Equipment state data analysis method based on spectral kurtosis
By constructing a speed-spectral kurtosis correlation curve using equipment status data analysis based on spectral kurtosis, the problem of insufficient correlation between features in traditional reducer testing is solved, and more accurate fault identification and calibration are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI KINGS AUTOTECH
- Filing Date
- 2025-12-08
- Publication Date
- 2026-04-10
AI Technical Summary
In traditional gearbox fault detection, feature extraction and individual judgment lead to low correlation and tightness between features, resulting in low accuracy of analysis results.
A device status data analysis method based on spectral kurtosis is adopted. By acquiring historical data of the reducer, a status dataset is constructed, invalid data is removed, and spectral kurtosis is calculated using correlation functions and spectral transformation. A speed-spectral kurtosis correlation curve is established to determine the device status.
It improves the accuracy of gearbox fault detection, can identify abnormal types of gears and rigid housings, and enhances the accuracy of detection through condition calibration, avoiding invalid data from affecting the results.
Smart Images
Figure CN121834431A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of equipment status data analysis technology, and specifically to a method for equipment status data analysis based on spectral kurtosis. Background Technology
[0002] The speed reducer includes a rigid housing and structures such as gear transmission, worm transmission, and gear-worm transmission.
[0003] Traditional gearbox fault detection typically involves extracting features separately and judging whether each feature exceeds a fixed warning value individually. This results in low correlation and tightness between features, leading to low accuracy of analysis results.
[0004] Based on this, the present invention designs a device status data analysis method based on spectral kurtosis to solve the above problems. Summary of the Invention
[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a device status data analysis method based on spectral kurtosis.
[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for analyzing device status data based on spectral kurtosis includes the following steps: Step 1: Obtain historical data of the reducer, including state type Y, time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid shell vibration signal A4; construct state dataset Bn (Y, T, A1, A2, A3, A4). The state type Y includes normal state, slightly abnormal state and severely abnormal state, and the state dataset Bn includes normal state dataset, slightly abnormal state dataset and severely abnormal state dataset.
[0007] Step 2: Obtain the correlation function L1 between the gear vibration signal A1 and the sound signal A2, and remove signals where L1 ≤ preset correlation threshold L. A The state dataset Bn; Obtain the correlation function L2 between the rotational speed signal A3 and the sound signal A2, and discard signals where L2 ≤ a preset correlation threshold L. B The state dataset Bn; Obtain the correlation function L3 between the rigid shell vibration signal A4 and the sound signal A2, and discard signals where L3 ≤ a preset correlation threshold L. C The state dataset Bn; Step 3: Perform spectrum conversion on gear vibration signal A1 and rigid shell vibration signal A4 respectively, output spectrum diagrams, and calculate spectral kurtosis; Construct the rotational speed-spectral kurtosis correlation curve under normal conditions based on the normal state dataset; Construct rotational speed-spectral kurtosis correlation curves under minor anomalous state conditions based on the minor anomalous state dataset; Construct rotational speed-spectral kurtosis correlation curves under severe abnormal conditions based on the severe abnormal condition dataset; Step 4: Based on the abnormal state dataset, establish the relationship between abnormal state type and spectral kurtosis and rotational speed, and generate the abnormal state-spectral kurtosis-rotational speed relationship curve; Step 5: Acquire real-time data of the reducer, including time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4; Step 6: Calculate the correlation function L1 using gear vibration signal A1 and sound signal A2; calculate the correlation function L2 using speed signal A3 and sound signal A2; calculate the correlation function L3 using rigid housing vibration signal A4 and sound signal A2. If any of the following exists, discard the real-time data of the reducer; otherwise, retain the real-time data of the reducer: L1 ≤ Preset association threshold L A ; L2 ≤ preset association threshold L B ; L3 ≤ Preset association threshold L C ; Step 7: Perform spectrum conversion on the gear vibration signal A1 of the reducer in the real-time data of the reducer, output the spectrum diagram, and calculate the spectral kurtosis; based on the speed and spectral kurtosis in the real-time data, determine whether the spectral kurtosis at this speed exceeds the corresponding spectral kurtosis critical value on the speed-spectral kurtosis correlation curve under slight abnormal conditions or under severe abnormal conditions. If not, it means the reducer is in normal condition, proceed to step 10; If so, then the situation is as follows: if it exceeds the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the slight abnormal state but does not exceed the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the severe abnormal state, then it belongs to the slight abnormal state; if it exceeds the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the severe abnormal state, then it belongs to the severe abnormal state. Step 8: Obtain the abnormal state through the abnormal state-spectral kurtosis-rotation speed relationship curve, and construct the state dataset Bn(Y, T, A1, A2, A3, A4). Step 9: Determine whether the abnormal states in Step 7 and Step 8 are consistent. If they are consistent, proceed to Step 10. If they are inconsistent, perform state calibration. Step 10: Output the reducer status Bn (Y, T, A1, A2, A3, A4).
[0008] Furthermore, in step 1, the gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4 of the reducer are all acquired by sensors, including vibration sensors, sound sensors, and speed sensors.
[0009] Furthermore, in step 3, the spectral kurtosis includes the spectral kurtosis 1 of the gear vibration signal A1 and the spectral kurtosis 2 of the rigid shell vibration signal A4.
[0010] Furthermore, in step 3, the correlation curves of rotational speed-spectral kurtosis 1 and rotational speed-spectral kurtosis 2 under normal conditions are constructed based on the normal state dataset.
[0011] Furthermore, in step 3, the correlation curves of speed-spectral kurtosis 1 and speed-spectral kurtosis 2 under the slight abnormal state are constructed based on the slight abnormal state dataset; the correlation curves of speed-spectral kurtosis 1 and speed-spectral kurtosis 2 under the severe abnormal state are constructed based on the severe abnormal state dataset.
[0012] Furthermore, in step 4, the relationship between the abnormal state type and spectral kurtosis 1 and rotational speed is established, generating an abnormal state-spectral kurtosis 1-rotational speed relationship curve; the relationship between the abnormal state type and spectral kurtosis 2 and rotational speed is established, generating an abnormal state-spectral kurtosis 2-rotational speed relationship curve.
[0013] Furthermore, the time T, the gear vibration signal A1 of the reducer, the sound signal A2, the speed signal A3, and the rigid shell vibration signal A4 can all be range values or point values.
[0014] Furthermore, the method for state calibration is as follows: Randomly extract an abnormal state dataset from the abnormal state database, obtain the abnormal state through steps 7 and 8, and determine whether it is consistent with the abnormal state in the abnormal state dataset. If they match, output the abnormal status of both step 7 and step 8 simultaneously; If there is a discrepancy, the following situations may occur: If the abnormal state in step 7 is inconsistent with the abnormal state in the state dataset, then repeat step 7. If the abnormal state in step 8 is inconsistent with the abnormal state in the state dataset, then step 8 is repeated. The system then checks whether the abnormal states in steps 7 and 8 are consistent. If they are consistent, it proceeds to step 10; otherwise, it issues an abnormal alarm.
[0015] Compared with the prior art, the beneficial effects of this invention are as follows: 1. The two curves in this invention correspond to the speed-spectral kurtosis correlation curves of gears and rigid housings under normal, slightly abnormal, and severely abnormal states, respectively, which helps to determine whether the abnormality is in the gears or the rigid housing, and also helps to determine the specific abnormal damage type.
[0016] 2. This invention comprehensively considers state type Y, gear vibration signal A1 of the reducer, speed signal A3, and rigid shell vibration signal A4, enhancing the connection and tightness between features, which is beneficial to improving the accuracy of the analysis results.
[0017] 3. The present invention can perform state calibration when the abnormal states of steps 7 and 8 are inconsistent, thereby further improving the accuracy of detection.
[0018] 4. This invention can pre-select invalid data to avoid invalid datasets affecting the accuracy of subsequent detection results. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0020] Figure 1 This is a flowchart of a device status data analysis method based on spectral kurtosis according to the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0022] Example 1: In some embodiments, please refer to the accompanying drawings. Figure 1 A method for analyzing device status data based on spectral kurtosis includes the following steps: Step 1: Obtain historical data of the reducer, including state type Y, time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid shell vibration signal A4; construct state dataset Bn (Y, T, A1, A2, A3, A4). Preferably, the state type Y includes normal state, slightly abnormal state and severely abnormal state, and the state dataset Bn includes normal state dataset, slightly abnormal state dataset and severely abnormal state dataset. The dataset for minor abnormal states includes minor damage to the gears of the reducer and minor damage to the rigid housing; the dataset for severe abnormal states includes severe damage to the gears of the reducer and severe damage to the rigid housing. Preferably, the gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4 of the reducer are all acquired by sensors, including vibration sensors, sound sensors, and speed sensors; Step 2: Obtain the correlation function L1 between the gear vibration signal A1 and the sound signal A2, and remove signals where L1 ≤ preset correlation threshold L. A The state dataset Bn; Obtain the correlation function L2 between the rotational speed signal A3 and the sound signal A2, and discard signals where L2 ≤ a preset correlation threshold L. B The state dataset Bn; Obtain the correlation function L3 between the rigid shell vibration signal A4 and the sound signal A2, and discard signals where L3 ≤ a preset correlation threshold L. C The state dataset Bn; This invention can pre-select invalid data to avoid invalid datasets affecting the accuracy of subsequent detection results.
[0023] Step 3: Perform spectrum conversion on gear vibration signal A1 and rigid shell vibration signal A4 respectively, output spectrum diagrams, and calculate spectral kurtosis; Construct the rotational speed-spectral kurtosis correlation curve under normal conditions based on the normal state dataset; Construct rotational speed-spectral kurtosis correlation curves under minor anomalous state conditions based on the minor anomalous state dataset; Construct rotational speed-spectral kurtosis correlation curves under severe abnormal conditions based on the severe abnormal condition dataset; In step 3, the spectral kurtosis includes the spectral kurtosis 1 of the gear vibration signal A1 and the spectral kurtosis 2 of the rigid shell vibration signal A4; In step 3, the correlation curves of rotational speed-spectral kurtosis 1 and rotational speed-spectral kurtosis 2 under normal conditions are constructed based on the normal state dataset. In step 3, the speed-spectral kurtosis 1 correlation curve and the speed-spectral kurtosis 2 correlation curve under the slight abnormal state are constructed based on the slight abnormal state dataset. In step 3, the speed-spectral kurtosis 1 correlation curve and the speed-spectral kurtosis 2 correlation curve under severe abnormal state conditions are constructed based on the severe abnormal state dataset. The two curves correspond to the speed-spectral kurtosis correlation curves of gears and rigid housings under normal, slightly damaged, and severely damaged conditions, respectively. This helps to determine whether the abnormality is in the gears or the rigid housing, and also helps to identify the specific type of abnormal damage (slight damage or severe damage). Step 4: Based on the abnormal state dataset, establish the relationship between abnormal state type (minor abnormality and severe abnormality) and spectral kurtosis and rotational speed, and generate the abnormal state-spectral kurtosis-rotational speed relationship curve; In step 4, the relationship between the abnormal state type (minor abnormality and severe abnormality) and spectral kurtosis 1 and rotational speed is established, and the abnormal state-spectral kurtosis 1-rotational speed relationship curve is generated; the relationship between the abnormal state type (minor abnormality and severe abnormality) and spectral kurtosis 2 and rotational speed is established, and the abnormal state (minor abnormality and severe abnormality)-spectral kurtosis 2-rotational speed relationship curve is generated. Step 5: Acquire real-time data of the reducer, including time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4; Step 6: Calculate the correlation function L1 using gear vibration signal A1 and sound signal A2; calculate the correlation function L2 using speed signal A3 and sound signal A2; calculate the correlation function L3 using rigid housing vibration signal A4 and sound signal A2. If any of the following exists, discard the real-time data of the reducer; otherwise, retain the real-time data of the reducer: L1 ≤ Preset association threshold L A ; L2 ≤ preset association threshold L B ; L3 ≤ Preset association threshold L C ; Step 7: Perform spectrum conversion on the gear vibration signal A1 of the reducer in the real-time data of the reducer, output the spectrum diagram, and calculate the spectral kurtosis; based on the speed and spectral kurtosis in the real-time data, determine whether the spectral kurtosis at this speed exceeds the corresponding spectral kurtosis critical value on the speed-spectral kurtosis correlation curve under slight abnormal conditions or under severe abnormal conditions. If not, it means the reducer is in normal condition, proceed to step 10; If so, then the situation is as follows: if it exceeds the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the slight abnormal state but does not exceed the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the severe abnormal state, then it belongs to the slight abnormal state; if it exceeds the spectral kurtosis threshold value corresponding to the speed-spectral kurtosis correlation curve under the severe abnormal state, then it belongs to the severe abnormal state. Step 8: By using the abnormal state-spectral kurtosis-rotation speed relationship curve, determine whether the abnormal state is a slight abnormal state or a severe abnormal state, and construct the state dataset Bn(Y, T, A1, A2, A3, A4). Abnormal states include minor abnormal states and severe abnormal states; Time T, gear vibration signal A1 of the reducer, sound signal A2, speed signal A3, and rigid shell vibration signal A4 can all be range values or point values; For example, a range value refers to a reducer in the same state within a certain time range, allowing for data accumulation. For instance, if a gear anomaly exists during time T = T1~Tn, then A1 = A11~A1n, A2 = A21~A2n, A3 = A31~A3n, and A4 = A41~A4n. A point value refers to a reducer in the same state at certain points in time. For instance, if a gear anomaly exists during T = (T1, T2, ..., Tn), then A1 = A11~A1n, A2 = A21~A2n, A3 = A31~A3n, and A4 = A41~A4n. This allows for a more intuitive understanding of the anomaly. This invention comprehensively considers state type Y, gear vibration signal A1 of the reducer, speed signal A3, and rigid shell vibration signal A4, enhancing the connection and tightness between features, which is beneficial to improving the accuracy of the analysis results.
[0024] Step 9: Determine if the abnormal states in Step 7 and Step 8 are consistent. If they are consistent, proceed to Step 10; otherwise, perform state calibration. The state calibration method is as follows: Randomly extract an abnormal state dataset from the abnormal state database, obtain the abnormal state through steps 7 and 8, and determine whether it is consistent with the abnormal state in the abnormal state dataset. If they match, output the abnormal status of both step 7 and step 8 simultaneously; If there is a discrepancy, the following situations may occur: If the abnormal state in step 7 is inconsistent with the abnormal state in the state dataset, then repeat step 7. If the abnormal state in step 8 is inconsistent with the abnormal state in the state dataset, then step 8 is repeated. The system then checks whether the abnormal states in steps 7 and 8 are consistent. If they are consistent, it proceeds to step 10; otherwise, it issues an abnormal alarm.
[0025] Step 10: Output the reducer status Bn (Y, T, A1, A2, A3, A4).
[0026] This invention can perform state calibration when the abnormal states in steps 7 and 8 are inconsistent, thereby further improving the accuracy of detection.
[0027] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for analyzing equipment status data based on spectral kurtosis, characterized in that, Includes the following steps: Step 1: Obtain historical data of the reducer, including state type Y, time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid shell vibration signal A4; construct state dataset Bn (Y, T, A1, A2, A3, A4). The state type Y includes normal state, slightly abnormal state and severely abnormal state, and the state dataset Bn includes normal state dataset, slightly abnormal state dataset and severely abnormal state dataset. Step 2: Obtain the correlation function L1 between the gear vibration signal A1 and the sound signal A2, and remove the state dataset Bn whose L1 ≤ preset correlation threshold LA; Obtain the correlation function L2 between the rotation speed signal A3 and the sound signal A2, and remove the state dataset Bn whose L2 is less than or equal to the preset correlation threshold LB; Obtain the correlation function L3 between the rigid shell vibration signal A4 and the sound signal A2, and remove the state dataset Bn where L3 ≤ the preset correlation threshold LC; Step 3: Perform spectrum conversion on gear vibration signal A1 and rigid shell vibration signal A4 respectively, output spectrum diagrams, and calculate spectral kurtosis; Construct the rotational speed-spectral kurtosis correlation curve under normal conditions based on the normal state dataset; Construct rotational speed-spectral kurtosis correlation curves under minor anomalous state conditions based on the minor anomalous state dataset; Construct rotational speed-spectral kurtosis correlation curves under severe abnormal conditions based on the severe abnormal condition dataset; Step 4: Based on the abnormal state dataset, establish the relationship between abnormal state type and spectral kurtosis and rotational speed, and generate the abnormal state-spectral kurtosis-rotational speed relationship curve; Step 5: Acquire real-time data of the reducer, including time T, gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4; Step 6: Calculate the correlation function L1 using gear vibration signal A1 and sound signal A2; calculate the correlation function L2 using speed signal A3 and sound signal A2; calculate the correlation function L3 using rigid housing vibration signal A4 and sound signal A2. If any of the following exists, discard the real-time data of the reducer; otherwise, retain the real-time data of the reducer: L1 ≤ preset association threshold LA; L2 ≤ preset association threshold LB; L3 ≤ preset association threshold LC; Step 7: Perform spectrum conversion on the gear vibration signal A1 of the reducer in the real-time data of the reducer, output the spectrum diagram, and calculate the spectral kurtosis; based on the speed and spectral kurtosis in the real-time data, determine whether the spectral kurtosis at this speed exceeds the corresponding spectral kurtosis critical value on the speed-spectral kurtosis correlation curve under slight abnormal conditions or under severe abnormal conditions. If not, it means the reducer is in normal condition, proceed to step 10; If so, then it depends on the situation: if it exceeds the critical value of spectral kurtosis on the speed-spectral kurtosis correlation curve under the slight abnormal state but does not exceed the critical value of spectral kurtosis on the speed-spectral kurtosis correlation curve under the severe abnormal state, then it belongs to the slight abnormal state. If the value exceeds the critical value of spectral kurtosis on the speed-spectral kurtosis correlation curve under severe abnormal conditions, then it is considered a severe abnormal condition. Step 8: Obtain the abnormal state through the abnormal state-spectral kurtosis-rotation speed relationship curve, and construct the state dataset Bn(Y, T, A1, A2, A3, A4). Step 9: Determine whether the abnormal states in Step 7 and Step 8 are consistent. If they are consistent, proceed to Step 10. If they are inconsistent, perform state calibration. Step 10: Output the reducer status Bn (Y, T, A1, A2, A3, A4).
2. The device status data analysis method based on spectral kurtosis according to claim 1, characterized in that, In step 1, the gear vibration signal A1, sound signal A2, speed signal A3, and rigid housing vibration signal A4 of the reducer are all acquired by sensors, including vibration sensors, sound sensors, and speed sensors.
3. The device status data analysis method based on spectral kurtosis according to claim 2, characterized in that, In step 3, the spectral kurtosis includes the spectral kurtosis 1 of the gear vibration signal A1 and the spectral kurtosis 2 of the rigid shell vibration signal A4.
4. The device status data analysis method based on spectral kurtosis according to claim 3, characterized in that, In step 3, the speed-spectral kurtosis 1 correlation curve and the speed-spectral kurtosis 2 correlation curve under normal conditions are constructed based on the normal state dataset.
5. The device status data analysis method based on spectral kurtosis according to claim 4, characterized in that, In step 3, the speed-spectral kurtosis 1 correlation curve and the speed-spectral kurtosis 2 correlation curve under the slight abnormal state are constructed based on the slight abnormal state dataset.
6. The device status data analysis method based on spectral kurtosis according to claim 5, characterized in that, In step 3, the speed-spectral kurtosis 1 correlation curve and the speed-spectral kurtosis 2 correlation curve under severe abnormal state conditions are constructed based on the severe abnormal state dataset.
7. The device status data analysis method based on spectral kurtosis according to claim 6, characterized in that, In step 4, the relationship between the abnormal state type and spectral kurtosis 1 and rotational speed is established, and the abnormal state-spectral kurtosis 1-rotational speed relationship curve is generated; the relationship between the abnormal state type and spectral kurtosis 2 and rotational speed is established, and the abnormal state-spectral kurtosis 2-rotational speed relationship curve is generated.
8. The device status data analysis method based on spectral kurtosis according to claim 7, characterized in that, Time T, gear vibration signal A1 of the reducer, sound signal A2, speed signal A3, and rigid shell vibration signal A4 can all be range values or point values.
9. The device status data analysis method based on spectral kurtosis according to claim 8, characterized in that, The method for condition calibration is as follows: Randomly extract an abnormal state dataset from the abnormal state database, obtain the abnormal state through steps 7 and 8, and determine whether it is consistent with the abnormal state in the abnormal state dataset. If they match, output the abnormal status of both step 7 and step 8 simultaneously; If there is a discrepancy, the following situations may occur: If the abnormal state in step 7 is inconsistent with the abnormal state in the state dataset, then repeat step 7. If the abnormal state in step 8 is inconsistent with the abnormal state in the state dataset, then step 8 is repeated. The system then checks whether the abnormal states in steps 7 and 8 are consistent. If they are consistent, it proceeds to step 10; otherwise, it issues an abnormal alarm.