Model adaptive adjustment displacement extraction method

Abstract

The invention belongs to the technical field of structure health monitoring, and discloses a model adaptive adjustment displacement extraction method. The method comprises the following steps: firstly, performing dynamic positioning displacement input through displacement sensors such as a displacement meter and a GNSS (Global Navigation Satellite System) to obtain initial displacement measurement; then performing frequency coarse identification to determine a vibration frequency range; then, sub-frequencies are selected within the frequency range to construct a set of state model sets. And finally, estimating parameters by using an interactive multi-model Kalman filter, and adaptively adjusting to extract displacement through multi-model combination. Simulation and field experiment results show that the accuracy of displacement estimation is improved under the condition that the models are not matched, and displacement estimation errors are effectively restrained under the static or low-frequency noise condition. According to the method, the problem of displacement data errors caused by measurement noise is solved, the accuracy and reliability of structural health monitoring are remarkably improved, and an effective technical means is provided for safety assessment of an engineering structure.

Description

Technical Field

[0001] The present invention relates to a displacement extraction method for model adaptive adjustment, and in particular to a method for adaptively extracting displacement from dynamic positioning using an interactive multi-model Kalman filter (IMM), belonging to the technical field of structural health monitoring (SHM). Background Art

[0002] Structural health monitoring (SHM) can ensure the safety, stability and reliability of buildings, bridges and equipment. Through real-time monitoring, potential problems can be detected early, the safety level of the structure can be improved, the service life can be extended, maintenance costs can be reduced, and instant data can be provided in emergency situations to support rapid and accurate emergency response. This technology helps to optimize the operating efficiency of the structure and improve the effectiveness of the overall project and equipment.

[0003] With the widespread application of high-frequency global navigation satellite systems (GNSS), GNSS has become an important means of obtaining displacement. GNSS technology can provide high-precision, real-time displacement data, which is of great significance for assessing structural health, detecting potential problems, and performing preventive maintenance. However, due to the influence of measurement noise, the displacement obtained from GNSS dynamic positioning has significant errors, which will affect the accuracy and reliability of structural monitoring and assessment.

[0004] Traditional displacement extraction methods mainly rely on single models or fixed parameter filtering methods, such as Kalman Filter. These methods perform well when dealing with linear systems, but they do not work well when faced with complex nonlinear dynamics and noise. For example, when dealing with nonlinear systems, the extended Kalman filter (EKF) and the unscented Kalman filter (UKF) rely on the first-order linear approximation of the system or use sigma points for estimation. As the state dimension of the system increases, the computational complexity increases dramatically. In addition, although the particle filter (PF) method has advantages in dealing with nonlinear systems, it requires processing a large number of particles and has high computational costs.

[0005] In order to improve the accuracy and robustness of displacement extraction, in recent years, interactive multi-model Kalman filtering (IMM) has gradually attracted attention as an adaptive state estimation algorithm. IMM can combine multiple models and flexibly adjust according to the changes in system state to provide more accurate estimates. However, in existing research, IMM is mainly used in the fields of target tracking and navigation, and its application in structural health monitoring is relatively small. Moreover, for vibration detection of structures, it is often necessary to accurately design a model set to cope with the multimodal characteristics and changes in natural frequencies of the structure.

[0006] Based on existing research, this patent uses an interactive multi-model Kalman filter to design a model set based on the natural frequency of the structure and perform interactive processing between multiple models. It combines the advantages of different models, optimizes displacement estimation, and significantly reduces the impact of measurement noise and signal interference, thereby improving the accuracy and reliability of displacement extraction. Summary of the invention

[0007] The present invention develops a displacement extraction method with model adaptive adjustment. It can improve the accuracy and robustness of displacement extraction in complex environments and variable conditions. The method realizes adaptive displacement extraction under uncertain model parameters by designing a multi-model set to adapt to the multi-modal characteristics and natural frequency changes of the structure.

[0008] The present invention provides a displacement extraction method based on model adaptive adjustment, comprising the following steps:

[0009] Step 1: Dynamic positioning displacement input: Get preliminary displacement measurement through displacement sensors such as displacement meter and GNSS.

[0010] Step 2: Rough frequency identification: Based on the preliminary displacement data, the vibration frequency of the structure is roughly identified.

[0011] Step 3: Determine the frequency range: Determine a frequency range based on the identified coarse frequency.

[0012] Step 4: State model set: Select sub-frequencies within the frequency range to construct a set of state model sets.

[0013] Step 5: Use the interactive multi-mode Kalman filter to process the state model set. Through the combination and interactive processing of multiple models, high-precision state estimation and displacement extraction can be achieved.

[0014] Preferably, in step 2, the frequency is roughly identified, and the identification step is to identify the main vibration frequency through the fast Fourier transform (FFT) method.

[0015] Assume that the initial displacement data is x(t), where t represents the time. Assume that the initial displacement data is x(t), and obtain its spectrum X(f) through FFT:

[0016]

[0017] f represents the frequency, j is the imaginary unit, and e is a natural constant. Next, calculate the amplitude of the spectrum:

[0018]

[0019] Where Re(X(f)) and Im(X(f)) are the real and imaginary parts of the spectrum X(f), respectively.

[0020] Finally, in the spectrum |X(f)|, find the frequency f0 with the largest amplitude as the main vibration frequency:

[0021] f0=arg max f |X(f)|

[0022] Preferably, in step 3, the frequency range is determined by setting a predetermined frequency window according to the identified coarse frequency. According to the identified main frequency f0, a frequency range [f min , f max ], the frequency range can be determined by empirical formula or statistical analysis:

[0023] f min =f0-Δf

[0024] f max =f0+Δf

[0025] Where Δf is the frequency range.

[0026] Preferably, in step 4, within the frequency range, sub-frequencies are divided with a certain step length to construct a state model set, ensuring that the model can cover all possible vibration frequencies. min , f max ], the sub-frequency {f i}:

[0027] f i =f min +i·δf,i=0,1,2,...,N

[0028] Where N is the number of sub-frequency, satisfying f i ≤f max .

[0029] The relationship between frequency f and spring coefficient k is:

[0030]

[0031] Among them, f i is the ith sub-frequency, k i is the corresponding spring coefficient, and m is the mass of the oscillator.

[0032] Preferably, in step 5, specific parameters of the constant position model and the damped vibration model are as follows:

[0033] (1) Constant position model, the state parameter is displacement [x], its state transfer matrix is ​​F = [1], and the design matrix is ​​H = [1];

[0034] When considering the spring coefficient k, the state parameter is [x, k] and the state transfer matrix is The design matrix is ​​H = [1, 0] T ;

[0035] (2) Damped vibration model, based on the sub-frequency in step 4, the state parameters are (displacement and velocity), its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0] T ;

[0036] When considering the spring coefficient k, the state parameter is Its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0, 0] T ;

[0037] Consider acceleration When , the state parameter is Its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0, 0] T ;

[0038] Considering the spring coefficient k and acceleration When , the state parameter is Its state transfer matrix is

[0039] The design matrix is ​​H = [1, 0, 0, 0] T ;

[0040] Preferably, in step 5, the state model set includes:

[0041] (1) Single frequency dual model (SFCD), assuming that the modal parameters remain unchanged, the structural state switches between damped vibration and static state, and the state transfer matrix is and Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval;

[0042] (2) Dual-frequency three-model (DFCT), assuming that the structure is affected by two main vibration modes and switches between vibration and static states. The state transfer matrix is and Where c is the damping coefficient, m is the mass of the oscillator, k1 and k2 are the corresponding spring coefficients, and Δt is the sampling interval;

[0043] (3) Multi-frequency multi-model (MFCM), assuming that the main vibration frequency is unclear or the structure is subjected to vibrations of multiple main frequencies, the state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, and k i is the corresponding spring coefficient, and Δt is the sampling interval.

[0044] In the present invention, first, preliminary displacement measurements are obtained through displacement sensors such as displacement meters and GNSS. Secondly, based on the preliminary displacement data, the approximate vibration frequency of the structure is roughly identified, and a frequency range is determined. Then, sub-frequencies are selected within the frequency range to construct a set of state model sets. Including constant position models and damped vibration models to adapt to different vibration modes and frequencies. Next, an interactive multi-model Kalman filter (IMM) is used to estimate displacement, and adaptive displacement extraction is achieved through mixing, filtering and combination steps. Finally, the final displacement estimate is output based on the filtering result. Verified by simulation experiments and field experiments, the method of the present invention exhibits good robustness and accuracy under different frequencies and vibration modes, and can effectively improve the reliability of structural health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic flow chart of a displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0046] Figure 2 A constant frequency simple vibration displacement simulation diagram of a displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0047] Figure 3 A graph showing the displacement extraction results of a simple vibration at a constant frequency according to a displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0048] Figure 4 Simulation diagrams of simple vibration displacements of different frequencies of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0049] Figure 5 A graph showing the displacement extraction results of simple vibrations of different frequencies using the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0050] Figure 6 RMS diagram of each method of simple vibration displacement simulation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0051] Figure 7 A constant frequency combined vibration displacement simulation diagram of a displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0052] Figure 8 A diagram showing the constant frequency combined vibration displacement extraction result of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0053] Fig. 9 Simulation diagram of vibration displacement with different frequency combinations of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0054] Fig.10 A diagram showing the extraction results of vibration displacements of different frequency combinations using the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0055] Fig.11 RMS diagram of each method of combined vibration displacement simulation of displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0056] Fig.12 A displacement result diagram of the RTK first vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0057] Fig.13 A displacement result diagram of the RTK second vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0058] Fig.14 A displacement result diagram of the RTK third vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0059] Fig.15 A displacement result diagram of the RTK fourth vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0060] Fig.16 The RMS of displacement extraction of each method in the RTK experiment of the displacement extraction method based on model adaptive adjustment provided by the embodiment of the present invention.

[0061] Fig.17 A displacement result diagram of the PPP first vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0062] Fig.18 A displacement result diagram of the PPP second vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0063] Fig.19 A displacement result diagram of the PPP third vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0064] Fig. 20 A displacement result diagram of the PPP fourth vibration estimation of the displacement extraction method based on model adaptive adjustment provided by an embodiment of the present invention;

[0065] Fig.21 The RMS of displacement extraction of each method in the PPP experiment of the displacement extraction method based on model adaptive adjustment provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0066] The present invention aims to explore a displacement extraction method with model adaptive adjustment, that is, a method of adaptive displacement extraction based on an interactive multi-model Kalman filter (IMM). The specific implementation process is to dynamically locate the displacement input and obtain preliminary displacement measurements through displacement sensors such as displacement meters and GNSS. Then, based on the preliminary displacement data, the approximate vibration frequency of the structure is roughly identified and a frequency range is determined. Then, sub-frequencies are selected within the frequency range to construct a set of state model sets. Finally, an interactive multi-model Kalman filter (IMM) is used for displacement estimation, and adaptive displacement extraction is achieved through mixing, filtering and combining steps.

[0067] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0068] This embodiment provides a displacement extraction method based on model adaptive adjustment, which mainly includes the following steps:

[0069] Step 1: Dynamic positioning displacement input: Get preliminary displacement measurement through displacement sensors such as displacement meter and GNSS.

[0070] Step 2: Rough frequency identification: Based on the preliminary displacement data, the vibration frequency of the structure is roughly identified.

[0071] The main vibration frequency is identified by the Fast Fourier Transform (FFT) method. Assume that the preliminary displacement data is x(t), where t represents the time. The spectrum X(f) is obtained by FFT:

[0072]

[0073] f represents frequency, j is an imaginary unit, and e is a natural constant. Next, calculate the amplitude of the spectrum:

[0074]

[0075] Where Re(X(f)) and Im(X(f)) are the real and imaginary parts of the spectrum X(f), respectively.

[0076] Finally, in the spectrum |X(f)|, find the frequency f0 with the largest amplitude as the main vibration frequency:

[0077] f0=arg max f |X(f)|

[0078] Step 3: Determine the frequency range: Determine a frequency range based on the identified approximate frequency.

[0079] According to the identified main frequency f0, a frequency range [f min , f max ], the frequency range can be determined by empirical formula or statistical analysis:

[0080] f min =f0-Δf

[0081] f max =f0+Δf

[0082] Where Δf is the frequency range.

[0083] Step 4: State model set: Select sub-frequencies within the frequency range to construct a set of state model sets.

[0084] In the frequency range [f min , f max ], the sub-frequency {f i}, used to construct the state model set to ensure that the model can cover all possible vibration frequencies.

[0085] f i =f min +i·δf,i=0,1,2,...,N

[0086] Where N is the number of sub-frequency, satisfying f i ≤f max .

[0087] The relationship between frequency f and spring coefficient k is:

[0088]

[0089] Among them, f i is the ith sub-frequency, k i is the corresponding spring coefficient, and m is the mass of the oscillator.

[0090] Step 5: Use the interactive multi-mode Kalman filter to process the state model set. Through the combination and interactive processing of multiple models, high-precision state estimation and displacement extraction can be achieved. The specific steps are as follows:

[0091] (1) Model interaction: Information is exchanged between different models to improve state prediction. The mixing probability of each model is calculated and the initial conditions of each model are generated.

[0092] (2) Filter combination: Based on the likelihood function and prior probability of each model, the posterior probability of each model is calculated. Using these posterior probabilities, the estimation results of each model are weighted and combined to generate the final state estimate.

[0093] (3) Kalman filter: Kalman filter is applied in each model for state estimation, including prediction and update steps.

[0094] The present invention is further described below.

[0095] See also Figure 1 The displacement extraction method of the model adaptive adjustment provided in this embodiment mainly includes the following steps:

[0096] In the invention, the first step is to dynamically locate the displacement input. Preliminary displacement measurements are obtained through displacement sensors such as displacement meters and GNSS. Then, based on the preliminary displacement data, the approximate vibration frequency of the structure is roughly identified, and the frequency range is determined. Then, the sub-frequencies are divided within the frequency range to construct a set of state model sets. Finally, an interactive multi-model Kalman filter (IMM) is used for displacement estimation, and adaptive displacement extraction is achieved through mixing, filtering and combining steps. The present invention can show good robustness and accuracy under different frequencies and vibration modes, and can effectively improve the reliability of structural health monitoring.

[0097] Several main parts of this embodiment are explained below.

[0098] 1. Dynamic positioning displacement input

[0099] Preliminary displacement measurements are obtained through displacement sensors such as displacement meters and GNSS.

[0100] 2. Rough frequency identification

[0101] Based on the preliminary displacement data, the vibration frequency of the structure is roughly identified. The purpose of this step is to determine the main vibration frequency range of the structure and provide a basis for the subsequent frequency range determination. The main vibration frequency is identified by the fast Fourier transform (FFT) method.

[0102] Assume that the initial displacement data is x(t), where t represents the time. The spectrum X(f) is obtained through FFT:

[0103]

[0104] f represents the frequency, j is the imaginary unit, and e is a natural constant. Next, calculate the amplitude of the spectrum:

[0105]

[0106] Where Re(X(f)) and Im(X(f)) are the real and imaginary parts of the spectrum X(f), respectively.

[0107] Finally, in the spectrum |X(f)|, find the frequency f0 with the largest amplitude as the main vibration frequency:

[0108] f0=arg max f |X(f)|

[0109] 3. Determine the frequency range

[0110] Determining the frequency range is of great significance in the model building process, mainly to improve the accuracy and robustness of the model and optimize the performance of the filter. In addition, determining an appropriate frequency range can ensure that the selected sub-frequency covers the main vibration frequency of the structure and its possible range of variation, thereby improving the accuracy of the state model.

[0111] According to the identified main frequency f0, a frequency range [f min , f max ], the frequency range can be determined by empirical formula or statistical analysis:

[0112] f min =f0-Δf

[0113] f max =f0+Δf

[0114] Where Δf is the frequency range.

[0115] 4. State Model Set

[0116] For the Interactive Multi-Model Kalman Filter (IMM), by selecting sub-frequencies within the frequency range to construct multiple state models, the filter performance can be optimized, the diversity of the state model set can be ensured, and the estimation accuracy and stability can be improved. In addition, the sub-frequency selection provides a complete frequency spectrum analysis framework, so that the analysis is not limited to the main frequency, but extends to the entire frequency range, thereby fully understanding the vibration characteristics and behavior of the structure.

[0117] This patent is in the frequency range [f min , f max ], the sub-frequency {f i}, used to construct the state model set to ensure that the model can cover all possible vibration frequencies.

[0118] f i =f min +i·δf,i=0,1,2,...,N

[0119] Where N is the number of sub-frequency, satisfying f i ≤f max .

[0120] The relationship between frequency f and spring coefficient k is:

[0121]

[0122] Among them, f i is the ith sub-frequency, k i is the corresponding spring coefficient, and m is the mass of the oscillator.

[0123] 5. Displacement estimation based on interacting multiple models

[0124] Use the IMM filter to process the state model set. Through the combination and interactive processing of multiple models, high-precision state estimation and displacement extraction are achieved. The specific steps are as follows:

[0125] (1) Model interaction: Information is exchanged between different models to improve state prediction. The mixing probability of each model is calculated and the initial conditions of each model are generated.

[0126] (2) Filter combination: Based on the likelihood function and prior probability of each model, the posterior probability of each model is calculated. Using these posterior probabilities, the estimation results of each model are weighted and combined to generate the final state estimate.

[0127] (3) Kalman filter: Kalman filter is applied in each model for state estimation, including prediction and update steps.

[0128] In order to verify the effect of the present invention, the following is a description in conjunction with experimental results.

[0129] According to the requirements, simulation experiments and field experiments were designed. The results of both simulation experiments and field experiments show that the present invention can realize adaptive displacement extraction, and can show good robustness and accuracy under different frequencies and vibration modes, and can effectively improve the reliability of structural health monitoring. The simulation experiment is divided into simple vibration displacement and combined vibration displacement simulation. Figure 2 and Figure 7 They are the simulation diagrams of simple vibration displacement and combined vibration displacement under constant frequency. Figure 3 and Figure 8 They are the corresponding displacement extraction result diagrams. Figure 3As shown, it can be seen that the displacement extracted by the CP and CP_K models still contains significant noise. The main reason is that the model does not match the actual motion state. DP2 and DP2_K are very close to the reference displacement, especially in the vibration stage, because the frequency and damping ratio adopted by the DP2 and DP2_K models are exactly what are used in the simulation. DP3 and DP3K are relatively unstable, mainly because the high-order parameters make the state equation satisfy both the displacement-acceleration relationship and the damped motion constraint. The model set IMM_SFCD and the model set IMM_MFCM achieve results consistent with the reference displacement through model adaptation.

[0130] Initially, the simulations focused on the superposition of two vibrations with the same frequency, with results such as Figure 7 The analysis results are shown in Figure 8 As shown, it can be observed that the displacement estimation obtained from the model group is better than other methods. In addition, for different frequencies, the combined displacement is simulated, such as Fig. 9 Similar to Figure 7 , except that here the two superimposed vibration frequencies are different. The displacements extracted by the model set have a superior advantage compared to the single model. In addition, the model sets IMM_DFCT and IMM_MFCM are superior to IMM_SFCD because they include state models corresponding to two vibration frequencies.

[0131] In order to verify the influence of the proposed method on the change of physical parameters, another set of frequency conversion data was simulated using a simulation software package. Figure 4 and Fig. 9 They are simulation diagrams of simple vibration displacement and combined vibration displacement at different frequencies. Figure 5 and Fig.10 They are the corresponding displacement extraction result diagrams. The simulation results are as follows Figure 4 As shown in the figure, the vibration frequencies at 5s and 35s are inconsistent, the displacement amplitudes are also inconsistent, and the displacement amplitudes are small. The displacement is finally extracted through the interactive Kalman filter, as shown in Figure 5 As shown. Results similar to those of the previous set of simulation experiments were obtained, but there are also some differences. For DP2, DP2_K, DP3, and DP3_K, the displacement extraction in the second half is not as smooth as that in the first half, mainly because the designed state model only considers a single main frequency. The model sets IMM_SFCD, IMM_DFCT, and IMM_MFCM are basically consistent with the reference displacement, and IMM_SFCD is slightly worse than the other two. The reason is that the model set IMM_SFCD lacks a state model corresponding to the vibration in the second half.

[0132] The root mean square values ​​of various methods in simple vibration displacement experiments are shown in Figure 6 , the results are similar to the above Figure 3 and Figure 5 The analysis results are consistent. Fig.11 It is the RMS table of combined vibration displacement simulation methods. Fig.11 The RMS statistics of the displacements extracted by various methods are summarized in Figure 8 and Fig.10 The accuracy of the extracted displacements.

[0133] In the field experiment, RTK and PPP kinematic experiments were conducted separately. During the experiment, two BD992OEM boards were used to collect GNSS data. The reference station was mounted on a tripod, while the rover was fixed on a vibration platform. Displacements were extracted using a nearby vision system as a reference. During the experiment, a total of eight vibrations were initiated. Among them, in the RTK experiment, after processing the first 4 vibrations, the displacements extracted by various methods were as follows Figure 12-15 The statistical results of the displacement estimation accuracy of various methods for all 8 vibrations are shown in Fig.16 As shown. The statistical results are not completely consistent with the results of the simulation experiment, indicating that some conditions or variables in the field experiment may be different from those in the simulation environment, resulting in differences between the actual results and the expected results. The CP model performs slightly better than the DP model here, mainly because the actual GNSS positioning results include low-frequency displacement components, which lead to changes in the equilibrium position. The results produced by using the physical quantity K as a parameter are slightly better than those of the corresponding model without K as a parameter. The possible reason is that this parameter alleviates the impact of model mismatch. However, it can be seen that the model set method shows significant improvement over the single model method, and the results obtained by the models IMM_SFCD, IMM_DFCT and IMM_MFCM are consistent.

[0134] When using PPP dynamic positioning mode for processing, it can be seen from the processing results that the accuracy of PPP dynamic processing is lower than the results of RTK processing. Various models and methods have been used to process the results from the kinematic positioning output. The results of the first four vibrations are Figure 17-20 The statistical results of all vibrations are summarized in Fig.21 Although the accuracy of the PPP kinematic processing results is slightly lower than that of RTK, the proposed method is still effective. The CP model outperforms the DP model, which may be due to the low-frequency unmodeled errors in the positioning results, resulting in the change of the equilibrium position being inconsistent with the actual vibration. But overall, the multi-model method based on interactive multimodality still achieved the best results.

[0135] The present invention aims to provide a displacement extraction method with adaptive model adjustment. The system can realize adaptive displacement extraction. It can show good robustness and accuracy in different frequencies and vibration modes, and can effectively improve the reliability of structural health monitoring. Compared with traditional methods, this method can effectively solve the problems faced by the existing technology in dynamic positioning data processing, such as large measurement noise, serious multipath effect and signal interference, high limitations of filtering methods, and single model, thereby improving the accuracy and reliability of displacement estimation.

[0136] Finally, it should be noted that the above specific implementation methods are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to examples, a person of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the present invention, which should be covered by the scope of the patent claims of the present invention.

Claims

1. A displacement extraction method with adaptive model adjustment, It is characterized in that The following steps are involved: Step 1: Dynamic positioning displacement input: Get preliminary displacement measurement through displacement sensors such as displacement meter and GNSS. Step 2: Rough frequency identification: Based on the preliminary displacement data, the vibration frequency of the structure is roughly identified. Step 3: Determine the frequency range: Determine a frequency range based on the identified coarse frequency. Step 4: State model set: Select sub-frequencies within the frequency range to construct a set of state model sets. Step 5: Use the interactive multi-mode Kalman filter to process the state model set. Through the combination and interactive processing of multiple models, high-precision state estimation and displacement extraction can be achieved.

2. A displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In the step 1, the displacement measurement method may be direct displacement measurement or indirect displacement measurement.

3. The displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In step 2, the frequency rough identification step is to identify the main vibration frequency by the fast Fourier transform (FFT) method. Assume that the preliminary displacement data is x(t), t represents the time. The spectrum X(f) is obtained by FFT: f represents the frequency, j is the imaginary unit, and e is a natural constant. Next, calculate the amplitude of the spectrum: Where Re(X(f)) and Im(X(f)) are the real and imaginary parts of the spectrum X(f), respectively. Finally, in the spectrum |X(f)|, find the frequency f with the largest amplitude 0 , as the main vibration frequency: f 0 =arg max f |X(f)| 4. The displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In step 3, the frequency range is determined by setting a predetermined frequency window according to the identified coarse frequency. According to the identified main frequency f 0 , determine a frequency range [f min , f max ], the frequency range can be determined by empirical formula or statistical analysis: f min =f 0 -Δf f max =f 0 +Δf Where Δf is the frequency range.

5. The displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In step 4, within the frequency range, sub-frequencies are divided with a certain step length to construct a state model set to ensure that the model can cover all possible vibration frequencies. min , f max ], the sub-frequency {f i }: f i =f min +i·δf,i=0,1,2,...,N Where N is the number of sub-frequency, satisfying f i ≤f max . The relationship between frequency f and spring coefficient k is: Among them, f i is the ith sub-frequency, k i is the corresponding spring coefficient, and m is the mass of the oscillator.

6. A displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In step 5, the state model set includes the following models: (1) Constant position model, used to describe the behavior of the structure in a static state; (2) Damped vibration model, which is used to describe the vibration behavior of the structure when subjected to external forces; (3) Multimodal vibration model, which is used to describe the dynamic characteristics of the structure under multiple vibration modes.

7. A GNSS displacement extraction system with model adaptive adjustment according to claim 6, It is characterized in that In step 5, the specific parameters of the constant position model and the damped vibration model are as follows: (1) Constant position model, the state parameter is displacement [x], its state transfer matrix is ​​F = [1], and the design matrix is ​​H = [1]; When considering the spring coefficient k, the state parameter is [x, k] and the state transfer matrix is The design matrix is ​​H = [1, 0] T ; (2) Damped vibration model, based on the sub-frequency in step 4, the state parameters are (displacement and velocity), its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0] T ; When considering the spring coefficient k, the state parameter is Its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0, 0] T ; Consider acceleration When , the state parameter is Its state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval. The design matrix is ​​H = [1, 0, 0] T ; Considering the spring coefficient k and acceleration When , the state parameter is Its state transfer matrix is The design matrix is ​​H = [1, 0, 0, 0] T。 8. The GNSS displacement extraction system with model adaptive adjustment according to claim 1, It is characterized in that In step 5, the state model set includes: (1) Single frequency dual model (SFCD), assuming that the modal parameters remain unchanged, the structural state switches between damped vibration and static state, and the state transfer matrix is and Where c is the damping coefficient, m is the mass of the oscillator, k is the corresponding spring coefficient, and Δt is the sampling interval; (2) Dual-frequency three-model (DFCT), assuming that the structure is affected by two main vibration modes and switches between vibration and static states. The state transfer matrix is and Where c is the damping coefficient, m is the mass of the oscillator, and k 1 , k 2 is the corresponding spring coefficient, Δt is the sampling interval; (3) Multi-frequency multi-model (MFCM), assuming that the main vibration frequency is unclear or the structure is subjected to vibrations of multiple main frequencies, the state transfer matrix is Where c is the damping coefficient, m is the mass of the oscillator, and k i is the corresponding spring coefficient, and Δt is the sampling interval.

9. The displacement extraction method with model adaptive adjustment according to claim 1, It is characterized in that In step 5, the state model set is processed using an interactive multi-mode Kalman filter. The following steps are included: (1) Pattern probability mixing: Calculate the mixing probability at the current moment based on the pattern probability and transition probability at the previous moment; (2) Independent filtering: Apply a Kalman filter to each state model for independent state estimation; (3) Pattern probability update: Calculate the posterior pattern probability based on the likelihood function and mixture probability of each model; (4) State estimation combination: The state estimates of different models are weighted combined to obtain the final state estimate.

Images