Pump station unit equipment predictive maintenance method based on dynamic threshold surface

Abstract

The invention discloses a pump station unit equipment predictive maintenance method based on a dynamic threshold surface. The method comprises the following steps: acquiring unit operation historical data information of a pump station unit; carrying out data cleaning and treatment on the obtained unit operation historical data information, and dividing the data into a plurality of data sets according to data types and working conditions; establishing a polynomial relationship between the state parameters and the operation condition based on the plurality of divided data sets to obtain dynamic thresholds of different state parameters; combining the obtained dynamic threshold value with an actual experience threshold value and a standard threshold value, displaying in a three-dimensional space, and constructing a dynamic threshold value model; and inputting the obtained real-time state monitoring data into the dynamic threshold model to obtain the running state of the current pump station unit equipment, and outputting a predictive maintenance result. According to the method, the dynamic threshold surface model combining historical health state sample data statistics and actual operation experience is used for evaluating the state of the pump set equipment in time, and predictive maintenance of the equipment is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of predictive maintenance of pump station unit equipment, and in particular to a predictive maintenance method for pump station unit equipment based on a dynamic threshold surface. Background Art

[0002] As a key component of the water diversion project, the health status of the pump station equipment plays a vital role in the stability and safety of the entire water diversion process. Therefore, real-time and accurate monitoring and evaluation of the health status of the pump group is of great significance for preventing failures, reducing maintenance costs, and improving production efficiency. Pump group equipment generally adopts regular maintenance and operation methods, but in the face of complex production and operation environments and equipment instability, equipment failures often occur, so the limitations of traditional maintenance methods are increasing. With the development of big data analysis technology, full use of the operating data of mechanical equipment and the collection of status parameters can achieve predictive maintenance of equipment.

[0003] Feng Fan et al. proposed an equipment health index model based on equipment health data, which includes equipment factory data, operation data, maintenance data, etc., to predict the health status of equipment and facilitate maintenance by business personnel. In order to reduce the operation and maintenance costs of equipment, Guan Yunquan et al. built an equipment big data health management system to analyze equipment monitoring data, warn and eliminate unstable factors in the safe operation of equipment, and greatly improve the level of equipment fault handling. Chen Weichong et al. took the operation status of pump stations as the research object, and improved the diagnostic efficiency of pump station operation faults by capturing abnormal data during the operation of large pump stations. Li Guoping designed a health monitoring and evaluation system for machine pumps based on the characteristics and needs of water equipment management, integrating equipment operation data and information in comprehensive monitoring, and improving the safety and feasibility of unit operation. Based on big data technology, Li Bao et al. combined the BP neural network method with the traditional expert system to establish a fault diagnosis system for sandy water flow pump stations to locate fault problems and diagnose fault causes in a timely manner.

[0004] Traditional pump health assessment methods often rely on empirical judgment or simple threshold analysis, which may not be accurate enough or respond slowly in complex and changing environments. In recent years, with the advancement of data acquisition technology and analysis methods, data-driven pump health assessment methods have received widespread attention. These methods analyze a large amount of monitoring data collected by pump equipment under different operating conditions and use statistics and machine learning algorithms to evaluate the health status of pumps. However, the multidimensionality and complexity of pump monitoring data, as well as the differences in data under different operating conditions, make it difficult to accurately assess the health of pumps.

[0005] Many existing studies use the comparison between status monitoring values ​​and single limit values ​​to achieve equipment status warning and evaluation. The status limit values ​​of unit equipment are usually determined by referring to relevant national standards, industry standards, and equipment manufacturer's recommended values. This single absolute value comparison method often cannot accurately reflect the differences in status parameter limit values ​​of unit equipment under different operating conditions. Moreover, most of the research is focused on the status of hydropower unit equipment, and there is less research on pump station unit equipment. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art. To achieve the above purpose, a predictive maintenance method for pump station unit equipment based on a dynamic threshold surface is adopted to solve the problems raised in the above background technology.

[0007] A predictive maintenance method for pump station equipment based on a dynamic threshold surface comprises the following steps:

[0008] Step S1, obtaining the unit operation history data information of the pump station unit;

[0009] Step S2: Clean and manage the acquired historical data of the unit operation, and divide the data into multiple data sets according to data type and operating conditions;

[0010] Step S3: establishing a polynomial relationship between the state parameter and the operating condition based on the divided multiple data sets to obtain dynamic thresholds of different state parameters;

[0011] Step S4, combining the obtained dynamic threshold with the actual experience threshold and the standard threshold, and displaying them in a three-dimensional space to construct a dynamic threshold model;

[0012] Step S5: input the acquired real-time status monitoring data into the dynamic threshold model to obtain the current operating status of the pump station unit equipment and output the predictive maintenance result.

[0013] As a further solution of the present invention: the specific steps in step S1 include:

[0014] Acquire the historical operation data information of the pump station unit, wherein the historical data information includes equipment status monitoring data, operation condition data, and maintenance records;

[0015] The equipment status monitoring data includes the monitoring temperature, vibration, and excitation current data of multiple sensors at different measuring points of the pump station unit equipment;

[0016] The operating condition data include the head of the device, active power, and blade placement angle, etc., and the equipment status monitoring data and the operating condition data correspond one to one in time;

[0017] The maintenance record data includes the start and stop time of the unit equipment, and component maintenance information.

[0018] As a further solution of the present invention: the specific steps in step S2 include:

[0019] Step S21, cleaning and managing the acquired unit operation history data information;

[0020] Detect abnormal data according to the 3σ criterion of the Laida criterion in statistics;

[0021] Step S22, dividing the state data and its corresponding operating condition data according to the state parameter type;

[0022] Step S23: sort the state parameters in the divided data set by size.

[0023] As a further solution of the present invention: the specific steps in step S3 include:

[0024] Step S31, establishing a polynomial relationship between the state parameters and the operating conditions in the divided and sorted data set, firstly performing a correlation analysis on the state parameters and the operating conditions, and selecting the operating conditions with a large correlation with the state parameters according to the correlation coefficient;

[0025] Step S32, determining the form of the polynomial after comparing the calculation result of the polynomial with the actual device state parameter monitoring value;

[0026] Step S33, solving the optimal parameters of the polynomial based on the optimization algorithm;

[0027] Step S34, calculating the state parameters on the data set according to the obtained polynomial to obtain the state thresholds under different operating conditions, wherein the state thresholds include a theoretical upper limit threshold and a theoretical reference threshold.

[0028] As a further solution of the present invention: the specific steps of step S32 include:

[0029] By testing the coefficient of certainty R 2 , and the statistical index of root mean square error RMSE, to test the accuracy of the form of the equipment state polynomial, and judge the polynomial between the equipment state parameters and the operating conditions according to the statistical index;

[0030] The specific calculation formula of statistical indicators is:

[0031]

[0032] In the formula, is the calculated value of the polynomial; z i is the actual state parameter monitoring value; is the actual state parameter monitoring mean; n is the number of actual state parameter monitoring values.

[0033] As a further solution of the present invention: the specific steps of step S33 include:

[0034] Set the historical monitoring data of the pump station equipment as discrete data points (x i ,y i ,z i ),(i=1,2,…,n);

[0035] Among them, z i is the target factor, i.e. the device state parameter, x i ,y i is the influencing factor, i.e., the operating condition;

[0036] The calculated value f(x i ,y i ) and the true value z i The residual sum of squares between is the smallest, and the formula is:

[0037]

[0038] Based on the least squares optimization algorithm, according to the historical monitoring data (x i ,y i ,z i ), (i=1,2,…,n), determine the coefficients of the binary multivariate polynomial a0,a1,…,a p The residual sum of squares is minimized to obtain the polynomial of the specific relationship between the operating conditions and the equipment state parameters.

[0039] As a further solution of the present invention: the specific steps in step S4 include:

[0040] Step S41: Determine the actual operation threshold of the state parameter according to the actual operation experience of the pump station management and maintenance personnel;

[0041] Step S42: determine the upper limit threshold of the state parameter according to the standard regulations and the equipment operation state limit of the pump group equipment.

[0042] Step S43, four threshold surfaces are drawn in a three-dimensional coordinate system according to the theoretical reference threshold, theoretical upper limit threshold, actual operation threshold, and specified upper limit threshold corresponding to different operating condition parameters to obtain a dynamic threshold model of the equipment status.

[0043] As a further solution of the present invention: the specific steps in step S5 include:

[0044] Step S51, marking the status monitoring data of the equipment during operation in the dynamic threshold model;

[0045] Step S52: According to the real-time status monitoring data values ​​being below different threshold ranges, predictive maintenance is performed according to the output predictive maintenance results.

[0046] Compared with the prior art, the present invention has the following technical effects:

[0047] The above technical solution is adopted, through in-depth analysis of the status data of the pump station unit during normal operation for many years, and mathematical statistics methods are used to fit a surface that accurately reflects the status change of the pump equipment in actual operation. The surfaces fitted under different status data are combined to obtain the spatial variation range of the state quantity of the pump equipment during normal operation, and all surfaces are intuitively displayed through a three-dimensional view, providing accurate health status assessment of the pump equipment for real-time monitoring data, which has certain guiding significance for timely discovery of potential equipment status problems and predictive maintenance. The dynamic threshold surface model that combines historical health status sample data statistics and actual operating experience is used to timely evaluate the status of the pump equipment and realize predictive maintenance of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings:

[0049] Figure 1 This is a schematic diagram of the structure of the embodiment disclosed in this application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Please refer to Figure 1 In an embodiment of the present invention, a predictive maintenance method for pump station equipment based on a dynamic threshold surface includes the following steps:

[0052] Step S1, obtaining the historical operation data information of the pump station unit, the specific steps include:

[0053] Obtain the historical operation data information of the pump station units, including equipment status monitoring data, operating condition data, and maintenance records;

[0054] The equipment status monitoring data includes the monitoring temperature, vibration, and excitation current data of multiple sensors at different measuring points of the pump station unit equipment;

[0055] The operating condition data includes the head of the device, active power, and blade placement angle, etc., and the equipment status monitoring data and the operating condition data correspond one to one in time;

[0056] Maintenance record data includes the start and stop time of the unit equipment, as well as component maintenance information.

[0057] Step S2: Clean and manage the acquired historical data of the unit operation, and divide the data into multiple data sets according to the data type and operating conditions. The specific steps include:

[0058] Step S21, cleaning and managing the acquired unit operation history data information;

[0059] Detect abnormal data according to the 3σ criterion of the Laida criterion in statistics;

[0060] Specifically, according to the 3σ criterion, abnormal data is detected. For a certain observation, if the deviation of the observed value from the overall expectation follows a normal distribution, then the interval (μ-3σ, μ+3σ) can be used as the basis for judging whether the observed value is out of limit. For the data that exceeds the limit, check the time when the data is generated and compare it with the historical records of the equipment operation. If there is no abnormal record of the equipment, the data that exceeds the limit is the result of abnormal sensor monitoring, which can be directly eliminated and the missing data can be filled using interpolation. The data monitored by different sensors at the same part of the equipment are analyzed and compared using the mean (μ) and mean square error (σ). After excluding data with large differences in the mean and mean square error of other monitoring data, the monitoring data of the remaining sensors at the same part are averaged as the status data of the equipment at that part.

[0061] Step S22: dividing the state data and its corresponding operating condition data according to the state parameter type, for example, dividing the temperature state data by month or quarter, dividing the original data set into 12 months or four quarters of data sets;

[0062] Step S23: sort the state parameters in the divided data set by size.

[0063] In this embodiment, each data set is further divided into two data sets, D5 and D7, according to the first 5% of the data and the last 95% of the data. 95 .

[0064] Step S3: establishing a polynomial relationship between the state parameters and the operating conditions based on the divided multiple data sets to obtain dynamic thresholds of different state parameters. The specific steps include:

[0065] Step S31, establishing a polynomial relationship between the state parameters and the operating conditions in the divided and sorted data set, firstly performing a correlation analysis on the state parameters and the operating conditions, and selecting the operating conditions with a large correlation with the state parameters according to the correlation coefficient;

[0066] Step S32, determining the form of the polynomial after comparing the calculation result of the polynomial with the actual device state parameter monitoring value;

[0067] In this embodiment, the specific steps of step S32 include:

[0068] By testing the coefficient of certainty R 2 , and the statistical index of root mean square error RMSE, to test the accuracy of the form of the equipment state polynomial, and judge the polynomial between the equipment state parameters and the operating conditions according to the statistical index;

[0069] The specific calculation formula of statistical indicators is:

[0070]

[0071] In the formula, is the calculated value of the polynomial; z i is the actual state parameter monitoring value; is the actual state parameter monitoring mean; n is the number of actual state parameter monitoring values.

[0072] Step S33, solving the optimal parameters of the polynomial based on the optimization algorithm;

[0073] Assume that the historical monitoring data of the pump station equipment is set as discrete data points (x i ,y i ,z i ),(i=1,2,…,n);

[0074] Among them, z i is the target factor, i.e. the device state parameter, x i ,y i is the influencing factor, i.e., the operating condition;

[0075] The calculated value f(x i ,y i ) and the true value z i The residual sum of squares between is the smallest, and the formula is:

[0076]

[0077] Based on the least squares optimization algorithm, according to the historical monitoring data (x i ,y i ,z i), (i=1,2,…,n), determine the coefficients of the binary multivariate polynomial a0,a1,…,a p The residual sum of squares is minimized to obtain the polynomial of the specific relationship between the operating conditions and the equipment state parameters.

[0078] Step S34: Calculate D5 and D according to the obtained polynomial 95 The state parameters on the data set are used to obtain state thresholds under different operating conditions, wherein the state thresholds include a theoretical upper limit threshold and a theoretical reference threshold.

[0079] In this embodiment, the state threshold of the D5 data set is the theoretical upper limit threshold, and D 95 The state threshold on the dataset is the theoretical benchmark threshold

[0080] Step S4: combining the obtained dynamic threshold with the actual experience threshold and the standard threshold, and displaying them in a three-dimensional space to construct a dynamic threshold model. The specific steps include:

[0081] Step S41: Determine the actual operation threshold of the state parameter based on the actual operation experience of the pump station management and maintenance personnel, which can be increased by 5% to 20% based on the theoretical upper limit threshold;

[0082] Step S42: determine the upper limit threshold of the state parameter according to the standard regulations and the equipment operating state limit on the factory nameplate of the pump set equipment.

[0083] Step S43: According to different operating condition parameters (x i ,y i ) corresponding to the theoretical benchmark threshold, theoretical upper limit threshold, actual operation threshold, and specified upper limit threshold are plotted in a three-dimensional coordinate system to obtain a dynamic threshold model of the equipment status.

[0084] Step S5: input the acquired real-time status monitoring data into the dynamic threshold model to obtain the current operating status of the pump station equipment and output the predictive maintenance results. The specific steps include:

[0085] Step S51, marking the status monitoring data of the equipment during operation in the dynamic threshold model;

[0086] Step S52: According to the real-time status monitoring data values ​​being below different threshold ranges, predictive maintenance is performed according to the output predictive maintenance results.

[0087] Specifically, if the real-time status monitoring data value is below the theoretical upper limit threshold, it indicates that the equipment status is normal; if the real-time status monitoring data value is between the actual operation threshold and the specified upper limit threshold, it indicates that the equipment may need to be inspected and maintained in advance; if the real-time status monitoring data value is above the specified upper limit threshold, it indicates that the equipment needs to be shut down immediately for inspection and maintenance.

[0088] In this embodiment, the sensor monitoring data is analyzed and processed to avoid the error of treating the sensor data as the device status data and to eliminate the situation where the device status is incorrectly judged due to sensor reasons;

[0089] Four thresholds are used to evaluate the equipment status, avoiding the rough limitation of a single threshold.

[0090] By making full use of the equipment operation history data and actual operation experience, the monitoring and evaluation of equipment status has more practical application value, reducing the possibility of excessive equipment maintenance and saving production operation costs;

[0091] The theoretical benchmark threshold, theoretical upper limit threshold, actual operation threshold and specified upper limit threshold are displayed in a three-dimensional coordinate system in combination with the operating conditions to form a dynamic threshold surface. The position of the real-time status monitoring data in the three-dimensional coordinate system can be clearly seen, and the current status of the equipment can be clearly displayed and judged, providing a basis for predictive maintenance of the equipment.

[0092] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents, and all should be included in the scope of protection of the present invention.

Claims

1. A predictive maintenance method for pump station equipment based on a dynamic threshold surface, characterized in that: The following steps are involved: Step S1, obtaining the unit operation history data information of the pump station unit; Step S2: Clean and manage the acquired historical data of the unit operation, and divide the data into multiple data sets according to data type and operating conditions; Step S3: establishing a polynomial relationship between the state parameter and the operating condition based on the divided multiple data sets to obtain dynamic thresholds of different state parameters; Step S4, combining the obtained dynamic threshold with the actual experience threshold and the standard threshold, and displaying them in a three-dimensional space to construct a dynamic threshold model; Step S5: input the acquired real-time status monitoring data into the dynamic threshold model to obtain the current operating status of the pump station unit equipment and output the predictive maintenance result.

2. According to claim 1, a method for predictive maintenance of pump station equipment based on a dynamic threshold surface, characterized in that: The specific steps in step S1 include: Acquire the historical operation data information of the pump station unit, wherein the historical data information includes equipment status monitoring data, operation condition data, and maintenance records; The equipment status monitoring data includes the monitoring temperature, vibration, and excitation current data of multiple sensors at different measuring points of the pump station unit equipment; The operating condition data include the head of the device, active power, and blade placement angle, etc., and the equipment status monitoring data and the operating condition data correspond one to one in time; The maintenance record data includes the start and stop time of the unit equipment, and component maintenance information.

3. According to claim 1, a method for predictive maintenance of pump station equipment based on a dynamic threshold surface, characterized in that: The specific steps in step S2 include: Step S21, cleaning and managing the acquired unit operation history data information; Detect abnormal data according to the 3σ criterion of the Laida criterion in statistics; Step S22, dividing the state data and its corresponding operating condition data according to the state parameter type; Step S23: sort the state parameters in the divided data set by size.

4. According to claim 1, a method for predictive maintenance of pump station equipment based on a dynamic threshold surface is characterized in that: The specific steps in step S3 include: Step S31, establishing a polynomial relationship between the state parameters and the operating conditions in the divided and sorted data set, firstly performing a correlation analysis on the state parameters and the operating conditions, and selecting the operating conditions with a large correlation with the state parameters according to the correlation coefficient; Step S32, determining the form of the polynomial after comparing the calculation result of the polynomial with the actual state monitoring value; Step S33, solving the optimal parameters of the polynomial based on the optimization algorithm; Step S34, calculating the state parameters on the data set according to the obtained polynomial to obtain the state thresholds under different operating conditions, wherein the state thresholds include a theoretical upper limit threshold and a theoretical reference threshold.

5. According to claim 1, a method for predictive maintenance of pump station equipment based on a dynamic threshold surface, characterized in that: The specific steps of step S32 include: By testing the coefficient of certainty R 2 , and the statistical index of root mean square error RMSE, to test the accuracy of the form of the equipment state polynomial, and judge the polynomial between the equipment state parameters and the operating conditions according to the statistical index; The specific calculation formula of statistical indicators is: In the formula, is the calculated value of the polynomial; z i is the actual state parameter monitoring value; is the actual state parameter monitoring mean; n is the number of actual state parameter monitoring values.

6. According to claim 1, a method for predictive maintenance of pump station equipment based on a dynamic threshold surface is characterized in that: The specific steps of step S33 include: Set the historical monitoring data of the pump station equipment as discrete data points (x i ,y i , z i ), (i=1,2,...,n); Among them, z i is the target factor, i.e. the device state parameter, x i ,y i is the influencing factor, i.e., the operating condition; The calculated value f(x i ,y i ) and the true value z i The residual sum of squares between is the smallest, and the formula is: Based on the least squares optimization algorithm, according to the historical monitoring data (x i ,y i , z i ), (i=1,2,...,n), determine the coefficients of the binary multivariate polynomial a0,a1,...,a p The sum of squares of the residuals is minimized and the polynomial of the specific relationship between the operating conditions and the equipment state parameters is obtained.

7. The method for predictive maintenance of pump station equipment based on a dynamic threshold surface according to claim 1, characterized in that: The specific steps in step S4 include: Step S41: Determine the actual operation threshold of the state parameter according to the actual operation experience of the pump station management and maintenance personnel; Step S42: determine the upper limit threshold of the state parameter according to the standard regulations and the equipment operating state limit of the pump group equipment. Step S43, four threshold surfaces are drawn in a three-dimensional coordinate system according to the theoretical reference threshold, theoretical upper limit threshold, actual operation threshold, and specified upper limit threshold corresponding to different operating condition parameters to obtain a dynamic threshold model of the equipment status.

8. The method for predictive maintenance of pump station equipment based on a dynamic threshold surface according to claim 1, characterized in that: The specific steps in step S5 include: Step S51, marking the status monitoring data of the equipment during operation in the dynamic threshold model; Step S52: According to the real-time status monitoring data values ​​being below different threshold ranges, predictive maintenance is performed according to the output predictive maintenance results.

Images