Calibration Method of Quartz Flexible Accelerometer

Through the linear regression model, the performance parameters of quartz flexible accelerometers are predicted, and the problems of frequent equipment use and high cost in the calibration method of quartz flexible accelerometer are solved, achieving a more efficient calibration process and cost savings.

CN116500300BActive Publication Date: 2024-11-08INSTITUTE OF GEOLOGY AND GEOPHYSICS CHINESE ACADEMY OF SCIENCES
View PDF 12 Cites 0 Cited by

Patent Information

Application Number
CN202211576272.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-12-09
Publication Date
2024-11-08
Estimated Expiration
2042-12-09

AI Technical Summary

Technical Problem

The calibration method of quartz flexible accelerometers requires a long time and a large number of special equipment, resulting in high costs and inefficiency.

Method used

The linear regression model is used to predict the performance parameters of the quartz flexible accelerometer based on the original data, reducing the number of use of special equipment and calibration test time, and improving calibration efficiency.

Benefits of technology

After the first few months, the quartz flexible accelerometer can complete calibration tests through a linear regression model, reducing the frequency and cost of special equipment and improving calibration efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116500300B_ABST
    Figure CN116500300B_ABST
Patent Text Reader

Abstract

The present invention discloses a calibration method for a quartz flexible accelerometer, comprising the following steps: S100 obtaining raw data of the quartz flexible accelerometer at a plurality of time nodes; S200 establishing a linear regression model based on the raw data; S300 forward predicting predicted values ​​of performance parameters of the quartz flexible accelerometer at a plurality of target time nodes based on the linear regression model. In the present application scheme, the calibration efficiency of the quartz flexible accelerometer is improved by establishing a linear regression model, and the number of times the special equipment is used is reduced, thereby saving costs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of sensors, and in particular to a calibration method for a quartz flexible accelerometer. Background Art

[0002] The calibration method of quartz flexible accelerometer is a core inertial sensor in the inertial navigation system. It is usually used to measure the linear acceleration of the carrier. After being solved by the navigation computer, the speed and position information of the carrier relative to the navigation coordinate system can be obtained.

[0003] In the related art, the calibration method of the quartz flexible accelerometer usually performs four-position flip calibration six times within a month, and obtains the measured values ​​of the performance parameters of the quartz flexible accelerometer each time. However, the quartz flexible accelerometer usually needs to be calibrated and tested on dedicated equipment for 1 month, 3 months, 6 months, 12 months or even longer, so a large amount of dedicated equipment is required, resulting in high costs. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a calibration method for a quartz flexible accelerometer. The calibration method for a quartz flexible accelerometer according to an embodiment of the present invention comprises the following steps:

[0005] S100 obtains the raw data of the quartz flexible accelerometer at several time points;

[0006] S200 builds a linear regression model based on the original data;

[0007] S300 forward predicts predicted values ​​of performance parameters of the quartz flexible accelerometer at several target time nodes according to the linear regression model.

[0008] The calibration method of the quartz flexible accelerometer according to the embodiment of the present invention has at least the following beneficial effects: after the initial few months, the quartz flexible accelerometer can complete the calibration test at each time node through the linear regression model. As can be seen from the above, the calibration method of the quartz flexible accelerometer in the present application scheme improves the calibration efficiency of the quartz flexible accelerometer, and the number of times the special equipment is used is reduced, saving costs.

[0009] According to some embodiments of the present invention, step S300 includes the following steps:

[0010] S310: obtaining the original data of the current time node of the quartz flexible accelerometer and using it as the variable value of the linear regression model;

[0011] S320 combines the linear regression model to forward predict the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes;

[0012] S330 loops step S310 and step S320 until the calibration of the quartz flexible accelerometer is completed.

[0013] According to some embodiments of the present invention, step S300 includes the following steps:

[0014] S310 obtains the raw data of the quartz flexible accelerometer calibration test;

[0015] S320 uses an interpolation method to calculate and obtain variable values ​​of the linear regression model;

[0016] S330 combines the linear regression model to forward predict the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes;

[0017] S340 loops step S310, step S320 and step S330 until the calibration of the quartz flexible accelerometer is completed.

[0018] According to some embodiments of the present invention, step S300 further includes the following steps:

[0019] Obtaining measured values ​​of the performance parameters at several time nodes in step S100, and calculating the measured standard deviation of the performance parameters;

[0020] Acquire data of the performance parameters at several time nodes in step S300, and calculate the predicted standard deviation of the performance parameters;

[0021] The measured standard deviation and the predicted standard deviation are compared and analyzed. If the difference between the two is equal to or greater than the first set value, in step S320, the interval between adjacent time nodes is reduced, and / or the number of predictions in step S320 is reduced.

[0022] According to some embodiments of the present invention, the characteristic variables of the linear regression model include at least IDorder and Vni270du, wherein IDorder is the number of days from the current target time node to the reference time node, and Vni270du is the original data generated by the quartz accelerometer during calibration testing.

[0023] According to some embodiments of the present invention, before step S300, the following step is further included: verifying the accuracy of the linear regression model, specifically including the following steps:

[0024] According to the original data at each time node in step S100, the measured value of the performance parameter at each time node in step S100 is calculated, and the measured square root of the performance parameter is calculated;

[0025] Combining the linear regression model with the original data at each time node in step S100, the predicted value of the performance parameter at each time node in step S100 is calculated, and the predicted square root of the performance parameter is calculated;

[0026] The measured square root and the predicted square root are compared and analyzed. If the difference between the two is at or below the second set value, step S300 is performed.

[0027] According to some embodiments of the present invention, step S200 includes the following steps:

[0028] Preprocess the original data according to the time series algorithm to obtain new feature variables;

[0029] Based on the stepAIC function in R language, some characteristic variables are eliminated;

[0030] Based on the multiple linear regression algorithm, the linear regression model is established.

[0031] According to some embodiments of the present invention, step S200 further includes the following step: eliminating the multicollinear characteristic variables according to the correlation matrix.

[0032] According to some embodiments of the present invention, the following steps are also included:

[0033] Determining the autocorrelation of various performance parameters of the quartz torsional accelerometer, if the autocorrelation of the performance parameters is greater than a third set value, the prediction requirement is met;

[0034] Wherein, in step S300, the predicted value of the performance parameter that meets the requirements is forward predicted based on the linear regression model.

[0035] According to some embodiments of the present invention, before determining the autocorrelation of various performance parameters of the quartz torsional accelerometer, the following steps are also included:

[0036] Get the raw data of the quartz flexible accelerometer;

[0037] Preprocess the original data according to the time series algorithm, obtain feature columns, and filter the feature columns;

[0038] Combined with the autocorrelation analysis function, the autocorrelation data of each type of performance parameter is calculated.

[0039] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments, wherein:

[0041] Figure 1 Schematic diagram of a flow chart of a calibration method for a quartz flexible accelerometer according to an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of a flow chart of step S300 of an embodiment of the present invention;

[0043] Figure 3 Another flowchart of step S300 of an embodiment of the present invention is shown;

[0044] Figure 4 Schematic diagram of the adjustment process of step S300 of an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of a flow chart for determining the accuracy of a linear regression model according to an embodiment of the present invention;

[0046] Figure 6 Schematic diagram of the process of step S200 of an embodiment of the present invention;

[0047] Figure 7 FIG. 4 is a schematic diagram of a flow chart of determining the autocorrelation of a performance parameter according to an embodiment of the present invention.

[0048] Figure 8 FIG. 4 is a schematic diagram of a preparation workflow for determining autocorrelation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0050] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0051] In the description of the present invention, "several" means more than one, "many" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0052] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0053] In the description of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0054] To facilitate the description of the present application, concepts related to the present disclosure are introduced below.

[0055] Quartz flexible accelerometer is a core inertial sensor in inertial navigation system. It is usually used to measure the linear acceleration of the carrier. After being solved by the navigation computer, the speed and position information of the carrier relative to the navigation coordinate system can be obtained. Due to its small size, high measurement accuracy and low cost, it is widely used in aerospace, aviation, navigation systems and various weapon models.

[0056] When evaluating the long-term stability of quartz flexible accelerometers, they usually need to be calibrated on dedicated equipment for 1 month, 3 months, 6 months, 12 months or even longer to obtain the measured values ​​of various performance parameters of the quartz flexible accelerometer, such as K0 (zero bias), K1 (scale factor), δ (installation error angle) and K2 (quadratic nonlinear coefficient). The time nodes for the two calibration tests of the quartz flexible accelerometer are generally between 3 and 8 days. The interval between any two adjacent time nodes can be the same or different. According to the actual situation, the number of days between the two time nodes should be reasonably arranged.

[0057] The present invention discloses a calibration method for a quartz flexible accelerometer, referring to Figure 1 , including the following steps:

[0058] S100 obtains the raw data of the quartz flexible accelerometer at several time points;

[0059] S200 builds a linear regression model based on the original data;

[0060] S300 forward predicts the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes according to the linear regression model.

[0061] In step S100, at each time node within the first few months, a special device is used to calibrate and test the quartz flexible accelerometer to obtain the original data of the corresponding time node. Based on the original data, the measured values ​​of the performance parameters of the quartz flexible accelerometer at each time node are calculated. For example, within the first 3 months, about 14 time nodes need to use special equipment to calibrate and test the quartz flexible accelerometer. Within the first 6 months, about 29 time nodes need to use special equipment to calibrate and test the quartz flexible accelerometer. In step S200, based on the data obtained from the calibration test in the first 3 months or 6 months, a linear regression model with accuracy that meets the prediction requirements is established. In step S300, after the first few months, for example, starting from the 4th or 7th month, the calibration test of the quartz flexible accelerometer at each time node can be calculated and obtained by the above-mentioned linear regression model. In order to ensure the accuracy of the prediction, a small number of time nodes need to use special equipment to complete the calibration test.

[0062] From the above, it can be seen that compared with the calibration test of the quartz flexible accelerometer at each time node in the prior art, it is necessary to use special equipment to perform the calibration test of the quartz flexible accelerometer. In the present application scheme, within the initial few months, the calibration test of the quartz flexible accelerometer needs to be completed by using special equipment at each time node. After the initial few months, the quartz flexible accelerometer can complete the calibration test at each time node through a linear regression model. From the above, it can be seen that the calibration method of the quartz flexible accelerometer in the present application scheme improves the calibration efficiency of the quartz flexible accelerometer, and the number of times the special equipment is used is reduced, saving costs.

[0063] In some embodiments, the characteristic variables of the linear regression model include at least IDorder and Vn i270du, for example: the formula of the linear regression model is set as: pred=2.267e+05+85.3-6.336e-02*IDorder+2.149e+05*Vni270du+4.167e-01*Huanzh i 180du;

[0064] Where pred is the predicted value of the performance parameter; IDorder is the number of days from the current target time node to the reference time node; Vn i270du and Huanzh i 180du are the original data generated during the calibration test of the quartz flexible accelerometer.

[0065] Specific description: Take a time node as the reference time node, at the reference time node, use special equipment to calibrate and test the quartz orbital accelerometer, and obtain the original data Vn i270du and Huanzh i 180du; when it is necessary to forward predict the predicted value of the performance parameter of a time node, for example, the reference time node is February 8, the target time node is set to February 14, and IDorder is set to 6 days, at this time, the above values ​​of Vn i270du and Huanzh i 180du and the value of IDorder are brought into the above linear regression model, so as to calculate the predicted value of the performance parameter on February 14. Similarly, when the target time node is set to February 18, and IDorder is set to 10 days, the above values ​​of Vn i270du and Huanzh i 180du and the value of IDorder are brought into the above linear regression model, so as to calculate the predicted value of the performance parameter on February 18.

[0066] In summary, the present application scheme uses a time node as a reference time node, and obtains the values ​​of Vn i270du and Huanzh i 180du according to the parameter time node calibration test, so as to forward predict the predicted values ​​of the performance parameters of multiple target time nodes.

[0067] In order to verify that there is no multicollinearity among the three independent variables IDorder, Vn i270du and Huanzh i 180du in the model, we used t-test and VIF value test (i.e. variance inflation factor) to verify.

[0068] The following are the results of the t-test. It can be seen that the pr(>|t|) of the three independent variable coefficients are all less than 0.05, so they all passed the t-test. In the figure below, the smaller the pr(>|t|), the more asterisks there are, indicating that the t-test is more significant. In addition, the multiple R 2 The value is 0.9062, indicating that these three independent variables can contribute 90.62% of the information for the accurate prediction of K0ug.

[0069]

[0070] The following is a VIF value test for the three independent variables. Their VIF values ​​are all less than 5, indicating that there is no multicollinearity.

[0071]

[0072] It should be noted that the linear regression model of the present application is not limited to the above-mentioned form, and needs to be combined with the original data of the quartz accelerometer during the calibration test to establish a model that can be used to accurately predict the numerical values ​​of the performance parameters of the quartz accelerometer. Therefore, the original data of the calibration test is different, and the linear regression model is also different, for example, pred = -0.06692*IDorder-98.42933*Vn i270du.

[0073] In some embodiments, reference Figure 1 and Figure 2 , step S200 includes the following steps:

[0074] S210 preprocesses the original data according to the time series algorithm to obtain new characteristic variables;

[0075] S220 is based on the stepAIC function in R language to eliminate some characteristic variables;

[0076] S230 builds a linear regression model based on the multiple linear regression algorithm.

[0077] Specifically, in step S100, the quartz flexible accelerometer is calibrated and tested by special equipment to obtain the original data of each time node within three or six months. In step S210, the original data is preprocessed by the time series algorithm to obtain new characteristic variables; in step S220, the characteristics of the stepAIC function in the R language can be used to automatically eliminate inappropriate characteristic variables, and the remaining characteristic variables are the characteristic variables required by the model; in step S230, based on the multivariate linear regression algorithm, a linear regression model can be established. After testing, the predicted values ​​of the performance parameters of the target time node calculated by the linear regression model have stability that does not exceed the requirements of the "Accelerometer Specification Requirements", and the prediction error is acceptable.

[0078] Furthermore, step S200 also includes the following steps: S221, based on the correlation matrix, eliminating multicollinear characteristic variables; this step can be set between S210 and S220, and can also be set between S220 and S230.

[0079] Specifically, there is serious multicollinearity between the characteristic variables. A linear regression model is established by using characteristic variables with multicollinearity, and the stability of the predicted value calculated by the linear regression model is poor. Through the setting of step S221, it can be found that there are strong correlations between many characteristic variables, and characteristic variables with multicollinearity are eliminated, and only characteristic variables without serious multicollinearity are retained. The remaining characteristic variables are all characteristic variables that have a strong correlation with the performance parameters, but there are no collinearity between them. Therefore, step S230 can establish an accurate linear regression model based on the multiple linear regression algorithm.

[0080] In some embodiments, reference Figure 1 and Figure 3 , before step S300, the following step is also included: verifying the accuracy of the linear regression model, which specifically includes the following steps:

[0081] S201 calculates the measured values ​​of the performance parameters at each time node in step S100 according to the original data at each time node in step S100, and calculates the measured square root of the performance parameters;

[0082] S202 combines the linear regression model with the original data at each time node in step S100 to calculate the predicted value of the performance parameter at each time node in step S100, and calculates the predicted square root of the performance parameter;

[0083] S203 compares and analyzes the measured square root and the predicted square root. If the difference between the two is at or below the second set value, proceed to step S300.

[0084] The specific analysis is as follows: within the initial 6 months in step S100, 29 time nodes are set in 6 months, so 29 groups of original data are obtained by calibration test using special equipment. According to the above 29 groups of original data, the measured values ​​of the performance parameters of the 29 time nodes within 6 months are calculated. Combining the square root formula with the measured values ​​of the above performance parameters, the measured square root of the performance parameters is calculated. Similarly, according to the values ​​of Vn i270du and Huanzh i180du in the 29 groups of original data, and the value of IDorder is set to 0, and the above linear regression model is used to calculate the predicted values ​​of the performance parameters of the 29 time nodes within 6 months. Combining the square root formula with the predicted values ​​of the above performance parameters, the predicted square root of the performance parameters is calculated. Comparing and analyzing the measured square root and the predicted square root, the difference between the measured square root and the predicted square root is at or below the second set value, indicating that the stability of the predicted 6-month data is not much different from the stability of the measured 6-month data, and the linear regression model established in step S200 meets the requirements. The calibration test of the quartz flexible accelerometer in the seventh month, the eighth month, etc. can be forward predicted by the linear regression model. Otherwise, the linear regression model needs to be readjusted.

[0085] Combined with the data explanation: the deviation between the predicted value and the measured K0 value (abspnct column), in 29 predictions, the maximum deviation is 6.49% (5.44ug), and the minimum deviation is 0.054% (0.05ug). The 6-month stability of the measured K0 is 6.97ug, and the 6-month stability of the predicted K0 is 6.636ug, with a deviation of 0.334ug, which is quite different from the 6-month K0 stability requirement of quartz flexible accelerometer of 15ug, so the predicted value can be used to evaluate the long-term stability of quartz flexible accelerometer.

[0086] Similarly, K1 and δ can also use the same method to train models and predict long-term stability.

[0087]

[0088] The calculation data and structure are shown in the figure above. K0ug is the true value of K0, and pred is the predicted value based on the original data. Because of linear regression, the small sample size is only 29, so the training data can be used to test the prediction accuracy.

[0089] d iff is the difference of K0ug–pred;

[0090] absd iff is the absolute value of di iff;

[0091] abspnct is: the absolute value of {(true value – predicted value) / true value}, that is, the degree of deviation;

[0092] Mean of K0ug: average value of K0ug;

[0093] The square of (pred-K0ug mean) is the square of the mean of the predicted value and K0ug, because the predicted value cannot deviate too far from the mean of the true value of K0ug;

[0094] Sum of column H: sum of the squares of (pred-K0ug mean);

[0095] A total of 29 values: There are a total of 29 measured values;

[0096] I2 / J2 value: first calculate the square of (pred-K0ug mean), then sum it, and then divide it by the number of measurements, that is, the average variance;

[0097] K2 square root: Take the square root of the calculation result of I2 / J2 (average variance), and the calculation result is the K0 stability value (i.e. 1σ) of the predicted data for 6 months, which is 6.636368, while the K0 stability of the measured data for 6 months (1σ in column B) is: 6.970757674.

[0098] In some embodiments, reference Figure 1 and Figure 4 , step S300 includes the following steps:

[0099] S310 obtains the raw data of the current time node of the quartz flexible accelerometer;

[0100] S320 combines the linear regression model with the original data obtained in step S310 to forward predict the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes;

[0101] S330 loops step S310 and step S320 until the calibration of the quartz flexible accelerometer is completed.

[0102] The specific analysis combined with the example is as follows: Referring to the table below, February 8 is set as the first time node, which is also the reference time node. At the first time node, the quartz flexible accelerometer is calibrated and tested using special equipment, and the measured values ​​(original data) of Vn i270du and Huanzh i 180du are obtained, and IDorder is set to 0;

[0103] February 14th is set as the second time node, the data of Vn i270du and Huanzh i 180du are the original data of the first time node, IDorder is set to 6 days, and the data of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (in the table).

[0104] February 18th is set as the third time node, the data of Vn i270du and Huanzh i 180du are the original data of the first time node, IDorder is set to 10 days, and the data of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (in the table).

[0105] February 22 is set as the fourth time node, the data of Vn i270du and Huanzh i 180du are the original data of the first time node, IDorder is set to 14 days, and the data of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (in the table).

[0106] Repeat the above steps:

[0107] February 28th was set as the fifth time node, which was also the reference time node. The quartz flexible accelerometer was calibrated and tested using special equipment, and the measured values ​​of Vn i270du and Huanzh i 180du were obtained. IDorder was set to 0.

[0108] March 5 is set as the sixth time node, the data of Vn i270du and Huanzh i 180du are the original data of the fifth time node, IDorder is set to 5 days, and the measured values ​​of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (in the table).

[0109] March 10 is set as the seventh time node, the data of Vn i270du and Huanzh i 180du are the original data of the fifth time node, IDorder is set to 10 days, and the data of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (in the table).

[0110] March 16 is set as the eighth time node, the data of Vn i270du and Huanzh i 180du are the original data of the fifth time node, IDorder is set to 16 days, and the data of Vn i270du, Huanzh i 180du and IDorder are brought into the above linear regression model to calculate the predicted values ​​of the performance parameters at this time node. Among them, the performance parameters can be K0, K1 and δ (K2 in the table).

[0111] Repeat the above steps until the 29th time node (refer to the table below).

[0112]

[0113] Through the above method, 29 predicted values ​​are obtained, and the results are shown in the following figure:

[0114]

[0115] As can be seen from the table above, the forward prediction results are calculated based on the data obtained from the actual calibration test every 3 calibrations (each calibration interval is about 6 days, and the 4 calibration intervals are about 20 days). The predicted 6-month K0 stability is 6.209ug, and the actual stability is 6.97ug. The difference between the two is 0.761ug, which accounts for 10.91% of the actual stability, that is, the error is 10.76%, indicating that the accuracy is about 90%.

[0116] Instead of the above step S300, in some embodiments, referring to Figure 1 , step S300 includes the following steps:

[0117] S310 obtains the raw data of the quartz flexible accelerometer calibration test;

[0118] S320 uses an interpolation method to calculate and obtain variable values ​​of the linear regression model;

[0119] S330 combines the linear regression model to forward predict the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes;

[0120] S340 loops step S310, step S320 and step S330 until the calibration of the quartz flexible accelerometer is completed.

[0121] The specific analysis is as follows: When calculating the predicted values ​​of the performance parameters at the target time node, the original data of the calibration test is directly applied to the above-mentioned linear regression model to obtain the predicted values ​​of the performance parameters. Instead of the above scheme, the interpolation method is used to calculate the variable values ​​of each target time node. Obviously, the calculated variable values ​​are used as the variable values ​​of the linear regression model, and the linear regression model can more accurately predict the predicted values ​​of the performance parameters.

[0122] In some embodiments, reference Figure 1 In step S300, the following steps are also included:

[0123] S301 obtains the measured values ​​of the performance parameters at several time nodes in step S100, and calculates the measured standard deviation of the performance parameters;

[0124] S302: acquiring data of performance parameters at several time nodes in step S300, and calculating predicted standard deviations of the performance parameters, wherein at least part of the data of the performance parameters are predicted values;

[0125] S303 compares and analyzes the measured standard deviation and the predicted standard deviation. If the difference between the two is equal to or greater than the first set value, in step S320, the number of days between adjacent time nodes is reduced, and / or the number of predictions in step S320 is reduced.

[0126] The specific analysis is as follows: within the initial 6 months in step S100, 29 time nodes are set in 6 months. According to the six consecutive groups of original data, the measured values ​​of the performance parameters of the 6 time nodes are calculated. Combined with the standard deviation formula and the measured values ​​of the above-mentioned performance parameters, the measured standard deviation of the performance parameters is calculated. Similarly, in steps S310 to S330, the data of the performance parameters of the 6 time nodes are obtained. Since the method of measuring one and calculating three is adopted, at least one of the data of the 6 performance parameters is the measured value of the performance parameters. Combined with the standard deviation formula and the predicted values ​​of the above-mentioned performance parameters, the predicted standard deviation of the performance parameters is calculated. Comparative analysis of the measured standard deviation and the predicted standard deviation, the difference between the measured standard deviation and the predicted standard deviation is above the first set value or the first set value, indicating that the stability of the predicted value is greatly different from the stability of the measured value, and it is necessary to reduce the number of days between adjacent time nodes, and / or reduce the number of predictions in step S320, so as to improve the accuracy of the predicted value of the performance parameter. On the contrary, there is no need to adjust the number of predictions or the number of days between the intervals.

[0127] In the above description, the starting time period is set to six months, and a linear regression model is established based on the original data of the calibration test of the quartz flexible accelerometer within six months. In the subsequent time, the linear regression model can be used to forward predict the values ​​of the performance parameters of the quartz flexible accelerometer at the target time node. It should be noted that the starting time period is not necessarily set to six months. For example, the starting time period is set to three months, and a linear regression model is established based on the original data of the calibration test of the quartz flexible accelerometer within three months. In the subsequent three months, the linear regression model can be used to forward predict the values ​​of the performance parameters of the quartz flexible accelerometer at the target time node in the subsequent three months. The specific analysis method is as follows:

[0128] At various time points within the first three months, the quartz coil accelerometer was calibrated and tested using special equipment. The interval between adjacent time points was 5 to 6 days, and a total of 16 time points were set within three months. The quartz coil accelerometer was calibrated and tested using special equipment to obtain 16 sets of raw data. A linear regression model (regression mathematical formula) was trained based on the above 16 sets of raw data.

[0129] At the first time node of the last three months, the quartz flexible accelerometer is calibrated and tested using special equipment to measure the original data of the quartz flexible accelerometer at the first time node; at multiple time nodes after the first time node (target time node), the values ​​of the performance parameters of the quartz flexible accelerometer at each time are predicted based on the original data of the first time node (Vn i270du), IDorder (the number of days from the current target time node to the first time node) and the linear regression model. For example, after the first time node, the values ​​of the performance parameters of the quartz flexible accelerometer at the subsequent 4 time nodes are predicted. By analogy, the quartz flexible accelerometer is calibrated and tested using special equipment again at a time node (the sixth time node of the last three months), and the original data of the quartz flexible accelerometer at this time node is measured. At multiple time nodes (target time nodes) after this time node, based on the original data (Vn i270du) of this time node (the sixth time node in the next three months), IDorder (the number of days from the current target time node to the sixth time node) and the linear regression model, the values ​​of the performance parameters of the quartz flexible accelerometer at each time are predicted. For example, after the sixth time node, the values ​​of the performance parameters of the quartz flexible accelerometer at the subsequent 7 time nodes are predicted, and exactly three months after the sixth time node.

[0130] In summary, on the one hand, the present application scheme can save a lot of actual measurement costs, especially after the third month, a total of 13 measurements are required, and only 2 measurements are actually required, and the other 11 measurements are predicted by the linear regression function; on the other hand, the present application scheme can compress the measurement cycle, which originally required 6 months of measurement, but now only requires 2 measurements, and can compress the actual measurement cycle by about 1.5 months to 2 months (for a 6-month stability test). If it is a one-year stability test, the actual measurement cycle can be compressed by 3 to 4 months.

[0131] In some embodiments, in step S300, predicted values ​​of performance parameters of the quartz flexible accelerometer are forward predicted according to the linear regression model until the calibration test of the quartz flexible accelerometer is completed.

[0132] Specific analysis: The calibration test of the quartz flexible accelerometer includes two time periods, namely the first time period and the second time period. At each time node in the first time period, step S100 uses special equipment to perform calibration test on the quartz flexible accelerometer and calculates the measured values ​​of the performance parameters; at each time node in the second time period, step 300 uses a linear regression model to predict the predicted values ​​of the performance parameters of the quartz flexible accelerometer. Finally, the stability of the quartz flexible accelerometer is demonstrated by combining the measured values ​​of the performance parameters in the first time period with the predicted values ​​of the performance parameters in the second time segment. For example, in the prior art, within a year, the quartz flexible accelerometer needs to be calibrated and tested using special equipment from January to December. The present application only needs to use special equipment to calibrate and test the quartz flexible accelerometer from January to October (the first time period), and the linear regression model can be used to predict the values ​​of the performance parameters of the quartz accelerometer from November to December (the second time period). As can be seen from the above, the present application solution only needs to use special equipment to calibrate and test the quartz flexible accelerometer for 10 months, in order to verify the stability of the quartz flexible accelerometer for one year. For another example, within half a year, the present application solution only needs to use special equipment to calibrate and test the quartz flexible accelerometer from January to April (the first time period), and from May to June (the second time period), a linear regression model can be used to predict the values ​​of the performance parameters of the quartz accelerometer. As can be seen from the above, the present application solution only needs to use special equipment to calibrate and test the quartz flexible accelerometer for 4 months, in order to verify the stability of the quartz flexible accelerometer for half a year.

[0133] In some embodiments, reference Figure 1 and Figure 7 Before step S100, the following steps are also included: determining the autocorrelation of various performance parameters of the quartz coil accelerometer; if the autocorrelation of the performance parameters is greater than a third set value, the prediction requirements are met; wherein, in step S300, based on the linear regression model, a predicted value of the performance parameter that meets the requirements is forward predicted.

[0134] Specifically, the various performance parameters are K0 (zero bias), K1 (scale factor), δ (installation error angle) and K2 (quadratic nonlinear coefficient), etc. The autocorrelations of K0 (zero bias), K1 (scale factor), δ (installation error angle) and K2 are analyzed, and the autocorrelations of K0, K1, K2 and δ are 77.9%, 65.6%, 18.3% and 86.6% respectively. Since the first set value is between 30% and 50%, K0, K1 and δ meet the prediction requirements. Since the parameters K0, K1 and δ meet the prediction requirements, in step S200, a linear regression model is established only for K0, K1 and δ (K3), and the predicted values ​​of the performance parameters of the quartz flexible accelerometer are forward predicted according to the linear regression model.

[0135] Further analysis: Before the step of combining the autocorrelation analysis function, the following steps are also included:

[0136] S10 obtains the raw data of the quartz flexible accelerometer;

[0137] S20 preprocesses the original data according to the time series algorithm, obtains the feature columns, and filters the feature columns;

[0138] S30 combines the autocorrelation analysis function to calculate the autocorrelation data of various performance parameters;

[0139] S40 determines the autocorrelation of various performance parameters of the quartz torsional accelerometer. If the autocorrelation of the performance parameters is greater than a third set value, the prediction requirement is met.

[0140] The specific analysis is as follows: Step S10 and step 20 refer to the following figure:

[0141]

[0142]

[0143]

[0144] In step S20, the original data is preprocessed according to the time series algorithm to obtain feature columns, and the feature columns are screened, that is, useless columns are removed and useful columns are retained (the feature names with "l ag" at the end of the table are all newly generated feature columns).

[0145] The K0ugl ag, K1mA.gl ag, K2ug.g2 l ag, and K3urad la in the above figure are based on the measured data K0ug, K1mA.g, K2ug.g2, and K3urad, using the l ag function (a function of autocorrelation analysis) to generate the autocorrelation results. The calculation method is to put the value of the previous moment at the current moment as a variable (parameter) to calculate the K0 value at the current moment. IDorder and d iffDays are calculated based on the actual calibration date. d iffDays is the number of days between the current calibration and the last calibration; IDorder is the number of days between the current calibration and the first calibration (February 8, 2022). The analysis results with strong autocorrelation are shown below. The autocorrelations of K0, K1, K2, and δ are 77.9%, 65.6%, 18.3%, and 86.6%, respectively, so K0, K1, and δ have strong autocorrelation.

[0146] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the purpose of the present invention. In addition, the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

Claims

1. The calibration method of the quartz flexible accelerometer includes the following steps: S100, obtaining raw data of a quartz flexible accelerometer at a plurality of time nodes; S200, establish a linear regression model based on the original data; S300, forward predicting the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes according to the linear regression model; characterized in that, Before step S200 establishes a linear regression model based on the original data, the following steps are also included: Preprocess the original data according to the time series algorithm, obtain feature columns, and filter the feature columns; Combined with the autocorrelation analysis function, the autocorrelation data of various performance parameters of the quartz flexible accelerometer are calculated; Determine the autocorrelation of various performance parameters of the quartz flexible accelerometer. If the autocorrelation of the performance parameters is greater than a set value, the prediction requirement is met; wherein the various performance parameters are zero bias, scale factor, installation error angle and quadratic nonlinear coefficient, and after analyzing the autocorrelation of the zero bias, scale factor, installation error angle and quadratic nonlinear coefficient, determine that the zero bias, scale factor and installation error angle meet the prediction requirement; Wherein, in step S200, a linear regression model is established only for performance parameters that meet the prediction requirements, and in step S300, the predicted values ​​of the performance parameters that meet the requirements are forward predicted based on the linear regression model; Furthermore, step S300 includes the following steps: S310, obtaining the original data of the current time node of the quartz flexible accelerometer and using it as the variable value of the linear regression model; S320, combining the linear regression model, forward predicting the predicted values ​​of the performance parameters of the quartz flexible accelerometer at several target time nodes; S330, looping step S310 and step S320 until the calibration of the quartz flexible accelerometer is completed. And, step S300 also includes the following steps: Obtaining measured values ​​of the performance parameters at several time nodes in step S100, and calculating the measured standard deviation of the performance parameters; Acquire data of the performance parameters at several time nodes in step S300, and calculate the predicted standard deviation of the performance parameters; The measured standard deviation and the predicted standard deviation are compared and analyzed. If the difference between the two is equal to or greater than the first set value, in step S320, the interval between adjacent time nodes is reduced, and / or the number of predictions in step S320 is reduced.

2. The calibration method of the quartz flexible accelerometer according to claim 1, characterized in that: The characteristic variables of the linear regression model include at least IDorder and Vni270du, wherein IDorder is the number of days between the current target time node and the reference time node, and Vni270du is the original data generated by the quartz accelerometer during the calibration test.

3. The calibration method of the quartz flexible accelerometer according to claim 1, characterized in that: Before step S300, the following step is also included: verifying the accuracy of the linear regression model, specifically including the following steps: According to the original data at each time node in step S100, the measured value of the performance parameter at each time node in step S100 is calculated, and the measured square root of the performance parameter is calculated; Combining the linear regression model with the original data at each time node in step S100, the predicted value of the performance parameter at each time node in step S100 is calculated, and the predicted square root of the performance parameter is calculated; The measured square root and the predicted square root are compared and analyzed. If the difference between the two is at or below the second set value, step S300 is performed.

4. The calibration method of a quartz flexible accelerometer according to claim 1, characterized in that: Step S200 includes the following steps: Preprocess the original data according to the time series algorithm to obtain new feature variables; Based on the stepAIC function in R language, some characteristic variables are eliminated; Based on the multiple linear regression algorithm, the linear regression model is established.

5. The calibration method of the quartz flexible accelerometer according to claim 4, characterized in that: Step S200 also includes the following step: according to the correlation matrix, eliminating the multicollinear characteristic variables.

6. A prediction system for a quartz flexible accelerometer, characterized in that The prediction system predicts the predicted values ​​of the performance parameters of the quartz flexible accelerometer at the target time node forward according to the linear regression model; wherein the linear regression model is established by multiple groups of data from the calibration test of the quartz flexible accelerometer.

Citation Information

Patent Citations

  • Rock quartz flexibility accelerometer measuring component parameter identification calibration method

    CN101290326A

  • Self-adaptation three-dimensional attitude positioning method based on microinertia and geomagnetic technology

    CN101915580A

  • Device for measuring gravitational acceleration in rotating state and method for measuring and extracting gravitational acceleration in rotating state

    CN106437683A

  • Deep learning method for automatic calibration of MEMS (Micro Electro Mechanical Systems) accelerometer

    CN109188026A

  • Rotary equipment fault prediction method and system based on regression model and readable storage medium

    CN112200048A