Trend alarm method and system for motor vibration data

Abstract

The invention discloses a trend alarm method and system for motor vibration data, and relates to the technical field of motor vibration data alarm. According to the method, the adaptability of threshold value judgment is improved, a dynamic threshold value setting method based on statistical analysis is adopted, a fixed static threshold value is not depended on any more, statistical analysis is conducted on half-year data, 99% quantiles are selected as the threshold value, and the system can more accurately adapt to changes of equipment in different operation states; compared with the prior art, the method has the advantages that the occurrence rate of misinformation and missed information is remarkably reduced, the data missing processing method is improved, the nearest neighbor filling technology (KNN) is used for data complementation, and compared with a traditional simple linear interpolation method, the integrity and accuracy of the data can be better kept. The KNN filling method can effectively avoid data distortion, thereby improving the precision of trend analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of motor vibration data alarm, and in particular to a trend alarm method and system for motor vibration data. Background Art

[0002] In the industrial manufacturing process, especially in high-temperature and high-pressure operating environments such as metallurgy and electrolysis, motor vibration (speed) monitoring plays a key role in the safe operation of equipment. Traditional motor vibration monitoring systems mainly rely on sensors to collect real-time data and determine whether an abnormality occurs based on preset thresholds. When the vibration amplitude or speed exceeds the preset range, the system will trigger an alarm to prompt the operator to perform inspection and maintenance.

[0003] Currently, these systems usually use experience-based threshold setting or simple statistical analysis methods, such as taking data within a certain period of time and setting a fixed threshold to determine the status of the equipment. The working principle of these methods is to identify abnormal operating status of the equipment by comparing the difference between real-time data and historical data.

[0004] The existing motor vibration monitoring system has exposed the following major problems and shortcomings in application:

[0005] Limitations of threshold setting: The thresholds used in the prior art are usually static, that is, they remain unchanged after being set once during the operation of the equipment. However, the normal vibration characteristics of the equipment may change under different operating stages and working conditions, and static thresholds are difficult to adapt to such dynamic changes, resulting in false alarms or missed alarms.

[0006] Inadequate processing of missing data: In actual industrial environments, sensor data is often missing due to equipment failure, communication interruption, etc. Existing technologies usually simply ignore these missing data or use basic methods such as linear interpolation to fill them, which can easily cause data distortion and affect subsequent analysis and judgment.

[0007] Lack of trend analysis and prediction capabilities: Most existing systems only focus on data anomalies at a single moment, but lack the ability to analyze the overall trend of the data. This means that the system cannot identify potential failure risks in advance, and usually issues an alarm only after a problem occurs, affecting the timeliness of maintenance.

[0008] Insufficient computing efficiency: As the amount of industrial data increases, existing systems often face the problem of low computing efficiency when processing large amounts of data, making it difficult to achieve real-time monitoring and analysis.

[0009] Therefore, a new solution to the above problems needs to be proposed. Summary of the invention

[0010] The object of the present invention is to provide a trend alarm method and system for motor vibration data to solve the technical problems raised in the background technology.

[0011] To achieve the above object, the present invention provides the following technical solutions: a trend alarm system for motor vibration data, comprising at least a data input module, a data preprocessing module, a threshold determination module, a change rate calculation module, a trend analysis module, a parallel computing module, an alarm module and a Flask interface module;

[0012] The data input module is used to receive motor vibration data and store it for subsequent processing;

[0013] The data preprocessing module includes a missing value filling processing unit, which uses the nearest neighbor filling method to complete the data to ensure the integrity and consistency of the data;

[0014] The threshold determination module sets a threshold through a statistical analysis method to determine whether the latest data requires further trend analysis;

[0015] The change rate calculation module is responsible for calculating the change rate of the data and identifying the time-dependent dynamic changes in the data;

[0016] The trend analysis module includes a local weighted regression analysis unit, which identifies the overall trend of the data through smoothing processing and provides a basis for alarm judgment;

[0017] The parallel computing module is used to speed up processing during data filling and trend analysis to improve the operating efficiency of the system;

[0018] The alarm module is used to determine whether an alarm needs to be issued based on the trend analysis results and transmit the alarm information to the operator;

[0019] The Flask interface module is responsible for connecting with the back-end system to realize data input and output of analysis results.

[0020] A trend alarm method for motor vibration data, comprising at least the following steps:

[0021] S1: Data preprocessing, preprocessing the collected motor vibration data;

[0022] S2: Perform threshold filtering, the pre-processed data will be compared with the set threshold to obtain the threshold determination;

[0023] S3: Change rate analysis: For valid data that passes the threshold, the algorithm calculates the change rate of the data to capture the dynamic changes of the data and identify potential abnormal fluctuations;

[0024] S4: Local weighted regression analysis, further using local weighted regression technology to perform trend analysis on the data. Local weighted regression technology effectively smooths the noise in the data, captures the overall trend of the data, and helps the system make more accurate judgments;

[0025] S5: Trend output and alarm. Based on the trend analysis results, the system determines whether an alarm needs to be issued. When an abnormal trend is detected, the system sends an alarm to the operator through the alarm module to remind him to intervene in time.

[0026] S6: Parallel computing;

[0027] S7: Connecting with the backend system through the Flask interface module, responsible for receiving input data and outputting analysis results. The Flask interface module receives requests from the external system, passes the data to the algorithm module for processing, and returns the final analysis results to the caller or stores them in the backend database.

[0028] Furthermore, the data preprocessing is used to process the data of the last five weeks to ensure sensitivity to the latest status. The nearest neighbor filling technology is used for missing data during the data preprocessing process. The nearest neighbor filling technology fills the missing values ​​by finding the closest sample data to ensure data integrity.

[0029] Furthermore, the threshold in S2 is obtained through statistical analysis of half-year data, and the 99% quantile is used as the judgment standard for the shutdown state of the device.

[0030] Furthermore, the processing method in S2 includes at least two cases: when the latest data is greater than the threshold and when the latest data is less than the threshold;

[0031] When the latest data is greater than the threshold, the system considers that the device may be in an abnormal state and continues to perform subsequent trend analysis;

[0032] When the latest data is less than the threshold, the system determines that the device is in shutdown state and stops further analysis.

[0033] Further, the S6 at least includes the following steps:

[0034] In the nearest neighbor filling method and local weighted regression analysis process, parallel computing technology is used to speed up the calculation and ensure that the system can respond quickly;

[0035] The parallel computing technology uses the Parallel and delayed modules in Python's joblib library.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The adaptability of threshold determination is improved. The present invention adopts a dynamic threshold setting method based on statistical analysis, and no longer relies on a fixed static threshold. By statistically analyzing the data for half a year and selecting the 99% quantile as the threshold, the system can more accurately adapt to the changes of the equipment under different operating conditions, significantly reducing the incidence of false positives and false negatives.

[0038] 2. Improved data missing processing method. The present invention uses the nearest neighbor filling technology (KNN) for data completion. Compared with the traditional simple linear interpolation method, the present invention can better maintain the integrity and accuracy of the data. The KNN filling method can effectively avoid data distortion, thereby improving the accuracy of trend analysis.

[0039] 3. Enhanced dynamic trend analysis capability. By introducing rate of change analysis and local weighted regression (LOESS) technology, the present invention enables the system to not only identify abnormalities in instantaneous data, but also analyze the overall trend of equipment operation in real time. Through this trend analysis, the system can identify potential failure risks in advance, provide timely warnings, and improve the foresight and effectiveness of equipment maintenance.

[0040] 4. The computational efficiency is greatly improved. In the process of KNN filling and LOESS regression analysis, the present invention adopts parallel computing technology (through the Parallel and delayed modules in Python's joblib library), which effectively improves the speed of data processing and calculation. This enables the system to maintain fast response when processing large-scale data and meet the needs of real-time monitoring.

[0041] 5. Simplified system integration and operation. The present invention realizes seamless connection with the back-end system by integrating a Flask-based interface module. The interface module makes the input of data and the output of analysis results easier, greatly simplifying the integration and operation process of the system. In addition, the alarm function provided by the system can immediately notify the operator when an abnormal trend is detected, so that the operator can take countermeasures quickly.

[0042] 6. Improved system reliability and stability. By introducing a series of advanced data processing and analysis technologies, the present invention significantly improves system reliability and stability. The system can accurately identify and warn of abnormal equipment conditions in complex industrial environments, reducing downtime caused by equipment failures, thereby improving overall production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of the process of the present invention;

[0045] Figure 2 It is a schematic diagram showing the original time domain signal of the motor vibration data collected over 5 weeks;

[0046] Figure 3 yes Figure 2 The motor vibration data samples in the figure are smoothed by LOESS regression;

[0047] Figure 4 yes Figure 2 The trend chart of the motor vibration data samples after the variation rate analysis;

[0048] Figure 5 yes Figure 2 The trend chart of the motor vibration data samples after LOESS regression analysis. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0050] Embodiment 1:

[0051] See also Figure 1-Figure 5 , a trend alarm method for motor vibration data, comprising at least the following steps:

[0052] Collect motor vibration (speed) data. The sampling frequency of the data is fixed and the collection period is 5 weeks. Please refer to Figure 2 .

[0053] The data is acquired through the sensor and transmitted to the data input module. In an embodiment of the present invention, the data input module is connected to a data storage system, and the collected motor vibration data is stored in a database for further analysis.

[0054] For the missing data in each sample, the nearest neighbor filling (KNN) technology is used to fill it. KNN filling fills the missing values ​​by finding the closest neighboring data points to ensure the integrity and consistency of the data. The filled samples are as follows:

[0055]

[0056] in, represents the filled data point, x(j) is the adjacent data point, and w j is the weight of the neighboring points;

[0057] The threshold is determined by statistically analyzing the historical data for half a year. This threshold is the 99% quantile of the data and serves as the criterion for determining the status of the device. The system compares the latest sampled data with the threshold: if the data is greater than the threshold, the subsequent trend analysis step is entered; if the data is less than or equal to the threshold, the device is determined to be in the shutdown state and no further analysis is performed.

[0058] The change rate of valid data that passes the threshold judgment is calculated, and the calculation formula of the change rate R is:

[0059]

[0060] Where x is the vibration amplitude data and t is the time. The rate of change is used to capture dynamic changes in the data and is an important basis for subsequent trend analysis. Figure 4 .

[0061]

[0062] Where x is the vibration amplitude data and t is the time. The rate of change is used to capture dynamic changes in the data and is an important basis for subsequent trend analysis. Figure 4 .

[0063] Perform local weighted regression (LOESS) analysis on the rate of change data. LOESS regression extracts the overall trend of the data by smoothing the noise, providing a reliable basis for trend judgment. Figure 3 The regression results determine whether the system triggers an alarm. For each sample that has undergone LOESS regression, the following are the results:

[0064]

[0065] Among them, y i is the trend value after regression, θ j is the look-back coefficient, K is the weight function, and λ is the smoothing parameter.

[0066] Based on the results of LOESS regression, the trend of motor vibration is determined. If the trend is abnormal, such as a continuous increase, the system triggers an alarm through the alarm module to notify the operator to intervene. Figure 5 .

[0067] The system outputs the analysis results to the backend system through the Flask interface module. The Flask interface module receives requests from the external system, passes the input data to the above processing modules, and returns the final analysis results to the caller in JSON format or stores them in the backend database.

[0068] Embodiment 2:

[0069] A trend alarm system for motor vibration data, comprising at least a data input module, a data preprocessing module, a threshold determination module, a change rate calculation module, a trend analysis module, a parallel calculation module, an alarm module and a Flask interface module;

[0070] The data input module is used to receive the motor vibration data and store it for subsequent processing;

[0071] The data preprocessing module includes a missing value filling processing unit, which uses the nearest neighbor filling method to complete the data to ensure the integrity and consistency of the data;

[0072] The threshold determination module sets the threshold through statistical analysis methods to determine whether the latest data requires further trend analysis;

[0073] The change rate calculation module is responsible for calculating the change rate of data and identifying the time-dependent dynamic changes in the data;

[0074] The trend analysis module includes a local weighted regression analysis unit, which identifies the overall trend of the data through smoothing processing and provides a basis for alarm judgment;

[0075] The parallel computing module is used to speed up processing during data filling and trend analysis, improving the operating efficiency of the system;

[0076] The alarm module is used to determine whether an alarm needs to be issued based on the trend analysis results and transmit the alarm information to the operator;

[0077] The Flask interface module is responsible for connecting with the back-end system to realize data input and output of analysis results.

[0078] The parallel computing technology uses the Parallel and delayed modules in Python's joblib library.

[0079] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A trend alarm system for motor vibration data, characterized in that: At least includes a data input module, a data preprocessing module, a threshold determination module, a change rate calculation module, a trend analysis module, a parallel computing module, an alarm module and a Flask interface module; The data input module is used to receive motor vibration data and store it for subsequent processing; The data preprocessing module includes a missing value filling processing unit, which uses the nearest neighbor filling method to complete the data to ensure the integrity and consistency of the data; The threshold determination module sets a threshold through a statistical analysis method to determine whether the latest data requires further trend analysis; The change rate calculation module is responsible for calculating the change rate of the data and identifying the time-dependent dynamic changes in the data; The trend analysis module includes a local weighted regression analysis unit, which identifies the overall trend of the data through smoothing processing and provides a basis for alarm judgment; The parallel computing module is used to speed up processing during data filling and trend analysis to improve the operating efficiency of the system; The alarm module is used to determine whether an alarm needs to be issued based on the trend analysis results and transmit the alarm information to the operator; The Flask interface module is responsible for connecting with the back-end system to realize data input and output of analysis results.

2. A trend alarm method for motor vibration data according to claim 1, used in a trend alarm system for motor vibration data according to the above claim, characterized in that: At least the following steps are included: S1: Data preprocessing, preprocessing the collected motor vibration data; S2: Perform threshold filtering, the pre-processed data will be compared with the set threshold to obtain the threshold determination; S3: Change rate analysis: For valid data that passes the threshold, the algorithm calculates the change rate of the data to capture the dynamic changes of the data and identify potential abnormal fluctuations; S4: Local weighted regression analysis, further using local weighted regression technology to perform trend analysis on the data. Local weighted regression technology effectively smooths the noise in the data, captures the overall trend of the data, and helps the system make more accurate judgments; S5: Trend output and alarm. Based on the trend analysis results, the system determines whether an alarm needs to be issued. When an abnormal trend is detected, the system sends an alarm to the operator through the alarm module to remind him to intervene in time. S6: Parallel computing; S7: Connecting with the backend system through the Flask interface module, responsible for receiving input data and outputting analysis results. The Flask interface module receives requests from the external system, passes the data to the algorithm module for processing, and returns the final analysis results to the caller or stores them in the backend database.

3. A trend alarm method for motor vibration data according to claim 2, characterized in that: The data preprocessing is used to process the data of the last five weeks to ensure sensitivity to the latest status. The nearest neighbor filling technology is used for missing data in the data preprocessing process. The nearest neighbor filling technology fills the missing values ​​by finding the closest sample data to ensure the integrity of the data.

4. A trend alarm method for motor vibration data according to claim 2, characterized in that: The threshold in S2 is obtained through statistical analysis of half-year data, and the 99% quantile is used as the judgment standard for the shutdown state of the device.

5. A trend alarm method for motor vibration data according to claim 4, characterized in that: The processing method in S2 includes at least two cases: when the latest data is greater than the threshold and when the latest data is less than the threshold; When the latest data is greater than the threshold, the system considers that the device may be in an abnormal state and continues to perform subsequent trend analysis; When the latest data is less than the threshold, the system determines that the device is in shutdown state and stops further analysis.

6. A trend alarm method and system for motor vibration data according to claim 2, characterized in that: The S6 at least comprises the following steps: In the nearest neighbor filling method and local weighted regression analysis process, parallel computing technology is used to speed up the calculation and ensure that the system can respond quickly; The parallel computing technology uses the Parallel and delayed modules in Python's joblib library.

Images