A bridge health monitoring ship collision alarm method

Abstract

The present invention relates to the field of ship collision alarm of bridge health monitoring system, and more specifically, to a ship collision alarm method for bridge structure health monitoring based on vibration acceleration response signal, which specifically comprises the following steps: S1: adopting a sliding window algorithm to segmentally intercept an acceleration signal to obtain a corresponding acceleration response subsequence; S2: adopting a natural excitation technology to extract a vibration response curve of the acceleration subsequence; S3: using Hilbert change to extract the envelope of the vibration response signal, and then fitting the envelope; S4: determining the signal type through an attenuation function, and on this basis, calculating the Mahalanobis distance mean ratio under different states; S5: using a Chauville method to determine the abnormal value of the Mahalanobis distance mean ratio of the acceleration vibration response at different bridge pier tops, and determining the ship collision time and position according to the time and size of the abnormal value.

Description

Technical Field

[0001] The present invention relates to the field of bridge ship collision alarm, and more specifically, to a bridge health monitoring ship collision alarm method. Background Art

[0002] With the rapid development of bridges and water transport in my country, the volume of water traffic is increasing, and ships are constantly developing towards large-scale and high-speed. However, most of the existing bridges are existing bridges built in the last century. Due to the restrictions on net width and height, they face an increasingly prominent risk of ship collisions. In recent years, the frequent occurrence of ship collisions with bridges has brought tremendous pressure to bridge maintenance units. The main reason is that bridge maintenance faces the following difficulties: there are many bridges to maintain, and bridge inspections are difficult; most bridge piers are located on the river surface, making it difficult to inspect at close range. In addition, small damage and deformation of bridge piers caused by ships scratching piers and anti-collision platforms are difficult to be discovered in time, which may make it difficult to know the accident in time, delay the status assessment and maintenance time of the bridge, and cause the damage to continue to develop, and even cause the collapse of the bridge, endangering the safety of life on the bridge.

[0003] In the existing technology, the common emergency monitoring method is the threshold alarm method. When a bridge is hit by a ship, the measured cable force, deflection, acceleration and other monitoring data can produce changes that exceed the warning value to alarm. However, the measured data is easily affected by environmental factors and has a large noise, which makes the measured value often exceed the warning value, thereby causing false alarms and reducing the effectiveness and credibility of the system alarm. Most of the warning values ​​are set according to the design values. However, the actual abnormal data response caused by the ship's scratches is far from reaching the warning value, resulting in the phenomenon of false warning values. If the warning value is reduced, there will be frequent false alarms. Moreover, for different types of piers, there are differences in their stiffness, structural form, and force form of the piers, and the installation position of the sensor, threshold setting, etc. are all different. Therefore, how to reasonably set the warning value faces a huge problem. Furthermore, some monitoring data are affected by temperature effects in addition to the environment, such as dynamic deflection and dynamic strain. The long-term monitoring data are affected by temperature effects more than the value changes caused by vehicle loads or other impact loads. Therefore, it is necessary to separate the temperature effect before an early warning can be issued. Otherwise, the early warning value setting range is too large and there is a phenomenon that no early warning can be issued. Compared with acceleration monitoring, dynamic deflection monitoring data is limited by monitoring conditions, the deflection changes before and after the ship collision are small, the data accuracy is low, and it is easy to be masked by environmental noise. Although the cointegration and Kalman filtering methods can identify abnormal dynamic deflection signals, it is impossible to determine whether the signal type is a ship collision signal. Similarly, although the method based on strain regression residuals can identify abnormal strain signals, the signal is not obvious, and the type of external load cannot be determined. It must be combined with other monitoring data such as the beam end angle for judgment. Although the comprehensive alarm index of the beam end horizontal rotation angle self-vibration characteristics composed of the beam end horizontal rotation angle amplitude, horizontal rotation angle peak and valley, and the first-order horizontal rotation angle self-power spectrum density amplitude can realize the bridge ship collision alarm, this method over-relies on the model information of the bridge itself, and there is dependence between the features. If the rotation angle amplitude is not matched, the latter two features will directly fail, and there is a problem of no effective redundancy mechanism.

[0004] Therefore, in order to solve the above problems, the present invention proposes a timely, economical and practical bridge health monitoring ship collision alarm method, which can timely detect the time and position of the ship collision and effectively perform the ship collision alarm. It solves the problems of generally using acceleration measured data for bridge ship collision abnormal alarm, such as different threshold settings, difficulty in signal type identification, and identification results being easily affected by environmental noise. It realizes real-time alarm of bridges under abnormal events such as ship collisions, which has important practical significance for the operation and management of long bridges. Summary of the invention

[0005] The present invention aims to overcome at least one defect (shortcoming) of the above-mentioned prior art and provide a bridge health monitoring ship collision alarm method for solving the problems of abnormal bridge ship collision alarm, such as different threshold settings, difficulty in signal type identification, and susceptibility of identification results to environmental noise. The present invention realizes real-time alarm of bridges under abnormal events such as ship collisions, and timely discovers damage and deformation of bridge piers caused by ship collisions, so as to evaluate and repair the bridge status, thereby reducing the problem of high difficulty in bridge maintenance management.

[0006] The technical solution adopted by the present invention is a bridge health monitoring ship collision alarm method, which specifically includes the following steps:

[0007] S1: Use the sliding window algorithm to segment the acceleration signal to obtain the corresponding acceleration response subsequence;

[0008] S2: Extract vibration response curves of acceleration subsequences using natural excitation technology;

[0009] S3: extracting the envelope of the vibration response signal using Hilbert transform, and then fitting the envelope;

[0010] S4: Determine the signal type through the attenuation function, and on this basis, calculate the mean ratio of the Mahalanobis distance under different states;

[0011] S5: The Chauville method is used to determine the abnormal values ​​of the Mahalanobis distance mean ratio of the acceleration vibration response at different pier tops, and the time and location of the ship collision are determined by the time and size of the abnormal values.

[0012] This method can identify the time and location of ship collision and effectively issue ship collision alarm, overcoming the disadvantage that the general algorithm based on vibration time domain parameter feature vector is difficult to identify. At the same time, the algorithm uses the impact vibration attenuation characteristics to judge the attenuation function, solving the problem of difficulty in signal type identification in general bridge ship collision algorithms.

[0013] Preferably, step S1 also includes measuring the acceleration time-history response a(t) of the bridge, setting the length of the measured acceleration time-history response to L, and setting a sliding window with a fixed window length of W to slide sequentially on a(t).

[0014] Further preferably, step S1 further includes setting basic parameters of the sliding window, setting the fixed window length W = mn*60*fs, the sliding interval S = ms*60*fs, wherein fs is the sampling frequency, the fixed window sampling time is mn minutes, and the recursive time is ms minutes; the sliding interval is less than the fixed window length, that is, S <W。

[0015] For the sliding window algorithm, the sliding spacing and fixed window length have three forms: if the sliding spacing is equal to the fixed window length, it is equivalent to intercepting data with a rolling window; if the sliding spacing is greater than the fixed window length, data will be lost; if the sliding spacing is less than the fixed window length, the window will overlap. The selection of sliding spacing and fixed window length has an important impact on the characteristic changes of the intercepted subsequence. If it is necessary to reflect the changing characteristics of the subsequence in a short period of time, the sliding spacing cannot be too large. For some algorithms that calculate the changing characteristics of subsequences, they also have certain requirements for the window width length of the subsequence, that is, the fixed window length cannot be too small. Therefore, this type of problem is the third form mentioned above, that is, the sliding spacing is less than the fixed window length.

[0016] Further preferably, in step S1, the acceleration response subsequence [X t1 , X t2 , …, X tk , …, X tM ] is obtained by intercepting M = (LW) / S sliding windows, where X tk for:

[0017] X tk =a(kS:kS+W)

[0018] In the formula: k = 1, 2, ..., tk = kS + W / 2, and W / 2≤tk≤LW / 2.

[0019] Since the acceleration response of the bridge under general environmental loads is different from that under ship collision excitation, it is necessary to use a sliding window algorithm to obtain the acceleration response subsequence of the bridge when a ship collision occurs, so as to provide data support for the subsequent judgment of the time and location of the ship collision.

[0020] Preferably, in step S2, the vibration response curve is extracted by obtaining the cross-correlation function of the structural acceleration response instead of the impulse response function. When the structure is excited by the input signal at point k, the formula of the cross-correlation function of the displacement response at points i and j is:

[0021]

[0022] Where: x(t) is the displacement response vector of the structure, is the rth order vibration mode, ω r and r are the rth order natural frequency and modal damping ratio, m r is the rth order modal mass, G ijr is a constant related to the measuring point and the modal order.

[0023] Since the response of the bridge after the ship collision excitation is not obvious in the acceleration time history curve, it is difficult to distinguish the ship collision event. However, when the bridge is subjected to impact load, its vibration has a free attenuation characteristic. The free vibration curve under the impact load is quite different from the vibration curve under environmental excitation. Therefore, the ship collision event can be distinguished based on the difference in the changing characteristics of the free vibration curve. In addition, under impact excitation or Gaussian white noise, the cross-correlation function of the structural acceleration response has a similar mathematical expression to the impulse response function, so the cross-correlation function can be used instead of the impulse response function.

[0024] Preferably, step S3 further includes performing a convolution operation on the signal x(t) and 1 / πt, and then analyzing the signal to obtain an envelope spectrum of x(t).

[0025] Further preferably, in step S3, the envelope curve of the single-degree-of-freedom acceleration attenuation signal is fitted using the envelope curve equation thereof, and the envelope curve fitting equation is:

[0026]

[0027] Where: C is the amplitude, ω n is the natural frequency, φ is the phase angle, and ζ is the system damping ratio.

[0028] Further preferably, the step S3 includes evaluating the fitting equation using a determination coefficient:

[0029]

[0030] Where: y i , are original data, predicted data, and original data mean, respectively. When -0.1 is satisfied <b<0,C> 0.8, and the fitting determination coefficient R s When >0.85, it is determined as a decay function.

[0031] When a bridge is hit by a ship, an attenuated pulse force will be generated, causing high-frequency resonance of the structure. Since damping lowers the resonance amplitude, the vibration signal shows attenuation characteristics, and there will be short-term amplitude and phase changes. It is essentially a damped free vibration attenuation process. The time and location of the ship collision can be found by judging the attenuation function.

[0032] Preferably, step S4 further includes setting the sample populations of the two subsequence acceleration responses to U and Z, each containing W samples:

[0033] U=[u1,u2,…,u W ] Τ

[0034] Z=[z1,z2,…,z W ] Τ

[0035] The Mahalanobis distance from vector U to Z is:

[0036]

[0037] Where: is the mean of vector Z, C Z is the covariance of vector Z;

[0038] Different subscripts 0, r, and t are used to represent the benchmark state, reference state, and test state, respectively. The first two states are the response states under conventional environmental excitation (i.e., normal states), and the test state is the measured free vibration response state. The abnormal signal discrimination index of the mean ratio of the Mahalanobis distance under different states is constructed as follows:

[0039]

[0040] Where: mean(·) represents the average value, parameter IFR tk The free vibration attenuation response signal of the acceleration response subsequence obtained from the measured acceleration response; parameter IFR r represents the free vibration attenuation response signal under the reference state; parameter IFR0 represents the free vibration attenuation response signal under the reference state.

[0041] The acceleration response under impact load is significantly different from the response under conventional environmental excitation. The difference between the measured free vibration response and the free vibration response under normal conditions can be used as an important means of distinguishing ship collisions. When the free attenuation curve of the bridge structure does not change significantly under conventional environmental excitation, that is, I k ≈1, when the bridge is subjected to impact load, its free vibration response will be significantly greater than 1. Under the impact load of ship collision, its Mahalanobis distance mean ratio will have an outlier, so the presence or absence of ship collision can be judged according to the outlier judgment method. This method can reduce the influence of outliers caused by environmental excitation and highlight the outliers under impact load by comparing the Mahalanobis distance mean ratio under different states.

[0042] Preferably, in step S5, the absolute value of the difference between a certain Mahalanobis distance mean ratio and the sample mean in the Chauvelet method is greater than the sum of the sample standard deviation and the Chauvelet coefficient ω M The product of , then this value is the abnormal value caused by the ship collision, that is: Among them, the Chauvelet coefficient ω M =1+λln(M1).

[0043] The vibration response of the bridge under environmental excitation is random, and its derived indicator, the mean ratio of the Mahalanobis distance, is also random and obeys the normal distribution. The Chauville method is more reasonable for processing limited data. Therefore, it is selected to make an abnormal judgment on the mean ratio of the Mahalanobis distance, so as to identify whether a ship collision occurs.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) It can realize real-time alarm of the bridge health monitoring system in abnormal events, reduce the difficulty of bridge operation and maintenance management, and timely evaluate and repair the small damage and deformation of bridge piers caused by ships scraping against bridge piers and anti-collision platforms, so as to prevent bridge collapse and endanger the safety of life on the bridge.

[0046] (2) By using NExT technology to extract the measured vibration acceleration signal, we can observe the obvious free attenuation characteristics under different excitation forces. By analyzing the abnormalities of the free attenuation signal characteristics of ship collision and the free attenuation signal characteristics caused by vehicle load and wind load excitation, we can distinguish the ship collision signal from the vibration signals caused by other abnormal vibrations such as vehicle load and wind load, realize ship collision signal recognition, and overcome the false alarms caused by other abnormal vibrations.

[0047] (3) The abnormal acceleration vibration caused by ship collision is judged based on the mean ratio of Mahalanobis distance under different states, which reduces the changes caused by environmental excitation to a certain extent, highlights the recognition degree of impact vibration signals, and overcomes the difficulties of setting warning values, false warning values, frequent false alarms and different threshold settings. By comparing the size of the abnormal value of the mean ratio of Mahalanobis distance at different locations, an alarm can be issued at the time and location of the ship collision. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the acceleration response sliding window of the present invention.

[0049] Figure 2 The figure is a flow chart of the ship collision alarm algorithm of the present invention.

[0050] Figure 3 This is a diagram of the measurement point layout of the Jiujiang Bridge vibration monitoring system in an embodiment of the present invention.

[0051] Figure 4 It is a schematic diagram of measured data of lateral acceleration on the pier top of Jiujiang Bridge (expansion) during a ship collision in an embodiment of the present invention.

[0052] Figure 5 It is a schematic diagram of measured data of lateral acceleration on the pier top of Jiujiang Bridge (old bridge) during a ship collision in an embodiment of the present invention.

[0053] Figure 6It is a schematic diagram of measured data of vertical acceleration at the mid-span of the main span of Jiujiang Bridge (expansion) during a ship collision in an embodiment of the present invention.

[0054] Figure 7 It is a schematic diagram of the free vibration attenuation signal extraction of the pier top lateral acceleration NExT in the present invention (expanded bridge).

[0055] Figure 8 It is a schematic diagram of the main span vertical acceleration NExT free vibration attenuation signal extraction (expanded bridge) in the present invention.

[0056] Fig. 9 It is a schematic diagram of the outlier discrimination of the mean ratio of the Mahalanobis distance of the lateral acceleration of the pier top in the present invention (bridge expansion).

[0057] Fig.10 It is a schematic diagram of the outlier discrimination (bridge expansion) of the main span vertical Mahalanobis distance mean ratio in the present invention.

[0058] Fig.11 This is the free vibration attenuation signal extraction of the pier top lateral acceleration NExT in the present invention (old bridge).

[0059] Fig.12 It is a schematic diagram of outlier discrimination (old bridge) of the mean ratio of the transverse Mahalanobis distance of the pier top in the present invention.

[0060] Fig.13 It is a schematic diagram of the measured data of the lateral acceleration of the pier top of Jiujiang Bridge (expansion) at the moment of abnormal vibration in the present invention.

[0061] Fig.14 Schematic diagram of the present invention. DETAILED DESCRIPTION

[0062] The drawings of the present invention are only for illustrative purposes and should not be construed as limiting the present invention. In order to better illustrate the following embodiments, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; it is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0063] Example 1

[0064] A bridge health monitoring ship collision alarm method is used to monitor whether a bridge has a ship collision, realize real-time alarm of the bridge under abnormal events such as ship collision, promptly discover the damage and deformation of the bridge pier caused by the ship collision, and prevent the damage from continuing to develop, so as to evaluate and repair the bridge status, thereby reducing the difficulty of bridge maintenance management.

[0065] Preferably, in the present invention, a ship collision alarm algorithm is provided for extracting acceleration vibration free vibration attenuation signal based on natural excitation technology:

[0066] Further preferably, the sliding window algorithm is first used to segment and intercept the acceleration signal to obtain corresponding acceleration response subsequences, specifically:

[0067] The measured acceleration time history response is a(t), its sampling frequency is fs, its length is set to L, and a sliding window with a fixed window length of W (W < L) slides in sequence. Assuming the fixed window is mn minutes, then the window width length W = mn * 60 * fs. A recursive sliding window with a length of S is used, and the recursive time is ms minutes, then the recursive interval length S = ms * 60 * fs. M = (L - W) / S acceleration response subsequences [X t1 , X t2 , …, X tk , …, X tM can be obtained. The specific schematic diagram is shown in Figure 1 as shown, where X tk is:

[0068] X tk = a(kS:kS + W) (1)

[0069] In the formula: k = 1, 2, …, M; M = (L - W) / S; tk = kS + W / 2, where W / 2 ≤ tk ≤ L - W / 2.

[0070] For the sliding window algorithm, the sliding distance and the window width have three forms: If the sliding distance is equal to the window width, it is equivalent to rolling window data interception; if the sliding distance is greater than the window width, data will be lost; if the sliding distance is less than the window width, the windows will overlap.

[0071] The selection of the sliding distance and the window width size has an important impact on the characteristic changes of the intercepted subsequences. If it is necessary to reflect the change characteristics of the subsequences in a short time, the sliding distance cannot be too large. For some algorithms for calculating the change characteristics of subsequences, there are also certain requirements for the window width length of the subsequences, that is, the window width cannot be too small. Therefore, this kind of problem is the above third form, that is, the sliding distance is less than the sliding window (ms < mn).

[0072] The acceleration response of a bridge under general environmental loads is different from the acceleration response of a bridge under ship collision excitation. However, the response is not obvious in the acceleration time history curve. Due to the influence of environmental noise, it is difficult to distinguish ship collision events using threshold algorithms and algorithms based on vibration time domain parameter characteristic vectors (such as: acceleration effective value, peak value, peak factor, absolute average value, root mean square amplitude, etc.), making it difficult to issue an alarm in a timely manner. When a bridge is subjected to impact loads, its vibration has a free attenuation characteristic. The free vibration curve under impact loads is quite different from the vibration curve under environmental excitation. Therefore, ship collision events can be distinguished based on the difference in the change characteristics of the free vibration curve. The present invention performs a method for distinguishing ship collision events based on the acceleration sub-response sequence X. tk NExT is used to extract vibration response curves.

[0073] Preferably, natural excitation technology is used to extract vibration response curves of acceleration subsequences;

[0074] The core idea of ​​NExT is that under impact excitation or Gaussian white noise, the cross-correlation function of the structural acceleration response has a similar mathematical expression to the impulse response function, and the impulse response function can be replaced by obtaining the cross-correlation function of the structural acceleration response.

[0075] The equation of motion for a linear time-invariant structure with n degrees of freedom is:

[0076]

[0077] Where: M, C and K are the mass, damping and stiffness matrices of the structure, x(t) is the displacement response vector of the structure, and f(t) is the external excitation force. The displacement vector is expressed in modal coordinates as:

[0078]

[0079] Where: Φ is the modal matrix, q(t) is the modal coordinate vector, is the rth order vibration mode, q r (t) is the r-th order modal coordinate component.

[0080] Substitute (3) into equation (2) and multiply both sides of the equation by Φ T , using the modal orthogonality, we get the decoupled motion equations:

[0081]

[0082] Where: r and r are the rth order natural frequency and modal damping ratio, m r is the r-th order modal mass.

[0083] Using Duhamel integral to solve the above equation (4) we can get:

[0084]

[0085] Where: g r (t) is the unit impulse response function:

[0086]

[0087] Where: is the r-th order modal damped natural frequency.

[0088] Substituting formula (5) into formula (3), we can obtain:

[0089]

[0090] When the structure is excited at point k, the impulse response function x at point i is ik (t) can be expressed as:

[0091]

[0092] When the input signal excitation is at point k, the displacement response correlation function of points i and j is:

[0093]

[0094] Where: G ijr is a constant related to the measuring point and the modal order.

[0095] Obviously, equation (8) and equation (9) have similar mathematical expressions, and the correlation function can be used to replace the impulse response function.

[0096] Preferably, the free vibration attenuation signal is used for discrimination:

[0097] Further preferably, the envelope of the vibration response signal is extracted using Hilbert transform, and then the envelope is fitted, specifically:

[0098] The bridge structure is a multi-degree-of-freedom system, which can be decoupled into several multi-degree-of-freedom systems according to modal theory. For a damped single-degree-of-freedom system, its free vibration response curve has an attenuation characteristic. Its acceleration response expression is:

[0099]

[0100] Where C is the amplitude, ω n is the natural frequency, φ is the phase angle, and ζ is the system damping ratio. It can be seen that the envelope equation of the acceleration attenuation signal is:

[0101]

[0102] When a bridge is hit by a ship, an attenuated pulse force will be generated, causing high-frequency resonance of the structure. Since the damping reduces the resonance amplitude, the vibration signal shows attenuation characteristics, and there are short-term amplitude changes and phase changes. It is essentially a damped free vibration attenuation process. Therefore, the envelope equation of the single-degree-of-freedom acceleration attenuation signal can be used to fit its envelope curve. The determination coefficient is used to evaluate the fitting method equation:

[0103]

[0104] Where: y i , are original data, predicted data, and original data mean, respectively. When -0.1 is satisfied <b<0,C> 0.8, and the fitting determination coefficient R s When >0.85, it is judged as a decay function.

[0105] The Hilbert transform can be used to extract the envelope curve of the acceleration response curve by convolving the signal x(t) with 1 / πt:

[0106]

[0107] The analytical signal x(t) can be expressed as:

[0108] q(t)=x(t)+ix h (t) (14)

[0109] The Hilbert envelope of x(t) is defined as:

[0110]

[0111] The envelope spectrum of x(t) is:

[0112]

[0113] Preferably, the signal type is determined by the attenuation function, and on this basis, the mean ratio of the Mahalanobis distances under different states is calculated, specifically:

[0114] The Mahalanobis distance formula is the distance between the vector and the mean divided by the covariance. The size of the covariance reflects the correlation of the data. It is an important indicator for measuring the distance between a point and a distribution. It is often used for multivariate outlier detection, classification of highly unbalanced data, and one-class classification. In this embodiment, the Mahalanobis distance is used to determine the proximity of the free vibration attenuation signals caused by two acceleration response subsequences to determine the abnormal acceleration response caused by the excitation force.

[0115] Assume that the sample populations of the two subsequence acceleration responses are U and Z, each containing W samples:

[0116] U=[u1,u2,…,u W ] Τ (17)

[0117] Z=[z1,z2,…,z W ] Τ (18)

[0118] The Mahalanobis distance from vector U to Z is:

[0119]

[0120] Where: is the mean of vector Z, C Z is the covariance of the vector Z.

[0121] The acceleration response under impact load is significantly different from the response under conventional environmental excitation. The difference between the measured free vibration response and the free vibration response in the normal state can be used as an important means to distinguish ship collisions. For the convenience of explanation, different subscripts 0, r, and t are used to represent the benchmark state, reference state, and test state, respectively. The first two states are the response states under conventional environmental excitation (i.e., the normal state), and the test state is the measured free vibration response state.

[0122] The acceleration responses a0(1:W) and a0(1:W) of a sample length W under the baseline state and the reference state are intercepted respectively. r (1:W), and the free vibration attenuation response signals are obtained by NExT using formula (9) as IFR0, IFR r .

[0123] The reference state uses the sliding window algorithm (1) and NExT (9) to obtain the free vibration attenuation response signal IFR of the acceleration response subsequence obtained from the measured acceleration response. tk .

[0124] The abnormal signal discrimination index of the mean ratio of Mahalanobis distance under different states is constructed as follows:

[0125]

[0126] Where: mean(·) represents the average value. Obviously, when the free attenuation curve of the bridge structure does not change significantly under the normal environmental excitation, that is, I k≈1, when the bridge is subjected to impact load, its free vibration response will be significantly greater than 1. Under the impact load of ship collision, its Mahalanobis distance mean ratio will have an outlier, so the presence or absence of ship collision can be judged according to the outlier judgment method. This method can reduce the influence of outliers caused by environmental excitation and highlight the outliers under impact load by comparing the Mahalanobis distance mean ratio under different states.

[0127] Preferably, the Chauville method is used to judge the abnormal values ​​of the Mahalanobis distance mean ratios of the acceleration vibration responses at different pier tops, and the time and position of the ship collision are judged by the time and size of the abnormal values.

[0128] Common methods for outlier identification include: Laida criterion, Chauville method, Grubbs criterion, box plot analysis, etc. Among them, the box plot method has no restrictions on the data format, while the other methods require the data to obey the normal distribution and have a certain limit on the number of data, but the Chauville method is more reasonable for limited data processing results. The vibration response of the bridge under environmental excitation is random, and its derived indicator, the Mahalanobis distance mean ratio, is also random and obeys the normal distribution. Therefore, in this embodiment, the Chauville method is used to make an abnormal judgment on the Mahalanobis distance mean ratio, so as to identify whether a ship collision has occurred. The essence of the Chauville method is that in M1 Mahalanobis distance mean ratio samples, if the number of times the error may occur is less than half a time, then the value is considered to be an outlier, that is, it has a confidence probability of 1-1 / 2M. The absolute value of the difference between a certain Mahalanobis distance mean ratio and the sample mean is greater than the sample standard deviation and the Chauville coefficient ω M The value may be an outlier caused by a ship collision:

[0129]

[0130] Where: Chauvelet coefficient ω M =1+λln(M1), in this embodiment, λ=2, the specific parameters are determined according to the historical data of the bridge. Order I warm =4.

[0131] In summary, the basic process of the ship collision signal alarm algorithm based on natural excitation technology and random reduction technology for free vibration attenuation signal extraction is as follows: Figure 2 The alarm condition for a ship collision is that the frequency response function is a free decay function, that is, the fitting coefficient of its envelope curve satisfies -0.1. <b<0,C> 0.8, and the fitting determination coefficient R s >0.85, in addition, the mean Mahalanobis distance is greater than I k >I warm Generally, there is the following rule: the smaller the absolute value of the fitting coefficient b, the smaller the fitting coefficient C, and the smoother the free decay function; the determination coefficient Rs The larger the value, the better the fitting effect; the mean Mahalanobis distance is greater than I k The larger the value, the greater the difference between the acceleration signal in the test state and the acceleration signal in the reference state.

[0132] Further preferably, in this embodiment, the Jiujiang Bridge (old bridge) of Fokai Expressway is taken as an example to measure the data.

[0133] The structural health monitoring system of the Jiujiang Bridge (old bridge) and the expanded bridge of Fokai Expressway includes monitoring contents such as deflection, strain, acceleration, structural temperature, temperature and humidity, bearing displacement, pier inclination, anti-ship collision monitoring system, and vehicle load monitoring system. In this embodiment, only the alarm algorithm of ship collision bridge is studied based on vibration monitoring data, so only the vibration monitoring system of Jiujiang Bridge and the expanded bridge is introduced. Figure 3 As shown. 14 acceleration sensors are arranged in the concrete box girders of Jiujiang Bridge (old bridge) and the expansion bridge, respectively, in the concrete box girders in the middle of the two main spans and the top of the three main piers in the middle. Jiujiang Bridge (old bridge) and the expansion bridge are each equipped with a main acquisition station, and the acceleration sensors are collected in a centralized manner and connected to the main acquisition stations respectively. The vibration acceleration of Jiujiang Bridge (old bridge) uses a velocity sensor, and the expansion bridge uses an acceleration sensor, and the sampling frequency is 50Hz.

[0134] On May 13, 2022, at around 18:47, a ship scraped against a bridge pier at Jiujiang Bridge. The ship Changxin, which was leaving Bohai from the war preparation dock, turned left. Affected by the current, the starboard side of the bow hit the right pier of the old bridge of Jiujiang Bridge (the first navigation hole on the right side of the lower navigation hole), and then the starboard side of the ship continued to scrape against the pier, and then the port side of the bow hit the pier of the new bridge. The ship first hit the pier of the Jiujiang Bridge on the national highway and then hit the upper left corner of the 22# pier cap of the Fokai Expressway.

[0135] Figure 4 , Figure 5 These are the measured data of the lateral acceleration of the pier top during the ship collision between the expanded bridge and the old bridge of Jiujiang Bridge. As can be seen from the figure, during the ship collision, the lateral acceleration curve of the pier top has obvious peak mutations, and the peak mutations have vibration attenuation characteristics. Figure 6 These are the measured data of vertical acceleration at the mid-span of the main span of the Jiujiang Bridge expansion project. It is obvious that the vertical acceleration curve has no obvious peak.

[0136] Further preferably, in this embodiment, algorithm identification is performed based on the above data.

[0137] Figure 7 , Figure 8They are respectively the free vibration decay signals of the lateral acceleration at the pier top and the vertical acceleration of the main span extracted by NExT. It can be seen that the free vibration decay signal caused by ship impact is greater than that under normal environmental excitation. Moreover, the decay characteristics that cannot be observed in the original vertical acceleration time history can be observed to a certain extent by using NExT, overcoming the disadvantage that it is difficult to identify by general algorithms based on vibration time-domain parameter eigenvectors. Among them, during the ship impact, the envelope fitting curve of the influence curve of the lateral acceleration at the pier top of the impacted pier has obvious free vibration decay characteristics, and its determination coefficient R s > 0.85, and the fitting coefficient exponent -0.1 < b < 0, and the fitting parameter C > 0.8. The envelope fitting curves of other non-directly impacted piers have certain decay characteristics.

[0138] To judge the closeness of the vibration decay signals before and after the sliding window, the mean ratio of Mahalanobis distances in different states is used to judge the abnormal acceleration vibration caused by ship impact. The reason for using the Mahalanobis distances in different states is that the Mahalanobis distance between the free vibration decay signals in the impact state and the environmental excitation state is much greater than the Mahalanobis distance between the free vibration decay signals in the same environmental excitation state. Its ratio reduces the changes caused by environmental excitation and highlights the recognition degree of the impact vibration signal. Fig. 9 、 Fig.10 For the judgment of abnormal signals using the mean ratio of Mahalanobis distances, there are relatively large abnormal peaks around 18:47. By observing the peak values of the Mahalanobis distance ratio, it is found that the position with a large peak value is the pier position impacted by the ship. Therefore, the ship impact position can be judged by the peak value of the Mahalanobis distance ratio. Fig.11 It is the acceleration frequency response at the alarm moment of the lateral acceleration at the pier top of Jiujiang Bridge (old bridge). The fitting coefficient and determination coefficient of the envelope curve of the middle pier of the main span meet the alarm conditions of the decay function. Fig.12 Using the mean ratio of Mahalanobis distances to judge the lateral acceleration at the pier top of Jiujiang Bridge (old bridge), obvious peaks can be observed within the same time, reflecting the applicability of the proposed method to continuous rigid frame bridges.

[0139] Fig.13 、 Fig.14They are respectively the measured data of the lateral acceleration of the pier top at the time of historical abnormal vibration of Jiujiang Bridge (expanded bridge) and the attenuation response signal of the free vibration of the lateral acceleration of the pier top extracted by NExT. It can be seen that the attenuation response signal is quite different from the attenuation response signal during the ship collision. The attenuation response signal during the ship collision is relatively gentle, and the amplitude is relatively sparse, that is, the frequency is small, while the attenuation response signal at the abnormal moment is steeper, the amplitude is dense, the frequency is large, and the curve is not monotonically decreasing. The mean ratio of the Mahalanobis distance is greater than the threshold, indicating that there is abnormal vibration. At present, the proposed method has been used in other super-large bridges, and can identify abnormal vibrations caused by typhoon excitation in cable-stayed bridges, such as the abnormal vibrations of Tongming Super Bridge and Zhanjiang Bay Super Bridge caused by Typhoon "Tali" on July 17, 2013. This response is different from the ship collision response. Using the method proposed in the present invention, there will be no false alarm. Obviously, the ship collision signal can be identified based on the attenuation response signal of the free vibration of the lateral acceleration of the pier top extracted by NExT.

[0140] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solution of the present invention, and are not intended to limit the specific implementation methods of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the claims of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A bridge health monitoring ship collision alarm method, specifically comprising the following steps: S1: Use the sliding window algorithm to segment the acceleration signal to obtain the corresponding acceleration response subsequence; S2: Extract vibration response curves of acceleration subsequences using natural excitation technology; In step S2, the vibration response curve is extracted by obtaining the cross-correlation function of the structural acceleration response instead of the impulse response function. When the structure is excited by the input signal at point k, the formula of the cross-correlation function of the displacement response at points i and j is: Where: x(t) is the displacement response vector of the structure, is the rth order vibration mode, ω r and r are the rth order natural frequency and modal damping ratio, m r is the rth order modal mass, G ijr is a constant related to the measuring point and the modal order; S3: extracting the envelope of the vibration response signal using Hilbert transform, and then fitting the envelope; The step S3 also includes performing a convolution operation on the signal x(t) and 1 / πt, and then analyzing the signal to obtain the envelope spectrum of x(t), specifically: Use Hilbert transform to extract the envelope curve of the acceleration response curve, and perform convolution operation on the signal x(t) and 1 / πt: The analytical signal x(t) can be expressed as: q(t)=x(t)+x h (t) The Hilbert envelope of x(t) is defined as: The envelope spectrum of x(t) is: S4: Determine the signal type through the attenuation function, and on this basis, calculate the mean ratio of the Mahalanobis distance under different states; S5: The Chauville method is used to determine the abnormal values ​​of the Mahalanobis distance mean ratio of the acceleration vibration response at different pier tops, and the time and location of the ship collision are determined by the time and size of the abnormal values.

2. A bridge health monitoring ship collision alarm method according to claim 1, characterized in that: The step S1 also includes measuring the acceleration time-history response a(t) of the bridge, setting the length of the measured acceleration time-history response to L, and setting a sliding window with a fixed window length of W to slide sequentially on a(t).

3. A bridge health monitoring ship collision alarm method according to claim 2, characterized in that: The step S1 also includes setting the basic parameters of the sliding window, setting the fixed window length W = mn*60*fs, the sliding interval S = ms*60*fs, where fs is the sampling frequency, the fixed window sampling time is mn minutes, and the recursive time is ms minutes; the sliding interval is less than the fixed window length, that is, S <W。 4. A bridge health monitoring ship collision alarm method according to claim 3, characterized in that: In step S1, the acceleration response subsequence [X t1 , X t2 , …, X tk , …, X tM ] is obtained by intercepting M = (LW) / S sliding windows, where X tk for: X tk =a(kS:kS+W) In the formula: k = 1, 2, ..., tk = kS + W / 2, and W / 2≤tk≤LW / 2.

5. A bridge health monitoring ship collision alarm method according to claim 4, characterized in that: In step S3, the envelope curve of the single-degree-of-freedom acceleration attenuation signal is fitted using the envelope curve equation, and the envelope curve fitting equation is: Where: C is the amplitude, ω n is the natural frequency, φ is the phase angle, and ζ is the system damping ratio.

6. A bridge health monitoring ship collision alarm method according to claim 5, characterized in that: The step S3 includes evaluating the fitting equation using a determination coefficient: Where: y i , are original data, predicted data, and original data mean, respectively. When -0.1 is satisfied <b<0,C> 0.8, and the fitting determination coefficient R s When >0.85, it is determined as a decay function.

7. A bridge health monitoring ship collision alarm method according to claim 6, characterized in that: Step S4 also includes setting the sample populations of the two subsequence acceleration responses to U and Z, each containing W samples: U=[u1,u2,…,u W ] Τ Z=[z1,z2,…,z W ] Τ The Mahalanobis distance from vector U to Z is: Where: is the mean of vector Z, C Z is the covariance of vector Z; Different subscripts 0, r, and t are used to represent the benchmark state, reference state, and test state, respectively. The first two states are the response states under conventional environmental excitation (i.e., normal states), and the test state is the measured free vibration response state. The abnormal signal discrimination index of the mean ratio of the Mahalanobis distance under different states is constructed as follows: Where: mean(·) represents the average value; parameter IFR tk The free vibration attenuation response signal of the acceleration response subsequence obtained from the measured acceleration response; parameter IFR r represents the free vibration attenuation response signal under the reference state; parameter IFR0 represents the free vibration attenuation response signal under the reference state.

8. A bridge health monitoring ship collision alarm method according to claim 7, characterized in that: In step S5, the absolute value of the difference between a Mahalanobis distance mean ratio and the sample mean in the Chauvelet method is greater than the sum of the sample standard deviation and the Chauvelet coefficient ω M The product of , then this value is the abnormal value caused by the ship collision, that is: Among them, the Chauvelet coefficient ω M =1+λln(M1).

Images