A Recursive Random Subspace Identification Method Based on Ramanujan Subspace Projection
By employing Ramanujan subspace projection and sliding window recursion algorithms, the problems of harmonic frequency misidentification and high computational complexity are solved, enabling accurate tracking of modal parameters of time-varying systems. This method is applicable to fields such as rotating machinery and aerospace.
Patent Information
- Application Number
- CN202611543262.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional random subspace identification methods are prone to misidentifying harmonic frequencies as structural mode frequencies in responses containing harmonic components, and the computational burden of tracking modal parameters of time-varying systems is large, making it difficult to reflect them in real time.
The Ramanujan subspace projection method is adopted. By determining the harmonic frequency and mapping it to the periodic domain, an orthogonal projection coefficient matrix is constructed to remove the harmonic response. Combined with the sliding window forgetting mechanism and the subspace tracking algorithm, the extended observable subspace matrix is updated in real time to extract time-varying modal parameters.
It effectively removes the interference of harmonic components, reduces the risk of misidentification of modal parameters, realizes online real-time identification of time-varying systems, reduces the amount of calculation, and is suitable for engineering practice.
Smart Images

Figure CN123115602A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of structural dynamics and working mode identification technology, and in particular to a recursive random subspace identification method based on Ramanujan subspace projection. Background Technology
[0002] In engineering fields such as rotating machinery, mechanical manufacturing, and aerospace, the mass, stiffness, boundary conditions, or load states of structural systems may change over time, causing their dynamic characteristics to exhibit time-varying characteristics. The measured response may also simultaneously contain random environmental excitations as well as harmonic excitations generated by rotational imbalance, periodic loads, gear meshing, etc., resulting in the superposition of random and periodic components.
[0003] Traditional stochastic subspace identification is typically based on the Gaussian stochastic excitation assumption. When harmonic components are mixed into the response, harmonic frequencies may be misidentified as structural modal frequencies. For time-varying systems, modal frequencies broaden in the frequency domain, and harmonic frequencies may be close to or overlap with natural frequencies. Using conventional frequency domain peak separation or fixed bandwidth filtering can easily lead to the accidental deletion of structural response components. On the other hand, batch identification of all historical data at each time step increases the computational load continuously, making it difficult to reflect changes in modal parameters in a timely manner.
[0004] Therefore, there is a need for a method that utilizes only the structural output response, can identify and remove harmonic components during the recursive identification process, and can track time-varying modal parameters with a finite data window. Summary of the Invention
[0005] The purpose of this invention is to overcome the problems of traditional recursive subspace identification, which easily misidentifies harmonic components as structural modes, suffers from the increasing computational burden due to data growth, and is susceptible to damage to true modal components by fixed filtering, under conditions of mixed random and harmonic excitation and time-varying system dynamic parameters. This invention aims to achieve accurate and continuous tracking of the natural frequencies and damping ratios of time-varying structures. To solve these technical problems, this invention provides a recursive random subspace identification method based on Ramanujan subspace projection. This method first establishes a strictly periodic subspace based on known or detected harmonic frequencies, extracts harmonic responses from the output responses of each measurement point, updates the extended observable subspace matrix using the deharmonicized responses within a fixed-length window, and obtains the time-varying modal parameters from the eigenvalues of the state matrix.
[0006] The objective of this invention is achieved through the following technical solution: A recursive random subspace identification method based on Ramanujan subspace projection includes: S1. Obtain the output response data of each measuring point of the structural system in the working state; S2. Determine the harmonic excitation frequency mixed in the output response data, and map the harmonic excitation frequency from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace; S3. Construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-deharmonic response data. S4. Based on the sliding window forgetting mechanism, a stacked vector matrix is constructed using the deharmonic response data, and the extended observable subspace matrix is updated in real time through a subspace tracking algorithm. S5. Extract the state matrix based on the recursively updated extended observable subspace matrix, and obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
[0007] Furthermore, in step S2, when mapping the harmonic excitation frequency from the frequency domain to the periodic domain, the following coprime integer parameters are used. Perform mapping: in, In response to the sampling frequency of the data, Let be the harmonic excitation frequency; select parameter pairs that satisfy the above mapping relationship and are coprime to calculate the normalized circular frequency in the Ramanujan subspace.
[0008] Furthermore, in step S3, for each normalized circular frequency, the corresponding Ramanujan subspace is spanned using conjugate complex pairs as basis functions, and the orthogonal projection coefficient matrix is calculated. The output response data at each measuring point is multiplied by the orthogonal projection coefficient matrix to obtain the harmonic response at the corresponding frequency. Subtracting the harmonic response from the original output response data yields the deharmonic response data.
[0009] Furthermore, in step S4, the method for constructing the stacked vector matrix includes: The output response data is divided into multiple fixed-length RSP windows. After introducing the response data at a new time step, an extended RSP window is created. Stacked vectors are constructed using the deharmonic response data within the extended RSP window, and multiple stacked vectors within the same extended RSP window are combined to form a stacked matrix.
[0010] Furthermore, in step S4, the subspace tracking algorithm is the Projection Approximation Subspace Tracking Method (PAST). Based on the stacked matrix, constraint equations are established using a sliding window forgetting mechanism, and the extended observable subspace matrix and covariance matrix at each time step are obtained recursively.
[0011] Furthermore, in step S5, the extraction of the state matrix based on the recursively updated extended observable subspace matrix specifically includes: The extended observable subspace matrix is divided into a front block row matrix and a rear block row matrix; The state matrix at the current moment is obtained by multiplying the pseudo-inverse of the front block row matrix with the rear block row matrix.
[0012] Furthermore, before step S1, the method includes: performing batch subspace identification based on historical response data of a fixed length, and obtaining recursive initial values for the extended observable subspace matrix and the covariance matrix.
[0013] The present invention also provides a recursive random subspace identification device based on Ramanujan subspace projection, comprising: The data acquisition unit is used to acquire the output response data of each measuring point of the structural system in the working state; Ramanujan subspace units are used to determine the harmonic excitation frequencies mixed in the output response data and map the harmonic excitation frequencies from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace. The harmonic removal unit is used to construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-removed response data. The update unit is used to construct a stacked vector matrix using the deharmonic response data based on the sliding window forgetting mechanism, and to recursively update the extended observable subspace matrix in real time through a subspace tracking algorithm. The output unit is used to extract the state matrix based on the recursively updated extended observable subspace matrix, and to obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the recursive random subspace identification method.
[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the recursive random subspace identification method.
[0016] Compared with the prior art, the beneficial effects of the technical solution of the present invention are as follows: 1. This invention employs a Ramanujan subspace for projection onto strictly periodic components. Traditional frequency domain filtering or peak separation can inadvertently remove true structural signals when frequencies are close together. This invention maps harmonic frequencies to the periodic domain and establishes a strictly orthogonal subspace, projecting and extracting only strictly periodic components. Therefore, even if harmonics and natural frequencies highly overlap in the frequency domain, they can be completely separated in the periodic domain, protecting the true random response data and reducing the risk of harmonic components being misidentified as structural modes.
[0017] 2. This invention introduces a fixed-length RSP window and a sliding window forgetting mechanism, combined with the PAST recursive method for projective approximation subspace tracking. Traditional batch identification suffers from an explosive increase in matrix dimension over time. This invention limits the data size involved in the calculation at each time step through a sliding window and directly reuses the subspace basis from the previous time step for recursion through the PAST constraint equation. This avoids the repeated decomposition of all historical data, controlling the data size involved in the update at each time step, thus meeting the needs of online real-time identification for time-varying systems. It is applicable to monitoring operational status in engineering practice where input is difficult to obtain.
[0018] 3. This invention utilizes only structural output response data for stacking matrix construction and state matrix identification. In practical engineering, environmental wind loads, road surface excitations, or complex internal mechanical excitations are often difficult to measure accurately (due to lack of input data). This invention, based on a state-space model and the assumption of output-only operation, eliminates the dependence on input excitation signals, thus improving the feasibility of the method in practical engineering applications (such as aerospace and rotating machinery). Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the method of the present invention.
[0020] Figure 2 This is a schematic diagram of a three-degree-of-freedom gradually varying mass structure system.
[0021] Figure 3a The displacement signal of the numerical system under 3.5Hz harmonic and random mixed excitation.
[0022] Figure 3b The displacement signal of the numerical system under 3Hz harmonic and random mixed excitation.
[0023] Figure 4a The results are spectral analysis of the first degree of freedom displacement signal under 3.5Hz harmonic and random mixed excitation.
[0024] Figure 4b The results are spectral analysis of the first degree of freedom displacement signal under 3Hz harmonic and random mixed excitation.
[0025] Figure 5aThe results show the intrinsic frequency identification using the method of this invention under 3.5Hz harmonic and random mixed excitation.
[0026] Figure 5b The results show the intrinsic frequency identification using the method of this invention under 3.5Hz harmonic and random mixed excitation. Detailed Implementation
[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.
[0028] Example 1 This embodiment provides a recursive random subspace identification method based on Ramanujan subspace projection, see [link to documentation]. Figure 1 The details are as follows: S1. Obtain the output response data of each measuring point of the structural system in the working state; For a multi-degree-of-freedom structural system under random and harmonic mixed excitation, its dynamic equations can be expressed as follows: (1) In the formula, Let be the displacement vector of the structural system over continuous time. , , These are the mass matrix, damping matrix, and stiffness matrix, respectively. As the activation vector, it can be decomposed into ,in For random incentives, It is a harmonic excitation.
[0029] Based on the above definition, the discrete-time state-space model can be established as follows: (2) In the formula, the subscript Represents the sampling time. for The state vector at time t, The sampling time interval, for The output vector at time step, i.e., the measured response data. Let be the state matrix of the structural system. The output matrix of the structural system, It is a state noise vector. It measures the noise vector, superscript and These correspond to random excitation and harmonic excitation, respectively.
[0030] S2. Determine the harmonic excitation frequency mixed in the output response data, and map the harmonic excitation frequency from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace. When the excitation is random, the state matrix of the structural system can be obtained based on measured response data using an output-only identification method. and output matrix This leads to the following eigenvalue decomposition operation: (3) In the formula, It is a diagonal matrix composed of eigenvalues. For the discrete-time system, the first i eigenvalues of order 1 The eigenvector matrix is used to identify the first eigenvalue of the structural system. i First natural frequency Damping ratio as follows: (4) Traditional output-only identification methods mostly assume that the system is subjected to Gaussian random excitation. However, when harmonic excitation is mixed with random excitation, the assumption of Gaussian random excitation is no longer satisfied. If the state matrix is still obtained using the method based on the above state-space model... Thus, identification and This introduces harmonic frequency interference into the identified modal parameters, so the harmonic response is processed using the RSP principle.
[0031] When random excitation is mixed with a frequency of When performing harmonic excitation, this frequency is first mapped from the frequency domain to the periodic domain. The corresponding mapping formula is: (5) In the formula, and These are coprime integer parameter pairs, through which the harmonic frequency can be determined. Normalized frequency in Ramanujan subspace The mapping, For the corresponding angular frequency, The sampling frequency for the response data.
[0032] It should be noted that, due to and All are integers and coprime, target normalized frequency Generally difficult to Completely identical, therefore it needs to be within the maximum allowed range. Within, select the parameter whose frequency is closest to the target harmonic. The specific constraint equations are as follows: (6) In the formula, This means the two parameters are coprime. For parameters The maximum integer that can be selected should be such that it can select the parameter pair with the smallest error while minimizing the amount of calculation. It is recommended to select 1000. If it does not meet the requirements, it should be adjusted accordingly.
[0033] For length of The response data, assuming the stimulus is mixed with Each harmonic excitation, then each normalized circumfrequency The corresponding Ramanujan subspaces can all be spanned by the following conjugate complex pairs as basis functions: (7) In the formula, for The corresponding Ramanujan subspace, basis functions ,in The imaginary unit, Let it be its complex conjugate vector. This is the conjugate transpose. To normalize the parameters, the basis functions must be unit vectors.
[0034] S3. Construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-deharmonic response data. Define the orthogonal projection coefficient matrix Thus, the frequency is extracted from the response data. The harmonic response is as follows: (8) In the formula, For the response data of a certain measuring point, Normalized circumfrequency Orthogonal projection in the Ramanujan subspace, i.e., frequency The corresponding harmonic response.
[0035] Then the deharmonic response It can be obtained in the following ways (9) Based on the above-mentioned response after harmonic removal, modal parameters can be obtained through batch identification methods.
[0036] S4. Based on the sliding window forgetting mechanism, a stacked vector matrix is constructed using the deharmonic response data, and the extended observable subspace matrix is updated in real time through a subspace tracking algorithm. The aforementioned projection method can obtain modal parameters based on response data of finite length. However, for structural systems whose dynamic characteristics change over time, if all historical data are collected again and batch-computed at every moment, not only will the computational load increase continuously with the data length, but it will also be difficult to reflect the changes in the system's modal parameters in a timely manner. Therefore, a sliding window forgetting mechanism is introduced to update the modal parameters based on the response data within a fixed window length.
[0037] The entire response data is divided into multiple RSP windows, where the first... RSP window at time Composition as follows (10) In the formula, the subscript represents the sampling time. The length of the RSP window. for The response data at each measuring point at time t, i.e., the response column vector, when the t is introduced... After receiving the response data at each moment, establish the first Extended RSP window at time as follows: (11) At this time, the extended RSP window contains a total of The response data of the group is used to apply a harmonic removal method based on the RSP strategy to the response data of each measuring point within this extended window, obtaining the data from the first group of response data. Time to the The deharmonic response data at time 1. To update the system's state matrix using this response data, the stacked vectors are... Construct it in the following form: (12) In the formula, the subscript To respond to the sampling time corresponding to the data, The response data after removing harmonic components, This represents the number of moments involved in the stacking process. For this time, it includes... An extended RSP window for group response data can be constructed. Stacked vectors are grouped into a stacked matrix by combining stacked vectors within the same extended RSP window. as follows: (13) The formula for calculating the length of the sliding window is as follows: To achieve the tracking of modal parameters of a time-varying system, the following constraint equations can be established based on the Projection Approximation Subspace Tracking method: (14) In the formula, To extend the observable subspace matrix, it contains the state matrix. Therefore, tracking time-varying modal parameters can be transformed into real-time tracking of the extended observable subspace matrix. (Subscript) For any given moment within the sliding window The stacked vector at this moment, Given the projection vector, based on the "time freeze" assumption, the extended observable subspace at this moment can be... Selected as the estimated value from the previous time step. ,and then It can be obtained in the following ways: (15) In the formula, for The extended observable subspace matrix at time step [time]. Based on the above definition, the recursive formulas for the extended observable subspace matrix and covariance matrix under the sliding window forgetting mechanism can be obtained as follows: (16) In the formula, the subscript and superscript represent the initial and final times of the data within the sliding window, respectively. and Based on the final sliding window length The extended observable subspace matrix and covariance matrix obtained from the data estimation within the range, wherein the extended observable subspace has the following form: (17) S5. Extract the state matrix based on the recursively updated extended observable subspace matrix, and obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
[0038] The state matrix can be obtained in the following way. : (18) In the formula, For the front block line, For the Empress The number of rows in each block is equal to the number of measurement points. The pseudo-inverse of the matrix is as follows: (19) Therefore, we obtain the extended observable subspace at each time step. Then, the state matrix at each time step can be obtained through equation (18). Update the modal parameters based on (3) and (4), and then update the modal parameters based on (4). The extended RSP window at time 1 is updated to the 1st time. RSP window at time as follows: (20) Before recursive identification, a fixed length is taken. After performing the deharmonicization operation in S3 on the response data, an initial stacking matrix is constructed. Furthermore, recursive initial values for the extended observable subspace matrix and covariance matrix can be obtained through singular value decomposition. and The details are as follows: (twenty one) In the formula, The length of the initial response data can be selected after the system stabilizes. Given an initial stacked matrix, after performing singular value decomposition on it, It is a left singular matrix. It is a right singular matrix. Let be a singular value matrix, where It is a diagonal matrix composed of non-zero singular values. and These are the left and right singular matrices corresponding to the non-zero singular values, respectively. and These are the left and right singular matrices corresponding to the zero singular values, respectively. The initial values for the recursion are obtained through the above operations. and After that, the identification process can be repeated to track modal parameters.
[0039] Example 2 This embodiment supplements the above-described recursive random subspace identification method based on Ramanujan subspace projection with specific applications and data. Specifically, it uses a numerical example of a three-degree-of-freedom time-varying system in conjunction with the accompanying drawings. The structure system is as follows: Figure 2 As shown: The dynamic equations of this structural system are shown in (22): (twenty two) In the formula, For displacement response, The external excitation vector and the mass matrix are shown. Damping matrix and stiffness matrix It has the following forms: (twenty three) In the formula, the stiffness coefficient ( ) are respectively , and Damping coefficient ( ) are respectively , and And quality ( It has the following forms: (twenty four) A random excitation was applied to the first degree of freedom, and harmonic excitations with frequencies of 3.5 Hz and 3 Hz were applied to the third degree of freedom, respectively. The sampling frequency was set to 100 Hz, and the displacement response in the three degrees of freedom from 0 to 800 seconds was calculated using the Runge-Kutta method. Figure 3a and Figure 3b As shown.
[0040] In this example, a recursive random subspace identification method based on the RSP strategy is used to identify the above-mentioned... Figure 2 The natural frequency of the time-varying system shown is identified, and the calculation process is as follows: First, a Fourier transform is performed on the displacement response signal of the first degree of freedom. The transform result is as follows: Figure 4a and Figure 4b As shown, the range of harmonic frequencies can be determined by the position of the peak, and the harmonic frequencies can be mapped from the frequency domain to the periodic domain to obtain parameter pairs. In this example, the parameter pairs corresponding to harmonic frequencies of 3.5Hz and 3Hz are respectively and .
[0041] Next, the RSP window length is determined. The RSP window length should ideally include an integer number of harmonic components corresponding to their periods; this is an adjustable parameter. In this example, the RSP window length is... This includes 28 3.5Hz harmonic cycles and 24 3Hz harmonic cycles. The response data is divided into multiple RSP windows, and harmonic removal and vector matrix stacking are performed within each RSP window. The construction of the stacking matrix is as follows. Considering this example has three degrees of freedom, corresponding to a state-space model order of 6, it is necessary to ensure that the rank of the stacking matrix is greater than or equal to 6, meaning the number of stacking events must be greater than or equal to 2. In this example, the number of stacking events is set to [value missing]. This can be adjusted based on the results. Simultaneously, it is based on the stacking vector matrix at each time step. Subspace recursive identification based on a sliding window forgetting mechanism is performed to estimate the extended observable subspace matrix at each time step. With covariance matrix .
[0042] Step 3: Based on the extended observable subspace matrix Obtain the state matrix The natural frequency and damping ratio at each moment are obtained through operations such as eigenvalue decomposition, and the identification results of the natural frequency are plotted.
[0043] Based on the above steps Figure 5a and Figure 5b The results of natural frequency identification for the time-varying structural system under the aforementioned harmonic and random excitation disturbances are presented. The solid red line represents the theoretical value, and the dotted blue line represents the identified value. It can be seen that the method of this invention can effectively suppress the interference of harmonic excitation on the modal parameter identification process of the time-varying structural system. Even when the harmonic frequency is close to or even intersects with the system's natural frequency, the method of this invention can still guarantee the accuracy of the identification.
[0044] To address the need for online identification of modal parameters in time-varying structural systems under complex vibration environments, this invention proposes an output-only recursive identification method for time-varying systems under harmonic excitation. This method combines the Recursive Stochastic Subspace (RSP) strategy with sliding window recursive random subspace identification. By utilizing only the structural output response, it identifies and removes harmonic components within the extended RSP window, and obtains the time-varying natural frequencies and damping ratios through recursive updates of the extended observable subspace matrix. Numerical examples based on three-degree-of-freedom gradually varying mass systems under mixed harmonic and random excitations of 3.5Hz and 3Hz show that this method can reduce the interference of harmonic excitation on modal parameter identification, achieve continuous tracking of time-varying modal parameters, and provide an effective approach for online identification of structural dynamic characteristics under complex operating conditions.
[0045] Example 3 Based on the same inventive concept, this application also provides a recursive random subspace identification device based on Ramanujan subspace projection, which can be used to implement the method described in the above embodiments, specifically including the following: The data acquisition unit is used to acquire the output response data of each measuring point of the structural system in the working state. In this embodiment, displacement data is measured by displacement sensors. The sensors need to be deployed on each degree of freedom to ensure the accuracy of mode shape estimation. Ramanujan subspace units are used to determine the harmonic excitation frequencies mixed in the output response data and map the harmonic excitation frequencies from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace. The harmonic removal unit is used to construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-removed response data. The update unit is used to construct a stacked vector matrix using the deharmonic response data based on the sliding window forgetting mechanism, and to recursively update the extended observable subspace matrix in real time through a subspace tracking algorithm. The output unit is used to extract the state matrix based on the recursively updated extended observable subspace matrix, and to obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
[0046] Preferably, embodiments of this application also provide a specific implementation of an electronic device capable of implementing all steps in the recursive random subspace identification method based on Ramanujan subspace projection in the above embodiments. The electronic device specifically includes the following: Processor, memory, communications interface, and bus; The processor, memory, and communication interface communicate with each other via a bus; the communication interface is used to realize information transmission between server-side devices, metering devices, and user-side devices.
[0047] The processor is used to call the computer program in the memory. When the processor executes the computer program, it implements all the steps in the recursive random subspace identification method based on Ramanujan subspace projection in the above embodiments.
[0048] Preferably, embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the recursive random subspace identification method based on Ramanujan subspace projection in the above embodiments. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements all steps of the recursive random subspace identification method based on Ramanujan subspace projection in the above embodiments. The computer-readable storage medium may be a disk, optical disk, read-only memory, random access memory, flash memory, portable hard disk, solid-state drive, or other tangible storage medium capable of storing program code and being read and executed by a processor.
[0049] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, hardware + program embodiments are relatively simple in description because they are fundamentally similar to method embodiments; relevant parts can be referred to the descriptions in the method embodiments.
[0050] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0052] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0053] This invention is not limited to the embodiments described above. The above description of specific embodiments is intended to illustrate and explain the technical solutions of this invention. The specific embodiments described above are merely illustrative and not restrictive. Without departing from the spirit and scope of the claims, those skilled in the art can make many specific modifications based on the teachings of this invention, and these modifications all fall within the scope of protection of this invention.
Claims
1. A recursive random subspace identification method based on Ramanujan subspace projection, characterized in that, include: S1. Obtain the output response data of each measuring point of the structural system in the working state; S2. Determine the harmonic excitation frequency mixed in the output response data, and map the harmonic excitation frequency from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace; S3. Construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-deharmonic response data. S4. Based on the sliding window forgetting mechanism, a stacked vector matrix is constructed using the deharmonic response data, and the extended observable subspace matrix is updated in real time through a subspace tracking algorithm. S5. Extract the state matrix based on the recursively updated extended observable subspace matrix, and obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
2. The recursive random subspace identification method according to claim 1, characterized in that, In step S2, when mapping the harmonic excitation frequency from the frequency domain to the periodic domain, the following coprime integer parameter pairs are used. Perform mapping: in, In response to the sampling frequency of the data, Let be the harmonic excitation frequency; select parameter pairs that satisfy the above mapping relationship and are coprime to calculate the normalized circular frequency in the Ramanujan subspace.
3. The recursive random subspace identification method according to claim 2, characterized in that, In step S3, for each normalized circular frequency, the corresponding Ramanujan subspace is spanned using conjugate complex pairs as basis functions, and the orthogonal projection coefficient matrix is calculated. The output response data at each measuring point is multiplied by the orthogonal projection coefficient matrix to obtain the harmonic response at the corresponding frequency. Subtracting the harmonic response from the original output response data yields the deharmonic response data.
4. The recursive random subspace identification method according to claim 1, characterized in that, In step S4, the method for constructing the stacked vector matrix includes: The output response data is divided into multiple fixed-length RSP windows. After introducing the response data at a new time step, an extended RSP window is created. Stacked vectors are constructed using the deharmonic response data within the extended RSP window, and multiple stacked vectors within the same extended RSP window are combined to form a stacked matrix.
5. The recursive random subspace identification method according to claim 4, characterized in that, In step S4, the subspace tracking algorithm is the Projection Approximation Subspace Tracking Method (PAST). Based on the stacked matrix, constraint equations are established using a sliding window forgetting mechanism, and the extended observable subspace matrix and covariance matrix at each time step are obtained recursively.
6. The recursive random subspace identification method according to claim 1, characterized in that, In step S5, extracting the state matrix based on the recursively updated extended observable subspace matrix specifically includes: The extended observable subspace matrix is divided into a front block row matrix and a rear block row matrix; The state matrix at the current moment is obtained by multiplying the pseudo-inverse of the front block row matrix with the rear block row matrix.
7. The recursive random subspace identification method according to any one of claims 1-6, characterized in that, Before step S1, the method further includes: performing batch subspace identification based on historical response data of a fixed length, and obtaining recursive initial values for the extended observable subspace matrix and the covariance matrix.
8. A recursive random subspace identification device based on Ramanujan subspace projection, characterized in that, include: The data acquisition unit is used to acquire the output response data of each measuring point of the structural system in the working state; Ramanujan subspace units are used to determine the harmonic excitation frequencies mixed in the output response data and map the harmonic excitation frequencies from the frequency domain to the periodic domain to construct the corresponding Ramanujan subspace. The harmonic removal unit is used to construct an orthogonal projection coefficient matrix based on the Ramanujan subspace, project the output response data of each measurement point onto the Ramanujan subspace, extract and remove harmonic response components, and obtain harmonic-removed response data. The update unit is used to construct a stacked vector matrix using the deharmonic response data based on the sliding window forgetting mechanism, and to recursively update the extended observable subspace matrix in real time through a subspace tracking algorithm. The output unit is used to extract the state matrix based on the recursively updated extended observable subspace matrix, and to obtain the natural frequency and damping ratio of the structure system as a function of time through eigenvalue decomposition.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the recursive random subspace identification method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the recursive random subspace identification method according to any one of claims 1 to 7.
