] The present invention relates to a vibration data processing and analysis technology for industrial equipment, and in particular to a sliding data dilution method and system for alleviating the burden of data storage and processing. Background Art

[0002] In modern industrial environments, with the improvement of automation and the rapid development of sensor technology, vibration data acquisition has become an important tool for equipment monitoring and maintenance management. During the operation of industrial equipment, its vibration signal can provide valuable operating status information, which is essential for fault detection, health monitoring and performance optimization. However, in the actual data acquisition and processing process, it is often affected by noise and other interference signals, which may come from the external environment, the mechanical vibration of the equipment itself, and electromagnetic interference. The influence of noise and interference signals makes the original vibration data complicated, and the ratio between effective information and noise in the signal (i.e., signal-to-noise ratio) decreases, which in turn affects the accuracy of data analysis and fault diagnosis. Data with low signal-to-noise ratio is likely to lead to misjudgment, so that maintenance personnel fail to identify the potential failure risk of equipment in time, which in turn affects the stability and safety of production. Therefore, how to effectively remove noise and improve the data signal-to-noise ratio has become a key issue in improving the reliability of data analysis.

[0003] In addition to noise, as data is continuously collected and stored, a large amount of redundant data will be generated in industrial systems. These redundant data will take up a lot of storage space, causing storage pressure on the server, which in turn affects the efficiency of data processing and analysis. Traditionally, some companies usually take the approach of clearing old data. Although it is simple and direct, it is likely to cause the loss of some important historical data. These historical data are of great significance for tracking performance changes of equipment, analyzing failure trends, and formulating maintenance plans.

[0004] Therefore, the current challenge is how to effectively remove noise during the data acquisition process and improve the signal-to-noise ratio of the data. At the same time, adopt a reasonable data dilution method to reduce the amount of data and alleviate storage pressure while ensuring data quality, so as to avoid reduced efficiency in the subsequent data processing and analysis stages. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a sliding data dilution method and system based on the sparrow optimization algorithm (SSA) optimization of successive variational mode decomposition (SVMD) denoising and the integration of multiple principles, so as to reduce the burden of data storage and processing and ensure the efficiency and accuracy of data processing and analysis.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions: A sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles, characterized by comprising the following steps: S1 collects vibration signal X and current, voltage and speed parameter signals; S2 uses a composite index based on SSA combined with mutual information entropy and energy difference as the fitness function to optimize the SVMD parameter maxAlpha; S3 uses the optimized SVMD algorithm to decompose the original vibration signal X to obtain several modal components IMF, discards the IMF components containing noise, and reconstructs the signal to obtain the denoised signal Y; S4 calculates the characteristic values ​​of the denoised signal Y: effective value of the through-frequency speed and vibration impact value; S5 identifies transient and steady-state conditions for each IMF component and eigenvalue data, retains all transient condition data with changed conditions, and dilutes steady-state condition data; S6 dilutes and retains the overall data other than the retained data according to the time rule.

[0007] In the above technical solution, step S2 uses mutual information entropy and energy difference as fitness functions, and the sparrow search algorithm SSA is set to adjust the parameter maxAlpha of SVMD according to the value of the function. The optimization goal is to minimize the mutual information entropy and energy difference.

[0008] The optimization algorithm adjusts the SVMD parameter maxAlpha to make the mutual information entropy and energy difference as small as possible. Smaller mutual information entropy means less information redundancy between the modal component and the original signal, while smaller energy difference means that the decomposed modal component can better reconstruct the original signal and retain the main energy characteristics of the signal. Through the optimization of this composite index, a more reasonable SVMD parameter maxAlpha can be obtained, thereby improving the quality of signal decomposition and denoising effect. As a result, the decomposed modal component has lower information redundancy with the original signal and can better retain the energy characteristics of the signal.

[0009] In the above technical solution, step S3 calculates the Pearson correlation coefficient of each IMF component, and selects the IMF component whose Pearson correlation coefficient is greater than a set threshold for denoising and reconstruction.

[0010] The present invention selects IMF components with a Pearson correlation coefficient greater than 0.5, discards IMF components containing noise, and reconstructs the signal to obtain a denoised signal Y.

[0011] In the above technical solution, step S5 also includes at least one of the following steps: S51 For the first type of steady-state operating data that meets the steady-state operating conditions, when the frequency-throughput speed effective value and vibration impact value parameters have an over-threshold alarm, at least the full amount of data for the alarm time period shall be retained; S52 For the second type of steady-state operating condition data that meets the steady-state operating conditions and excludes the threshold crossing alarm, for the rare data in the current, voltage, and speed parameters, the data of a set number of points before and after the current value of the parameter is retained, and the remaining parameters are retained synchronously according to the timestamp of the parameter retained data to ensure the consistency of the timestamp.

[0012] Therefore, when the amount of steady-state data storage is very large, the data dilution method that follows multiple principles can effectively dilute a large amount of historical data, retain important historical data, and reduce storage pressure.

[0013] In the above technical solution, step S5 calculates the absolute error between the current value and the previous value of each parameter for the effective value of the through-frequency speed, the vibration impact value, the current, the voltage, and the speed parameter change rate parameters. If and only if the absolute errors of all parameters fall within the set range, it is a steady-state condition, otherwise it is a transient condition.

[0014] In the above technical solution, in step S5, the dilution data is grouped and retained according to the time period of year, month and week.

[0015] For example, weekly data is retained for 5 minutes, monthly data is retained for 30 minutes, and annual data is retained for 6 hours. The time can be set according to actual needs.

[0016] In the above technical solution, in step S51, all data of the alarm time period, the set time before the alarm triggering moment, and the set time after the alarm ending moment are retained.

[0017] In the above technical solution, in step S51, the data between 15 seconds before the alarm triggering time and the data between 15 seconds after the alarm ending time are retained.

[0018] In the above technical solution, step S52 uses a gradient ratio detection method to determine whether the current, voltage, and speed parameters are rare data.

[0019] In the above technical solution, the gradient ratio detection method of step S52 is as follows: assuming that the data has 2N points, if the ratio of the Nth point to the N-1th point, the ratio of the N+1th point to the N-2th point, the ratio of the N+2th point to the N-3th point, the ratio of the N+3th point to the N-4th point, ... are greater than the set threshold for ten consecutive times, they are regarded as rare data.

[0020] In the above technical solution, in step S52, if any of these key parameters is rare data, the data of 10 points before and after the current value of the parameter is retained, and the remaining parameters are retained synchronously according to the timestamp of the retained data of the parameter to ensure the consistency of the timestamp.

[0021] A sliding data dilution system based on SSA optimization SVMD noise reduction and fusion of multiple principles, characterized by comprising: Signal acquisition module, used to collect vibration signal X and current, voltage, and speed parameter signals; The signal denoising and reconstruction module is used to optimize the parameter maxAlpha of SVMD based on the composite index of SSA combined with mutual information entropy and energy difference as the fitness function; the original vibration signal X is decomposed using the optimized SVMD algorithm to obtain several modal components IMF, the IMF components containing noise are discarded, and the signal is reconstructed to obtain the denoised signal Y; The working condition identification and data dilution module is used to calculate the characteristic values ​​of the denoised signal Y: the effective value of the full-frequency speed, the vibration impact value; identify the transient working condition and steady-state working condition of each IMF component and characteristic value data, retain all the transient working condition data with changed working conditions, dilute the steady-state working condition data, and then dilute the overall data according to the time series and retain them in groups.

[0022] In the above technical solution, a first type of steady-state working condition data storage unit is set in the working condition identification and data dilution module. For the full data that meets the steady-state working condition, when the effective value of the through-frequency speed and the vibration impact value parameters have an over-threshold alarm, at least the full data of the alarm time period is retained.

[0023] In the above technical solution, a second type of steady-state operating condition data storage unit is set in the operating condition identification and data dilution module. For the full amount of data that meets the steady-state operating condition and eliminates the threshold crossing alarm, for the rare data in the current, voltage, and speed parameters, the data of a set number of points before and after the current value of the parameter is retained, and the remaining parameters are retained synchronously according to the timestamp of the parameter retained data to ensure the consistency of the timestamp.

[0024] In addition, the present invention also provides a computer-readable storage medium, in which a computer program is stored. When the program is executed, it is used to implement the above-mentioned sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles.

[0025] Therefore, the present invention optimizes the successive variational mode decomposition (SVMD) algorithm based on the sparrow optimization algorithm (SSA), and uses the composite index of mutual information entropy and energy difference as the fitness function to optimize the parameter maxAlpha, so as to perform noise reduction on the signal to improve the signal-to-noise ratio of the data. At the same time, it is proposed to use a sliding dilution method with multi-principle fusion to dilute the stored historical data, so as to retain valid data, eliminate invalid data, and reduce the burden of data storage and processing.

[0026] Compared with the prior art, the beneficial effects of the present invention are: Based on the sparrow optimization algorithm (SSA) combined with the composite index of mutual information entropy and energy difference as the fitness function, the successive variational mode decomposition (SVMD) algorithm is optimized to achieve denoising of the signal and improve the signal-to-noise ratio of the data; when the collected data contains a lot of noise and interference signals, the denoising performance of the algorithm can be improved, the noise in the vibration signal can be effectively removed, and the signal-to-noise ratio of the signal can be improved.

[0027] The absolute error accumulation verification method is proposed to identify transient and steady-state conditions, retain all transient condition data, and only process steady-state condition data. When the operating state changes, the transient and steady-state conditions can be effectively identified, and the transient condition data can be retained, while the steady-state condition data can be diluted to ensure the quality of the data.

[0028] It is proposed to use a multi-principle method based on the fusion of alarm principle, dilution principle and time principle to dilute the steady-state operating condition data. When the steady-state data storage volume is very large, the multi-principle fusion data dilution method is followed to effectively dilute a large amount of historical data, thereby retaining valid data, eliminating invalid data, and reducing the burden of data storage and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 It is a flow chart of a sliding data dilution method based on SSA optimization of SVMD noise reduction and integration of multiple principles in an embodiment of the present invention, the left half is data noise reduction, and the right half is data dilution.

[0030] Figure 2 It is the original vibration acceleration data X collected in one embodiment of the present invention.

[0031] Figure 3 yes Figure 2 The signal Y after denoising. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0033] like Figure 1 As shown, in order to make the purpose, technical solutions and advantages of the present invention more clear, the sliding data dilution method based on SSA optimization SVMD denoising and fusion of multiple principles of the embodiment of the present invention will be described in detail with reference to the accompanying drawings.

[0034] In this embodiment, the sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles includes the following steps: Collect vibration signal X and current, voltage and speed parameter signals of rotating equipment; Based on the sparrow optimization algorithm (SSA), a composite index combining mutual information entropy and energy difference is used as the fitness function to optimize the parameter maxAlpha of the successive variational mode decomposition (SVMD); The optimized SVMD algorithm is used to decompose the original vibration signal X to obtain several modal components IMF, and the Pearson correlation coefficient of each IMF is calculated. The IMF components with a Pearson correlation coefficient greater than 0.5 are selected, the IMF components containing noise are discarded, and the signal is reconstructed to obtain the denoised signal Y; Calculate the characteristic values ​​of signal Y: effective value of full-frequency speed, vibration impact value; The formula of effective value of frequency speed is: Step 1: Integrate the signal Y to obtain the speed signal Y(n); Step 2: Filter the speed signal Y(n) within the filtering range of [3,1000Hz] to obtain the filtered signal y(n); Step 3: Calculate the effective value of the pass frequency speed:

[0035] Vibration shock value calculation steps: Step 1: Bandpass filter the signal Y in the filtering range of [5k, 10kHz] to obtain the filtered signal y1; Step 2: Perform Hilbert envelope transform on the filtered signal y1 to obtain envelope signal y2; Step 3: Perform bandpass filtering on the envelope signal y2 again, with the filtering range of [5, 1000 Hz], and obtain the filtered signal y3; Step 4: Calculate the vibration impact value;

[0036] In order to distinguish the steady-state and transient conditions of the equipment and to reasonably process the data under different conditions, a steady-state operating model is created based on the logical “and” relationship between the effective value of the through-frequency speed, vibration impact value, current, voltage, and speed parameter change rate.

[0037] The following are the steps to build a steady-state model: (1) Data collection and preprocessing: Collect parameter data such as effective value of frequency speed, vibration impact value, current, voltage, speed, etc.

[0038] The collected data is preprocessed, including filtering, denoising, normalization and other operations, to improve data quality and consistency.

[0039] (2) Calculation of rate of change: For each parameter, calculate the absolute error between its current value and its previous value , Represents the i-th parameter (such as effective value of the through-frequency speed, vibration impact value, etc.).

[0040] (3) Construction of steady-state judgment logic: According to the set error range (such as [0, 0.5]), determine whether the change rate of each parameter meets the steady-state condition. Specifically, the current working condition is considered to meet the steady-state condition only when the absolute errors of all parameters fall within this range.

[0041] A logical "and" relationship is adopted, that is, when all parameters meet the conditions, it is determined to be a steady-state condition.

[0042] (4) Absolute error accumulation verification: Continuously monitor the changes in parameters and accumulate the number of times the steady-state conditions are met. When the cumulative number reaches 5, it is finally determined to be a steady-state condition; otherwise, it is determined to be a transient condition.

[0043] This method can effectively avoid misjudgment caused by accidental factors and improve the reliability of steady-state condition judgment.

[0044] (5) Data processing and model building: All data judged to be transient conditions are retained for further analysis.

[0045] For the data of steady-state conditions, dilution processing is performed to reduce the data volume and improve data processing efficiency.

[0046] Through the above steps, a model that can accurately distinguish between steady-state and transient conditions is constructed, providing a basis for subsequent equipment status monitoring and fault diagnosis.

[0047] For the five parameters of frequency-pass speed effective value (V), vibration impact value (A), current (I), voltage (U), and speed parameter change rate (Ra), the steady-state operating model can be expressed as follows: Steady state condition = ; By monitoring the change rate of each parameter and combining it with the absolute error accumulation verification method, the model can accurately identify the operating status of the equipment, providing strong support for equipment operation status assessment and fault warning.

[0048] The absolute error accumulation verification method is used to identify transient and steady-state conditions. That is, the absolute error between the current value and the previous value of each operating condition parameter falls within the range of [0, 0.5] (the error range can be set according to the actual situation of the equipment), and the condition is met for 5 cumulative times. It can be determined as a steady-state condition, otherwise it is a transient condition. All transient condition data with changes in the working conditions are retained, and the steady-state condition data are diluted.

[0049] For the full data that meets the steady-state working conditions, when the effective value of the through-frequency speed and the vibration impact value parameters have an over-the-line threshold alarm, the full data of the alarm time period is retained, and the data between 15 seconds before the alarm trigger time and 15 seconds after the alarm end time are retained.

[0050] For all data that meet steady-state conditions and exclude threshold crossing alarms, the gradient ratio detection method is used to determine whether the key parameters of current, voltage, and speed are rare data. If any of these key parameters is rare data, the data of 10 points before and after the current value of the parameter is retained (the number of points can be set according to actual needs), and the remaining parameters are retained synchronously according to the timestamp of the parameter's retained data to ensure the consistency of the timestamp.

[0051] Gradient ratio detection method: Assume that there are 2N data points. If the ratio of the Nth point to the N-1th point, the ratio of the N+1th point to the N-2th point, the ratio of the N+2th point to the N-3th point, the ratio of the N+3th point to the N-4th point, ... are all greater than 1.5 for ten consecutive times (the threshold can be set according to the actual situation of the device), it is regarded as rare data.

[0052] The overall data is diluted and retained according to the time series. Weekly data is retained in a group of 5 minutes, monthly data is retained in a group of 30 minutes, and annual data is retained in a group of 6 hours (the time can be set according to actual needs). The time principle dilution method is as follows: 1) Weekly data dilution: When the 5-minute time segment contains any of transient data, alarm data, and rare data, the data is retained according to steps S5, S6, and S7, and the remaining data is deleted; if none of them are included, a set of data is retained for 5 minutes; 2) Monthly data dilution: When the full amount of key parameter data contains any of transient data, alarm data and rare data within the 30-minute time segment, the data is retained according to steps S5, S6 and S7, and the remaining data is deleted; if none of them are included, a set of data is retained for 30 minutes; 3) Annual data dilution: When the 6-hour time segment contains any of transient data, alarm data, and rare data, the data is retained according to steps S5, S6, and S7, and the remaining data is deleted; if none of them are included, a set of data is retained for 6 hours.

[0053] Complete data dilution.

[0054] Embodiment 2: According to Figure 1 Process, select the original vibration acceleration data X of 1s as Figure 2 As shown; and simulate the current, voltage and speed parameter values.

[0055] The 1 second data is denoised, and the Pearson correlation coefficient values ​​of the decomposed IMF components are shown in Table 1. The IMF components greater than 0.5 are selected to reconstruct the signal and obtain the denoised signal Y, as shown in Figure 3 shown.

[0056] Calculate the characteristic values ​​of signal Y: effective value of full-frequency speed and vibration impact value.

[0057] Table 1

[0058] Determine transient and steady-state conditions, and retain transient condition data; Determine whether the characteristic value exceeds the alarm line and retain the alarm data; Determine the scarcity of data and retain rare data; According to the time rules, the overall data other than the retained data is diluted and retained. Weekly data is retained as a group every 5 minutes, monthly data is retained as a group every 30 minutes, and annual data is retained as a group every 6 hours.

[0059] At this point, data dilution is complete.

[0060] Example 3 The sliding data dilution system based on SSA optimization SVMD noise reduction and fusion of multiple principles implemented by the present invention includes: Signal acquisition module, used to collect vibration signal X and current, voltage, and speed parameter signals; The signal denoising and reconstruction module is used to optimize the parameter maxAlpha of SVMD based on the composite index of SSA combined with mutual information entropy and energy difference as the fitness function; the original vibration signal X is decomposed using the optimized SVMD algorithm to obtain several modal components IMF, the IMF components containing noise are discarded, and the signal is reconstructed to obtain the denoised signal Y; The working condition identification and data dilution module is used to calculate the characteristic values ​​of the denoised signal Y: the effective value of the full-frequency speed, the vibration impact value; identify the transient working condition and steady-state working condition of each IMF component and characteristic value data, retain all the transient working condition data with changed working conditions, dilute the steady-state working condition data, and then dilute the overall data according to the time series and retain them in groups.

[0061] In the working condition identification and data dilution module, a first type of steady-state working condition data storage unit is set to store the full amount of data that meets the steady-state working condition. When the effective value of the through-frequency speed and the vibration impact value parameters have an over-threshold alarm, at least the full amount of data of the alarm time period is retained; A second type of steady-state operating condition data storage unit is set up to store all data that meet the steady-state operating conditions and exclude threshold crossing alarms. For rare data in current, voltage, and speed parameters, data of a set number of points before and after the current value of the parameter is retained; the remaining parameters are retained synchronously according to the timestamp of the parameter's retained data to ensure the consistency of the timestamp; All transient operating condition data of the changed operating condition are retained in the third storage unit.

[0062] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A sliding data dilution method based on SSA optimization SVMD denoising and fusion of multiple principles, characterized by The steps include: S1 collects vibration signal X and current, voltage and speed parameter signals; S2 uses a composite index based on SSA combined with mutual information entropy and energy difference as the fitness function to optimize the SVMD parameter maxAlpha; S3 uses the optimized SVMD algorithm to decompose the original vibration signal X to obtain several modal components IMF, discards the IMF components containing noise, and reconstructs the signal to obtain the denoised signal Y; S4 calculates the characteristic values ​​of the denoised signal Y: effective value of the through-frequency speed and vibration impact value; S5 identifies transient and steady-state conditions for each IMF component and eigenvalue data, retains all transient condition data with changed conditions, and dilutes steady-state condition data; S6 dilutes and retains the overall data other than the retained data according to the time rule.

2. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 1 is characterized in that Step S2: Mutual information entropy and energy difference are used as fitness functions, and the sparrow search algorithm SSA is set to adjust the parameter maxAlpha of SVMD according to the value of the function. The optimization goal is to minimize the mutual information entropy and energy difference.

3. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 1 is characterized in that Step S3: Calculate the Pearson correlation coefficient of each IMF component, and select the IMF component whose Pearson correlation coefficient is greater than a set threshold for denoising and reconstruction.

4. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 1 is characterized in that Step S5 also includes at least one of the following steps: S51 For the first type of steady-state operating data that meets the steady-state operating conditions, when the frequency-throughput speed effective value and vibration impact value parameters have an over-threshold alarm, at least the full amount of data for the alarm time period shall be retained; S52 For the second type of steady-state operating condition data that meets the steady-state operating conditions and excludes the threshold crossing alarm, for the rare data in the current, voltage, and speed parameters, the data of a set number of points before and after the current value of the parameter is retained, and the remaining parameters are retained synchronously according to the timestamp of the parameter retained data to ensure the consistency of the timestamp.

5. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 4 is characterized in that In step S51, all data of the alarm time period, the set time before the alarm triggering time, and the set time after the alarm ending time are retained.

6. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 4 is characterized in that Step S52 uses a gradient ratio detection method to determine whether the current, voltage, and speed parameters are rare data.

7. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 1 is characterized in that Step S5: For the effective value of the through-frequency speed, the vibration impact value, the current, the voltage, and the speed parameter change rate parameters, for each parameter, the absolute error between the current value and the previous value is calculated. If and only if the absolute errors of all parameters fall within the set range, it is a steady-state condition, otherwise it is a transient condition.

8. The sliding data dilution method based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 1 is characterized in that In step S6, the dilution data is grouped and retained according to the time period of year, month and week.

9. A sliding data dilution system based on SSA optimization SVMD noise reduction and fusion of multiple principles, characterized by include: Signal acquisition module, used to collect vibration signal X and current, voltage, and speed parameter signals; Signal denoising and reconstruction module, used to optimize the parameter maxAlpha of SVMD based on the composite index of SSA combined with mutual information entropy and energy difference as the fitness function; The optimized SVMD algorithm is used to decompose the original vibration signal X to obtain several modal components IMF, the IMF components containing noise are discarded, and the signal is reconstructed to obtain the denoised signal Y; The working condition identification and data dilution module is used to calculate the characteristic values ​​of the denoised signal Y: the effective value of the full-frequency speed, the vibration impact value; identify the transient working condition and steady-state working condition of each IMF component and characteristic value data, retain all the transient working condition data with changed working conditions, dilute the steady-state working condition data, and then dilute the overall data according to the time series and retain them in groups.

10. The sliding data dilution system based on SSA optimization SVMD noise reduction and fusion of multiple principles according to claim 9 is characterized in that include: In the working condition identification and data dilution module, a first type of steady-state working condition data storage unit is set to store the full amount of data that meets the steady-state working condition. When the effective value of the through-frequency speed and the vibration impact value parameters have an over-threshold alarm, at least the full amount of data of the alarm time period is retained; A second type of steady-state operating condition data storage unit is set up to store all data that meets the steady-state operating conditions and excludes threshold crossing alarms. For rare data in current, voltage, and speed parameters, data of a set number of points before and after the current value of the parameter is retained; the remaining parameters are retained synchronously according to the timestamp of the parameter's retained data to ensure the consistency of the timestamp.

Images