Early warning method for hoisting irregular structures based on multi-angle IMU and time-frequency dynamic constraints

Abstract

This invention discloses an early warning method for hoisting irregular structures based on multi-angle IMU and time-frequency dynamic constraints, belonging to the field of safety monitoring technology for hoisting large components and irregular structures. The method involves deploying inertial measurement units at at least four preset locations on the irregular structure being hoisted, and accelerometers, strain sensors, and displacement sensors at key stress points to collect Euler angles, acceleration, angular velocity, strain, and displacement data in real time. The collected data is preprocessed synchronously to construct attitude-derived features and time-frequency derived features. The hoisting stage is identified based on the attitude-derived features. Dynamic constraints are generated based on the hoisting stage identification results and the finite element dynamic model. A comprehensive risk coefficient is calculated based on the attitude-derived features, time-frequency derived features, and dynamic constraints, and the risk type is determined. A graded early warning and handling suggestions are then output. This method can improve the real-time performance, accuracy, and interpretability of safety early warning throughout the entire hoisting process of irregular structures.

Description

Technical Field

[0001] This invention belongs to the field of safety monitoring technology for hoisting construction of large components and irregular structures, and relates to an early warning method for hoisting irregular structures based on multi-angle IMU and time-frequency dynamic constraints. Background Technology

[0002] Large bridge segments, building steel structures, equipment shells, and irregularly shaped components in hydropower projects often exhibit irregular geometric shapes, off-center centers of gravity, complex lifting point arrangements, and significant local stiffness differences during hoisting. If the hoisted structure experiences attitude instability, amplified local vibrations, excessive strain, or uncoordinated forces at the lifting points during hoisting, horizontal transport, fine-tuning in position, and locking, it may lead to component collisions, sling slippage, single-point unhooking, local buckling, or even a complete collapse.

[0003] Existing hoisting monitoring methods mainly include manual visual observation, single-point theodolite or total station measurement, sling tension monitoring, and single displacement or strain monitoring. These methods have the following shortcomings: First, manual experience-based judgment is highly subjective and it is difficult to continuously capture transient changes throughout the hoisting process; second, single-point measurement is difficult to characterize the attitude coordination relationship between the four corners or multiple hoisting points of irregular structures, and cannot effectively identify whether the hoisted structure maintains a rigid body attitude; third, relying solely on displacement, tension, or strain thresholds is insufficient to reflect the time-frequency changes of vibration, impact, rotation, and local dynamic response; fourth, sensors may detach, have their links interrupted, or become loose at the hoisting site, and existing methods are prone to misjudging measurement point failures as overall structural instability, or misjudging actual instability as sensor malfunction. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to solve the problems of traditional hoisting monitoring, such as difficulty in quantifying attitude throughout the entire process, difficulty in determining the synchronization of four corners, difficulty in capturing dynamic time-frequency risks, and difficulty in distinguishing between actual structural anomalies and measurement point faults. It provides a hoisting early warning method for irregular structures based on multi-angle IMU and time-frequency dynamic constraints.

[0005] Technical solution: The present invention provides an early warning method for hoisting irregularly shaped structures based on multi-angle IMU and time-frequency dynamic constraints, comprising the following steps:

[0006] S1. Establish a hoisting operation coordinate system, install inertial measurement units at at least four preset positions on the hoisted irregular structure, collect triaxial acceleration, triaxial angular velocity data, roll angle, pitch angle, and yaw angle data at each measuring point in real time, and install vibration acceleration sensors, strain sensors, and displacement sensors at key stress-bearing parts of the irregular structure to collect vibration acceleration, strain, and displacement data at each measuring point in real time.

[0007] S2. Perform synchronous preprocessing on the data collected in step S1 to obtain cleaned synchronous data;

[0008] S3. Based on the cleaned synchronization data, construct attitude-derived features and time-frequency-derived features. The attitude-derived features include the tilt angle, acceleration amplitude, angular velocity amplitude, cumulative heading angle change, and four-corner attitude consistency index of each inertial measurement unit measurement point. The time-frequency-derived features include the key frequencies, amplitudes, and energy proportions of vibration acceleration data, strain data, and displacement data.

[0009] S4. Identify the current hoisting stage based on angular velocity amplitude, tilt angle change rate, and four-corner attitude consistency index;

[0010] S5. Based on the hoisting stage, construct a finite element dynamic model of the hoisted irregular structure, input the time-frequency derived features into the finite element dynamic model, correct the model parameters, and obtain the dynamic constraint conditions corresponding to the current hoisting stage.

[0011] S6. Calculate the comprehensive risk coefficient based on the attitude-derived features, time-frequency-derived features and dynamic constraints, and distinguish the overall attitude anomaly, local dynamic response anomaly and measurement point failure anomaly by using the four-corner attitude consistency index and the measurement point reliability risk item.

[0012] S7. Output graded early warnings based on the comprehensive risk coefficients, and generate corresponding response measures based on the risk type of the early warning.

[0013] Furthermore, the at least four preset positions mentioned in step S1 include the four corner points, four hanging points, or four boundary representative positions of the irregular structure.

[0014] Furthermore, the synchronization preprocessing in step S2 includes: converting the absolute time of each measuring point into a relative time based on the first valid record of each measuring point; performing continuous unfolding processing on the heading angle output by the inertial measurement unit to obtain continuous heading angles; and performing outlier removal and filtering smoothing on the unfolded data to obtain cleaned synchronization data.

[0015] Furthermore, the continuous unfolding process for the heading angle includes: calculating the heading angle difference Δi=yaw between adjacent sampling points. i -yaw i-1 When Δ i When the angle is greater than +180°, the compensation amount will be expanded. i Updated to o i-1 -360°; when Δ i When the angle is less than -180°, the compensation amount will be expanded. i Updated to o i-1 +360°; otherwise keep o i =o i-1 ; and with yaw unwrap_i =yaw i +oi As a continuous heading angle sequence; where, Δ i Let yaw be the heading angle difference at the i-th sampling point. i Let yaw be the original heading angle of the i-th sampling point. i-1 Let o be the original heading angle of the (i-1)th sampling point. i o is the expansion compensation amount for the i-th sampling point. i-1 yaw is the expansion compensation amount for the (i-1)th sampling point. unwrap_i Let be the continuous heading angle of the i-th sampling point.

[0016] Furthermore, the outlier removal includes: calculating the mean μ and standard deviation σ for each column of triaxial acceleration, triaxial angular velocity, roll angle, pitch angle, and continuous heading angle for each measuring point; when any channel data exceeds [μ-3σ, μ+3σ], the entire row of data corresponding to that moment is marked as an outlier sample and removed.

[0017] Furthermore, the formula for calculating the tilt angle in step S3 is as follows: ,in The tilt angle is used to characterize the levelness of the suspended structure. This is the roll angle. The pitch angle is given; the formula for calculating the acceleration amplitude is: ,in The acceleration amplitude is used to characterize the tendency of impact, sudden stop, or free fall. It is a triaxial acceleration; the formula for calculating the amplitude of the angular velocity is: ,in The angular velocity amplitude is used to characterize the intensity of swaying and rotation. The angular velocity is the three-axis velocity; the formula for calculating the cumulative change in heading angle is: Δyaw = yaw unwrap_end -yaw unwrap_start Where Δyaw is the cumulative change in heading angle, used to characterize the cumulative change in direction during the lifting phase; yaw unwrap_end The continuous heading angle at the end of the current phase; yaw unwrap_start The continuous heading angle at the start time.

[0018] Further, the four-corner attitude consistency index in step S3 includes calculating the root mean square of the tilt angle difference between the remaining inertial measurement unit (IMU) measurement points and the reference measurement point at the synchronization moment, using one of the IMU measurement points as the reference measurement point; or calculating the dispersion of the tilt angle, angular velocity amplitude, and continuous heading angle changes of multiple IMU measurement points. When at least three measurement points remain consistent while one measurement point deviates significantly, it is determined to be a candidate single measurement point fault; when multiple measurement points deviate simultaneously and the direction or trend of deviation satisfies the rigid body motion relationship of the structure, it is determined to be a candidate overall attitude anomaly.

[0019] Furthermore, the time-frequency derived features mentioned in step S3 are obtained through short-time Fourier transform, including: using the Hanning window function to window the preprocessed vibration acceleration, strain, and displacement data; performing a discrete Fourier transform on each time window; sliding the time window according to a preset overlap rate to form the vibration acceleration time spectrum matrix, strain time spectrum matrix, and displacement time spectrum matrix respectively; and extracting key frequencies, vibration acceleration amplitude, and energy percentage from each time spectrum matrix.

[0020] Furthermore, the length of the Hanning window is 128 to 512 points, preferably 256 points; the overlap rate of adjacent windows is 25% to 75%, preferably 50%; and the data sampling frequency is not less than 10Hz, preferably 100Hz.

[0021] Further, the hoisting stage in step S4 includes at least a pre-hoisting stationary stage, a main hoisting movement stage, and a placement stationary stage; the main hoisting movement stage includes at least one sub-stage of hoisting, horizontal movement, pre-positioning fine-tuning, placement locking, and unhooking; the hoisting stage identification includes: searching forward from the head of the time series for sample segments that continuously satisfy the angular velocity amplitude being less than the stationary angular velocity threshold as the pre-hoisting stationary stage; searching backward from the tail of the time series for sample segments that continuously satisfy the angular velocity amplitude being less than the stationary angular velocity threshold as the placement stationary stage; the period between the two is the main hoisting movement stage; wherein the stationary angular velocity threshold is 0.3° / s to 2° / s, preferably 1.5° / s.

[0022] Furthermore, during the positioning and locking phase, if the tilt angle is consistently less than 2°, the angular velocity amplitude is consistently less than 0.3° / s, and the duration is not less than 3 minutes, the suspended irregular structure is determined to have entered a stable positioning state; otherwise, a positioning fine-tuning prompt is output.

[0023] Furthermore, the dynamic constraints in step S5 include at least one of the following constraints associated with the hoisting stage: tilt angle constraint, angular velocity constraint, acceleration deviation constraint, cumulative heading angle constraint, attitude consistency constraint, vibration acceleration amplitude constraint, frequency constraint, energy proportion constraint, strain constraint, and displacement constraint.

[0024] Furthermore, the comprehensive risk coefficient in step S6 is calculated using the following formula:

[0025] R=αR p +βR f +γR h Where R is the comprehensive risk coefficient; R pThe overall attitude risk term is calculated from the normalized values ​​of the bank angle, angular velocity amplitude, acceleration amplitude, cumulative heading angle change, and four-corner attitude consistency index relative to the bank angle constraint, angular velocity constraint, acceleration deviation constraint, cumulative heading angle constraint, and attitude consistency constraint, respectively; R f The local time-frequency response risk term is calculated from the normalized values ​​of the key frequency, vibration acceleration amplitude, energy proportion, strain, and displacement relative to the frequency constraint, vibration acceleration amplitude constraint, energy proportion constraint, strain constraint, and displacement constraint, respectively; R h The reliability risk term for the measurement point is calculated from the data interruption duration of the measurement point of the inertial measurement unit, the abrupt change amplitude of a single measurement point, and the degree of consistency deviation between a single measurement point and the other measurement points; α, β, and γ are weights, and α+β+γ=1.

[0026] Furthermore, when R h If the tilt angle of the measuring point exceeds the fault threshold and the tilt angle of at least three other inertial measurement unit measuring points is less than the overall attitude threshold, while the four-corner attitude consistency index remains within the attitude consistency constraint range, the anomaly is determined to be a measuring point fault, and a warning message for re-fixing or replacing the corresponding sensor is output; when R p When the corresponding warning threshold is exceeded and multiple measuring points are synchronously abnormal, the abnormality is judged as an overall attitude abnormality; when R f When the corresponding warning threshold is exceeded and multiple measuring points are synchronously abnormal, the abnormality is judged as a local dynamic response abnormality.

[0027] Furthermore, step S7 specifically includes:

[0028] When the comprehensive risk coefficient R is less than the first threshold, a low-risk warning is output, prompting a reduction in lifting speed and enhanced monitoring.

[0029] When R is greater than or equal to the first threshold and less than the second threshold, a risk warning is output, prompting the hoisting to be suspended and the hoisting points, slings and key structural parts to be inspected.

[0030] When R is greater than or equal to the second threshold, a high-risk warning is output, prompting the work to be stopped, the danger zone to be evacuated, and the emergency response procedure to be initiated.

[0031] Both the first threshold and the second threshold are preset values ​​or dynamically adjusted values.

[0032] Beneficial effects: (1) By synchronously monitoring at least four corners of the distributed inertial measurement units (IMUs), the levelness, sway intensity, impact degree, cumulative rotation of direction and the consistency of the four corners of the suspended irregular structure can be continuously quantified, making up for the deficiency of single-point monitoring in expressing the overall attitude;

[0033] (2) By continuously expanding the heading angle, the statistical threshold is not lost due to the jump of Euler angle at ±180°, so that the cumulative rotation and anomaly identification are more consistent with the physical process of hoisting;

[0034] (3) By using the four-corner attitude consistency and measurement point reliability risk criteria, it is possible to identify individual IMU detachment, link interruption or installation failure, thereby reducing the risk of misreporting measurement point failure as structural instability;

[0035] (4) By combining the time-frequency characteristics of short-time Fourier transform (STFT) with the dynamic constraints of finite element, the risks of vibration, strain and displacement can be associated with structural design, construction conditions and hoisting stage, thus avoiding the problem that fixed thresholds cannot adapt to the differences in different stages and irregular structures.

[0036] (5) By outputting the comprehensive risk coefficient and risk type, on-site personnel can not only know the warning level, but also know whether the source of the anomaly is an overall attitude anomaly, a local dynamic response anomaly or a measurement point fault anomaly, so as to take targeted measures. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method of the present invention;

[0038] Figure 2 This is a schematic diagram of the continuous unfolding processing of the heading angle according to the present invention;

[0039] Figure 3 The images show a comparison of the heading angle before and after unwrap processing, where (a) is the original data image and (b) is the data image after unwrap processing. Detailed Implementation

[0040] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0041] like Figure 1 As shown, this embodiment provides a method for early warning of hoisting of irregular structures based on multi-angle IMU and time-frequency dynamic constraints, including:

[0042] (1) Sensor placement and data acquisition:

[0043] On a hoisted irregular structure, a three-dimensional coordinate system is established with a fixed point in the hoisting operation plane as the origin, a reference direction in the operation plane as the x-axis, a horizontal direction perpendicular to the x-axis as the y-axis, and the hoisting height direction as the z-axis. Measuring points are located near at least four corner points or representative hoisting points of the irregular structure. Nine-axis inertial measurement units (IMUs), including a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, are installed at each measuring point. Each IMU outputs a three-axis acceleration. Triaxial angular velocity And roll angle, pitch angle and yaw angle.

[0044] Vibration acceleration sensors, strain sensors, and displacement sensors are installed at key stress-bearing parts of the irregular structure. In specific implementations, key stress-bearing parts include lugs, webs, flanges, stress concentration areas, support transition parts, or large displacement sensitive parts.

[0045] In specific implementations, the data acquisition frequency can be from 10Hz to 100Hz, preferably 100Hz; when used only for attitude accuracy evaluation, the IMU sampling frequency should not be lower than 10Hz. All data is transmitted to the data analysis terminal via a wireless gateway, Bluetooth, serial port, or industrial Ethernet.

[0046] (2) Data synchronization preprocessing:

[0047] Since each IMU and sensor may have initial clock offsets and sampling frequency differences, this embodiment uses the first valid record time T0 of each measurement point as a reference and sets the timestamp T of the i-th record. i Convert to relative time t i =(T i -T0) / s, while retaining millisecond-level time accuracy. For features requiring synchronous comparison across measurement points, interpolation or nearest neighbor time matching methods are used to align multi-point data to a unified time axis.

[0048] The yaw angle in Euler angles is a periodic angle with a 360° jump at ±180°. Directly using the original yaw for outlier removal and cumulative rotation calculations can lead to misjudgments. This embodiment employs a continuous unfolding algorithm to obtain the yaw. unwrap Continuous sequence, and in yaw unwrap Used for accumulating changes in heading angle and detecting anomalies. For example... Figure 2 As shown, the continuous unfolding process for the heading angle includes: calculating the heading angle difference Δi=yaw between adjacent sampling points. i -yaw i-1 When Δ i When the angle is greater than +180°, the compensation amount will be expanded. i Updated to o i-1 -360°; when Δ i When the angle is less than -180°, the compensation amount will be expanded. i Updated to o i-1 +360°; otherwise keep o i =o i-1 ; and with yaw unwrap_i =yaw i +o i As a continuous heading angle sequence; where, Δ i Let yaw be the heading angle difference at the i-th sampling point.i Let yaw be the original heading angle of the i-th sampling point. i-1 Let o be the original heading angle of the (i-1)th sampling point. i o is the expansion compensation amount for the i-th sampling point. i-1 yaw is the expansion compensation amount for the (i-1)th sampling point. unwrap_i Let be the continuous heading angles of the i-th sampling point. For example... Figure 3 As shown, the original heading angle data has a jump at the boundary. After continuous unfolding processing by the present invention, a continuous heading angle sequence is obtained, which eliminates the jump and can truly reflect the cumulative rotation angle of the suspended structure.

[0049] For triaxial acceleration, triaxial angular velocity, roll angle, pitch angle and yaw unwrap 3σ outlier identification is performed column by column. If any channel exceeds the range [μ-3σ, μ+3σ] at any given moment, the entire row at that moment is marked as an outlier and removed. Samples with empty timestamps are also removed. It should be noted that the purpose of 3σ processing is not to delete real impact events, but to remove obvious acquisition anomalies and non-physical jumps. Therefore, the impact, rollover, and sensor detachment features that can be identified after cleaning are still retained in the result analysis. Further, low-pass filtering and moving average filtering can be used to remove high-frequency noise. The low-pass filter cutoff frequency fc can be taken as 1.5 to 2 times the upper limit of the main vibration frequency of the hoisting, and the moving window M can be taken as 5 to 15.

[0050] (3) Calculation of attitude-derived features:

[0051] The tilt angle, acceleration amplitude, angular velocity amplitude, cumulative heading angle change, and four-corner attitude consistency index of each inertial measurement unit measurement point are calculated based on the cleaned data. The calculation method is as follows:

[0052] The formula for calculating the tilt angle is: ,in The tilt angle is used to characterize the levelness of the suspended structure. This is the roll angle. The pitch angle;

[0053] The formula for calculating the acceleration amplitude is: ,in The acceleration amplitude is used to characterize the tendency of impact, sudden stop, or free fall. It is a triaxial acceleration;

[0054] The formula for calculating the amplitude of angular velocity is: ,in The angular velocity amplitude is used to characterize the intensity of swaying and rotation. () represents the angular velocity of the three axes;

[0055] The formula for calculating the cumulative change in heading angle is: Δyaw = yaw unwrap_end -yaw unwrap_start Where Δyaw is the cumulative change in heading angle, used to characterize the cumulative change in direction during the lifting phase; yaw unwrap_end The continuous heading angle at the end of the current phase; yaw unwrap_start The continuous heading angle at the start time.

[0056] The consistency of attitude at four corners can be calculated as follows: Taking the installation of inertial measurement units (IMUs) at four measuring points as an example, using one IMU measuring point as the reference measuring point, square the difference in tilt angle at the other three IMU measuring points at the synchronization moment, take the average, and then take the square root to obtain the RMS consistency index. Alternatively, the standard deviations of tilt angle, angular velocity amplitude, and cumulative changes in heading angle at the four IMU measuring points can be weighted. The smaller the index, the more synchronized the attitude at four corners, and the closer the suspended structure is to a rigid body stable state.

[0057] When the tilt angles of the three measuring points are all less than 8° and the trend of change is consistent, while one measuring point suddenly changes to more than 75° in a short period of time and maintains a high tilt angle, and its angular velocity or heading angle shows independent abnormality, it is determined that a single measuring point has fallen off or failed to be installed; when multiple measuring points simultaneously show an increase in tilt angle, an increase in angular velocity, an acceleration deviation of 1g, and a synchronous deterioration in the consistency of the four corners, it is determined that the overall attitude is abnormal.

[0058] (4) Identification of the hoisting stage:

[0059] This embodiment sets the static angular velocity threshold |ω| quiet The value is 1.5° / s. Starting from the head of the data sequence, search in the positive direction for consecutive values ​​satisfying |ω| < |ω|. quiet The sample is used as the static placement stage before lifting; the search continues from the end of the sequence in reverse order, continuously satisfying |ω|<|ω|. quiet The sample is used as the initial placement and static stage; the main hoisting motion stage is between the two. For hoisting operations with high control precision requirements, the main hoisting motion stage can be further divided into five sub-stages based on the convergence trend of the tilt angle and the amplitude of the angular velocity: lifting, horizontal movement, fine-tuning before placement, placement and locking, and unhooking.

[0060] During the positioning and locking phase, if the tilt angle remains less than 2° and the angular velocity amplitude remains less than 0.3° / s for at least 3 minutes, the positioning is considered stable; if the tilt angle does not converge or the angular velocity continues to fluctuate, a positioning fine-tuning prompt will be output.

[0061] (5) Time-frequency derived feature extraction:

[0062] Short-time Fourier transform was applied to the vibration acceleration, strain, and displacement data. The Hanning window function was selected as w(n) = 0.5 - 0.5cos(2πn / (N)). w -1), where n=0,1,...,N w -1, N w The window length is preferably 256 points. A 50% overlap rate is set, and the preprocessed data is segmented according to the time window. Each segment is multiplied by the Hanning window and then subjected to a Discrete Fourier Transform to obtain the spectrum at different times. These spectra are then used to construct the spectrum matrix STFTa(t,f) for vibration acceleration, the spectrum matrix STFTε(t,f) for strain, and the spectrum matrix STFTd(t,f) for displacement, respectively.

[0063] Key frequencies, peak amplitudes, band energy proportions, and frequency drift characteristics for each stage are extracted from the time-frequency matrix. The vibration acceleration time-frequency spectrum can be used to identify shock and abnormal vibration sources, strain time-frequency characteristics can be used to identify abrupt changes in local force, and displacement time-frequency characteristics can be used to identify low-frequency large displacements and gradual instability trends.

[0064] (6) Generation of dynamic constraints for finite element methods:

[0065] A finite element dynamic model is established based on the irregular structure design drawings, material properties, hoisting scheme, sling arrangement, hoisting point locations, sling stiffness, boundary conditions, and hoisting stages. Key frequencies obtained from time-frequency analysis are input as external excitation parameters. Vibration amplitude and energy percentage are used to correct damping parameters. Measured strain and displacement are compared with the model calculation results along the time and frequency dimensions. When the model calculation results are inconsistent with the measured data, material parameters, connection methods, boundary conditions, or equivalent sling stiffness are corrected until the model response and field data meet the preset error requirements.

[0066] The finite element dynamic model outputs dynamic constraints for each hoisting stage, including upper limits for tilt angle, angular velocity, acceleration deviation, cumulative heading angle, four-corner consistency, critical frequency amplitude, strain, displacement, and energy percentage. Compared to fixed thresholds, dynamic constraints can reflect the risk tolerance differences for different irregular structures and different hoisting stages.

[0067] (7) Comprehensive risk calculation and early warning output:

[0068] The overall risk coefficient R is derived from the overall attitude risk term R. p Local time-frequency response risk item R f and measurement point reliability risk item R h Composition. Overall attitude risk term R pIt can be calculated from the tilt angle, angular velocity amplitude, acceleration amplitude deviation, cumulative heading angle change, and the normalized values ​​of the four-corner attitude consistency index relative to the corresponding dynamic constraints; the local time-frequency response risk term R f It can be obtained by weighting the normalized results of key frequencies, amplitudes, energy proportions, strains, and displacements relative to dynamic constraints; the measurement point reliability risk term R h It can be calculated from the duration of data interruption, sampling missing rate, abrupt change amplitude of a single measurement point, and the degree to which a single measurement point deviates from the other measurement points.

[0069] When the comprehensive risk coefficient R is less than the first threshold, a low-risk warning is output, prompting a reduction in lifting speed and enhanced monitoring.

[0070] When R is greater than or equal to the first threshold and less than the second threshold, a risk warning is output, prompting the hoisting to be suspended and the hoisting points, slings and key structural parts to be inspected.

[0071] When R is greater than or equal to the second threshold, a high-risk warning is output, prompting the work to be stopped, the danger zone to be evacuated, and the emergency response procedure to be initiated.

[0072] Both the first and second thresholds are preset values ​​or dynamically adjusted values. In one specific implementation, the first threshold is 1.5 and the second threshold is 2.0. That is, when R is in [1.0, 1.5), a low-risk warning is output, suggesting reducing the lifting speed and strengthening observation; when R is in [1.5, 2.0), a medium-risk warning is output, suggesting suspending the lifting and checking the lifting points, slings, connection parts, and key structural parts; when R ≥ 2.0, a high-risk warning is output, suggesting stopping the operation, evacuating the danger zone, and activating the emergency plan.

[0073] When R h When the fault threshold for the measurement point is exceeded and at least three other IMU measurement points remain stable and consistent, the system outputs a measurement point fault warning, such as a suspected detachment of measurement point P4; please check the fixation status or replace the sensor. When R... p or R f When the threshold is exceeded and multiple measuring points are synchronously abnormal, the system outputs a structural risk warning, such as an abnormal overall posture during the main hoisting movement, and recommends that the hoisting be stopped immediately.

Claims

1. A method for early warning of hoisting irregular structures based on multi-angle IMU and time-frequency dynamic constraints, characterized in that, Includes the following steps: S1. Establish a hoisting operation coordinate system, install inertial measurement units at at least four preset positions on the hoisted irregular structure, collect triaxial acceleration, triaxial angular velocity data, roll angle, pitch angle, and yaw angle data at each measuring point in real time, and install vibration acceleration sensors, strain sensors, and displacement sensors at key stress-bearing parts of the irregular structure to collect vibration acceleration, strain, and displacement data at each measuring point in real time. S2. Perform synchronous preprocessing on the data collected in step S1 to obtain cleaned synchronous data; S3. Based on the cleaned synchronization data, construct attitude-derived features and time-frequency-derived features. The attitude-derived features include the tilt angle, acceleration amplitude, angular velocity amplitude, cumulative heading angle change, and four-corner attitude consistency index of each inertial measurement unit measurement point. The time-frequency-derived features include the key frequencies, amplitudes, and energy proportions of vibration acceleration data, strain data, and displacement data. S4. Identify the current hoisting stage based on angular velocity amplitude, tilt angle change rate, and four-corner attitude consistency index; S5. Based on the hoisting stage, construct a finite element dynamic model of the hoisted irregular structure, input the time-frequency derived features into the finite element dynamic model, correct the model parameters, and obtain the dynamic constraint conditions corresponding to the current hoisting stage. S6. Calculate the comprehensive risk coefficient based on the attitude-derived features, time-frequency-derived features and dynamic constraints, and distinguish the overall attitude anomaly, local dynamic response anomaly and measurement point failure anomaly by using the four-corner attitude consistency index and the measurement point reliability risk item. S7. Output graded early warnings based on the comprehensive risk coefficients, and generate corresponding response measures based on the risk type of the early warning.

2. The early warning method according to claim 1, characterized in that, The synchronization preprocessing in step S2 includes: converting the absolute time of each measuring point into a relative time based on the first valid record of each measuring point; performing continuous unfolding processing on the heading angle output by the inertial measurement unit to obtain continuous heading angles; performing outlier removal and filtering smoothing on the triaxial acceleration, triaxial angular velocity, roll angle, pitch angle, and continuous heading angle of each measuring point to obtain cleaned synchronization data; the continuous unfolding processing of the heading angle includes: calculating the heading angle difference Δi=yaw between adjacent sampling points. i -yaw i-1 When Δ i When the angle is greater than +180°, the compensation amount will be expanded. i Updated to o i-1 -360°; when Δ i When the angle is less than -180°, the compensation amount will be expanded. i Updated to o i-1 +360°; otherwise keep o i =o i-1 ; and with yaw unwrap_i =yaw i +o i As a continuous heading angle sequence; where, Δ i Let yaw be the heading angle difference at the i-th sampling point. i Let yaw be the original heading angle of the i-th sampling point. i-1 Let o be the original heading angle of the (i-1)th sampling point. i o is the expansion compensation amount for the i-th sampling point. i-1 yaw is the expansion compensation amount for the (i-1)th sampling point. unwrap_i Let be the continuous heading angle of the i-th sampling point.

3. The early warning method according to claim 2, characterized in that, The formula for calculating the cumulative change in heading angle in step S3 is: Δyaw = yaw unwrap_end -yaw unwrap_start Where Δyaw is the cumulative change in heading angle, used to characterize the cumulative change in direction during the lifting phase; yaw unwrap_end The continuous heading angle at the end of the current phase; yaw unwrap_start The continuous heading angle at the start time.

4. The early warning method according to claim 1, characterized in that, The four-corner attitude consistency index mentioned in step S3 includes taking one inertial measurement unit (IMU) measurement point as the reference measurement point and calculating the root mean square of the tilt angle difference between the other IMU measurement points and the reference measurement point at the synchronization moment, or calculating the dispersion of the tilt angle, angular velocity amplitude and continuous heading angle change of multiple IMU measurement points.

5. The early warning method according to claim 1, characterized in that, The time-frequency derived features mentioned in step S3 are obtained through short-time Fourier transform, including: using the Hanning window function to window the preprocessed vibration acceleration, strain, and displacement data; performing a discrete Fourier transform on each time window; sliding the time window according to a preset overlap rate to form the vibration acceleration time spectrum matrix, strain time spectrum matrix, and displacement time spectrum matrix respectively; and extracting key frequencies, vibration acceleration amplitude, and energy percentage from each time spectrum matrix.

6. The early warning method according to claim 1, characterized in that, The hoisting stage in step S4 includes at least a pre-hoisting stationary stage, a main hoisting movement stage, and a placement stationary stage; the main hoisting movement stage includes at least one sub-stage of hoisting, horizontal movement, pre-positioning fine-tuning, placement locking, and unhooking; the hoisting stage identification includes: searching forward from the beginning of the time series for sample segments that continuously satisfy the angular velocity amplitude being less than the stationary angular velocity threshold as the pre-hoisting stationary stage; searching backward from the end of the time series for sample segments that continuously satisfy the angular velocity amplitude being less than the stationary angular velocity threshold as the placement stationary stage; the period between the two is the main hoisting movement stage; wherein the stationary angular velocity threshold is 0.3° / s to 2° / s.

7. The early warning method according to claim 1, characterized in that, The dynamic constraints in step S5 include at least one of the following constraints associated with the hoisting stage: tilt angle constraint, angular velocity constraint, acceleration deviation constraint, cumulative heading angle constraint, attitude consistency constraint, vibration acceleration amplitude constraint, frequency constraint, energy proportion constraint, strain constraint, and displacement constraint.

8. The method according to claim 7, characterized in that, The comprehensive risk coefficient in step S6 is calculated using the following formula: R=αR p +βR f +γR h Where R is the comprehensive risk coefficient; R p The overall attitude risk term is calculated from the normalized values ​​of the bank angle, angular velocity amplitude, acceleration amplitude, cumulative heading angle change, and four-corner attitude consistency index relative to the bank angle constraint, angular velocity constraint, acceleration deviation constraint, cumulative heading angle constraint, and attitude consistency constraint, respectively; R f The local time-frequency response risk term is calculated from the normalized values ​​of the key frequency, vibration acceleration amplitude, energy proportion, strain, and displacement relative to the frequency constraint, vibration acceleration amplitude constraint, energy proportion constraint, strain constraint, and displacement constraint, respectively; R h The reliability risk term for the measurement point is calculated from the data interruption duration of the measurement point of the inertial measurement unit, the abrupt change amplitude of a single measurement point, and the degree of consistency deviation between a single measurement point and the other measurement points; α, β, and γ are weights, and α+β+γ=1.

9. The early warning method according to claim 8, characterized in that, When R h If the tilt angle of the measuring point exceeds the fault threshold and the tilt angle of at least three other inertial measurement unit measuring points is less than the overall attitude threshold, while the four-corner attitude consistency index remains within the attitude consistency constraint range, the anomaly is determined to be a measuring point fault, and a warning message for re-fixing or replacing the corresponding sensor is output; when R p When the corresponding warning threshold is exceeded and multiple measuring points are synchronously abnormal, the abnormality is judged as an overall attitude abnormality; when R f When the corresponding warning threshold is exceeded and multiple measuring points are synchronously abnormal, the abnormality is judged as a local dynamic response abnormality.

10. The early warning method according to claim 1, characterized in that, Step S7 specifically includes: When the comprehensive risk coefficient R is less than the first threshold, a low-risk warning is output, prompting a reduction in lifting speed and enhanced monitoring. When R is greater than or equal to the first threshold and less than the second threshold, a risk warning is output, prompting the hoisting to be suspended and the hoisting points, slings and key structural parts to be inspected. When R is greater than or equal to the second threshold, a high-risk warning is output, prompting the work to be stopped, the danger zone to be evacuated, and the emergency response procedure to be initiated. Both the first threshold and the second threshold are preset values ​​or dynamically adjusted values.

Images