Mechanical equipment operation life timing method based on multi-feature fusion

Abstract

The invention relates to a mechanical equipment operation life timing method based on multi-feature fusion. The method comprises the steps of 1, data collection and classification; 2, frequency band screening; step 3, feature extraction; step 4, data fusion; 5, setting a threshold value; and step 6, starting and stopping recognition. According to the invention, the service life timer sensor is used for collecting the operation state of the electromechanical equipment in a timing manner, and starting and stopping time information is wirelessly transmitted to the upper computer on the panel for display in a Bluetooth manner. In the aspect of judging a start-stop algorithm, start-stop data of each working condition is collected in an earlier stage, data fusion of time-frequency domain parameters is combined to form a more representative index after dimension reduction, and a threshold configuration mode is newly added to personalize start-stop thresholds of each electromechanical device, so that spatial dependence is reduced, and the algorithm judgment precision is improved.

Description

Technical Field

[0001] The invention relates to the field of mechanical equipment status monitoring, and in particular to a method for timing the operating life of mechanical equipment. Background Art

[0002] As electromechanical equipment is used for a long time and is affected by the environment, its reliability will gradually decrease. Therefore, it is very important to keep statistics on the operating life of electromechanical equipment. This can help the maintenance team monitor the use of the equipment, predict the time for equipment repair and replacement, avoid sudden failures, and help optimize equipment maintenance plans and cost management. The current mainstream methods for keeping statistics on the operating life of electromechanical equipment mainly include manual regular inspection records and automatic data collection.

[0003] The manual regular inspection and recording method is relatively simple. The maintenance personnel only need to use a log book to record the start and stop time of each equipment startup and shutdown. This method is suitable for equipment management of various sizes. However, this method has the disadvantages of strong subjectivity, inaccuracy, low reliability, and high labor costs. Currently, automatic data collection methods are beginning to replace traditional manual recording methods.

[0004] The automatic data acquisition system uses advanced methods to automatically monitor and record equipment operation data, capture the operating status of the equipment, and automatically record the operating life of electromechanical equipment, thereby improving reliability and reducing labor costs. However, this method often requires large-scale laying of status monitoring systems, which is not suitable for scenarios with limited space and difficult wiring. In addition, the single preset start-stop judgment index will affect the accuracy of life statistics. Therefore, this method still needs to be improved in reducing spatial dependence and improving judgment accuracy. Summary of the invention

[0005] In view of the spatial dependence and single start-stop judgment index problems when the traditional automatic data acquisition system performs statistics on the equipment operating life, the present invention proposes a mechanical equipment operating life timing method based on multi-feature fusion, which collects the operating status of electromechanical equipment through the life timer sensor at regular intervals, and wirelessly transmits the start-stop time information to the upper computer on the tablet via Bluetooth for display. In the start-stop judgment algorithm, the initial data collects the start-stop data of each working condition, and then combines the data fusion of the time-frequency domain parameters to form a more representative index after dimensionality reduction, and adds a threshold configuration mode to personalize the start-stop threshold of each electromechanical equipment, thereby reducing spatial dependence and improving the algorithm judgment accuracy.

[0006] To achieve the above purpose, the technical solution of the present invention is: a method for timing the operating life of mechanical equipment based on multi-feature fusion, the steps of which are:

[0007] Step 1: Data collection and classification

[0008] Combined with the scenario in which the sensor is used, the interference conditions that may occur in the scenario are considered, and the start-up and shutdown vibration data under various conditions are collected for subsequent data analysis and feature extraction;

[0009] Step 2: Frequency band screening

[0010] After collecting the vibration data under various working conditions, the data is preprocessed and then feature extracted. In vibration analysis, the self-spectrum is used to analyze and interpret the vibration signal of the mechanical equipment, revealing the frequency components of the signal and its corresponding energy distribution. By analyzing these energy distributions, the frequency area where the vibration signal energy is concentrated when the mechanical equipment is started and stopped can be identified, and combined with the typical distribution range of high-frequency and low-frequency noise in the ship environment, the required frequency band range can be screened for the life timing method;

[0011] Step 3: Feature Extraction

[0012] In terms of frequency domain indicators, the total vibration value is selected as the start-stop judgment indicator;

[0013] In terms of time domain indicators, it is necessary to consider the interference conditions in the application scenario, and use the time domain indicators as the conditions for screening out interference conditions. By selecting typical time domain indicators, combined with the physical characteristics and comparison of the time domain indicators, the root mean square value and root square amplitude can be obtained to exclude the impact of vibration data on the time domain waveform and the influence of abnormal glitches. The root mean square value represents the vibration amplitude in the full frequency band, and the root square amplitude represents the overall energy and statistical stability of the signal to prevent false triggering. The calculation formulas for the root mean square value and root square amplitude are as follows. Therefore, the root mean square value and root square amplitude are selected as the start-stop judgment indicators in the time domain;

[0014] RMS value:

[0015] Square root amplitude:

[0016] Where x(n) is the value of each data point, E represents the mathematical expectation, and the mean is taken. rms is the RMS value of this data, X r is the square root amplitude of this data;

[0017] Step 4: Data Fusion

[0018] PCA principal component analysis is used as a data fusion method to simplify the data structure by reducing the dimension. After PCA dimension reduction, the first principal component is selected as the fused index. This index combines the information of the three original indicators and can effectively improve the reliability of the timing method.

[0019] Step 5: Threshold Setting

[0020] Different electromechanical equipment will have different time-frequency domain characteristics of vibration due to different models and different environments. Therefore, a threshold configuration mode is used to personalize the threshold of the start-stop index of different electromechanical equipment; a threshold configuration protocol is added to the communication protocol between the tablet host computer and the sensor. After the sensor receives the instruction packet and enters the threshold configuration mode, it will collect the vibration data at the start and stop time respectively according to the current state selected by the user on the tablet: start or stop, and calculate the index after the data fusion at the start and stop time through the algorithm, and then transmit the characteristic value data packet to the tablet host computer via Bluetooth to form two curves for display respectively. The user selects a value between the start curve and the stop curve as the threshold for judging the start and stop, and as the standard for subsequent algorithm judgment;

[0021] Step 6: Start and stop identification

[0022] In order to save installation space and simplify the layout difficulty, the sensor eliminates the signal line and power supply line; in order to make the sensor work for a long time, in actual scenarios, the sensor is normally in a dormant state, and the start and stop judgment is performed regularly according to the wake-up interval configured by the user; after the sensor is awakened, three sets of vibration acceleration data are continuously collected, and the PCA dimension reduction indicators of the three sets of data are calculated by an algorithm, and compared with the threshold set in step 5. When the indicators of the three sets of data are all greater than the set threshold, it is judged that the state of the electromechanical equipment detected by the wake-up sensor this time is started, otherwise it is stopped; at the same time, after each judgment of the start and stop state, the total equipment operation time will be updated once, and the new total equipment operation time = the old total equipment operation time + the length of this switch record time.

[0023] Furthermore, in step 1, a detailed operating condition table is first established to list the working environment and conditions that the sensor in this scenario will encounter to ensure the comprehensiveness and representativeness of the data, and the start-up and shutdown vibration data under the corresponding working conditions are collected according to the operating condition table, including normal interference-free, knocking interference, human walking interference, and hull shaking interference conditions when the mechanical equipment is started and shut down.

[0024] Furthermore, in step 2, considering the particularity of the onboard environment, such as the background noise that may be generated by the operation of the engine and the shaking of the hull caused by the impact of waves, the vibration data needs to be pre-filtered by frequency bands.

[0025] Furthermore, in step 3, in actual application scenarios, the vibration data of electromechanical equipment, such as the conditions with large interference to the total vibration value, include knocking and walking interference, so these interference conditions are mainly used when collecting sample data.

[0026] The beneficial effects of the present invention are:

[0027] The present invention specifically uses the time-frequency characteristics of vibration data, combined with wireless communication and host computer software to time the operating life of electromechanical equipment. The method is applicable to most electromechanical equipment, and has the following specific beneficial effects:

[0028] (1) Improve the accuracy of start-stop identification: By adding a "threshold configuration" mode on the sensor side and the host computer side, the thresholds of different electromechanical equipment can be personalized, thereby significantly improving the accuracy of start-stop identification.

[0029] (2) Enhanced robustness under interference conditions: Through feature extraction and data fusion, the square root amplitude, root mean square value, and total vibration value are reduced by PCA to obtain a new index. This index can not only characterize the vibration intensity within the selected frequency band, but also serve as a condition for screening out interference conditions (such as knocking and walking interference), making this method still effective in complex and harsh environments.

[0030] (3) Applicable to most scenarios: The command transmission and data communication of the present invention are based on Bluetooth sending and receiving, which is a wireless communication solution. This makes the method highly practical, does not require wiring, is easy to install, and is applicable to most scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flow chart of the method of the present invention;

[0032] Figure 2 is the band screening graph;

[0033] Figure 3 This is the data fusion recognition effect diagram;

[0034] Figure 4 It is the algorithm judgment logic block diagram. DETAILED DESCRIPTION

[0035] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0036] The mechanical equipment operation life timing method based on multi-feature fusion of the present invention includes the following steps. The specific process is as follows: Figure 1 As shown:

[0037] Step 1: Data collection and classification

[0038] In order to ensure the accuracy of the mechanical equipment operating life timing method proposed in this invention, it is necessary to combine the scenario in which the sensor is used, consider the interference conditions that may occur in the scenario, and collect the start-stop vibration data under each condition for subsequent data analysis and feature extraction. Therefore, first establish a detailed working condition table, list the working environment and conditions that the sensor in this scenario will encounter, to ensure the comprehensiveness and representativeness of the data, and collect the start-stop vibration data under the corresponding working conditions according to the working condition table. The working condition table is as follows:

[0039] Table 1 Working conditions

[0040]

[0041] Step 2: Frequency band screening

[0042] After collecting the vibration data under various working conditions, the data must be preprocessed before feature extraction can be performed. Taking into account the particularity of the onboard environment, such as the background noise that may be generated by engine operation and hull shaking caused by wave impact, this method requires pre-band filtering of the vibration data. In vibration analysis, the use of autospectral graphs can be used to analyze and interpret the vibration signals of mechanical equipment, revealing the frequency components of the signal and its corresponding energy distribution. By analyzing these energy distributions, the frequency areas where the vibration signal energy is concentrated when the mechanical equipment is started and stopped can be identified, and combined with the typical distribution range of high-frequency and low-frequency noise in the onboard environment, a suitable frequency band range can be screened for the life timing method.

[0043] Step 3: Feature extraction. In terms of frequency domain indicators, the total vibration value is a comprehensive representation of all frequency components of the equipment vibration signal in the selected frequency band. The total vibration value of the equipment is high when it is running and significantly reduced when it is shut down. Therefore, the total vibration value is selected as the start-stop judgment indicator in the frequency domain.

[0044]

[0045] Among them, A i It represents the amplitude corresponding to the i-th frequency point, where the value range of i is 1 to k, k means that this frequency domain data has k frequencies, and RMS represents the total vibration value of this frequency domain data.

[0046] In terms of time domain indicators, it is necessary to consider the interference conditions in the application scenario, and use the time domain indicators as conditions for screening out interference conditions. In actual application scenarios, the vibration data of electromechanical equipment, such as the total vibration value, has greater interference conditions including knocking and walking interference, so these interference conditions are mainly used when collecting sample data. By selecting typical time domain indicators, combining the physical characteristics and comparison of the time domain indicators, the root mean square value and root square amplitude can be obtained to exclude the impact of vibration data on the time domain waveform and the influence of abnormal glitches. The root mean square value represents the vibration amplitude in the full frequency band, and the root square amplitude represents the overall energy and statistical stability of the signal to prevent false triggering. The calculation formulas for the root mean square value and root square amplitude are as follows. Therefore, in the time domain, the root mean square value and root square amplitude are selected as the start-stop judgment indicators.

[0047] RMS value:

[0048] Square root amplitude:

[0049] Among them, x(n) is the value of each data point, E represents the mathematical expectation (i.e., the mean), X rms is the RMS value of this data, X r is the square root amplitude of this data.

[0050] Step 4: Data fusion. Data fusion is a process that combines information from multiple data sources to improve the accuracy and efficiency of decision-making. In the application scenario of this method, data fusion can integrate different measurement indicators in the time and frequency domains, reduce the uncertainty and bias caused by a single data source, and enhance the accuracy of algorithm judgment.

[0051] In the present invention, principal component analysis (PCA) is used as a method for data fusion. PCA is a statistical technique used to extract key information from multiple related variables and simplify the data structure by reducing the dimension. PCA calculates the covariance matrix of the data and then finds the eigenvectors and eigenvalues ​​of the matrix. These eigenvectors represent the most important directions in the data, while the eigenvalues ​​represent the importance of these directions. After PCA dimensionality reduction, the first principal component is selected as the fused index. This index combines the information of the three original indicators and can effectively improve the reliability of the timing method.

[0052] Step 5: Threshold setting. Different electromechanical equipment will have different time-frequency domain characteristics of vibration due to different models and different environments. Therefore, this scheme intends to design a threshold configuration mode to personalize the threshold of the start-stop index of different electromechanical equipment. A new threshold configuration protocol is added to the communication protocol between the tablet host computer and the sensor. After the sensor receives the instruction packet and enters the threshold configuration mode, it will collect the vibration data at the start and stop time respectively according to the current state (start or stop) selected by the user on the tablet, and calculate the index after the data fusion at the start and stop time through the algorithm, and then transmit the characteristic value data packet to the tablet host computer via Bluetooth to form two curves for display. The user can select a value between the start curve and the stop curve as the threshold for judging the start and stop, and as the standard for subsequent algorithm judgment.

[0053] Step 6: Start and stop identification. In order to save installation space and simplify the layout difficulty, the sensor cancels the signal line and power supply line. In order to make the sensor work for a long time, in actual scenarios, the sensor is normally in a dormant state, and the start and stop judgment is performed regularly according to the wake-up interval configured by the user. After the sensor is awakened, it will continuously collect 3 sets of vibration acceleration data, and use the algorithm to calculate the PCA dimension reduction indicators of the 3 sets of data, and compare them with the threshold set in step 5. When the indicators of the 3 sets of data are all greater than the set threshold, it is judged that the state of the electromechanical equipment detected by the wake-up sensor this time is started, otherwise it is stopped. At the same time, after each judgment of the start and stop status, the total operation time of the equipment will be updated once, and the total operation time of the equipment (new) = the total operation time of the equipment (old) + the length of the time of this switch record. Specific embodiment:

[0055] The present invention is a method for timing the life of equipment. The present invention is described in detail below by taking the life timing of a diesel generator as an example:

[0056] Step 1: Data collection and classification. The sensor is installed at the engine foot of the diesel generator, and the vibration data of the start-up and shutdown of the generator under four working conditions are collected and classified, including normal without interference, knocking interference, human walking interference, and hull shaking interference. This lays a data foundation for subsequent frequency band screening and feature extraction.

[0057] Step 2: Frequency band screening. Take the data collected in step 1, perform FFT calculation on the data according to startup and shutdown, and then draw the self-spectrum diagram, such as Figure 2 As shown, the energy distinction between the vibrations in the start-up and stop states is most obvious within the selected frequency band. Considering that the background noise caused by the hull shaking is generally in the low frequency band below 10 Hz, 10-1000 Hz is selected as the analysis frequency band. Specifically, a pre-processing algorithm of a bandpass filter is embedded in the sensor to achieve frequency band screening.

[0058] Step 3-4: Feature extraction and data fusion

[0059] The selected time domain index root mean square value, root square amplitude and frequency domain index total vibration value are fused into a new index after PCA dimension reduction as the index basis for judging the start and stop of the equipment. The PCA dimension reduction formula is as follows:

[0060] Var(Xproject)=m1∑i=1m(X1(i)ω1+X2(i)ω2+……+Xn(i)ωn)2

[0061] Where Xproject represents the result of projecting the original data X to reduce its dimension, X1(i), X2(i), …, Xn(i) represent the eigenvalues ​​(variables) in the original data matrix X. Assuming that the data set contains m samples and n features (dimensions), then Xj(i) represents the value of the i-th sample on the j-th feature. ω1, ω2, ωn represent the weight coefficients of the projection matrix, and each ωj represents the direction of a principal component, usually a unit vector, which defines the variance contribution in that direction.

[0062] The data collected in step 1 under the four working conditions can clearly distinguish the start and stop status of the electromechanical equipment and are not affected by the interference conditions. The specific effects are as follows: Figure 3 shown.

[0063] Step 5: Threshold Setting

[0064] After entering the host computer configuration interface, select the current device status as "off", and the diesel generator will not start. Then the host computer sends a collection instruction to the sensor. At this time, the sensor continuously collects 10 sets of vibration acceleration data at a sampling frequency of 2560Hz, with 1024 points in each set of data, and calculates the data fusion index of these 10 sets of data and sends it to the host computer. Then select the current device status as "on" in the configuration interface, start the diesel generator, and the sensor also collects, calculates and sends 10 sets of data fusion indicators of the startup state. Then the user sets the appropriate indicator threshold according to the level of the start-stop indicator curve on the configuration interface and sends it to the sensor.

[0065] Step 6: Start and stop identification

[0066] After configuring the threshold, the sensor enters the normal working state according to the 5-minute timed wake-up interval (configurable). The sensor wakes up every 5 minutes. Each time it wakes up, it will collect 3 groups of 1024 points of data and call the algorithm to calculate the data fusion index. It is compared with the preset threshold. When all are greater than the corresponding threshold, it is judged to be in the startup state and the RTC time of the current sensor is written into the startup time of the ledger. Otherwise, it is in the shutdown state and the total operation time of the equipment is updated at the same time.

[0067] The user can control the host computer at any time to obtain the start and stop records and the total operation time of the equipment recorded by the sensor. After receiving the command, the sensor will send the records to the host computer in the form of data packets via the Bluetooth protocol. The user can intuitively understand the start and stop records of the equipment since the sensor was installed. The algorithm judgment logic is as follows Figure 4 shown.

Claims

1. A method for timing the operating life of mechanical equipment based on multi-feature fusion, characterized in that: The steps are: Step 1: Data collection and classification Combined with the scenario in which the sensor is used, the interference conditions that may occur in the scenario are considered, and the start-up and shutdown vibration data under various conditions are collected for subsequent data analysis and feature extraction; Step 2: Frequency band screening After collecting the vibration data under various working conditions, the data is preprocessed and then feature extracted. In vibration analysis, the self-spectrum is used to analyze and interpret the vibration signal of the mechanical equipment, revealing the frequency components of the signal and its corresponding energy distribution. By analyzing these energy distributions, the frequency area where the vibration signal energy is concentrated when the mechanical equipment is started and stopped can be identified, and combined with the typical distribution range of high-frequency and low-frequency noise in the ship environment, the required frequency band range can be screened for the life timing method; Step 3: Feature Extraction In terms of frequency domain indicators, the total vibration value is selected as the start-stop judgment indicator; In terms of time domain indicators, it is necessary to consider the interference conditions in the application scenario, and use the time domain indicators as the conditions for screening out interference conditions. By selecting typical time domain indicators, combined with the physical characteristics and comparison of the time domain indicators, the root mean square value and root square amplitude can be obtained to exclude the impact of vibration data on the time domain waveform and the influence of abnormal glitches. The root mean square value represents the vibration amplitude in the full frequency band, and the root square amplitude represents the overall energy and statistical stability of the signal to prevent false triggering. The calculation formulas for the root mean square value and root square amplitude are as follows. Therefore, the root mean square value and root square amplitude are selected as the start-stop judgment indicators in the time domain; RMS value: Square root amplitude: Among them, x(n) is the value of each data point, E represents the mathematical expectation, and the mean is taken. rms is the RMS value of this data, X r is the square root amplitude of this data; Step 4: Data Fusion PCA principal component analysis is used as a data fusion method to simplify the data structure by reducing the dimension. After PCA dimension reduction, the first principal component is selected as the fused index. This index combines the information of the three original indicators and can effectively improve the reliability of the timing method. Step 5: Threshold Setting Different electromechanical equipment will have different time-frequency domain characteristics of vibration due to different models and different environments. Therefore, a threshold configuration mode is used to personalize the threshold of the start-stop index of different electromechanical equipment; a threshold configuration protocol is added to the communication protocol between the tablet host computer and the sensor. After the sensor receives the instruction packet and enters the threshold configuration mode, it will collect the vibration data at the start and stop time respectively according to the current state selected by the user on the tablet: start or stop, and calculate the index after the data fusion at the start and stop time through the algorithm, and then transmit the characteristic value data packet to the tablet host computer via Bluetooth to form two curves for display respectively. The user selects a value between the start curve and the stop curve as the threshold for judging the start and stop, and as the standard for subsequent algorithm judgment; Step 6: Start and stop identification In order to save installation space and simplify the layout difficulty, the sensor eliminates the signal line and power supply line; in order to make the sensor work for a long time, in actual scenarios, the sensor is normally in a dormant state, and the start and stop judgment is performed regularly according to the wake-up interval configured by the user; after the sensor is awakened, three sets of vibration acceleration data are continuously collected, and the PCA dimension reduction indicators of the three sets of data are calculated by an algorithm, and compared with the threshold set in step 5. When the indicators of the three sets of data are all greater than the set threshold, it is judged that the state of the electromechanical equipment detected by the wake-up sensor this time is started, otherwise it is stopped; at the same time, after each judgment of the start and stop state, the total equipment operation time will be updated once, and the new total equipment operation time = the old total equipment operation time + the length of this switch record time.

2. The method for timing the operating life of mechanical equipment based on multi-feature fusion according to claim 1 is characterized in that: In step 1, a detailed operating condition table is first established, listing the working environment and conditions that the sensor in this scenario will encounter to ensure the comprehensiveness and representativeness of the data, and the start-up and shutdown vibration data under the corresponding working conditions are collected according to the operating condition table, including normal interference-free, knocking interference, human walking interference, and hull shaking interference conditions when the mechanical equipment is started and shut down.

3. The method for timing the operating life of mechanical equipment based on multi-feature fusion according to claim 1 is characterized in that: In step 2, considering the particularity of the onboard environment, such as the background noise that may be generated by engine operation and hull shaking caused by waves, the vibration data needs to be pre-filtered by frequency bands.

4. The method for timing the operating life of mechanical equipment based on multi-feature fusion according to claim 1 is characterized in that: In step 3, the vibration data of electromechanical equipment in actual application scenarios, such as the conditions with large interference to the total vibration value, include knocking and walking interference, so these interference conditions are mainly used when collecting sample data.

Images