An intelligent detection tower crane lifting process control system

Abstract

The present invention discloses an intelligent detection tower crane lifting process control system, comprising a data acquisition module, an adaptive spectrum filtering module, an impact retention module, a motion timing feature construction module, a linear drift monitoring module, a nonlinear disturbance separation module, a threshold statistical decision module, and a lifting process control module. The present invention belongs to the field of intelligent control, and specifically refers to an intelligent detection tower crane lifting process control system. This solution adopts an adaptive sampling method based on gradient triggering to improve the accuracy of fault detection; applies an optimized wavelet threshold function to improve the accuracy of lifting process control; extracts the linear dynamic characteristics of the linear relationship of load current through a linear drift monitoring module; constructs residual differences in the residual subspace, calculates nonlinear energy indicators and errors, is extremely sensitive to nonlinear disturbances and non-Gaussian noise, and strengthens the blind spots of linear analysis; thereby improving the control effect of the tower crane lifting process.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control, and in particular to an intelligent detection tower crane lifting process control system. Background Art

[0002] The control system for detecting the tower crane's jacking process is a system used to ensure the safe and stable operation of tower cranes during the jacking process. The system monitors, analyzes, and controls various parameters and states during the jacking process in real time, promptly discovers potential faults and abnormal conditions, and takes appropriate measures to ensure the smooth progress of the jacking operation. However, general control systems for detecting the tower crane's jacking process miss key transient information during the structural resonance or impact mutation stage, and have difficulty distinguishing the pulsating noise of the hydraulic pump from the main motion signal, resulting in delayed fault detection and increased safety risks. General control systems for detecting the tower crane's jacking process have difficulty detecting linear drift faults caused by mechanical wear or insufficient lubrication, and ignore nonlinear factors such as wind load, swing, and impact in the tower crane's jacking process, which in turn leads to poor control of the tower crane's jacking process. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent detection tower crane jacking process control system. In view of the problem that the general detection tower crane jacking process control system misses key transient information in the structural resonance or impact mutation stage, and it is difficult to distinguish the hydraulic pump pulsation noise from the main motion signal, resulting in untimely fault detection and increased safety risks, this solution adopts an adaptive sampling method based on gradient triggering. When there is an impact or structural resonance mutation, the system automatically increases the sampling rate to improve the accuracy of fault detection; modes with different signal-to-noise ratios are classified and processed through adaptive spectrum filtering. For modes judged to have a low signal-to-noise ratio, an optimized wavelet threshold function is applied to introduce harmonic parameters into the threshold to adapt to the time-varying characteristics of the hydraulic pump pulsation and the structural vibration background noise. Based on high sensitivity, it captures the subtle fault trends of seal leakage and loose rigid connection, and timely discovers potential safety hazards, thereby improving Accuracy of jacking process control; The general tower crane jacking process control system has the problem of difficulty in detecting linear drift faults caused by mechanical wear or insufficient lubrication, and ignores nonlinear factors such as wind load, swing and impact in the tower crane jacking process, which leads to poor control effect of the tower crane jacking process. This solution constructs past and future vectors based on reconstructed signals, extracts the linear dynamic characteristics of the linear relationship between load and current through the linear drift monitoring module, quickly detects linear drift faults caused by mechanical wear or insufficient lubrication, and is sensitive to the tiny drift of linear parameters caused by abnormal wire rope tension; constructs residual differences in the residual subspace, calculates nonlinear energy indicators and errors, is extremely sensitive to nonlinear disturbances and non-Gaussian noise, and strengthens the blind spots of linear analysis; uses historical normal jacking data and adopts kernel density estimation to non-parametrically determine the thresholds of each indicator to achieve jacking process control; thereby improving the control effect of the tower crane jacking process.

[0004] The technical solution adopted by the present invention is as follows: the present invention provides an intelligent detection tower crane jacking process control system, including a data acquisition module, an adaptive spectrum filtering module, an impact retention module, a motion time series feature construction module, a linear drift monitoring module, a nonlinear disturbance separation module, a threshold statistical decision module and a jacking process control module;

[0005] The data acquisition module collects the lifting signal during the tower crane lifting process and dynamically adjusts the sampling rate using an adaptive sampling mechanism based on gradient triggering;

[0006] The adaptive spectrum filtering module filters the noise mode;

[0007] The impact retention module applies a non-jump optimized wavelet threshold function for low signal-to-noise ratio modes, combines dynamic thresholds with harmonic parameters for threshold processing, and finally reconstructs the signal;

[0008] The motion time series feature construction module arranges the reconstructed signal into an original data matrix, constructs past vectors and future vectors, and normalizes and splices them to form a time series feature matrix;

[0009] The linear drift monitoring module constructs a scaled Hankel matrix based on the load current signal covariance matrix and performs singular value decomposition, and calculates the residual vector and detection statistics through canonical variable projection;

[0010] The nonlinear disturbance separation module constructs and centers a kernel matrix in the residual subspace using kernel distance for whitening, and then extracts independent signal components and calculates energy indices through kernel independent component analysis;

[0011] The threshold statistical decision module uses historical normal data to perform Gaussian kernel density estimation on each detection indicator and calculates the corresponding threshold according to the set confidence level;

[0012] The lifting process control module implements lifting process control for each lifting sample based on detection indicators and corresponding thresholds.

[0013] Furthermore, the data acquisition module collects the lifting signal during the tower crane lifting process; performs variational modal decomposition on the original lifting signal to obtain K intrinsic mode functions; adopts adaptive sampling based on gradient triggering, and when the first-order derivative of the load signal Or the current signal change rate When the threshold is exceeded, the sampling rate is automatically increased, which is expressed as: ;in, and They are the upper and lower limits of the sampling rate; is the sigmoid function; and is the normalized weight coefficient; is the trigger threshold.

[0014] Furthermore, the adaptive spectrum filtering module uses approximate entropy Evaluate the random complexity of the eigenmodes; by permutation entropy Evaluation signal perturbation; respectively expressed as: ; ; Normalized synthetic harmonic parameters , expressed as: ; Calculate the mean P and confidence interval σ, expressed as: ; σ ;in, and are the sample sequence similarity statistics of dimension m and dimension m+1 respectively, r is the tolerance threshold, i is the modal index, are all possible permutations of symbols of length m; In the time series of the ith mode, the arrangement pattern is observed probability; and After normalization, and ; K is the total number of modes; for The high noise mode corresponds to the hydraulic pump pulsation or motor switching interference and is discarded; The high signal-to-noise ratio modes corresponding to the main motion displacement and structural resonance are retained; The low signal-to-noise ratio mode is used to perform signal denoising.

[0015] Furthermore, the impact retention module applies an optimized wavelet threshold function to perform signal denoising on modes determined to have a low signal-to-noise ratio, and introduces harmonic parameters into the threshold. The formula used is: ; ;in, It is the output of the optimized wavelet threshold function; is the original wavelet coefficient, j is the wavelet decomposition series index, and u is the position index; is the threshold parameter; sign(·) is the sign function; is the standard deviation of the background noise at the jth level of the wavelet; N is the signal length; is the threshold adjustment factor; the high signal-to-noise ratio mode and the low signal-to-noise ratio mode are reconstructed and expressed as: ;in, is the reconstructed signal, is the signal component of the i-th eigenmode at time k.

[0016] Furthermore, the motion time series feature construction module arranges the reconstructed signals in time series to form an original data matrix; constructs past and future vectors, normalizes them and splices them into a matrix, which is expressed as: ;in, and are the past and future sample vectors respectively; q is the number of delay steps; 、 、 and are the original vectors of the reconstructed signals at the k-1th, kqth, kth and k+q-1th moments respectively.

[0017] Furthermore, the linear drift monitoring module calculates the covariance matrix of the past sample vector and itself, the covariance matrix of the future sample vector and itself, and the cross covariance of the future sample vector and the past sample vector, which are expressed as ; Construct the scaled Hankel matrix and perform singular value decomposition, expressed as: ; where H is the scaled Hankel matrix; 、 and are the first r left singular vectors, corresponding to the future space projection basis, the first r right singular vectors, corresponding to the past space projection basis and the diagonal singular value matrix; r is the truncation rank; the projection obtains the typical variable, which is expressed as: ;in, and are the future typical variable vector and the past typical variable vector at the kth moment respectively; and They are and Normalization of ; Calculate the difference and indicators , expressed as: ; ; Where I is the identity matrix; is the sample residual vector; is the test statistic.

[0018] Furthermore, the nonlinear disturbance separation module constructs residual difference in the residual subspace of the linear drift monitoring module, which is expressed as: ;in, is the difference vector of the sample in the residual subspace; and are the future projection matrix and the past projection matrix of the residual subspace respectively; It is the mapping matrix of the residual subspace, which describes the linear mapping relationship between the residual of the past sample vector and the residual of the future sample vector. The kernel distance is defined as: ;in, is the difference vector in the residual subspace and the distance between them; is the Gaussian kernel width; the kernel matrix is ​​constructed for the residual vector set of the linear drift monitoring module and then centered and normalized to obtain the whitened data Z. The kernel independent component s is extracted from Z. The formula used is: ; ;in, and are the demixing vectors before and after updating, respectively; g(·) is the tanh function; is the derivative of g(·); is the expectation; the calculation index is expressed as: ; ; ;in, It is a nonlinear energy index; is the error; is a significant independent component of s, taking the d-dimensional column vector of the first d rows and k columns of s; is the generalized inverse matrix; is a matrix composed of the first d demixed vectors extracted in rows; is the eigenvector of the kth sample after whitening in the kernel space.

[0019] Furthermore, the threshold statistical decision module uses historical normal lifting data and adopts kernel density estimation to non-parametrically determine the threshold of each indicator to ensure accuracy and robustness; for each indicator : ; ;in, It is a non-parametric estimate of the probability density function of indicator J under historical normal jacking data; is the Gaussian kernel function; M is the total number of data, i is the index of historical normal lifting data; is the index value of the i-th data; x is the index value position during density estimation; h is the bandwidth; is the indicator threshold; is the confidence level; hence the indicator threshold 、 and ,correspond .

[0020] Furthermore, the lifting process control module calculates the detection index for each new lifting sample. 、 and , the abnormal degree of the detection index is mapped to the warning level and corresponds to different control strategies; when any detection index exceeds the corresponding index threshold five times in a row, a fault alarm is triggered immediately.

[0021] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0022] (1) Aiming at the problem that the control system of the general detection tower crane jacking process misses key transient information during the structural resonance or impact mutation stage, and it is difficult to distinguish the hydraulic pump pulsation noise from the main motion signal, resulting in delayed fault detection and increased safety risks, this scheme adopts an adaptive sampling method based on gradient triggering. When there is an impact or structural resonance mutation, the system automatically increases the sampling rate to improve the accuracy of fault detection; the modes with different signal-to-noise ratios are classified and processed through adaptive spectrum filtering. For the modes judged to have low signal-to-noise ratios, the optimized wavelet threshold function is applied to introduce harmonic parameters into the threshold to adapt to the time-varying characteristics of the hydraulic pump pulsation and the structural vibration background noise. Based on high sensitivity, the small fault trends of seal leakage and loose rigid connection are captured, and potential safety hazards are discovered in time, thereby improving the accuracy of jacking process control.

[0023] (2) In view of the fact that the general tower crane lifting process control system has difficulty in detecting linear drift faults caused by mechanical wear or insufficient lubrication, and ignores the nonlinear factors such as wind load, swing and impact in the tower crane lifting process, which leads to poor control effect of the tower crane lifting process, this scheme constructs past and future vectors based on the reconstructed signal, extracts the linear dynamic characteristics of the linear relationship between load and current through the linear drift monitoring module, quickly detects linear drift faults caused by mechanical wear or insufficient lubrication, and is sensitive to the small drift of linear parameters caused by abnormal wire rope tension; constructs residual differences in the residual subspace, calculates nonlinear energy indicators and errors, is extremely sensitive to nonlinear disturbances and non-Gaussian noise, and strengthens the blind spots of linear analysis; uses historical normal lifting data and adopts kernel density estimation to non-parametrically determine the thresholds of each indicator to achieve lifting process control; thereby improving the control effect of the tower crane lifting process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a schematic diagram of the module structure of an intelligent detection tower crane jacking process control system provided by the present invention.

[0025] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0027] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0028] Example 1, see Figure 1 The present invention provides an intelligent detection tower crane lifting process control system, which includes a data acquisition module, an adaptive spectrum filtering module, an impact retention module, a motion time series feature construction module, a linear drift monitoring module, a nonlinear disturbance separation module, a threshold statistical decision module and a lifting process control module;

[0029] The data acquisition module collects the lifting signal during the tower crane lifting process, and dynamically adjusts the sampling rate using an adaptive sampling mechanism based on gradient triggering; and sends the data to the adaptive spectrum filtering module;

[0030] The adaptive spectrum filtering module filters the noise mode and sends the data to the impact retention module;

[0031] The impact retention module applies a non-jump optimized wavelet threshold function for low signal-to-noise ratio modes, combines dynamic thresholds with harmonic parameters for threshold processing, and finally reconstructs the signal; and sends the data to the motion time series feature construction module;

[0032] The motion time series feature construction module arranges the reconstructed signal into an original data matrix, constructs past vectors and future vectors, normalizes and splices them, and forms a time series feature matrix; and sends the data to the linear drift monitoring module;

[0033] The linear drift monitoring module constructs a scaled Hankel matrix based on the load current signal covariance matrix and performs singular value decomposition, calculates the residual vector and detection statistics through typical variable projection, and sends the data to the nonlinear disturbance separation module;

[0034] The nonlinear disturbance separation module constructs and centralizes the kernel matrix in the residual subspace using kernel distance for whitening, then extracts independent signal components and calculates energy indicators through kernel independent component analysis; and sends the data to the threshold statistical decision module;

[0035] The threshold statistical decision module uses historical normal data to perform Gaussian kernel density estimation on each detection indicator, and calculates the corresponding threshold according to the set confidence level; and sends the data to the jacking process control module;

[0036] The lifting process control module implements lifting process control for each lifting sample based on detection indicators and corresponding thresholds.

[0037] Example 2, see Figure 1 This embodiment is based on the above embodiment. The data acquisition module collects the lifting signal during the tower crane lifting process, including the main hook load, the lifting motor current, the wire rope angle and the vibration acceleration; the original lifting signal is subjected to variational mode decomposition to obtain K intrinsic mode functions; adaptive sampling based on gradient triggering is adopted. When the first-order derivative of the load signal Or the current signal change rate When the threshold is exceeded, the sampling rate is automatically increased, which is expressed as: ;in, and They are the upper and lower limits of the sampling rate; is the sigmoid function; and is the normalized weight coefficient; It is the trigger threshold; it reduces the amount of data and alleviates communication and storage pressure under stable working conditions; it automatically increases the frequency when there is an impact or structural resonance mutation to ensure high-fidelity capture.

[0038] Example 3, see Figure 1 This embodiment is based on the above embodiment. The adaptive spectrum filtering module uses the approximate entropy Evaluate the random complexity of the eigenmodes; by permutation entropy Evaluation signal perturbation; respectively expressed as: ; ; Normalized synthetic harmonic parameters , expressed as: ; Calculate the mean P and confidence interval σ, expressed as: ; σ ;in, and are the sample sequence similarity statistics of dimension m and dimension m+1 respectively, r is the tolerance threshold, i is the modal index, are all possible permutations of symbols of length m; In the time series of the ith mode, the arrangement pattern is observed probability; and After normalization, and ; K is the total number of modes; for The high noise mode corresponds to the hydraulic pump pulsation or motor switching interference and is discarded; The high signal-to-noise ratio modes corresponding to the main motion displacement and structural resonance are retained; The low signal-to-noise ratio mode is used to perform signal denoising.

[0039] Example 4, see Figure 1 This embodiment is based on the above embodiment. The impact retention module applies the optimized wavelet threshold function to perform signal denoising for the mode determined to have a low signal-to-noise ratio. There is no jump at the threshold to avoid ringing interference control at the threshold; when the coefficient is large, it is approximate to a hard threshold to retain the mutation information of the transient impact of the top sliding hook; the harmonic parameter is introduced into the threshold to adapt to the time-varying characteristics of the hydraulic pump pulsation and structural vibration background noise. The formula used is: ; ;in, It is the output of the optimized wavelet threshold function; is the original wavelet coefficient, j is the wavelet decomposition series index, and u is the position index; is the threshold parameter; sign(·) is the sign function; is the standard deviation of the background noise at the jth level of the wavelet; N is the signal length; is the threshold adjustment factor; it accurately separates hydraulic pulse noise from structural vibration and retains key mutations to the maximum extent; it has no constant offset and does not affect the displacement and velocity peak monitoring; it reconstructs the high signal-to-noise ratio mode and the low signal-to-noise ratio mode, which can be expressed as: ;in, is the reconstructed signal, It is the signal component of the i-th eigenmode at time k; it can capture small fault trends such as seal leakage and loose rigid connection with high sensitivity.

[0040] By performing the above operations, the general tower crane jacking process control system misses key transient information during the structural resonance or impact mutation stage, and it is difficult to distinguish the hydraulic pump pulsation noise from the main motion signal, resulting in untimely fault detection and increased safety risks. This solution adopts an adaptive sampling method based on gradient triggering. When there is an impact or structural resonance mutation, the system automatically increases the sampling rate to improve the accuracy of fault detection; modes with different signal-to-noise ratios are classified and processed through adaptive spectrum filtering. For modes judged to have low signal-to-noise ratios, the optimized wavelet threshold function is applied, and harmonic parameters are introduced into the threshold to adapt to the time-varying characteristics of the hydraulic pump pulsation and the structural vibration background noise. Based on high sensitivity, it captures the subtle fault trends of seal leakage and loose rigid connection, and promptly discovers potential safety hazards, thereby improving the control accuracy of the jacking process.

[0041] Example 5, see Figure 1 This embodiment is based on the above embodiment. The motion time series feature construction module arranges the reconstructed signals in time series to form an original data matrix; constructs past and future vectors, normalizes them and splices them into a matrix, which is expressed as: ;in, and are the past and future sample vectors respectively; q is the number of delay steps; 、 、 and are the original vectors of the reconstructed signals at the k-1th, kqth, kth and k+q-1th moments respectively; T is the transpose operation; q is used to match different hook oscillation periods to ensure the timing consistency of feature extraction.

[0042] Example 6, see Figure 1 Based on the above embodiment, this embodiment extracts the linear dynamic characteristics of the load-current linear relationship during the lifting process, which is used to quickly detect linear drift faults caused by mechanical wear or insufficient lubrication. The covariance matrix of the past sample vector and itself, the covariance matrix of the future sample vector and itself, and the cross covariance of the future sample vector and the past sample vector are calculated respectively, which are expressed as ; Construct the scaled Hankel matrix and perform singular value decomposition, expressed as: ; where H is the scaled Hankel matrix; 、 and are the first r left singular vectors, corresponding to the future space projection basis, the first r right singular vectors, corresponding to the past space projection basis and the diagonal singular value matrix; r is the truncation rank; the projection obtains the typical variable, which is expressed as: ;in, and are the future typical variable vector and the past typical variable vector at the kth moment respectively; and They are and Normalization of ; Calculate the difference and indicators , expressed as: ; ; Where I is the identity matrix; is the sample residual vector, which represents the error of the future canonical variable based on the past canonical variable; It is a detection statistic that measures the energy of the difference vector in the weighted space. The larger it is, the more serious the deviation from normality. It is sensitive to tiny drifts of linear parameters caused by abnormal wire rope tension and provides intuitive positioning.

[0043] Example 7, see Figure 1 This embodiment is based on the above embodiment. The nonlinear disturbance separation module captures the nonlinear characteristics and non-Gaussian noise of wind load, swing and impact during the jacking process, compensating for the insufficient nonlinear detection of the linear drift monitoring module. The residual difference is constructed in the residual subspace of the linear drift monitoring module, which is expressed as: ;in, is the difference vector of the sample in the residual subspace; and are the future projection matrix and the past projection matrix of the residual subspace respectively; It is the mapping matrix of the residual subspace, which describes the linear mapping relationship between the residual of the past sample vector and the residual of the future sample vector. The kernel distance is defined as: ;in, is the difference vector in the residual subspace and the distance between them; is the Gaussian kernel width; the kernel matrix is ​​constructed for the residual vector set of the linear drift monitoring module and then centered and normalized to obtain the whitened data Z. The kernel independent component s is extracted from Z. The formula used is: ; ;in, and are the demixing vectors before and after updating, respectively; g(·) is the tanh function; is the derivative of g(·); is the expectation; the calculation index is expressed as: ; ; ;in, It is a nonlinear energy index; is the error, and the larger it is, the more abnormal disturbances that cannot be covered by the kernel independent components have occurred; is a significant independent component of s, taking the d-dimensional column vector of the first d rows and k columns of s; is the generalized inverse matrix; is a matrix composed of the first d demixed vectors extracted in rows; It is the eigenvector of the kth sample after whitening in the kernel space. Each moment corresponds to a sample. It is extremely sensitive to nonlinear disturbances such as wind swing and impact, as well as non-Gaussian noise, and fills the blind spot of linear analysis.

[0044] Example 8, see Figure 1 This embodiment is based on the above embodiment. The threshold statistical decision module uses historical normal lifting data and adopts kernel density estimation to non-parametrically determine the threshold of each indicator to ensure accuracy and robustness. : ; ;in, It is a non-parametric estimate of the probability density function of indicator J under historical normal jacking data; is the Gaussian kernel function; M is the total number of data, i is the index of historical normal lifting data; is the index value of the i-th data; x is the index value position during density estimation; h is the bandwidth; is the indicator threshold; is the confidence level; hence the indicator threshold 、 and ,correspond ; Through non-parametric estimation, it adapts to any distribution form, and the threshold is more in line with the actual operation statistical characteristics.

[0045] Example 9, see Figure 1 This embodiment is based on the above embodiment. The lifting process control module calculates the detection index for each new lifting sample. 、 and , the abnormal degree of the detection index is mapped to the warning level and corresponds to different control strategies; when any detection index exceeds the corresponding index threshold five times in a row, a fault alarm is immediately triggered; the mapping is expressed as: ; ; Where L is the warning level; and Any detection indicator exceeds the indicator threshold three times or five times in a row; is the control strategy mapping; is the current boost speed.

[0046] By performing the above operations, the general tower crane jacking process control system has the problem of difficulty in detecting linear drift faults caused by mechanical wear or insufficient lubrication, and ignores nonlinear factors such as wind load, swing and impact in the tower crane jacking process, which leads to poor control effect of the tower crane jacking process. This scheme constructs past and future vectors based on reconstructed signals, extracts the linear dynamic characteristics of the linear relationship between load and current through the linear drift monitoring module, quickly detects linear drift faults caused by mechanical wear or insufficient lubrication, and is sensitive to the tiny drift of linear parameters caused by abnormal wire rope tension; constructs residual differences in the residual subspace, calculates nonlinear energy indicators and errors, is extremely sensitive to nonlinear disturbances and non-Gaussian noise, and strengthens the blind spots of linear analysis; uses historical normal jacking data and adopts kernel density estimation to non-parametrically determine the thresholds of each indicator to achieve jacking process control, thereby improving the control effect of the tower crane jacking process.

[0047] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0048] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An intelligent detection tower crane lifting process control system, characterized by: The system includes a data acquisition module, an adaptive spectrum filtering module, an impact retention module, a motion timing feature construction module, a linear drift monitoring module, a nonlinear disturbance separation module, a threshold statistical decision module, and a lifting process control module; The data acquisition module collects the lifting signal during the tower crane lifting process and dynamically adjusts the sampling rate using an adaptive sampling mechanism based on gradient triggering; The adaptive spectrum filtering module filters the noise mode; The impact retention module applies a non-jump optimized wavelet threshold function for low signal-to-noise ratio modes, combines dynamic thresholds with harmonic parameters for threshold processing, and finally reconstructs the signal; The motion time series feature construction module arranges the reconstructed signal into an original data matrix, constructs past vectors and future vectors, and normalizes and splices them to form a time series feature matrix; The linear drift monitoring module constructs a scaled Hankel matrix based on the load current signal covariance matrix and performs singular value decomposition, and calculates the residual vector and detection statistics through canonical variable projection; The nonlinear disturbance separation module constructs and centers a kernel matrix in the residual subspace using kernel distance for whitening, and then extracts independent signal components and calculates energy indices through kernel independent component analysis; The threshold statistical decision module uses historical normal data to perform Gaussian kernel density estimation on each detection indicator and calculates the corresponding threshold according to the set confidence level; The lifting process control module implements lifting process control for each lifting sample based on detection indicators and corresponding thresholds.

2. The intelligent detection tower crane lifting process control system according to claim 1 is characterized in that: The data acquisition module collects the lifting signal during the tower crane lifting process; performs variational mode decomposition on the original lifting signal to obtain K intrinsic mode functions; adopts adaptive sampling based on gradient triggering, and when the first-order derivative of the load signal Or the current signal change rate When the threshold is exceeded, the sampling rate is automatically increased, which is expressed as: ;in, and They are the upper and lower limits of the sampling rate; is the sigmoid function; and is the normalized weight coefficient; is the trigger threshold.

3. The intelligent detection tower crane lifting process control system according to claim 2 is characterized in that: The adaptive spectrum filtering module is based on the approximate entropy Evaluate the random complexity of the eigenmodes; by permutation entropy Evaluation signal perturbation; respectively expressed as: ; ; Normalized synthetic harmonic parameters , expressed as: ; Calculate the mean P and confidence interval σ, expressed as: ; σ ;in, and are the sample sequence similarity statistics of dimension m and dimension m+1 respectively, r is the tolerance threshold, i is the modal index, are all possible permutations of symbols of length m; In the time series of the ith mode, the arrangement pattern is observed probability; and After normalization, and ; K is the total number of modes; for The high noise mode corresponds to the hydraulic pump pulsation or motor switching interference and is discarded; The high signal-to-noise ratio modes corresponding to the main motion displacement and structural resonance are retained; The low signal-to-noise ratio mode is used to perform signal denoising.

4. The intelligent detection tower crane lifting process control system according to claim 3 is characterized in that: The impact retention module applies the optimized wavelet threshold function to perform signal denoising for modes determined to have a low signal-to-noise ratio, and introduces harmonic parameters into the threshold. The formula used is: ; ;in, It is the output of the optimized wavelet threshold function; is the original wavelet coefficient, j is the wavelet decomposition series index, and u is the position index; is the threshold parameter; sign(·) is the sign function; is the standard deviation of the background noise at the jth level of the wavelet; N is the signal length; is the threshold adjustment factor; the high signal-to-noise ratio mode and the low signal-to-noise ratio mode are reconstructed and expressed as: ;in, is the reconstructed signal, is the signal component of the i-th eigenmode at time k.

5. The intelligent detection tower crane lifting process control system according to claim 4 is characterized in that: The motion time series feature construction module arranges the reconstructed signals in time series to form an original data matrix; constructs past and future vectors, normalizes them and splices them into a matrix, which is expressed as: ;in, and are the past and future sample vectors respectively; q is the number of delay steps; 、 、 and are the original vectors of the reconstructed signals at the k-1th, kqth, kth and k+q-1th moments respectively.

6. The intelligent detection tower crane lifting process control system according to claim 5, characterized in that: The linear drift monitoring module calculates the covariance matrix between the past sample vector and itself, the covariance matrix between the future sample vector and itself, and the cross covariance between the future sample vector and the past sample vector, which are expressed as ; Construct the scaled Hankel matrix and perform singular value decomposition, expressed as: ; where H is the scaled Hankel matrix; 、 and are the first r left singular vectors, corresponding to the future space projection basis, the first r right singular vectors, corresponding to the past space projection basis and the diagonal singular value matrix; r is the truncation rank; the projection obtains the typical variable, which is expressed as: ;in, and are the future typical variable vector and the past typical variable vector at the kth moment respectively; and They are and Normalization of ; Calculate the difference and indicators , expressed as: ; ; Where I is the identity matrix; is the sample residual vector; is the test statistic.

7. The intelligent detection tower crane lifting process control system according to claim 6, characterized in that: The nonlinear disturbance separation module constructs the residual difference in the residual subspace of the linear drift monitoring module, which is expressed as: ;in, is the difference vector of the sample in the residual subspace; and are the future projection matrix and the past projection matrix of the residual subspace respectively; It is the mapping matrix of the residual subspace, which describes the linear mapping relationship between the residual of the past sample vector and the residual of the future sample vector. The kernel distance is defined as: ;in, is the difference vector in the residual subspace and the distance between them; is the Gaussian kernel width; the kernel matrix is ​​constructed for the residual vector set of the linear drift monitoring module and then centered and normalized to obtain the whitened data Z. The kernel independent component s is extracted from Z. The formula used is: ; ;in, and are the demixing vectors before and after updating, respectively; g(·) is the tanh function; is the derivative of g(·); is the expectation; the calculation index is expressed as: ; ; ;in, It is a nonlinear energy index; is the error; is a significant independent component of s, taking the d-dimensional column vector of the first d rows and k columns of s; is the generalized inverse matrix; is a matrix composed of the first d demixed vectors extracted in rows; is the eigenvector of the kth sample after whitening in the kernel space.

8. The intelligent detection tower crane lifting process control system according to claim 7 is characterized in that: The threshold statistical decision module uses historical normal jacking data and adopts kernel density estimation to non-parametrically determine the threshold of each indicator to ensure accuracy and robustness; : ; ;in, It is a non-parametric estimate of the probability density function of indicator J under historical normal jacking data; is the Gaussian kernel function; M is the total number of data, i is the index of historical normal lifting data; is the index value of the i-th data; x is the index value position during density estimation; h is the bandwidth; is the indicator threshold; is the confidence level; hence the indicator threshold 、 and ,correspond .

9. The intelligent detection tower crane lifting process control system according to claim 8, characterized in that: The lifting process control module calculates the detection index for each new lifting sample 、 and , the abnormal degree of the detection index is mapped to the warning level and corresponds to different control strategies; when any detection index exceeds the corresponding index threshold five times in a row, a fault alarm is triggered immediately.

Images