A High-Precision Identification Method for Dynamic Model Parameters of a Digital Accelerometer
The proposed method for digital quartz flexure accelerometer model parameter identification addresses system delays and noise by using closed-loop identification and VMD noise separation, enhancing precision in dynamic model parameter estimation.
Patent Information
- Application Number
- CN202410216000.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-27
- Publication Date
- 2025-07-15
- Estimated Expiration
- 2044-02-27
AI Technical Summary
The traditional accelerometer dynamic model parameter identification method is susceptible to system pure delay, nonlinearity and noise, resulting in insufficient identification accuracy.
A nonlinear dynamic model containing pure delay terms is used, combined with intelligent optimization algorithm and variational modal decomposition (VMD) to process the accelerometer response data, separate noise and reconstruct data, and use a standard second-order system for high-precision identification.
It effectively improves the identification accuracy of the accelerometer dynamic model parameters, weakens the impact of pure delay, nonlinearity and noise on model identification, and improves the system sensitivity and identification accuracy.
Smart Images
Figure CN118069992B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of digital accelerometer dynamic model parameter identification, and particularly to a method for high-precision identification of dynamic model parameters of a digital accelerometer. Background Art
[0002] Due to characteristics such as high precision, small volume, and high stability, digital quartz flexure accelerometers are currently widely used in fields such as inertial navigation, attitude control, and vibration monitoring. The identification results of their dynamic model parameters directly affect the dynamic measurement performance of the accelerometer. Therefore, improving the identification accuracy of accelerometer dynamic model parameters is of great significance for the practical application of the accelerometer system.
[0003] Traditional methods for identifying dynamic model parameters of accelerometers can be divided into time-domain identification and frequency-domain identification. Time-domain identification mainly collects the response data of the accelerometer system under standard pulse shocks according to ISO 16063-12:2001 "Calibration of Shock and Vibration Sensors - Part 12: Shock Calibration by the Absolute Method", and further uses algorithms such as the least squares method and intelligent optimization algorithms to obtain model parameters; frequency-domain identification mainly obtains the frequency response function of the accelerometer by using steady-state sinusoidal vibration and laser interferometry measurement according to ISO 16063-11:1999 "Calibration Methods for Vibration and Shock Sensors", and then uses methods such as sine approximation method, fringe counting method, and minimum point method.
[0004] In the process of implementing the above application, the inventor found that there are at least the following problems in this technology. When identifying the dynamic model of the accelerometer, it is usually affected by system pure delay, nonlinearity, and noise factors. Moreover, traditional accelerometer dynamic models often use system characteristic parameters such as moment of inertia, damping, and stiffness to describe. In the identification process, the identification accuracy of the stiffness parameter is more susceptible to system noise due to low parameter sensitivity. Therefore, there is still a large room for improvement in the identification accuracy of accelerometer dynamic model parameters identified by traditional methods. Summary of the Invention
[0005] In order to improve the identification accuracy of accelerometer dynamic model parameters, the present application provides a method for high-precision identification of dynamic model parameters of a digital accelerometer.
[0006] The method for high-precision identification of dynamic model parameters of a digital accelerometer provided by the present application adopts the following technical solutions:
[0007] A method for high-precision identification of dynamic model parameters of a digital accelerometer includes the following steps:
[0008] Step 1: Collect the response data of the digital accelerometer closed-loop system;
[0009] Step 2: Establish a non-linear dynamic mathematical model of the accelerometer including a pure delay term;
[0010] Step 3: Using the collected response data as the identification data, initially identify the parameters of the accelerometer dynamic mathematical model using an intelligent optimization algorithm.
[0011] Step 4: Process the data residuals after the initial identification. Use variational mode decomposition (VMD) to separate the noise in the residual data and reconstruct new identification data.
[0012] Step 5: Based on the reconstructed identification data processed by VMD, use an intelligent optimization algorithm to accurately identify the wn parameter in the accelerometer dynamic model.
[0013] Step 6: Analyze whether the residual data after the high-precision identification in Step 5 conforms to the closed-loop noise characteristics of the digital accelerometer. If it conforms to the characteristics, it proves that the VMD data processing is successful and the parameter identification is accurate; if it does not conform to the characteristics, it is necessary to return to Step 5 for data processing until the identification residual conforms to the closed-loop noise characteristics.
[0014] By adopting the above technical solution, a nonlinear dynamic model of the accelerometer including a pure time-delay term is established. The digital closed-loop is used to expand the system bandwidth and weaken the nonlinear influence. Further, an intelligent optimization algorithm is used to initially identify the model parameters in the system, then VMD is used to separate the noise in the identification residuals, and finally the reconstructed data after separating the noise is used to accurately identify the natural frequency parameter in the model until the residual data after identification conforms to the closed-loop noise characteristics, thereby realizing the high-precision identification of the dynamic model parameters of the digital closed-loop accelerometer.
[0015] Optionally, in Step 1, the frequency band range of the response input covers the frequency band required by the accelerometer; the system closed-loop controller should be a controller that can be expressed by a simple formula to avoid increasing the complexity of model identification.
[0016] By adopting the above technical solution, it is necessary to collect the response data of the accelerometer system under closed-loop control. The frequency band range of the response input should cover the frequency band required by the accelerometer as much as possible to make the identification result more accurate, and a digital controller that can be clearly expressed by a simple formula should be used for closed-loop control.
[0017] Optionally, in Step 2, the accelerometer dynamic mathematical model and the nonlinear mathematical model are respectively expressed as:
[0018]
[0019]
[0020] where J, C, K, K c are respectively the moment of inertia of the accelerometer sensing element, the head damping, the head stiffness, and the head static gain parameter; d is the pure time-delay parameter of the accelerometer system;
[0021] f(cap) is a function describing the capacitance non-linearity; cap is the capacitance collected by the system; k1 and k2 are both coefficients describing the capacitance non-linearity function of the accelerometer.
[0022] By adopting the above technical solution, it is necessary to establish a non-linear dynamic mathematical model of the accelerometer including a pure time-delay term. The non-linear term combines the tests of each link of the actual accelerometer to locate that the main influencing link of the accelerometer non-linearity is the capacitance measurement link, and a mathematical model that can describe the non-linearity is established with the least parameters.
[0023] Optionally, in step three, using the collected response data as the identification data, an intelligent optimization algorithm is used to preliminarily identify the parameters of the accelerometer dynamic mathematical model; the identification criterion of the intelligent algorithm adopts the general sum of squared errors criterion
[0024]
[0025] Among them, fit_data is the fitting data of the identified parameters, and response is the response data of the accelerometer system.
[0026] By adopting the above technical solution, in step three, using the collected response data as the identification data, an intelligent optimization algorithm is used to preliminarily identify the parameters of the accelerometer dynamic mathematical model; the preliminarily identified model is a dynamic model described by the parameters of the accelerometer system, that is, a dynamic model composed of coefficients such as the moment of inertia, damping, and stiffness of the accelerometer system, and the identification criterion of the intelligent algorithm adopts the general sum of squared errors criterion.
[0027] Optionally, in step four, the criterion for noise separation and data reconstruction after VMD processing is based on the average energy entropy
[0028]
[0029]
[0030] where p i = E i / E is the proportion of the energy of the i-th component in the total energy ; H E , are the signal energy entropy and the average energy entropy respectively.
[0031] By adopting the above technical solutions, in Step 4, data processing is performed on the data residuals after preliminary identification. Variational mode decomposition (VMD) is used to separate the noise in the residual data and reconstruct new identification data. Among them, the noise separation and data reconstruction criteria after VMD processing adopt the average energy entropy criterion. Among the intrinsic mode functions (IMFs) decomposed by VMD, the more noise components there are, the smaller the average energy entropy; on the contrary, the more useful components there are, the larger the average energy entropy.
[0032] Optionally, in Step 4, the reconstructed identification data is the preliminary identification fitting data plus the reconstructed response data, or the response data minus the separated noise data.
[0033] fit_data - response = cap_error = cap_noise + cap_response
[0034] fit_data - cap_noise = response + cap_response
[0035] Where cap_error, cap_noise, and cap_response are the preliminary identification residuals, the separated noise in the identification residuals, and the reconstructed response data in the identification residuals, respectively.
[0036] By adopting the above technical solutions, it can be obtained by comparison that fit_data - cap_noise = response + cap_response, which is convenient for the subsequent steps.
[0037] Optionally, in Step 5, a standard second-order system dynamic model is used to replace the dynamic model described by the accelerometer system parameters for identification
[0038]
[0039] Where are the damping, natural frequency, and steady-state gain parameters of the second-order system, respectively.
[0040] By adopting the above technical solutions, since the identification accuracy of the stiffness parameter in the original model is limited by the parameter sensitivity and is easily affected by the closed-loop noise, while the sensitivity of the natural frequency parameter in the standard second-order system is much higher than that of the stiffness parameter, a standard second-order system dynamic model is used to replace the dynamic model described by the accelerometer system parameters for identification.
[0041] Optionally, on the basis of the preliminary identification parameters, it is necessary to limit the t parameter range.
[0042] By adopting the above technical solutions, because ξ and K tThe parameters are all related to the stiffness K parameter. Therefore, when using the standard second-order system model for identification, it is easy to fall into a local optimal solution. Therefore, on the basis of the preliminary identification of the parameters, it is necessary to limit ξ and K t parameter range, which is more conducive to high-precision identification of w n parameters.
[0043] In summary, the present application includes at least one of the following beneficial technical effects:
[0044] 1. The high-precision identification method for the dynamic model parameters of the digital accelerometer proposed by the present invention. Aiming at the fact that the traditional accelerometer dynamic model identification method is easily affected by system pure delay and nonlinearity, a dynamic model including pure delay and nonlinear terms is proposed, and a closed-loop identification method is used to expand the system bandwidth and improve the system sensitivity to utilize pure delay identification. At the same time, the closed-loop control also controls the controlled object near the zero position, greatly weakening the influence of nonlinearity on model identification; effectively improving the identification accuracy of the dynamic model parameters of the digital accelerometer;
[0045] 2. The high-precision identification method for the dynamic model parameters of the digital accelerometer proposed by the present invention. Aiming at the fact that the stiffness parameter in the dynamic model described by the traditional accelerometer system parameters is vulnerable to noise due to its low parameter sensitivity, it is proposed to first use an intelligent optimization algorithm to identify the dynamic model of the accelerometer, further use VMD to separate noise and reconstruct data from the identification residual data, and finally based on the fact that the sensitivity of the natural frequency parameter in the standard second-order system is higher than that of the stiffness parameter, use an intelligent optimization algorithm to identify the natural frequency parameter in the standard second-order system dynamic model of the accelerometer with high precision, greatly weakening the influence of noise on the identification of system dynamic model parameters; effectively improving the identification accuracy of the dynamic model parameters of the digital accelerometer;
[0046] 3. The high-precision identification method for the dynamic model parameters of the digital accelerometer proposed by the present invention is applicable to the dynamic model identification of various closed-loop accelerometers, and the proposed identification method makes full use of the advantages of the accelerometer closed-loop, and effectively separates the closed-loop noise in the response data by using VMD, greatly weakening the influence of accelerometer pure delay, nonlinearity and noise on model parameter identification, and effectively improving the identification accuracy of the dynamic model parameters of the digital accelerometer. Description of the Drawings
[0047] Figure 1 is the flow chart of the high-precision identification method for the dynamic model parameters of the digital accelerometer;
[0048] Figure 2 is a schematic diagram of the response data of the digital accelerometer closed-loop system collected;
[0049] Figure 3 is a schematic diagram of the preliminary identification curve of the digital accelerometer dynamic model;
[0050] Figure 4 It is the VMD processing diagram of the identification residual data of the digital accelerometer;
[0051] Figure 5 It is the noise separation and data reconstruction diagram of the identification residual data of the digital accelerometer;
[0052] Figure 6 It is the schematic diagram of the accurate identification curve of the digital accelerometer dynamic model;
[0053] Figure 7 It is the data residual diagram before and after the accurate identification of the digital accelerometer dynamic model;
[0054] Figure 8 It is Figure 4 The partial enlarged view of the residual data in
[0055] Figure 9 It is Figure 4 The partial enlarged view of the VMD intrinsic mode function in
[0056] Figure 10 It is Figure 5 The partial enlarged view of the data separation and reconstruction in Specific implementation manner
[0057] The following is a further detailed description of this application in conjunction with the attached Figures 1 - 10 drawings.
[0058] The embodiment of this application discloses a method for high-precision identification of the dynamic model parameters of a digital accelerometer. Referring to Figure 1 , the method for high-precision identification of the dynamic model parameters of a digital accelerometer includes the following steps:
[0059] Step 1: Collect the response data of the digital accelerometer closed-loop system;
[0060] Step 2: Establish an accelerometer nonlinear dynamic mathematical model including a pure delay term;
[0061] Step 3: Use the intelligent optimization algorithm to preliminarily identify the parameters of the accelerometer dynamic mathematical model with the collected response data as the identification data;
[0062] Step 4: Process the data residuals after preliminary identification, and use variational mode decomposition (VMD) to separate the noise in the residual data and reconstruct new identification data;
[0063] Step 5: Based on the reconstructed identification data processed by VMD, use the intelligent optimization algorithm to perform high-precision identification of the wn parameter in the accelerometer dynamic model;
[0064] Step 6: Analyze whether the residual data after high-precision identification in Step 5 conforms to the closed-loop noise characteristics of the digital accelerometer. If it conforms to the characteristics, it proves that the VMD data processing is successful and the parameter identification is accurate; if it does not conform to the characteristics, it is necessary to return to Step 5 for data processing until the identification residual conforms to the closed-loop noise characteristics.
[0065] Refer to Figure 1 and Figure 2 , in Step 1, collect the response data of the digital accelerometer closed-loop system. Under the condition of stable control of the closed-loop system, a standard excitation signal needs to be generated internally or externally, such as: standard step signal, standard pulse signal, etc. As Figure 2 shown, in this example, the standard unit step signal generated inside the digital controller is adopted, and the output of the controlled object (cap) of the digital accelerometer and the output of the controller (pid) are synchronously collected;
[0066] The collected controller output stabilizes with the stabilization of the step signal due to the setback tracking effect in the closed-loop circuit, while the output of the accelerometer controlled object gradually returns to zero due to the zero position setting and integral control synchronization in the closed-loop circuit. In essence, both can be used to effectively identify the dynamic model of the accelerometer. However, considering that the regression dynamic process of the output of the controlled object is longer and the dynamic characteristics are more obvious, the data to be identified is selected as the output of the controlled object in the accelerometer closed-loop circuit;
[0067] Due to the zero position regression requirement of the controlled object, if the traditional pid controller structure is adopted, pi control must be used. Using only proportional p control cannot achieve the zero position regression function of the controlled object. Using pid control will make the system a high-order system due to the differential action and will amplify the noise in the closed-loop system, greatly increasing the difficulty of high-precision identification of the dynamic model parameters of the accelerometer.
[0068] In Step 2, it is necessary to establish a nonlinear dynamic mathematical model of the accelerometer including a pure time delay term. The nonlinear term is combined with the actual tests of each link of the accelerometer to locate that the main link affecting the nonlinearity of the accelerometer is the capacitance measurement link, and a mathematical model that can describe the nonlinearity with the least parameters is established. The dynamic mathematical model and the nonlinear mathematical model of the accelerometer are respectively expressed as:
[0069]
[0070]
[0071] Among them, J, C, K, K cThey are the moment of inertia, head damping, head stiffness, and head static gain parameters of the accelerometer sensitive element respectively; d is the pure time delay parameter of the accelerometer system; f(cap) is a function describing the capacitance nonlinearity; cap is the capacitance collected by the system; k1 and k2 are both coefficients describing the capacitance nonlinearity function of the accelerometer. Combining the test positioning of each link of the digital accelerometer system open-loop system, the system nonlinearity mainly comes from the capacitance measurement nonlinearity. Therefore, based on the principle of describing the nonlinear characteristics with the least parameters and combining the measured data, an accelerometer capacitance nonlinear function f(cap) based on the sigmod nonlinear model is established. And since the function is only related to the measured capacitance, after identifying the nonlinear parameters, online compensation can be performed in the digital controller.
[0072] Refer to Figure 1 , Figure 3 , Figure 4 , Figure 8 and Figure 9 , in step three, using the collected response data as the identification data, an intelligent optimization algorithm is used to preliminarily identify the parameters of the accelerometer dynamic mathematical model. The preliminarily identified model is a dynamic model described by the parameters of the accelerometer system, that is, a dynamic model composed of coefficients such as the moment of inertia, damping, and stiffness of the accelerometer system. The identification criterion of the intelligent algorithm uses the general sum of squared errors criterion, that is
[0073]
[0074] where fit_data is the fitting data of the identified parameters and response is the response data of the accelerometer system. As Figure 3 shown, the output of the accelerometer preliminary identification model has a high fitting degree under large-range disturbances, but has a low fitting degree under small-range disturbances affected by noise.
[0075] Refer to Figure 1 , Figure 4 , Figure 5 and Figure 10 , in step four, as Figure 4 shown, data processing is performed on the data residuals after preliminary identification. The variational mode decomposition (VMD) is used to separate the noise in the residual data and reconstruct new identification data. The noise separation and data reconstruction criterion after VMD processing uses the average energy entropy criterion. Among the intrinsic mode functions (IMFs) decomposed by VMD, the more noise components, the smaller the average energy entropy; conversely, the more useful components, the larger the average energy entropy. That is, the average energy entropy is used as the basis for error separation and data reconstruction.
[0076]
[0077]
[0078] where p i = E i / E is the proportion of the energy of the i-th component in the total energy ; H E and are the signal energy entropy and the average energy entropy respectively.
[0079] Referring to Figure 4 , in step four, the preliminary identification parameter data is subjected to intrinsic mode separation, and a total of 6 modes are separated. Then, the average energy entropy of each intrinsic mode is calculated respectively. It can be seen from the calculation results that the average energy entropy indexes of modes 1, 2, and 6 are significantly higher than those of the other modes, indicating that these three modes contain more effective response data.
[0080] Referring to Figure 5 , in step four, based on Figure 4 the VMD data processing results, the modes 1, 2, and 6 are superimposed to obtain the effective data to be reconstructed, and the remaining modes are superimposed to obtain the separated noise data. Then, spectral analysis is performed on both of them respectively. According to the spectral characteristics, it can be seen that the spectrum of the data to be reconstructed still conforms to the frequency concentration characteristics caused by resonance in the closed-loop system response, while the spectrum of the noise data does not have obvious resonance characteristics, further proving the effective separation of the data.
[0081] In step four, the reconstructed identification data is the preliminary identification fitting data plus the reconstructed response data, or the response data minus the separated noise data.
[0082] fit_data - response = cap_error = cap_noise + cap_response
[0083] fit_data - cap_noise = response + cap_response
[0084] where cap_error, cap_noise, and cap_response are the preliminary identification residual, the noise separated from the identification residual, and the reconstructed response data in the identification residual respectively.
[0085] Referring to Figure 1 , Figure 6 and Figure 7 , in step five, based on the reconstructed identification data processed by VMD, an intelligent optimization algorithm is used to perform high-precision identification on the wn parameter in the dynamic model of the accelerometer standard second-order system.
[0086]
[0087] where They are the damping, natural frequency, and steady-state gain parameters of the second-order system, respectively. As Figure 6 shown, the output of the accurate identification model of the accelerometer has a higher fitting degree than that of the preliminary identification model under both large-range and small-range disturbances, further proving the effectiveness of the proposed dynamic model accurate identification method.
[0088] Since the identification accuracy of the stiffness parameter in the original model is limited by the parameter sensitivity and is easily affected by the closed-loop noise, and the sensitivity of the natural frequency parameter in the standard second-order system is greatly improved compared with the stiffness parameter, the dynamic model described by the standard second-order system dynamic model is used to replace the accelerometer system parameters for identification. Also, because the ξ and K t parameters are both related to the stiffness K parameter, when using the standard second-order system model for identification, it is easy to fall into a local optimal solution. Therefore, on the basis of the preliminary identification parameters, it is necessary to limit the ξ and K t parameter range, which is more conducive to high-precision identification of the w n parameter.
[0089] In step six, analyze whether the residual data after high-precision identification in step five conforms to the characteristics of the digital accelerometer closed-loop noise. If it conforms to the characteristics, it proves that the forced response data of the system in the data is fully utilized and the identified model parameters are accurate; if it does not conform to the characteristics, it proves that the noise and response data are not fully separated after VMD processing. Therefore, it is necessary to return to step five for data processing until the identification residual conforms to the closed-loop noise characteristics. As Figure 7 shown, the residual of the preliminary identification data and the residual of the accurate identification data are compared, and the frequency spectra of the two are analyzed. It can be found that the output error of the model after accurate identification is significantly smaller than that of the preliminary identification model, indicating that the accuracy of the accurate model is higher. There is no obvious frequency concentration characteristic due to system resonance in the frequency spectrum of the residual data after accurate identification compared with the frequency spectrum of the residual data after preliminary identification, indicating that the accurate identification model makes more efficient use of the system closed-loop response data, which further proves that the accuracy of the digital accelerometer dynamic model after accurate identification is higher.
[0090] The implementation principle of a method for high-precision identification of dynamic model parameters of a digital accelerometer in an embodiment of this application is as follows: In view of the influence of system pure time delay, nonlinearity, and noise on the identification of accelerometer dynamic model parameters through the methods of Step 1 to Step 6, first, a nonlinear dynamic model of the accelerometer including a pure time delay term is established. Secondly, the closed-loop identification is used to expand the system bandwidth and improve the system sensitivity to facilitate the pure time delay identification. At the same time, the sensitive element of the accelerometer is controlled at the zero position to weaken the influence of nonlinearity on parameter identification. Finally, since the identification of the stiffness parameter or natural frequency parameter of the digital closed-loop system is susceptible to closed-loop noise, it is proposed to use VMD to separate the noise from the residual data after identification, and further use the processed reconstructed data to adopt an intelligent optimization algorithm to perform high-precision identification of the natural frequency parameter. If it meets the characteristics, it proves that the forced response data of the system in the data is fully utilized and the identified model parameters are accurate. If it does not meet the characteristics, it proves that the noise and response data are not fully separated after VMD processing. Therefore, it is necessary to return to Step 5 for data processing until the identification residual meets the closed-loop noise characteristics.
[0091] The above are all preferred embodiments of this application. The protection scope of this application is not limited accordingly. Therefore, all equivalent changes made according to the structure, shape, and principle of this application shall be covered by the protection scope of this application.
Claims
1. A method for high-precision identification of dynamic model parameters of a digital accelerometer, characterized in that, It includes the following steps: Step 1: Collect the response data of the digital accelerometer closed-loop system; Step 2: Establish a nonlinear dynamic mathematical model of the accelerometer that includes a pure time-delay term; Step 3: Using the collected response data as identification data, adopt an intelligent optimization algorithm to preliminarily identify the parameters of the accelerometer dynamic mathematical model; Step 4: Process the data residuals after preliminary identification. Use variational mode decomposition (VMD) to separate the noise in the residual data and reconstruct new identification data; Step 5: Based on the reconstructed identification data processed by VMD, use an intelligent optimization algorithm to perform high-precision identification of the w n parameter in the accelerometer dynamic model; Step 6: Analyze whether the residual data after high-precision identification in Step 5 conforms to the closed-loop noise characteristics of the digital accelerometer. If it conforms to the characteristics, it proves that the VMD data processing is successful and the parameter identification is accurate; if it does not conform to the characteristics, it is necessary to return to Step 5 for data processing until the identification residuals conform to the closed-loop noise characteristics; In Step 2, the dynamic mathematical model and the nonlinear mathematical model of the accelerometer are respectively expressed as: ； ； Among them, , , , are the moment of inertia, head damping, head stiffness, and head static gain parameters of the accelerometer sensing element respectively; is the pure delay parameter of the accelerometer system; A function for describing capacitance non-linearity; The capacitance acquired by the system; and are all coefficients of the function for describing the capacitance non-linearity of the accelerometer.
2. The high-precision identification method for dynamic model parameters of the digital accelerometer according to claim 1, characterized in that: In Step 1, the frequency band range of the response input covers the frequency band required by the accelerometer; the system closed-loop controller uses a controller that can be expressed by a simple formula to avoid increasing the complexity of model identification.
3. The high-precision identification method for the dynamic model parameters of the digital accelerometer according to claim 1, characterized in that: In Step 3, using the collected response data as identification data, adopt an intelligent optimization algorithm to preliminarily identify the parameters of the accelerometer dynamic mathematical model; the criterion for intelligent algorithm identification adopts the general sum of squared errors criterion ； Among them, is the identification parameter fitting data, is the accelerometer system response data.
4. The high-precision identification method for dynamic model parameters of a digital accelerometer according to claim 1, characterized in that: In Step 4, the criterion for noise separation and data reconstruction after VMD processing is based on the average energy entropy ； ； wherein, is the proportion of the energy of the i-th component in the total energy ; , are the signal energy entropy and the average energy entropy respectively.
5. The high-precision identification method for dynamic model parameters of the digital accelerometer according to claim 1, characterized in that: In Step 4, the reconstructed identification data is the preliminary identification fitting data plus the reconstructed response data, or the response data minus the separated noise data ； ； Among them, , e, and are the preliminary identification residual, the noise separated from the identification residual, and the reconstructed response data in the identification residual, respectively.
6. The high-precision identification method for dynamic model parameters of the digital accelerometer according to claim 1, characterized in that: In Step 5, a standard second-order system dynamic model is used to replace the dynamic model described by the accelerometer system parameters for identification ； Among them, are respectively the damping, natural frequency, and steady-state gain parameters of the second-order system.
7. The high-precision identification method for dynamic model parameters of the digital accelerometer according to claim 6, wherein: Based on the preliminary identification of parameters, it is necessary to limit and the parameter range.