Structural Modal Identification System Based on Edge Computing Nodes

Abstract

This invention discloses a structural modal identification system based on edge computing nodes. A data acquisition module collects charge signals generated during the operation of a wind turbine structure in real time and transmits them as voltage signals to a data receiving and processing unit. The data receiving and processing unit receives the voltage signals and converts them into vibration response signals, which are then transmitted to a data storage module. The data storage module stores the vibration response signals. A modal identification unit downloads the vibration response signals from the data storage module, removes noise from the vibration response signals, separates the free vibration response signals from the noise-removed vibration response signals, obtains the modal parameters of the wind turbine structure based on the free vibration response signals, and transmits them to a monitoring module. The monitoring module displays the modal parameters of the wind turbine structure in real time. This system collects and processes data at the data source, enabling low-cost, fast, efficient, safe, and accurate identification of modal parameters to monitor the health status of the wind turbine structure.

Description

Technical Field

[0001] This invention relates to the field of structural modal identification technology, and in particular to a structural modal identification system based on edge computing nodes. Background Technology

[0002] Because of their exposure to harsh natural environments, offshore wind turbines are particularly vulnerable to a variety of disasters, such as blade damage, fire, and structural instability. These are all causes of turbine failure. Therefore, monitoring the structural health of offshore wind turbines is of great significance for disaster prevention and timely maintenance.

[0003] Structural health monitoring (SHM) utilizes field sensing systems and related analysis techniques to monitor the health status of structures and evaluate their safety, durability, and reliability. It is primarily used for structures with high reliability requirements, such as nuclear power plants, dams, and offshore drilling platforms. However, due to the large number and high cost of monitoring equipment and sensors, the cost of structural monitoring remains high. Furthermore, the large amount of response data generated by the structure requires lengthy transmission and processing at a computing center, resulting in low efficiency and insufficient security. Therefore, it is necessary to develop a low-cost, fast, efficient, safe, and accurate offshore wind turbine health monitoring system that can monitor the modal parameters of wind turbines in real time and efficiently with low energy consumption, thereby determining the structural health status. Summary of the Invention

[0004] This invention provides a structural modal identification system based on edge computing nodes to overcome the problems of low efficiency and insufficient security in monitoring the health status of wind turbine structures, where a large amount of response data generated by the wind turbine structure must be transmitted and processed by the computing center for a long time.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A structural modal identification system based on edge computing nodes is characterized in that it includes an edge computing module, a data acquisition module, a data storage module, and a monitoring module, wherein the edge computing module is electrically connected to the data acquisition module, the data storage module, and the monitoring module, respectively.

[0007] The edge computing module includes a data receiving and processing unit and a modality recognition unit. The modality recognition unit includes a noise removal mechanism, a signal separation mechanism, and a recognition mechanism.

[0008] The data acquisition module is used to acquire the charge signal generated during the operation of the wind turbine structure in real time, and transmit the acquired charge signal to the data receiving and processing unit in the form of a voltage signal; the data receiving and processing unit is used to receive the voltage signal and convert the voltage signal into a vibration response signal and transmit it to the data storage module; the data storage module is used to store the vibration response signal; the modal recognition unit is used to download the vibration response signal from the data storage module, remove noise from the vibration response signal through a noise removal mechanism, separate the free vibration response signal from the noise-removed vibration response signal through a signal separation mechanism, and finally obtain the modal parameters of the wind turbine structure based on the free vibration response signal through the recognition mechanism, and transmit the obtained modal parameters to the monitoring module; the monitoring module is used to display the modal parameters of the wind turbine structure in real time, and judge the current health status of the wind turbine structure based on the displayed modal parameters.

[0009] Furthermore, the data acquisition module includes a sensor unit, a signal conversion unit, and an electrical signal acquisition unit. The sensor unit is electrically connected to the signal conversion unit, and the signal conversion unit is electrically connected to the electrical signal acquisition unit.

[0010] The sensor unit is used to acquire charge signals of the wind turbine structure, including acceleration signals, velocity signals, displacement signals and strain signals of the wind turbine structure, and transmits the acquired charge signals to the signal conversion unit; the signal conversion unit is used to amplify the charge signals and convert them into voltage signals, and then transmit the voltage signals to the electrical signal acquisition unit; the electrical signal acquisition unit receives voltage signal data packets based on the UDP protocol and transmits the voltage signal data packets to the data receiving and processing unit.

[0011] Furthermore, the data receiving and processing unit converts the voltage signal into a vibration response signal according to the sensitivity parameters of the sensor unit, and stores the vibration response signal in the form of a queue and uploads it to the data storage module.

[0012] Furthermore, the noise removal mechanism uses Kalman filtering to remove noise from the acquired vibration response signal; the signal separation mechanism uses random subtraction to separate the free vibration response signal from the noise-removed vibration response signal; and the identification mechanism uses random subspace method to process the free vibration response signal to obtain the frequency, damping ratio, and mode shape of the wind turbine structure.

[0013] Furthermore, the signal separation mechanism uses a random subtraction method to separate the free vibration response signal from the noise-removed vibration response signal, and sets the vibration response model of the linear system as follows:

[0014]

[0015] x(0), These are the system's initial displacement and initial velocity, respectively. This is the free vibration response of the system with an initial displacement of 1 and an initial velocity of 0. Let be the free vibration response of the system with an initial displacement of 0 and an initial velocity of 1; h(t) is the system's unit impulse response function; u(τ) is the external excitation;

[0016] Take constant Draw a straight line The vibration response x(t) intersects with the vibration response x(t) to form N intersection points. Starting from each intersection point, samples of the vibration response signal for the same time length are extracted, where x(t) = ... i ) is the subsample function, t i The response at time t includes the initial displacement x(t) i Free vibration response caused by ) initial velocity Caused free vibration and t i The forced vibration response at time t is given by the following formula;

[0017]

[0018] The starting point of the subsample function is changed from t i Moving the value to 0 yields the vibration response at each intersection point:

[0019]

[0020] x i The average value of (t) is:

[0021]

[0022] The free vibration response signal is obtained as follows:

[0023]

[0024] Furthermore, the identification mechanism expresses the free vibration response signal using discrete state-space equations to obtain a discrete vibration response signal; a Hankel matrix is ​​constructed based on the discrete vibration response signal; the Hankel matrix is ​​decomposed using QR decomposition and optimized to obtain a system matrix A and an output matrix C; and eigenvalues ​​μ are obtained by decomposing the system matrix A using eigenvalue decomposition. i Based on eigenvalue μ i The modal parameters of the wind turbine structure were obtained.

[0025] Beneficial effects: This invention sets up a data acquisition module on the wind turbine to be monitored, which transmits the acquired structural voltage signal to the edge computing module in real time. The edge computing module receives the voltage signal through the data receiving and processing unit and converts it into a vibration response signal, which is then uploaded to the data storage module. The edge computing module downloads the vibration response signal from the data storage module through the modal identification unit, processes it, and obtains the modal parameters of the wind turbine structure, which are then transmitted to the monitoring module in real time. By collecting and processing data at the data source, this system can identify modal parameters in a low-cost, fast, efficient, safe, and accurate manner, thereby monitoring the health status of the wind turbine structure. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a structural diagram of the edge computing node-based structural mode identification system in this invention;

[0028] Figure 2 The acceleration response data obtained by the edge computing node-based structural modal identification system in this invention;

[0029] Figure 3a This is a first-order vibration mode diagram of the wind turbine model in an embodiment of the present invention;

[0030] Figure 3b This is a second-order vibration mode diagram of the wind turbine model in an embodiment of the present invention;

[0031] Figure 4 This is a time-domain comparison diagram of the free vibration response data obtained by striking the wind turbine tower with a small hammer in an embodiment of the present invention and the free vibration response data obtained during the modal identification process;

[0032] Figure 5 This is a frequency domain comparison diagram of the free vibration response data obtained by striking the wind turbine tower with a small hammer in an embodiment of the present invention and the free vibration response data obtained during the modal identification process;

[0033] Figure 6 This is a stability graph used to verify the frequency authenticity obtained by the modal identification unit in this embodiment of the invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Edge computing refers to the technology of performing computation at the network edge, where the edge is defined as any computing and network resource node between the data source and the cloud data center. Edge computing nodes are data processing-oriented; they are distributed open platforms that integrate core capabilities of networking, computing, storage, and applications at the network edge, close to the source of the object or data source. They provide edge intelligence services locally, meeting the key needs of industry digitalization in areas such as agile connectivity, real-time business, data optimization, application intelligence, and security and privacy protection. In structural health monitoring, the large amount of response data generated by the structure must undergo long-term transmission and processing through the computing center, resulting in low efficiency and insufficient security. Furthermore, considering the complex marine environment of offshore wind turbines, long-distance data transmission should be avoided as much as possible. Therefore, combining edge computing theory, one of the core advantages of edge computing is that, compared to traditional cloud computing models where data needs to be transmitted from the device to the cloud and is susceptible to network latency, edge computing places some data processing tasks closer to the user at the edge node, thereby reducing data latency during transmission. This is of great value for scenarios like offshore wind turbines that require real-time responses.

[0036] Devices used for edge computing should be made of corrosion-resistant materials and have waterproof and sealing properties; they should have low power consumption to reduce energy consumption and extend the service life of the devices; and edge computing devices need to be able to process large amounts of data and run complex algorithms to cope with the complex working environment of offshore wind turbines; they should have reliable communication capabilities to interact and transmit data with the central control system or cloud server to ensure the safe and efficient operation of the wind turbines.

[0037] Based on the above research, this embodiment provides a structural modality identification system based on edge computing nodes, such as... Figure 1 As shown, it includes an edge computing module, a data acquisition module, a data storage module, and a monitoring module. The edge computing module is electrically connected to the data acquisition module, the data storage module, and the monitoring module, respectively. The edge computing module includes a data receiving and processing unit and a modal recognition unit. The modal recognition unit includes a noise removal mechanism, a signal separation mechanism, and a recognition mechanism.

[0038] The data acquisition module is used to acquire the charge signal generated during the operation of the wind turbine structure in real time, and transmit the acquired charge signal to the data receiving and processing unit in the form of a voltage signal; the data receiving and processing unit is used to receive the voltage signal and convert the voltage signal into a vibration response signal and transmit it to the data storage module; the data storage module is used to store the vibration response signal; the modal recognition unit is used to download the vibration response signal from the data storage module, remove noise from the vibration response signal through a noise removal mechanism, separate the free vibration response signal from the noise-removed vibration response signal through a signal separation mechanism, and finally obtain the modal parameters of the wind turbine structure based on the free vibration response signal through the recognition mechanism, and transmit the obtained modal parameters to the monitoring module; the monitoring module is used to display the modal parameters of the wind turbine structure in real time, and judge the current health status of the wind turbine structure based on the displayed modal parameters.

[0039] The edge computing module is a Raspberry Pi 4B. As a low-cost microcomputer, the Raspberry Pi can reduce the hardware costs for real-time monitoring and processing tasks, meeting the requirements of edge computing—which pushes computing closer to users and data sources—and the special environments of offshore wind turbines. It features small size, low price, high stability, corrosion resistance, and waterproof sealing. Furthermore, the Raspberry Pi uses a TF card as a storage node, and while data is uploaded to the MySQL database, a backup is also reserved locally on the Raspberry Pi to prevent signal disruption during disasters.

[0040] In this embodiment, the data processing tasks are distributed across various devices and analyzed and processed on the device side through the edge computing module. This can significantly reduce the amount of data that needs to be transmitted to the cloud, effectively alleviate bandwidth pressure, and keep the data processing tasks on the device side, which also helps to reduce the risk of data leakage and improve data security and privacy protection. At the same time, it has high scalability and offline processing capabilities, providing an efficient, low-cost and flexible solution for real-time monitoring of offshore wind turbine structural status and data processing.

[0041] The data acquisition module includes a sensor unit, a signal conversion unit, and an electrical signal acquisition unit. The sensor unit is electrically connected to the signal conversion unit, and the signal conversion unit is electrically connected to the electrical signal acquisition unit.

[0042] The sensor unit is used to acquire charge signals of the wind turbine structure, including acceleration signals, velocity signals, displacement signals and strain signals of the wind turbine structure, and transmits the acquired charge signals to the signal conversion unit; the signal conversion unit is used to amplify the charge signals and convert them into voltage signals, and then transmit the voltage signals to the electrical signal acquisition unit; the electrical signal acquisition unit receives voltage signal data packets based on the UDP protocol and transmits the voltage signal data packets to the data receiving and processing unit.

[0043] In this embodiment, the electrical signal acquisition unit is an electrical signal acquisition card, which is an eight-channel board based on the STM32 chip. It can acquire various electrical signals such as IEP, bridge, current, and voltage, and execute the UDP communication protocol. Since this system is an edge computing node, the selection of the communication protocol is crucial in edge computing. The UDP communication protocol used eliminates the process of establishing and tearing down connections and removes the retransmission verification mechanism. It offers high communication speed and meets the requirements for data transmission security, real-time performance, and stability in harsh monitoring environments.

[0044] In practice, acceleration, velocity, displacement, etc. can all represent the response data of a structure. In this embodiment, considering that most structures are subject to environmental excitation, the sensor module is preferably a piezoelectric accelerometer.

[0045] In this embodiment, the signal conversion unit is an IEPE bridge box. Since the signal directly obtained from the sensor is extremely weak and contains a lot of noise, the IEPE bridge box can amplify and convert the signal collected by the sensor.

[0046] The data receiving and processing unit is built using Python based on object-oriented programming. The data receiving and processing unit converts the voltage signal into a vibration response signal according to the sensitivity parameters of the sensor unit, and stores the vibration response signal in the form of a queue and uploads it to the data storage module, which is a MySQL database.

[0047] The data receiving and processing unit adopts the object-oriented approach. Object-oriented programming is a computer programming architecture whose basic principle is that a computer program is composed of individual objects and units that can function as subroutines. It has the characteristics of reusability, flexibility, and extensibility, making the system more flexible and easier to maintain, and meeting the requirements of fast, accurate, and secure edge computing.

[0048] The data receiving and processing unit connects to the electrical signal acquisition card via the UDP communication protocol to acquire and process data in real time and visualize it on a graphical interface. This unit achieves real-time data acquisition, analysis, and visualization by combining multiple classes and functions. It runs in a Tkinter main window and uses the matplotlib library to draw and update data graphs, enabling users to view device data and perform corresponding operations in real time.

[0049] The noise removal mechanism uses Kalman filtering to remove noise from the acquired vibration response signal; the signal separation mechanism uses random subtraction to separate the free vibration response signal from the noise-removed vibration response signal; the identification mechanism uses random subspace method to process the free vibration response signal to obtain the frequency, damping ratio and mode shape of the wind turbine structure.

[0050] Specifically, the noise removal mechanism removes noise signals from the acquired vibration response signal using a Kalman filter:

[0051] The vibration control equation is defined as follows:

[0052]

[0053] M is the mass matrix; C1 is the damping coefficient; K is the spring stiffness; and u(t) is the externally applied excitation. q(t) represents the acceleration, velocity, and displacement vectors, respectively. The equation can be expressed in state-space as follows:

[0054]

[0055] x(t) and y(t) represent the state response function and the output vector, respectively. Let x(t) be the first derivative. Assuming the time interval is Δt, the above equation can be written in discrete form:

[0056]

[0057] x k Let k be the state response under the optimal estimate at time k. Let y be the derivative of the state vector at time k+1. k For the measured response, y k+1 Let u be the output vector at time k+1. k When an external excitation is applied at time k, the measured response A represents the state matrix; B is the input matrix; C is the output matrix; and D is the feedthrough matrix. Considering noise and errors in establishing the dynamic model, equation (3) can be simplified to:

[0058] x k =Ax k-1 +Bu k +w k-1 (4)

[0059] z k =Hx k +v k (5)

[0060] In the formula z k The measurement result at time k; x k-1The state response at time k-1 is the optimal estimate, w k-1 The process noise at time k-1; v k Let H be the measurement noise at time k; H be the transformation matrix from state variables to measurements. Assuming both types of noise follow a normal distribution, then:

[0061]

[0062] Q1 is the covariance matrix of the process noise w; R1 is the covariance matrix of the measurement noise v; therefore, the prior estimate of the Kalman filter at time k is... and posterior estimation They are respectively:

[0063]

[0064]

[0065] u k-1 Apply an external stimulus at time k-1; For residuals; K k The Kalman gain is estimated by the posterior estimation of the error e. k To determine:

[0066]

[0067] Similarly, the prior estimation error is

[0068]

[0069] Define the prior error covariance matrix and the posterior estimation error covariance matrix P k They are respectively:

[0070]

[0071]

[0072] E represents expectation; for The inverse of a matrix is ​​the transpose of the matrix. for The transpose of a matrix; for K k Determine the derivative so that it equals zero:

[0073]

[0074]

[0075]

[0076] Pk-1 Let A be the Kalman filter gain at time k-1. T H is the transpose of matrix A. T Let H be the transpose of H, and I be the identity matrix;

[0077] The Kalman filter consists of a prediction function and an update function. The prediction function calculates the state and error covariance of the next time step based on the previous time step. In the update function, the Kalman gain is calculated using equation (14). Based on the measured data, the posterior estimate is calculated using equation (8), thus obtaining the vibration response data after processing by the Kalman filter method.

[0078] The signal separation mechanism separates the free vibration signal from the noise-removed signal using a random subtraction method. Specifically, the vibration response data processed by the Kalman filter is used to extract the free vibration signal using the random subtraction method. The vibration response model of the linear system is set as follows:

[0079]

[0080] x(0), These are the system's initial displacement and initial velocity, respectively. This is the free vibration response of the system with an initial displacement of 1 and an initial velocity of 0. Let be the free vibration response of the system with an initial displacement of 0 and an initial velocity of 1; h(t) is the system's unit impulse response function; u(τ) is the external excitation.

[0081]

[0082]

[0083] ξ is the damping ratio, ω is the system frequency, ω d The natural frequency is t, and time is t. This represents the phase difference.

[0084] Take constant Draw a straight line The vibration response x(t) intersects with the vibration response x(t) to form N intersection points. Starting from each intersection point, samples of the vibration response signal for the same time length are extracted, where x(t) = ... i ) is the subsample function, t i The response at time t includes the initial displacement x(t) i Free vibration response caused by ) initial velocity Caused free vibration and t i The forced vibration response at time t is given by the following formula;

[0085]

[0086] Since the excitation is stationary and the starting point of time does not affect the stochastic properties of the structure, the starting point of the subsample function is set from t. i Moving the value to 0 yields the vibration response at each intersection point:

[0087]

[0088] x i The average value of (t) is:

[0089]

[0090] Since the excitation is a stationary, purely random vibration with a mean of 0, the system vibration response is also a stationary random vibration with a mean of 0. Therefore, the free vibration response signal is:

[0091]

[0092] The identification mechanism processes the free vibration signal using the random subspace method to obtain the structure's frequency, damping ratio, and mode shape. Specifically, this embodiment employs a data-driven random subspace method for data processing, resulting in high computational accuracy. Furthermore, the Hankel matrix of the data-driven random subspace method is composed of input data, significantly reducing the computational load and making it more suitable for the low-power requirements of edge computing nodes. Based on the random subspace method, the discrete state space equation is defined as follows:

[0093]

[0094] w k v k These are measurement error and process error, respectively, both assumed to be white noise with a mean of 0; furthermore, u k Environmental excitations are also assumed to be random, and their intensity is difficult to measure, thus they are also considered noise. Therefore, the input term and the noise term are combined into the following discrete state-space equation:

[0095]

[0096] The dynamic characteristics of the system are represented by the eigenvalues ​​and eigenvectors of matrix A. A 2mi×j dimensional Hankel matrix H is constructed using the response data processed by the random subtraction method. i As output data of the system:

[0097]

[0098] x represents the response of all measurement points at time i; j is the number of sampling points; X p X is the first i-1 rows of the Hankel matrix; f It is the i to 2i-1 rows of the Hankel matrix, therefore X can be considered asf It was delayed by (2i-1)-(i-1) steps, i.e., i steps; X p For the "past" part, X f This is the "future" section. The next step is to perform QR decomposition on the Hankel matrix:

[0099] H i =RQ T (27)

[0100] Q T It is the transpose of an orthogonal matrix, R is an upper triangle matrix, and according to projection theory, the orthogonal projection of the row space O i It projects the output line space of the "future" onto the output line space of the "past":

[0101] O i =X f / X p =X f X p T (X p X p T ) + X p (28)

[0102] For O i Perform SVD decomposition:

[0103] O i =USV T (29)

[0104] U represents O i The left singular matrix; S represents O i Singular value matrix; V T O i The transpose of the right singular matrix;

[0105] O i Decomposed into the observability matrix Γ i and Kalman filter state sequence

[0106] Γ i =US 1 / 2 (30)

[0107]

[0108] (·) + It is the Moore-Penrose pseudo-inverse matrix; O i-1 The next time step i+1 can be defined as

[0109]

[0110] Therefore, we get:

[0111]

[0112] The state-space equations can be described as follows:

[0113]

[0114] The system matrix A and output matrix C are obtained; W i V i For process and measurement residuals; X i|j It is a Hankel matrix with only one block of rows.

[0115]

[0116] Using eigenvalue decomposition to decompose A, we can obtain...

[0117] A=ψΛψ -1 (36)

[0118] Where Λ=diag(λ) i ), i = 1, 2, ..., n are diagonal matrices; λ i It is the i-th th There are 1 eigenvalues; ψ is the system eigenvector matrix. The frequency ω can be obtained. i and mode shape Φ:

[0119] The system's modal parameters can be determined by the intermediate variable λ. i get:

[0120]

[0121] In the formula, μ i Let A be the eigenvalues ​​of matrix A.

[0122] get:

[0123] frequency:

[0124] ω i =|λ i | (38)

[0125] Damping ratio:

[0126]

[0127] Vibration shape:

[0128] Φ = CΨ. (40)

[0129] This embodiment uses a scaled-down model of an existing offshore wind turbine as an example to illustrate how edge computing can be used for structural modal identification, and to verify the accuracy and practicality of the modal identification unit. The scaled-down wind turbine model is 2m high, with the tower section made of low-carbon steel and the blades manufactured using 3D printing technology and secured with bolts. The bottom of the tower is a 6.69cm diameter steel pipe, used to simulate the large pipe pile foundation commonly used in offshore wind turbines. A geotechnical test chamber, measuring 1m × 1m × 1m, is used to simulate the foundation environment of the offshore wind turbine. The bottom of the chamber is lined with waterproof membrane to prevent seepage and errors. During the experiment, sand is layered and compacted in the geotechnical test chamber to simulate different types of seabed sites. The bottom of the wind turbine is filled with soil to a depth of 55cm and operated normally at a speed of 75r / min. Set up the experimental platform, configure the IP address, port, and sensor gain of the data acquisition card, set the acquisition time to 10 seconds, and the sampling frequency to 50 Hz. Configure the four channel types (current, voltage, bridge, IEPE) and sensor sensitivity parameters according to the sensor type. After setting the parameters, configure the MySQL database name, password, port, and other relevant parameters.

[0130] The acceleration response data was measured by starting a monitoring program on a Raspberry Pi 4B. Figure 2 As shown;

[0131] The relevant parameters of the monitoring algorithm are set. Since the monitoring algorithm is a modal parameter identification algorithm that combines Kalman filtering with random decrement technology and random subspace method, the parameters of the three algorithm parts need to be set separately. First, for the Kalman filter, the Kalman filter is set to the golden ratio. The tangent point value of the random decrement method is set to 1.5, the output data length is 512, and the sampling rate is 50Hz. The order of the random subspace method is set to 20. The latest 4000 rows of data are used in each calculation process. After the modal identification is completed, the process pauses for a period of time and repeats the previous identification process. The modal identification unit is started to measure the first mode shape of the wind turbine model, as shown in the figure. Figure 3a As shown, the second-order vibration mode is as follows Figure 3b As shown.

[0132] A time-domain comparison of the free vibration response data obtained by striking the wind turbine tower with the free vibration response data obtained during modal identification is shown below. Figure 4 The FFT modal identification algorithm was used to perform frequency domain transformation on two sets of free vibration response data to obtain the first-order frequency of the wind turbine model. The impact data was processed by FFT to obtain the frequency domain information of the wind turbine model's response. The response data of the wind turbine model under environmental excitation was processed by Kalman filtering and further processed using the RDT method. The final free vibration response data was also processed by FFT to obtain frequency domain information, such as... Figure 5As shown, comparing the two sets of frequency domain information reveals that the first-order frequency estimated by both methods is very close, both being 7.03 Hz. The frequency domain curve presented by the filtered response data is smoother and does not exhibit excessive spurious frequencies. Therefore, the KF-RDT-SSI method can accurately extract partial frequency domain information of the structure.

[0133] The frequency accuracy obtained by the modal identification unit is verified using the stability graph method, such as... Figure 6 As shown.

[0134] Specifically, the algorithm program for modality recognition is written using the Python language, and the specific steps are as follows:

[0135] (1) Define a function get_sql_data to connect to the MySQL database, import the latest 4000 rows of data from the MySQL database and close the database;

[0136] (2) Define a class for the Kalman filter, and use the parameters Q, R, K, PK, etc. of the Kalman filter as attributes of the class. Define the calculation process of the Kalman filter, ADC_Value, as a function and use it as a method of the Kalman filter class to return the result of the Kalman filter.

[0137] (3) Define the function RDT for the random decrement method, take the Kalman filter result data, sampling frequency, and output data length as parameters of the function RDT, and return the result of the random decrement method;

[0138] (4) Define the function clean_before_ssi to process the results of the random decrement method as parameters for the random subspace method;

[0139] (5) Define the ran_subsid class, whose main attributes are t data and whose method is the subsid function of the random subspace method, to obtain the final modal parameters of the structure.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An edge computing node based structural modal identification system, comprising: The edge computing module, the data acquisition module, the data storage module and the monitoring module are electrically connected; The edge computing module comprises a data receiving processing unit and a modal identification unit, and the modal identification unit comprises a noise elimination mechanism, a signal separation mechanism and an identification mechanism. The data acquisition module is used for collecting the charge signals generated by the operation of the fan structure in real time, and transmitting the collected charge signals in the form of voltage signals to the data receiving processing unit; the data receiving processing unit is used for receiving the voltage signals and converting the voltage signals into vibration response signals and transmitting the vibration response signals to the data storage module; the data storage module is used for storing the vibration response signals; the modal identification unit is used for downloading the vibration response signals from the data storage module, eliminating the noise in the vibration response signals through the noise elimination mechanism, separating the free vibration response signals from the vibration response signals after the noise is eliminated through the signal separation mechanism, and finally obtaining the modal parameters of the fan structure according to the free vibration response signals through the identification mechanism, and transmitting the obtained modal parameters to the monitoring module; the monitoring module is used for displaying the modal parameters of the fan structure in real time, and judging the health state of the current fan structure through the displayed modal parameters of the fan structure.

2. The edge-computing node-based structural modal identification system of claim 1, wherein, The data acquisition module comprises a sensor unit, a signal conversion unit and an electric signal acquisition unit, the sensor unit is electrically connected with the signal conversion unit, and the signal conversion unit is electrically connected with the electric signal acquisition unit; The sensor unit is used for obtaining the charge signals of the fan structure, including acceleration signals, speed signals, displacement signals and strain signals of the fan structure, and transmitting the obtained charge signals to the signal conversion unit; the signal conversion unit is used for amplifying and converting the charge signals into voltage signals, and then transmitting the voltage signals to the electric signal acquisition unit; the electric signal acquisition unit receives the voltage signal data packet based on the UDP protocol and transmits the voltage signal data packet to the data receiving processing unit.

3. The edge computing node based structural modal identification system of claim 2, wherein, The data receiving processing unit converts the voltage signals into vibration response signals according to the sensitivity parameters of the sensor unit, and stores and uploads the vibration response signals in the form of a queue to the data storage module.

4. The edge computing node based structural modal identification system of claim 1, wherein, The noise elimination mechanism adopts Kalman filtering method to eliminate the noise from the obtained vibration response signals; the signal separation mechanism adopts random decrement method to separate the free vibration response signals from the vibration response signals after the noise is eliminated; and the identification mechanism adopts random subspace method to process the free vibration response signals to obtain the frequency, damping ratio and modal shape of the fan structure.

5. The edge computing node based structural modal identification system of claim 4, wherein, The signal separation mechanism adopts random decrement method to separate the free vibration response signals from the vibration response signals after the noise is eliminated, and sets the vibration response model of the linear system as: x(0) = 0, x(0) = 0, x(0) = 0, x(0) = 0, Take constant Make straight line The N intersection points are formed by the intersection of the vibration response x(t) and the straight line, and the vibration response signal of the same time length is intercepted as a sample from each intersection point, where x(t-t i ) is a sub-sample function, and the response at time t i includes the free vibration response caused by the initial displacement x(t i ) Initial velocity Caused by free vibration And the forced vibration response at time t i , the formula is as follows; The sub-sample function starting point is moved from t i to 0 to get the vibration response at each intersection point: x i The average of (t) is: The free vibration response signal is obtained as:

6. The edge computing node based structural modal identification system of claim 4, wherein, The identification mechanism adopts the free vibration response signal to express in a discrete state space equation to obtain a discrete vibration response signal; constructs a Hankel matrix based on the discrete vibration response signal; performs QR decomposition on the Hankel matrix and performs optimization processing to obtain a system matrix A and an output matrix C; Based on the system matrix A, eigenvalue decomposition method is used to obtain the eigenvalue μ i , and the modal parameters of the fan structure are obtained based on the eigenvalue μ i .

Images