Intelligent signal filtering method and system based on cloud computing

Abstract

The present invention belongs to the field of signal filtering, and in particular to an intelligent signal filtering method and system based on cloud computing. The method first performs initial filtering and compressive sensing on monitoring signals through an algorithm built into a sensor and uploads them to a cloud platform. Secondly, a secondary adaptive filtering model is constructed through Kalman filtering and a particle swarm algorithm to further process the initial filtered signal to obtain a high-quality signal. After the signal quality is evaluated by a cloud platform evaluation model, if it meets a preset threshold, it is visualized; otherwise, the filtering parameters are adjusted until the requirements are met. Finally, the signal that meets the conditions is input into a cloud platform abnormal monitoring model, the abnormal results are analyzed and judged, and a corresponding emergency strategy is generated to achieve intelligent monitoring and early warning.

Description

Technical Field

[0001] The present invention belongs to the field of signal filtering, and in particular relates to an intelligent signal filtering method and system based on cloud computing. Background Art

[0002] With the acceleration of urbanization and the rapid development of industrial technology, the health monitoring and safety maintenance of infrastructure have become increasingly important. In particular, for key structures such as bridges, tunnels, and high-rise buildings, their safe and stable operation is directly related to public safety. As important monitoring methods, vibration monitoring and settlement monitoring are of immeasurable value for early detection of potential structural defects and prevention of major safety accidents. However, modern monitoring systems generate a large amount of monitoring data every moment through vibration monitors and settlement monitors installed in key parts of the structure. These data are not only large in volume, but also contain various noises and interferences, such as environmental noise, equipment noise, electromagnetic interference, etc. Secondly, traditional local processing methods are difficult to complete efficient filtering and analysis of large amounts of data in a short period of time, which affects the real-time alarm capability of the monitoring system. In addition, in complex and changeable monitoring environments, static filtering strategies often cannot adapt to changes in signal characteristics under different working conditions.

[0003] For example, the patent with authorization announcement number CN108886405B discloses a method, device and system for signal filtering, including: a first analog-to-digital converter, used to convert a first radio frequency signal output by an optical receiver into a first digital signal; a digital domain filter, used to filter out the background noise of an unused frequency band in the first digital signal converted by the first analog-to-digital converter; a DOCSIS data access unit, used to convert a first downstream optical signal received by a DCA device into a second digital signal, combine it with the first digital signal output by the digital domain filter and filtered out the background noise of the unused frequency band, and convert the obtained third digital signal into a second radio frequency signal for output.

[0004] For example, the patent with authorization publication number CN101536342A discloses a method and device for adaptive filtering of signals, wherein the level of a reference signal necessary for adaptive filtering is determined by evaluation with the aid of a numerical value equivalent to the reference signal within the scope of adaptation, wherein the filtering is performed based on the electrical power of at least one signal extracted from the transmission channel to be filtered, and the equivalent amount is determined based on the electrical power of at least the extracted signal by a mathematical function.

[0005] The above existing technologies have the following problems: 1) The existing filtering focuses on removing the background noise of specific unused frequency bands, and has limited adaptability to complex and changeable signal environments, resulting in residual low-frequency noise, and static filtering strategies often cannot adapt to changes in signal characteristics under different working conditions, resulting in signal analysis results that are far from the actual results. In order to solve the above problems, the present invention provides an intelligent signal filtering method and system based on cloud computing. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes an intelligent signal filtering method and system based on cloud computing. The method first performs initial filtering and compressive sensing on the monitoring signal through the algorithm built into the sensor and uploads it to the cloud platform. Secondly, the Kalman filter and particle swarm algorithm are used to construct a secondary adaptive filtering model, and the initial filtered signal is further processed to obtain a high-quality signal. After the signal quality is evaluated by the cloud platform evaluation model, if it meets the preset threshold, it is visualized; otherwise, the filtering parameters are adjusted until the requirements are met. Finally, the signal that meets the conditions is input into the cloud platform abnormal monitoring model, the abnormal results are analyzed and judged, and the corresponding emergency strategy is generated to achieve intelligent monitoring and early warning.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] Intelligent signal filtering method based on cloud computing, including:

[0009] S1: Acquire vibration or settlement monitoring signal data, use the built-in signal segmentation decomposition filtering algorithm in the sensor to perform initial filtering on the monitoring signal, obtain a multi-period initial filtering signal sequence, and upload the multi-period initial filtering signal data to the cloud platform database for storage through a compressed sensing algorithm;

[0010] S2: A secondary adaptive filtering model is constructed based on Kalman filtering and particle swarm algorithm. The obtained multi-period initial filtering monitoring signal sequence is input into the secondary adaptive filtering model, and the signal of each stage after segmented decomposition is filtered to obtain a multi-period secondary filtering monitoring signal sequence.

[0011] S3: splicing the acquired multi-period secondary filtering monitoring signal sequence through a signal splicing algorithm to obtain a complete secondary filtering signal;

[0012] S4: Set the quality evaluation index threshold, feed back the acquired complete secondary filtering signal to the signal quality evaluation model preset by the cloud platform through the corresponding interface for quality evaluation, compare the evaluation index results, and if the evaluation index threshold is met, display the secondary filtering signal through the cloud platform visualization interface; if not, feed back the comparison result to the secondary adaptive filtering model for filtering parameter adjustment until the evaluation index is met;

[0013] The specific steps of obtaining the multi-period initial filtering signal sequence in S1 include:

[0014] S101, assume that the monitoring signal sequence with a period length of T is acquired The acquired monitoring signal sequence is framed using the Haining window function, and the N frames of monitoring signals within the T time are obtained. The obtained N frames of monitoring signals are Fourier transformed to obtain the spectrum coefficient X(k, t) corresponding to each frame of monitoring signal, and the short-time energy value E of each frame within the time point t is calculated according to the obtained spectrum coefficient X(k, t) by the short-time energy formula. n (m); where x i Indicates obtaining the i-th signal sampling point in the monitoring signal sequence, E n represents the short-time energy value corresponding to the monitoring signal of the nth frame; m represents the total number of signal sampling points in the monitoring signal sequence obtained within T time; k represents the index of the kth frequency point in the discrete frequency domain after the monitoring signal sequence is transformed by Fourier transform;

[0015] S102, setting short-time energy dynamic segmentation thresholds h1 and h2, and setting monitoring signal segmentation conditions according to the short-time energy dynamic segmentation thresholds, specifically including: a first segmentation condition, a second segmentation condition and a third segmentation condition;

[0016] S103, according to the set monitoring signal segmentation conditions and the calculated N-frame monitoring signal short-time energy value E n (t), the monitoring signal is divided into a first monitoring signal period, a second monitoring signal period and a third monitoring signal period, specifically:

[0017]

[0018] in, represents the first monitoring signal period obtained according to the first segmentation condition, represents the second monitoring signal period obtained according to the second segmentation condition, represents the third monitoring signal period obtained according to the third segmentation condition; x(h1) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h1, w1 represents the extended length of the breakpoint of the first monitoring signal period on the x-axis, and is also the breakpoint identification area between the first monitoring signal period and the second monitoring signal period; x(h2) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h2, w2 represents the extended length of the breakpoint of the second monitoring signal period on the x-axis, that is, the identification interval of the second monitoring signal period and the third monitoring signal period, and x(·) represents the abscissa transformation function;

[0019] S104, inputting the obtained first monitoring signal period, second monitoring signal period and third monitoring signal period into the variational mode decomposition algorithm for decomposition, and obtaining u IMF components corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period respectively;

[0020] S105, retaining the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period from the u IMF components corresponding to the first monitoring signal period, the second monitoring signal period, and the third monitoring signal period, respectively, to obtain a multi-period initial filtered signal sequence;

[0021] The specific steps of constructing a secondary adaptive filtering model based on Kalman filtering and particle swarm algorithm in S2 include:

[0022] S201, initializing Kalman filter state estimation parameters based on multi-period initial filtering signal sequences Estimated error covariance matrix p 0|0 and state model parameters, wherein the state model parameters include: state transfer matrix A, control matrix B, process noise covariance matrix Q, observation model matrix H and observation noise covariance matrix R;

[0023] S202, according to the initialized Kalman filter state estimation parameters, the estimated error covariance matrix and the state model parameters are initialized to the number of particles N p The initial position and initial update speed of the particle in the particle swarm algorithm; wherein the initial position of the particle is composed of the state estimation parameters, the estimation error covariance matrix and the elements in the state model parameters in the Kalman filter;

[0024] S203, based on the multi-period initial filtering signal sequence and the configured target filtering effect, set the fitness function and constraint conditions corresponding to the particle swarm algorithm, specifically:

[0025]

[0026] in, represents the signal-to-noise ratio of the kth IMF component corresponding to the jth monitoring signal period, represents the purity of the kth IMF component corresponding to the jth monitoring signal period, C represents the complexity of monitoring signal decomposition; α1, β, γ are the weighting coefficients of the corresponding fitness function, SNR h represents the signal-to-noise ratio threshold, c h represents the optimal decomposition complexity;

[0027] S204, setting the maximum iteration cycle and the fitness function threshold, constructing a secondary adaptive filtering model for Kalman filtering and particle swarm algorithm for iterative training, stopping the iteration when the fitness function value is less than the fitness function threshold, otherwise continuing the iteration until the iteration cycle reaches the maximum iteration cycle, and obtaining the trained secondary adaptive filtering model;

[0028] Specifically, the specific steps of obtaining the multi-period secondary filtering monitoring signal sequence in S2 include:

[0029] S205, input the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period into pairs respectively to the secondary adaptive filtering model trained in S204 for secondary filtering processing, and obtain the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after the secondary filtering is completed;

[0030] S206, obtaining a multi-period secondary filtering monitoring signal sequence according to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after secondary filtering in represents the first monitoring signal period after secondary filtering, represents the second monitoring signal period after secondary filtering, Represents the third monitoring signal time period after secondary filtering.

[0031] Specifically, the specific steps of S3 are:

[0032] S301, according to the multi-period secondary filtering monitoring signal sequence obtained in S206 Through the breakpoint identification area corresponding to the secondary filtering monitoring signal period, the superposition average maximum value method is used to To splice, specifically:

[0033] S3011, according to the breakpoint identification area w1 set between the first monitoring signal period and the second monitoring signal period set in S103, the corresponding sampling points of the adjacent breakpoint areas w1 of the first monitoring signal period and the second monitoring signal period are superimposed and averaged, and the minimum value after the corresponding sampling points are summed and averaged is used as the splicing point of the first monitoring signal period and the second monitoring signal period, and the first monitoring signal period and the second monitoring signal period are spliced;

[0034] S3012. Repeat the process of S3011 to splice the second monitoring signal period and the third monitoring signal period to obtain a complete secondary filtering signal.

[0035] The intelligent signal filtering system based on cloud computing includes: data initial processing module, secondary filtering processing module and evaluation feedback module;

[0036] The data initial processing module includes a sensor unit; the sensor unit is used to obtain vibration or settlement monitoring signal data, and use the signal segmentation decomposition filtering algorithm built into the sensor to perform initial filtering processing on the monitoring signal to obtain a multi-time period initial filtering signal sequence.

[0037] Specifically, the secondary filtering processing module includes a secondary filtering processing unit and a splicing unit; the secondary filtering processing unit is used to perform secondary filtering processing on the uploaded multi-time period initial filtering signal sequence according to a secondary adaptive filtering model constructed based on Kalman filtering and particle swarm algorithm; the splicing unit is used to seamlessly splice the multi-time period initial filtering signal after secondary filtering through a signal splicing algorithm to obtain a complete secondary filtering signal.

[0038] Specifically, the evaluation feedback module includes a signal quality evaluation unit and a feedback adjustment unit; the signal quality evaluation unit is used to perform quality evaluation on the received complete secondary filtered signal according to a preset evaluation index and an evaluation signal quality evaluation model; the feedback adjustment unit is used to judge the signal filtering quality through the signal quality evaluation result and the quality evaluation index threshold, and trigger the feedback mechanism according to the judgment result, and continuously adjust the filtering parameters using the secondary adaptive filtering model.

[0039] A computer-readable storage medium, characterized in that computer instructions are stored thereon, and when the computer instructions are executed, an intelligent signal filtering method based on cloud computing is executed.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. In view of the deficiencies of the prior art, the present invention realizes the elastic expansion of signal processing capabilities by combining cloud computing resources, can adaptively adjust the secondary adaptive filtering model according to the real-time changes of signal characteristics, improve the accuracy of signal filtering under different working conditions, and make the signal analysis results closer to the actual situation. Secondly, by introducing the Hening window function framing, short-time energy analysis and variational mode decomposition (VMD), the signal can be intelligently segmented, which not only improves the recognition ability of complex signal structures, but also effectively filters out irrelevant information by retaining key IMF components, thereby enhancing the purity of the signal. Thirdly, by setting quality evaluation indicators and thresholds, the present invention realizes a closed-loop quality control system. When the signal quality does not meet the standard, it will automatically trigger the adjustment of the filtering parameters until the preset standard is met, which ensures the high quality and reliability of the output signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of an intelligent signal filtering method based on cloud computing according to Embodiment 1 of the present invention;

[0043] Figure 2 This is a flow chart of a process for obtaining a multi-period initial filtering signal sequence according to Embodiment 1 of the present invention;

[0044] Figure 3 This is a system analysis flow chart of an intelligent signal filtering method based on cloud computing according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0045] Example 1

[0046] See also Figure 1 , an embodiment of the present invention provides: an intelligent signal filtering method based on cloud computing, comprising the following steps:

[0047] S1: Acquire vibration or settlement monitoring signal data, use the built-in signal segmentation decomposition filtering algorithm in the sensor to perform initial filtering processing on the monitoring signal, obtain a multi-period initial filtering signal sequence, and upload the multi-period initial filtering signal data to the cloud platform database through a compressed sensing algorithm for storage; further, in this embodiment, the vibration or settlement monitoring signal data is obtained through a configured vibration detector or settlement detector;

[0048] S2: A secondary adaptive filtering model is constructed based on Kalman filtering and particle swarm algorithm. The obtained multi-period initial filtering monitoring signal sequence is input into the secondary adaptive filtering model, and the signal of each stage after segmented decomposition is filtered to obtain a multi-period secondary filtering monitoring signal sequence;

[0049] S3: splicing the acquired multi-period secondary filtering monitoring signal sequence through a signal splicing algorithm to obtain a complete secondary filtering signal;

[0050] S4: Setting a quality assessment index threshold, feeding back the acquired complete secondary filtering signal through a corresponding interface to a signal quality assessment model preset by the cloud platform for quality assessment, comparing the evaluation index results, and if the evaluation index threshold is met, displaying the secondary filtering signal through a cloud platform visualization interface, and if not, feeding back the comparison result to the secondary adaptive filtering model for filtering parameter adjustment until the evaluation index is met; further, in this embodiment, the signal quality assessment model is a support vector machine model trained by historical signal evaluation index data; the evaluation indicators used in this embodiment include signal-to-noise ratio and purity of the monitoring signal after filtering processing; in this embodiment, the threshold is set by a person skilled in the art according to the target requirements of the acquired signal;

[0051] S5. Input the secondary filtered signal that meets the threshold condition into the built-in abnormal monitoring model of the cloud platform, analyze the monitoring signal, and determine the abnormal monitoring result. According to the acquired abnormal monitoring result, the corresponding emergency strategy is generated through the built-in early warning subsystem and emergency strategy production algorithm of the cloud platform. In this embodiment, the emergency strategy production algorithm generates a large model through pre-trained text, generates corresponding early warning notices and emergency strategies, and sends them to users through mobile or fixed terminals.

[0052] Furthermore, the specific steps of obtaining the multi-period initial filtering signal sequence in S1 include:

[0053] S101, assume that the monitoring signal sequence with a period length of T is acquired The acquired monitoring signal sequence is framed using the Haining window function, and the N frames of monitoring signals within the T time are obtained. The obtained N frames of monitoring signals are Fourier transformed to obtain the spectrum coefficient X(k, t) corresponding to each frame of monitoring signal, and the short-time energy value E of each frame within the time point t is calculated according to the obtained spectrum coefficient X(k, t) by the short-time energy formula. n (m); where x i Indicates obtaining the i-th signal sampling point in the monitoring signal sequence, E n represents the short-time energy value corresponding to the monitoring signal of the nth frame; m represents the total number of signal sampling points in the monitoring signal sequence obtained within the T time, and k represents the index of the kth frequency point in the discrete frequency domain after the monitoring signal sequence is Fourier transformed; further, the short-time energy formula in this embodiment is the prior art and will not be repeated here;

[0054] S102, setting short-time energy dynamic segmentation thresholds h1 and h2, and setting monitoring signal segmentation conditions according to the short-time energy dynamic segmentation thresholds, specifically including: a first segmentation condition, a second segmentation condition and a third segmentation condition; further, the monitoring signal segmentation is performed in this embodiment because, in the process of electrical signal propagation, the electromotive force amplitude is large in the initial stage and the signal-to-noise ratio is high; in the transition stage, the electromotive force amplitude decays rapidly and the signal-to-noise ratio is relatively low; in the later stage, the electromotive force amplitude decays slowly and tends to be flat, the effective signal is submerged by the noise, and the signal-to-noise ratio is low; if fixed parameters and related functions are used to directly filter the entire signal, the filtered signal may have more residual noise or be distorted; therefore, the monitoring signal is dynamically segmented in this implementation process, and the above segmentation can adaptively perform more targeted processing on each corresponding stage of the monitoring signal to reduce the occurrence of residual noise and distortion;

[0055] Further, in this embodiment, the specific steps of obtaining the short-time energy dynamic segmentation threshold include:

[0056] Obtain historical monitoring signal sequence data and corresponding segmentation threshold data, and use the support vector machine algorithm and the dynamic segmentation model obtained by training the acquired data;

[0057] Input each monitoring signal sequence with a cycle length of T obtained at the current moment into the dynamic segmentation model, output the segmentation threshold pair h1, h2 of the corresponding mode, and set the obtained segmentation threshold pair h1, h2 as the short-time energy dynamic segmentation threshold h1, h2; This process can make the segmentation threshold more targeted and in line with the actual situation when segmenting the monitoring signal sequence by obtaining the corresponding segmentation threshold pair h1, h2 through the monitoring signal sequence, so that the segmentation operation within each cycle length is more accurate;

[0058] S103, according to the set monitoring signal segmentation conditions and the calculated N-frame monitoring signal short-time energy value E n (t), the monitoring signal is divided into a first monitoring signal period, a second monitoring signal period and a third monitoring signal period, specifically:

[0059]

[0060] in, represents the first monitoring signal period obtained according to the first segmentation condition, represents the second monitoring signal period obtained according to the second segmentation condition, represents the third monitoring signal period obtained according to the third segmentation condition; x(h1) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h1, w1 represents the extended length of the breakpoint of the first monitoring signal period on the x-axis, and is also the breakpoint identification area between the first monitoring signal period and the second monitoring signal period. The specific length of w1 is x(h2) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h2, w2 represents the extended length of the breakpoint of the second monitoring signal period on the x-axis, that is, the identification interval of the second monitoring signal period and the third monitoring signal period, and the specific length of w2 is x(·) represents the abscissa transformation function, which is used to calculate the abscissa of the signal sampling point corresponding to the short-time energy dynamic segmentation threshold. This process divides the monitoring signal into different time periods through the threshold segmentation process in S102, and uses the abscissa transformation function to transform the temporal transformation data into spatial change data, making the signal easier to observe and analyze. In addition, in order to facilitate the subsequent filtering signal splicing, the signal breakpoints of each time period are expanded forward and backward by w1 and w2, avoiding the signal information loss caused by signal splicing in the later stage, so that the signal information after splicing is retained more completely;

[0061] S104, inputting the obtained first monitoring signal period, second monitoring signal period and third monitoring signal period into the variational mode decomposition algorithm for decomposition, and obtaining u IMF components corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period respectively;

[0062] S105. From the u IMF components corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period respectively, retain the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period, and obtain a multi-period initial filtered signal sequence.

[0063] This process automatically adjusts the processing strategy according to the different stages of the signal (initial, transitional and late stages), which can better filter out noise and prevent signal distortion. The dynamic segmentation method makes signal processing more flexible and targeted, reducing the possibility of noise residue and signal distortion. In addition, this process uses the variational mode decomposition algorithm to extract the key components of each time period when the amplitude of the signal's electromotive force varies greatly, ensuring higher signal quality and obtaining a multi-time period initial filtered signal sequence with less noise and more complete information retention.

[0064] Furthermore, the specific steps of constructing a secondary adaptive filtering model based on Kalman filtering and particle swarm algorithm in S2 include:

[0065] S201, initializing Kalman filter state estimation parameters based on multi-period initial filtering signal sequences Estimated error covariance matrix p 0|0 and state model parameters, wherein the state model parameters include: state transfer matrix A, control matrix B, process noise covariance matrix Q, observation model matrix H and observation noise covariance matrix R;

[0066] S202, according to the initialized Kalman filter state estimation parameters, the estimated error covariance matrix and the state model parameters are initialized to the number of particles N p The initial position and initial update speed of the particle in the particle swarm algorithm; wherein the initial position of the particle is composed of the state estimation parameters, the estimation error covariance matrix and the elements in the state model parameters in the Kalman filter;

[0067] S203, based on the multi-period initial filtering signal sequence and the configured target filtering effect, set the fitness function and constraint conditions corresponding to the particle swarm algorithm, specifically: C=c h ,in, represents the signal-to-noise ratio of the kth IMF component corresponding to the jth monitoring signal period, represents the purity of the kth IMF component corresponding to the jth monitoring signal period, C represents the complexity of monitoring signal decomposition; α1, β, γ are the weighting coefficients of the corresponding fitness function, SNR h represents the signal-to-noise ratio threshold, c h represents the optimal decomposition complexity; the corresponding formula is: power(·) represents the power of the monitoring signal. Furthermore, in this embodiment, the spectrum peak energy ratio is used as a measure of purity. The specific formula is: in It represents the spectrum peak energy ratio of the kth IMF component corresponding to the jth monitoring signal period, It represents the energy of the main spectrum peak corresponding to the kth IMF component in the jth monitoring signal period; Indicates the total energy corresponding to the kth IMF component of the jth monitoring signal period; this embodiment uses the mode number and length ratio to measure the decomposition complexity of the corresponding monitoring signal, and the specific formula is: Wherein K represents the total number of IMF components, and in this embodiment, K=6, and the specific decomposition complexity setting is performed according to the actual situation in actual use; further, in this embodiment, the fitness function and the constraint conditions are constructed by the signal-to-noise ratio, purity and decomposition complexity, wherein a high signal-to-noise ratio means that the effective information content of the signal is large, and a high purity indicates that the signal has less interference, and a small decomposition complexity indicates that the signal is easy to process. Through the comprehensive consideration of these variables, the obtained secondary adaptive filtering model can reduce the computational complexity as much as possible while ensuring the signal quality and easy processing, thereby improving the efficiency and accuracy of the algorithm; in addition, this embodiment uses the spectrum peak energy ratio as a measure of purity, which can more accurately judge the purity of the signal, and measures the decomposition complexity of the corresponding monitoring signal by the mode number and length ratio, and can perform specific decomposition complexity settings according to the actual situation, thereby reducing the difficulty of the actual operation process;

[0068] S204, setting the maximum iteration cycle and the fitness function threshold, constructing a secondary adaptive filtering model for Kalman filtering and particle swarm algorithm for iterative training, stopping the iteration when the fitness function value is less than the fitness function threshold, otherwise continuing the iteration until the iteration cycle reaches the maximum iteration cycle, and obtaining the trained secondary adaptive filtering model;

[0069] Furthermore, the specific steps of obtaining the multi-period secondary filtering monitoring signal sequence in S2 include:

[0070] S205, input the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period into pairs respectively to the secondary adaptive filtering model trained in S204 for secondary filtering processing, and obtain the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after the secondary filtering is completed;

[0071] S206, obtaining a multi-period secondary filtering monitoring signal sequence according to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after secondary filtering in represents the first monitoring signal period after secondary filtering, represents the second monitoring signal period after secondary filtering, Represents the third monitoring signal time period after secondary filtering.

[0072] Furthermore, the specific steps of S3 are:

[0073] S301, according to the multi-period secondary filtering monitoring signal sequence obtained in S206 Through the breakpoint identification area corresponding to the secondary filtering monitoring signal period, the superposition average maximum value method is used to To splice, specifically:

[0074] S3011, according to the breakpoint identification area w1 set between the first monitoring signal period and the second monitoring signal period set in S103, the corresponding sampling points of the adjacent breakpoint areas w1 of the first monitoring signal period and the second monitoring signal period are superimposed and averaged, and the minimum value after the corresponding sampling points are summed and averaged is used as the splicing point of the first monitoring signal period and the second monitoring signal period, and the first monitoring signal period and the second monitoring signal period are spliced;

[0075] S3012. Repeat the process of S3011 to splice the second monitoring signal period and the third monitoring signal period to obtain a complete secondary filtering signal.

[0076] This process uses the superposition averaging method to not only ensure the smooth transition of signals between time periods and reduce or eliminate discontinuities at breakpoints, but also offsets random noise in the average. In addition, by selecting the minimum value as the splicing point, the quality degradation caused by signal mutation can be reduced, making the final signal more reliable and accurate.

[0077] Example 2

[0078] See also Figure 3 , another embodiment provided by the present invention: an intelligent signal filtering system based on cloud computing, comprising: a data initial processing module, a secondary filtering processing module, an evaluation feedback module and an abnormality analysis module;

[0079] The data initial processing module is used to obtain vibration or settlement monitoring signal data, perform preliminary filtering processing, and save the initially processed monitoring signal in the cloud platform database; the data initial processing module includes a sensor unit and a compression storage unit; the sensor unit is used to obtain vibration or settlement monitoring signal data, and use the built-in signal segmentation decomposition filtering algorithm in the sensor to perform initial filtering processing on the monitoring signal to obtain a multi-period initial filtering signal sequence; the compression storage unit is used to upload the obtained multi-period initial filtering signal sequence to the cloud platform database for storage through a compressed sensing algorithm;

[0080] The secondary filtering processing module is used to perform secondary filtering on the signal after initial processing; the secondary filtering processing module includes a secondary filtering processing unit and a splicing unit; the secondary filtering processing unit is used to perform secondary filtering processing on the uploaded multi-period initial filtering signal sequence according to the secondary adaptive filtering model constructed by Kalman filtering and particle swarm algorithm; the splicing unit is used to seamlessly splice the multi-period initial filtering signal after secondary filtering through a signal splicing algorithm to obtain a complete secondary filtering signal;

[0081] An evaluation feedback module is used to evaluate the quality of the final filtered signal and adjust the signal filtering parameters according to the evaluation results; the evaluation feedback module includes a signal quality evaluation unit and a feedback adjustment unit; the signal quality evaluation unit is used to evaluate the quality of the received complete secondary filtered signal according to a preset evaluation index and an evaluation signal quality evaluation model; the feedback adjustment unit is used to judge the signal filtering quality through the signal quality evaluation result and the quality evaluation index threshold, and trigger the feedback mechanism according to the judgment result, and continuously adjust the filtering parameters using the secondary adaptive filtering model until the signal quality meets the requirements;

[0082] The anomaly analysis module is used to perform monitoring anomaly analysis using the monitoring signal after secondary filtering processing and generate corresponding processing strategies; the anomaly analysis module includes an anomaly analysis unit and a strategy generation unit; the anomaly analysis unit is used to perform anomaly analysis on the monitoring signal based on the acquired secondary filtering signal that meets the threshold conditions and the built-in anomaly monitoring model of the cloud platform, and judge the monitoring anomaly results; the strategy generation unit is used to generate corresponding emergency strategies based on the judged anomaly results through the built-in early warning subsystem and emergency strategy production algorithm of the cloud platform.

[0083] Example 3

[0084] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of an intelligent signal filtering method based on cloud computing.

[0085] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.

Claims

1. An intelligent signal filtering method based on cloud computing, characterized in that: include: S1: Acquire vibration or settlement monitoring signal data, use the built-in signal segmentation decomposition filtering algorithm in the sensor to perform initial filtering on the monitoring signal, obtain a multi-period initial filtering signal sequence, and upload the multi-period initial filtering signal data to the cloud platform database for storage through a compressed sensing algorithm; S2: A secondary adaptive filtering model is constructed based on Kalman filtering and particle swarm algorithm. The obtained multi-period initial filtering monitoring signal sequence is input into the secondary adaptive filtering model, and the signal of each stage after segmented decomposition is filtered to obtain a multi-period secondary filtering monitoring signal sequence. S3: splicing the acquired multi-period secondary filtering monitoring signal sequence through a signal splicing algorithm to obtain a complete secondary filtering signal; S4: Set the quality evaluation index threshold, feed back the acquired complete secondary filtering signal to the signal quality evaluation model preset by the cloud platform through the corresponding interface for quality evaluation, compare the evaluation index results, and if the evaluation index threshold is met, display the secondary filtering signal through the cloud platform visualization interface; if not, feed back the comparison result to the secondary adaptive filtering model for filtering parameter adjustment until the evaluation index is met; The specific steps of obtaining the multi-period initial filtering signal sequence in S1 include: S101, assume that the monitoring signal sequence with a period length of T is acquired The acquired monitoring signal sequence is framed using the Haining window function, and the N frames of monitoring signals within the T time are obtained. The obtained N frames of monitoring signals are Fourier transformed to obtain the spectrum coefficient X(k, t) corresponding to each frame of monitoring signal, and the short-time energy value E of each frame within the time point t is calculated according to the obtained spectrum coefficient X(k, t) by the short-time energy formula. n (m); where x i Indicates obtaining the i-th signal sampling point in the monitoring signal sequence, E n represents the short-time energy value corresponding to the monitoring signal of the nth frame; m represents the total number of signal sampling points in the monitoring signal sequence obtained within T time; k represents the index of the kth frequency point in the discrete frequency domain after the monitoring signal sequence is transformed by Fourier transform; S102, setting short-time energy dynamic segmentation thresholds h1 and h2, and setting monitoring signal segmentation conditions according to the short-time energy dynamic segmentation thresholds, specifically including: a first segmentation condition, a second segmentation condition and a third segmentation condition; S103, according to the set monitoring signal segmentation conditions and the calculated N-frame monitoring signal short-time energy value E n (t), the monitoring signal is divided into a first monitoring signal period, a second monitoring signal period and a third monitoring signal period, specifically: in, represents the first monitoring signal period obtained according to the first segmentation condition, represents the second monitoring signal period obtained according to the second segmentation condition, represents the third monitoring signal period obtained according to the third segmentation condition; x(h1) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h1, w1 represents the extended length of the breakpoint of the first monitoring signal period on the x-axis, and is also the breakpoint identification area between the first monitoring signal period and the second monitoring signal period; x(h2) represents the index of the signal sampling point in the signal sampling sequence corresponding to the threshold h2, w2 represents the extended length of the breakpoint of the second monitoring signal period on the x-axis, that is, the identification interval of the second monitoring signal period and the third monitoring signal period, and x(·) represents the abscissa transformation function; S104, inputting the obtained first monitoring signal period, second monitoring signal period and third monitoring signal period into the variational mode decomposition algorithm for decomposition, and obtaining u IMF components corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period respectively; S105, retaining the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period from the u IMF components corresponding to the first monitoring signal period, the second monitoring signal period, and the third monitoring signal period, respectively, to obtain a multi-period initial filtered signal sequence; The specific steps of constructing a secondary adaptive filtering model based on Kalman filtering and particle swarm algorithm in S2 include: S201, initializing Kalman filter state estimation parameters based on multi-period initial filtering signal sequences Estimated error covariance matrix p 0|0 and state model parameters, wherein the state model parameters include: state transfer matrix A, control matrix B, process noise covariance matrix Q, observation model matrix H and observation noise covariance matrix R; S202, according to the initialized Kalman filter state estimation parameters, the estimated error covariance matrix and the state model parameters are initialized to the number of particles N p The initial position and initial update speed of the particle in the particle swarm algorithm; wherein the initial position of the particle is composed of the state estimation parameters, the estimation error covariance matrix and the elements in the state model parameters in the Kalman filter; S203, based on the multi-period initial filtering signal sequence and the configured target filtering effect, set the fitness function and constraint conditions corresponding to the particle swarm algorithm, specifically: C=ch, in, represents the signal-to-noise ratio of the kth IMF component corresponding to the jth monitoring signal period, represents the purity of the kth IMF component corresponding to the jth monitoring signal period, C represents the complexity of monitoring signal decomposition; α1, β, γ are the weighting coefficients of the corresponding fitness function, SNR h represents the signal-to-noise ratio threshold, c h represents the optimal decomposition complexity; S204, setting the maximum iteration cycle and the fitness function threshold, constructing a secondary adaptive filtering model for Kalman filtering and particle swarm algorithm for iterative training, stopping the iteration when the fitness function value is less than the fitness function threshold, otherwise continuing the iteration until the iteration cycle reaches the maximum iteration cycle, and obtaining the trained secondary adaptive filtering model.

2. The intelligent signal filtering method based on cloud computing according to claim 1, characterized in that: The specific steps of obtaining the multi-period secondary filtering monitoring signal sequence in S2 include: S205, input the first IMF component d1(1) and the first IMF component d1(2) corresponding to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period, the first IMF component d2(1) and the second IMF component d2(2) corresponding to the second monitoring signal period, and the first IMF component d3(1) and the first IMF component d3(2) corresponding to the third monitoring signal period into pairs respectively to the secondary adaptive filtering model trained in S204 for secondary filtering processing, and obtain the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after the secondary filtering is completed; S206, obtaining a multi-period secondary filtering monitoring signal sequence according to the first monitoring signal period, the second monitoring signal period and the third monitoring signal period after secondary filtering in represents the first monitoring signal period after secondary filtering, represents the second monitoring signal period after secondary filtering, Represents the third monitoring signal period after secondary filtering.

3. The intelligent signal filtering method based on cloud computing according to claim 2, characterized in that: The specific steps of S3 are: S301, according to the multi-period secondary filtering monitoring signal sequence obtained in S206 Through the breakpoint identification area corresponding to the secondary filtering monitoring signal period, the superposition average maximum value method is used to To splice, specifically: S3011, according to the breakpoint identification area w1 set between the first monitoring signal period and the second monitoring signal period set in S103, the corresponding sampling points of the adjacent breakpoint areas w1 of the first monitoring signal period and the second monitoring signal period are superimposed and averaged, and the minimum value after the corresponding sampling points are summed and averaged is used as the splicing point of the first monitoring signal period and the second monitoring signal period, and the first monitoring signal period and the second monitoring signal period are spliced; S3012. Repeat the process of S3011 to splice the second monitoring signal period and the third monitoring signal period to obtain a complete secondary filtering signal.

4. A cloud computing-based intelligent signal filtering system, which is implemented based on the cloud computing-based intelligent signal filtering method according to any one of claims 1 to 3, characterized in that: include: Data initial processing module, secondary filtering processing module and evaluation feedback module; The data initial processing module includes a sensor unit; the sensor unit is used to obtain vibration or settlement monitoring signal data, and use the signal segmentation decomposition filtering algorithm built into the sensor to perform initial filtering processing on the monitoring signal to obtain a multi-time period initial filtering signal sequence.

5. The cloud computing-based intelligent signal filtering system according to claim 4, characterized in that: The secondary filtering processing module includes a secondary filtering processing unit and a splicing unit; the secondary filtering processing unit is used to perform secondary filtering processing on the uploaded multi-time period initial filtering signal sequence according to a secondary adaptive filtering model constructed based on Kalman filtering and particle swarm algorithm; the splicing unit is used to seamlessly splice the multi-time period initial filtering signal after secondary filtering through a signal splicing algorithm to obtain a complete secondary filtering signal.

6. The cloud computing-based intelligent signal filtering system according to claim 5, characterized in that: The evaluation feedback module includes a signal quality evaluation unit and a feedback adjustment unit; the signal quality evaluation unit is used to perform quality evaluation on the received complete secondary filtering signal according to a preset evaluation index and an evaluation signal quality evaluation model; The feedback adjustment unit is used to judge the signal filtering quality through the signal quality evaluation result and the quality evaluation index threshold, trigger the feedback mechanism according to the judgment result, and continuously adjust the filtering parameters by using the secondary adaptive filtering model.

7. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the cloud computing-based intelligent signal filtering method described in any one of claims 1 to 3 is executed.

Images