Future blindness lost data complementation method for equipment real-time health monitoring

Abstract

The invention relates to the technical field of mechanical equipment intelligent operation and maintenance, in particular to an equipment real-time health monitoring-oriented future blindness lost data completion method, which comprises the following steps of: reading real-time operation data of mechanical equipment; using an indicator to classify whether the real-time operation data is lost or not; constructing a distance measurement matrix to quantify the correlation influence degree between the lost data and forward known data; constructing a lag degradation function according to the distance measurement matrix; reading self-correlation information of forward known data, and fitting a weight matrix and an offset vector parameter value in the lag degradation function; measuring the influence weight of the forward known data on the lost data through a lag degradation function; constructing a monotonic matrix, and determining the monotonicity of the forward known data; and complementing the lost data according to the monotonic matrix of the forward known data and the influence weight of the forward known data on the lost data. According to the invention, the loss of degradation information can be reduced, and the original degradation track can be maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance of mechanical equipment, and more specifically, to a method for completing missing data of future blindness for real-time health monitoring of equipment. Background Art

[0002] Modern manufacturing is inseparable from mechanical manufacturing. As an important part of the manufacturing industry, mechanical equipment plays a crucial role in improving production efficiency, reducing costs, and improving product quality. Predictive maintenance of mechanical equipment can avoid catastrophic failures, reduce maintenance costs, and effectively improve the safety, reliability, and economy of machines. Therefore, predictive maintenance of mechanical equipment is of great significance in the mechanical manufacturing industry. In the process of predictive maintenance of mechanical equipment, it is necessary to rely on real-time monitoring data for judgment. In actual industrial scenarios, data loss may occur due to reasons such as observation noise and sensor failure, and complete monitoring data cannot be obtained. The high data loss rate has brought an important obstacle to the predictive maintenance of mechanical equipment. How to complete the missing data during real-time monitoring is an important research issue.

[0003] In the existing health monitoring process, no effective method for completing missing data has been proposed. When encountering missing data, the main option is the direct elimination method, which will cause information loss and lead to deviation in data distribution, and is only applicable to cases with a small missing ratio. When the missing rate is high, the direct elimination method will cause a large amount of degraded information loss, which will directly affect the prediction effect and reduce the prediction accuracy. The commonly used method in the field of data processing is the filling method, which fills the missing data by using the statistical characteristics of the non-missing data before and after the missing data. The filling results obtained by this kind of method are stable and applicable to cases where the distribution characteristics of the missing sequence are simple and clear and the variable correlation is strong. The disadvantage of the filling method is that it needs to rely on the full-life operation data set for completion, that is, it needs to combine the forward and backward data relationships of the missing data for completion. However, only forward-known data can be obtained in real-time health monitoring, and data after the missing data cannot be obtained, that is, the future blindness situation. Therefore, the method for completing missing data in the field of data processing cannot be directly applied to the process of predictive maintenance of mechanical equipment.

[0004] Therefore, how to complete the data with as little loss of mechanical equipment degradation information as possible and with good completion performance under the condition of only knowing the forward data and a high data loss rate is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for completing missing data of future blindness for real-time health monitoring of equipment, which can reduce the loss of degraded information and maintain the original degradation trajectory.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for complementing missing data of future blindness for equipment real-time health monitoring, including the following steps:

[0008] Read the real-time operation data of the mechanical equipment;

[0009] Use an indicator to classify the status of whether the real-time operation data is missing;

[0010] When missing data appears, construct a distance metric matrix to quantify the degree of associated influence between the missing data and the forward known data;

[0011] Construct a lag degradation function according to the distance metric matrix to describe the degradation relationship between data;

[0012] Read the autocorrelation information of the forward known data and fit the weight matrix and bias vector parameter values in the lag degradation function;

[0013] Measure the influence weight of the forward known data on the missing data through the lag degradation function;

[0014] Construct a monotonic matrix to determine the monotonicity of the forward known data;

[0015] Complement the missing data according to the monotonic matrix of the forward known data and the influence weight of the forward known data on the missing data.

[0016] Further, the indicator is a binary vector used to represent whether the data is missing, indicating that the data at the time point is missing, indicating that the data at the time point is not missing.

[0017] Further, the construction process of the distance metric matrix includes:

[0018] When the expression of the distance metric matrix is: ;

[0019] When the expression of the distance metric matrix is: ;

[0020] wherein, is the previous absolute time interval, is the time position of the missing data, is the position of the previous non-lost data point; is the data feature loss pattern at the previous moment; is recursively extended to and the calculation of the distance between data is stopped. The smaller it is, the closer the data is in the time dimension, and the greater the degree of associated influence.

[0021] Furthermore, the autocorrelation of the forward preset number of non-lost running data is read through the autocovariance function. For the running data , the autocorrelation function of lag orders is expressed as:

[0022] ;

[0023] where is the autocovariance function of lag orders, is the autocovariance function of zero lag, is the number of running data, represents the selected time lag number, is the running data value at the moment, is the running data value at the moment, is the average value of the running data; combined with the missing data indicator , if and only if and are both 0 at the same time, has a calculated value.

[0024] Furthermore, the expression of the lag degradation function is:

[0025] ;

[0026] where ; is the weight matrix, the bias vector, which are the parameters to be fitted; represents the distance metric matrix.

[0027] Furthermore, the process of fitting the parameter values of the weight matrix and the bias vector in the lag degradation function includes:

[0028] is to combine the data autocorrelation information, take , read the data degradation information, that is, use the following model to fit the unknown parameters:

[0029] ;

[0030] where It is the distance metric vector corresponding to the operation data at the lag step;

[0031] Construct a loss function using the squared error and calculate the weight matrix by minimizing the loss and the bias vector , and the specific calculation formula is:

[0032] ;

[0033] Bring the calculated weight matrix and the bias vector into the lag degradation function.

[0034] Furthermore, the expression of the monotonic matrix is:

[0035] ;

[0036] Among them, is the time point position of the missing data. If the operation data is monotonically increasing as a whole, the monotonic matrix ; if the operation data is monotonically decreasing as a whole, the monotonic matrix .

[0037] Furthermore, the calculation formula for complementing the missing data is:

[0038] ;

[0039] Among them, is the data after complementing the missing data at the th time point, and are the influence weights of the forward known operation data and on the missing data ; is the global average value, that is, the average value of all forward known data; , indicating that the data at the th time point is missing.

[0040] In a second aspect, the present invention provides a future blind missing data complementing device for real-time health monitoring of equipment, including:

[0041] A data acquisition module for reading the real-time operation data of the mechanical equipment;

[0042] A data loss classification module for classifying the status of whether the real-time operation data is lost by using an indicator;

[0043] A distance metric module for constructing a distance metric matrix to quantify the associated influence degree between the missing data and the forward known data when missing data appears;

[0044] A data correlation description module, configured to construct a lag degradation function according to a distance metric matrix to describe the degradation relationship between data; read the autocorrelation information of forward known data, and fit the weight matrix and bias vector parameter values in the lag degradation function; measure the influence weight of the forward known data on the missing data through the lag degradation function.

[0045] A data monotonicity determination module, configured to construct a monotonic matrix to determine the monotonicity of forward known data.

[0046] A data completion module, configured to complete the missing data according to the monotonic matrix of the forward known data and the influence weight of the forward known data on the missing data.

[0047] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for completing future blind missing data for equipment real-time health monitoring as described above are implemented.

[0048] It can be seen from the above technical solutions that compared with the prior art, the present invention has the following beneficial effects:

[0049] 1) By constructing a distance metric matrix, the present invention quantifies the relationship between missing data and existing data, ensuring that the completion value will not affect the overall monotonicity of the degradation data of mechanical equipment.

[0050] 2) By constructing an autocovariance function to read the autocorrelation information of forward known data, and combining the autocorrelation information to fit the position parameters of the lag degradation function, and then calculating the influence weight of different forward known data on the missing data, and then completing the data in combination with the monotonicity of the forward known data. This process can capture the irregular lag relationship between the missing data and the forward observations to avoid losing valuable information. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0052] Figure 1 It is a flowchart of the method for completing future blind missing data for equipment real-time health monitoring provided by the present invention.

[0053] Figure 2 It is an overall architecture diagram of the method for completing future blind missing data for equipment real-time health monitoring provided by the present invention.

[0054] Figure 3 This is a schematic diagram for comparing the performance of the present invention with existing methods. Detailed implementation manners

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] As Figure 1 - Figure 2 shown, an embodiment of the present invention discloses a method for complementing missing data of future blindness for equipment real-time health monitoring, including the following steps:

[0057] S1. Read the real-time operation data of the mechanical equipment;

[0058] S2. Use an indicator to classify the status of whether the real-time operation data is missing;

[0059] S3. When missing data appears, construct a distance metric matrix to quantify the associated influence degree between the missing data and the forward known data;

[0060] S4. Construct a lag degradation function according to the distance metric matrix to describe the degradation relationship between data;

[0061] Read the autocorrelation information of the forward known data, and fit the weight matrix and bias vector parameter values in the lag degradation function;

[0062] S5. Measure the influence weight of the forward known data on the missing data through the lag degradation function;

[0063] Construct a monotonic matrix to determine the monotonicity of the forward known data;

[0064] Complement the missing data according to the monotonic matrix of the forward known data and the influence weight of the forward known data on the missing data.

[0065] The following further explains the above steps.

[0066] S1. Read the real-time operation data of the mechanical equipment. Obtain the real-time operation data of the mechanical equipment by using sensors, etc. Any data can be expressed as , where represents the operation data of the mechanical equipment, , is the dimension of the operation data, is a certain time point, Indicates the operating time of the mechanical equipment.

[0067] S2. Classify the real-time operation data.

[0068] Set a missing data indicator , the indicator is a binary vector used to represent whether the data is missing. Indicates the data at the time point is missing. Indicates the data at the time point is not missing, and the non-missing data is continued to be retained in the real-time operation data.

[0069] S3. Construct a distance metric matrix. The specific construction process includes:

[0070] The operation data of the mechanical equipment is time-varying. As the distance between data increases, the degree of associated influence between data will gradually decrease. And in the real situation, the missing positions are random and unevenly distributed. Therefore, a distance metric matrix is introduced to read the distance between the missing data and the non-missing data, which is used to reflect the degree of associated influence between data points.

[0071] When , the expression of the distance metric matrix is: ;

[0072] When , the expression of the distance metric matrix is: ;

[0073] Among them, is the previous absolute time interval, is the time position of the missing data, is the position of the previous non-missing data point; is the data feature missing pattern at the previous moment; is recursively calculated to , and the calculation of the distance between data is stopped. The smaller

[0074] S4. Obtain the temporal correlation between data, which specifically includes:

[0075] S41. Read the autocorrelation information of the non-missing data.

[0076] The operation data of mechanical equipment has autocorrelation. The autocovariance function is constructed to read the autocorrelation of the operation data. According to the position of the missing data, the operation data is grouped, and each group consists of the missing data and the data of the previous ten time points. For the operation data , the autocorrelation function with a lag of orders is expressed as:

[0077] ;

[0078] where is the autocovariance function with a lag of orders, is the autocovariance function with zero lag, is the number of operation data, represents the selected time lag number, is the operation data value at time is the operation data value at time is the average value of the operation data; combined with the missing data indicator , when and only when and are both 0, has a calculated value.

[0079] S42. Read the temporal correlation between the missing data and the forward known data.

[0080] is to read the temporal correlation between the non-missing data and the known data, maintain the monotonicity of the original data, and reduce the loss of degraded information. The lag degradation function is introduced to describe the degradation relationship between the data, and its expression is:

[0081] ;

[0082] where ; is the weight matrix, the bias vector, which is the parameter to be fitted, and are fitted using the results of the autocorrelation function; represents the distance metric matrix.

[0083] S43. Parameter fitting and weight calculation.

[0084] is to combine the autocorrelation information of the data. Let be the influence weight of the forward known data on the missing data. Take , read the data degradation information, that is, fit the unknown parameters using the following model:

[0085] ;

[0086] Among them, is the distance metric vector of the corresponding operation data at the lag step;

[0087] Using the squared error Construct a loss function, and by minimizing the loss, use the least squares method for fitting to calculate the weight matrix and the bias vector , and the specific calculation formula is:

[0088] ;

[0089] Substitute the calculated weight matrix and the bias vector into the lag degradation function.

[0090] S5. Complement the missing data based on the forward known data.

[0091] S51. Read the monotonicity of the operation data of the mechanical equipment.

[0092] Construct a monotonicity matrix , read the monotonicity of the forward known data, and avoid changing the trend of the original operation data when complementing the data. The specific expression is:

[0093] ;

[0094] Among them, is the time point position of the missing data. If the operation data is monotonically increasing as a whole, the monotonicity matrix ; if the operation data is monotonically decreasing as a whole, the monotonicity matrix .

[0095] Data loss may also occur in the initial stage of operation. When the second data is missing, it is impossible to determine the monotonicity matrix value only using the first data. The range of 3 - 5 is delimited because when the missing data points are within 3 - 5, there are at most 2 forward known data, and only 2 data can be used to determine. The range of 5 - 7 is delimited because there are only 4 forward known data. The more known data, the easier it is to determine the monotonicity of the operation data. However, too many data points are meaningless for determining the monotonicity. It is mainly necessary to focus on the data near the missing data. Therefore, it is set that the maximum number of data points used is 6.

[0096] S52. Combine the lag degradation function value representing the temporal correlation obtained above and the missing data indicator , and complement the missing data. Measure the missing data through the lag degradation function and the forward known data and The influence degree between them is defined as the weight and , and the missing data completion model is:

[0097] ;

[0098] Among them, is the data after missing data completion at the th time point, and are the forward known operating data and 's influence weights on the missing data ; is the global average value, that is, the average value of all forward known data; , indicating that the data at the th time point is missing.

[0099] The method of the present invention is a real-time missing data completion method. During the real-time monitoring process, when the first missing data point is encountered, the missing data completion mechanism is triggered. After completion, the completed data is incorporated into the known data set, and real-time monitoring continues until the next missing data is encountered.

[0100] When data is continuously missing, the previously completed data is incorporated into the known data. That is, after the first data is completed, the first completed data is regarded as known data, and then the next missing data is completed. That is, the completed data is used as a known data, combined with to complete the data.

[0101] To avoid the calculation result being too small due to continuous missing and affecting the correct summation calculation, a global is added for additional supplementation, and finally the missing data is completed by weighted summation.

[0102] S6. Real-time monitor the operating data. When missing data is encountered again, perform the next round of missing data completion, and repeat the above steps until all real-time monitored data is completed.

[0103] In other embodiments, the present invention also provides a future blind missing data completion device for equipment real-time health monitoring, including:

[0104] A data acquisition module for reading the real-time operating data of mechanical equipment;

[0105] A data loss classification module for classifying the status of whether the real-time operating data is lost by using an indicator;

[0106] ​A distance metric module, configured to construct a distance metric matrix when missing data occurs, so as to quantify the associated influence degree between the missing data and the forward-known data;

[0107] A data correlation description module, configured to construct a lag degradation function according to the distance metric matrix to describe the degradation relationship between data; read the autocorrelation information of the forward-known data, and fit the weight matrix and bias vector parameter values in the lag degradation function; measure the influence weight of the forward-known data on the missing data through the lag degradation function;

[0108] A data monotonicity determination module, configured to construct a monotonic matrix to determine the monotonicity of the forward-known data;

[0109] A data completion module, configured to complete the missing data according to the monotonic matrix of the forward-known data and the influence weight of the forward-known data on the missing data.

[0110] In another embodiment, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for completing future blind missing data for equipment real-time health monitoring as described above are implemented.

[0111] Next, taking the trend prediction of a rail transit bearing as an example, the effectiveness of the method of the present invention was verified through experimental bench prediction simulation data.

[0112] The test bearing used in this experiment is an HRB 352213 double-row tapered roller bearing. The bearing-related parameters and fault characteristic frequencies are shown in Table 1. The test bearing in the experimental bench used is driven by a belt driven by a drive motor (three-phase AC asynchronous motor). The axial and radial loads of the test bearing are applied to the bearing support through a hydraulic loading device. This embodiment is a constant-condition experiment, and the working conditions are set as the motor speed of 2500 r / min, the axial load of -20 kN, and the radial load of 60 kN. A three-axis acceleration sensor is used to monitor the test bearing and the support bearing, a current clamp is used to monitor the three-phase current of the motor, and a speed sensor is used to measure the speed (rotation frequency) in real time. The sampling strategy for the bearing vibration operation signal is set as follows: the sampling frequency is 25.6 kHz, the sampling interval is 30 s, and the duration of each sampling is 3.84 s, that is, each sample contains 98304 data points.

[0113] Table 1. Geometric parameters of HRB 352213 and its fault characteristic frequencies

[0114]

[0115] Table 2 gives the actual operating life and failure modes of each test bearing.

[0116] Table 2. Experimental bench bearing accelerated degradation experiment information

[0117]

[0118] The method proposed by the present invention was used to complete the bearing operation data in the experiment. The method proposed by the present invention was compared with three common methods, namely KNN, forward filling, and Random Forest. Before completing the missing data, for the data loss situation in the real scenario of the mode, the bearing operation data was grouped, and six missing modes were randomly set for each group of data. The specific missing scenario settings are shown in Table 3.

[0119] Table 3. Missing scenario settings

[0120]

[0121] The running data was set with different missing rates for comparison. Table 4 and Figure 3 lists the comparison effects of the poor prognosis and average relative error of the running data of three bearings with five missing rates respectively.

[0122] Table 4 Comparison table of average relative errors of four methods

[0123]

[0124] It can be seen from the experimental results that as the missing rate increases, the completion effect of the method of the present invention is significantly better than the other three comparison methods, and at the same time, the degradation information of the bearing is better retained.

[0125] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0126] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A future blind missing data completion method for real-time health monitoring of equipment, characterized in that: The following steps are involved: Read real-time operation data of mechanical equipment; Using indicators to classify the status of whether real-time operation data is lost; When missing data occurs, a distance metric matrix is ​​constructed to quantify the impact of the association between the missing data and the forward known data; A hysteresis degradation function is constructed based on the distance metric matrix to describe the degradation relationship between data; Read the autocorrelation information of the forward known data, and fit the weight matrix and bias vector parameter values ​​in the hysteresis degradation function; The influence weight of forward known data on missing data is measured by the hysteresis degradation function; Construct a monotone matrix to determine the monotonicity of the forward known data; The missing data is completed according to the monotone matrix of the forward known data and the influence weight of the forward known data on the missing data.

2. The future blind missing data completion method for real-time health monitoring of equipment according to claim 1 is characterized in that: Indicator is a binary vector used to indicate whether data is lost. Indicates Point in time data Lost, Indicates Point in time data Not lost.

3. The future blind missing data completion method for real-time health monitoring of equipment according to claim 1 is characterized in that: The construction process of the distance metric matrix includes: when When the distance metric matrix The expression is: ; when When the distance metric matrix The expression is: ; in, is the last absolute time interval, is the time position of the lost data, is the position of the last data point that was not lost; is the missing mode of data features at the previous moment; Recursion to , stop calculating the distance between data, The smaller it is, the closer the distance between the data in the time dimension is, and the greater the degree of correlation influence is.

4. The future blind missing data completion method for real-time health monitoring of equipment according to claim 1 is characterized in that: The autocorrelation function is used to read the autocorrelation of the forward preset number of non-missing running data. , hysteresis The autocorrelation function of order is expressed as: ； in, Lag The autocovariance function of order is is the zero-lag autocovariance function, is the number of running data, represents the selected number of time lags, for Running data value at all times, for Running data value at all times, is the average of the running data; combined with the missing data indicator , if and only if and At the same time, Only then will there be a calculated value.

5. The future blind missing data completion method for real-time health monitoring of equipment according to claim 4 is characterized in that: The expression of the hysteresis degradation function is: ； in, ; is the weight matrix, The bias vector is the parameter that needs to be fitted; Represents the distance metric matrix.

6. The future blind missing data completion method for real-time health monitoring of equipment according to claim 5 is characterized in that: The process of fitting the weight matrix and bias vector parameter values ​​in the hysteresis degradation function includes: In order to combine the data with relevant information, , read the data degradation information, that is, use the following model to fit the unknown parameters: ； in, is the distance metric vector corresponding to the running data under the lag step; The loss function is constructed using the square error, and the weight matrix is ​​calculated by minimizing the loss. and the bias vector , the specific calculation formula is: ； The calculated weight matrix and the bias vector Substitute into the hysteresis degradation function.

7. The future blind missing data completion method for real-time health monitoring of equipment according to claim 1 is characterized in that: The expression of the monotone matrix is: ； in, is the time point where the data is lost. If the overall running data is monotonically increasing, the monotonic matrix ; If the overall running data is monotonically decreasing, the monotonic matrix .

8. The future blind missing data completion method for real-time health monitoring of equipment according to claim 1 is characterized in that: The calculation formula for completing the missing data is: ； in, For the The data after the missing data at the time point is supplemented, and The forward known running data are and For lost data Impact weight; is the global average, that is, the average of all forward known data; , indicating the Data is lost at a certain point in time.

9. A future blind missing data completion device for real-time health monitoring of equipment, characterized in that: include: Data acquisition module, used to read real-time operation data of mechanical equipment; A data loss classification module, used for classifying the state of whether the real-time operation data is lost by using an indicator; The distance measurement module is used to construct a distance measurement matrix when missing data occurs, so as to quantify the degree of correlation between the missing data and the forward known data; The data correlation description module is used to construct a hysteresis degradation function based on the distance measurement matrix to describe the degradation relationship between data; read the autocorrelation information of the forward known data, and fit the weight matrix and bias vector parameter values ​​in the hysteresis degradation function; The influence weight of forward known data on missing data is measured by the hysteresis degradation function; A data monotonicity determination module is used to construct a monotonic matrix and determine the monotonicity of forward known data; The data completion module is used to complete the missing data according to the monotone matrix of the forward known data and the influence weight of the forward known data on the missing data.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the future blind missing data completion method for real-time health monitoring of equipment as described in any one of claims 1 to 8 are implemented.

Images