KR101218354B1 - An exaggerate error processing method in monitoring a bridge based on the global navigation satellite system - Google Patents

Abstract

PURPOSE: An exaggerate error processing method for monitoring a bridge based on a GNSS(Global Navigation Satellite System) is provided to prevent an accident due to management carelessness, thereby improving the stability of a bridge. CONSTITUTION: A position coordinate value of a receiver is obtained through an adjustment process of an output signal from a satellite. Multipath errors in the position coordinate value are corrected. A quality control process including a statistical verification process is implemented for the position coordinate value which deviates from a multiple path effect. The position coordinate value with excessive error removed is finally outputted. De-noising is implemented by a wavelet transformation technique. The wavelet transformation technique selectively separates noise into high frequency band and low frequency band.

Description

An exaggerate error processing method in monitoring a bridge based on the Global Navigation Satellite System}

The present invention relates to a technique for monitoring the behavior of a bridge using a Global Navigation Satellite System (hereinafter referred to as "GNSS"), in particular, to compensate for errors caused by multiple paths of GNSS data and to exaggerate through quality control. The present invention relates to a GNSS-based bridge monitoring over-error processing method that provides an algorithm with a function to remove errors.

In recent years, with the economic development and increasing traffic volume and material transportation, there is a growing demand for the construction of larger and more convenient long bridges in terms of size. However, bridges for this purpose are expensive, and the construction also involves time losses that can take years to decades. In addition, because it is a large infrastructure directly connected to life, accidents due to structural defects cause great casualties. Recently, a large number of long bridges have been constructed and are in progress. However, due to the construction-oriented policies, the maintenance of bridges under construction and public use has been considerably insufficient until now.

Bridge collapses have occurred frequently anywhere in the world. Most recently, the Kota Chambal bridge in India in December 2009 collapsed during construction, resulting in nine casualties and 45 missing persons. Korea has also been receiving unprecedented social interest and expectation in the field of maintenance in order to ensure the stability of long bridges after experiencing a series of bridge collapse accidents such as Sinhaengju Bridge and Seongsu Bridge in the 1990s.

However, due to the deterioration of domestic maintenance technology, the maintenance technique for structures that exhibit complex behavior such as long bridges is in the early stage. Therefore, in order to secure and maintain the stability and usability of the constructed structure reasonably, it is necessary to periodically evaluate the structural reactions. For this reason, a study on the behavior monitoring technique to constantly evaluate the stability and usability of the structure in the civil engineering field Is in progress.

At present, the measurement and monitoring of bridges are carried out by loading tests or by inserting a large number of measurement equipment such as accelerometers, stress gauges, sinkers, displacement meters, etc. inside the bridge to receive data directly by wire. However, in case of loading test, it is difficult to carry out the measurement work at any time due to a lot of manpower and cost, and it is necessary to restrict the usability of the bridge because it requires traffic control during measurement. However, there are difficulties due to problems such as data corruption due to environmental conditions (Yoon Hong-sik, 2001; Kim Sung-gon, 2006). In addition, due to the long span of bridges, existing behavior monitoring systems such as the effects of weather, difficulty in securing observation points, and measurement of relative displacement occur.

GNSS is a next generation system that can overcome the limitations of the existing monitoring system. It can output 3D absolute coordinates regardless of climatic conditions by using satellite navigation system, and it is not affected by the distance between the points and the visible distance. It is a system specialized in bridges. Already, overseas advanced countries have been using GNSS to continuously measure and report on major structures including bridges and dams (Park, 2006).

However, GNSS involves various errors in transmitting and receiving signals from outer space in the earth, and these errors affect the accuracy of observation data (Represented, 2008). Therefore, in order to increase the range of use of GNSS in the field of bridge measurement, in addition to data quality analysis, a technique for removing errors generated in bridges is required.

Accordingly, an object of the present invention is to provide a GNSS-based bridge monitoring over-error processing method that can provide the required accuracy desired by the user of the observed GNSS data.

Another object of the present invention is to provide a GNSS-based bridge monitoring over-error processing method that can remove the noise due to the effect of multi-path.

It is still another object of the present invention to provide a GNSS-based bridge monitoring overerror processing method capable of removing random errors out of statistics.

GNSS-based bridge monitoring over-error processing method according to the present invention for achieving these objects is characterized by comprising a process of correcting the multi-path error included in the GNSS position coordinate value.

As an example, the process of correcting the multipath error may include a de-nosing process using wavelet transformation.

As an example, the wavelet transformation is characterized by using a multi-resolution-analysis (MRA) method, which is a fast discrete wavelet algorithm.

As an example, the wavelet transformation is characterized by selectively separating noise into a high frequency band and a low frequency band.

As an example, the method may further include a quality control process of removing an excessive error included in the GNSS position coordinate value, wherein the removing of the excessive error comprises: a detection step of detecting an excessive error of observation data; Identifying excessive observations, identifying the observed values including the excessive errors; And an adaptation step of removing the contaminated observations identified after the detection and identification steps.

delete

As an example, the detection step is characterized by performing a statistical hypothesis test on the post-dispersion factor.

As one example, the quality control process includes estimating the GNSS receiver coordinates when an internal reliability, which is the minimum magnitude of the excess error that can be found in the identification of the excess error present in the observation, and an undetected excess error are used for the estimation. It is characterized by the performance of statistical verification using an external reliability that indicates its impact on the system.

GNSS-based bridge monitoring excessive error processing method according to the present invention can be expected the following effects.

First, it is possible to secure the stability of bridges by preventing accidents that may occur due to neglect of management using systematic and modern methods.

Second, it can reduce the manpower and long-term bridge maintenance costs in the field.

Third, it is possible to increase the applicability of the GNSS by removing the excessive errors generated during GNSS-based bridge monitoring to secure the reliability of the results.

Fourth, GNSS sensors can be used indoors to monitor behavior such as momentary displacement as well as ground subsidence through long-term deformation observation, and automate and informatize relevant information in real time.

1 is a flow chart showing the progress of the bridge monitoring excessive error processing method based on GNSS according to the present invention.
2 is an exemplary view showing a decomposition process according to an embodiment of the present invention.
3 is an exemplary view showing a reconstruction process according to an embodiment of the present invention.
4 is an exemplary view illustrating a noise removing process according to an embodiment of the present invention.
5 is an exemplary view comparing the position adjustment value and the wavelet application value of the high elevation satellite combinations 22, 14, and 18.
6 is an exemplary view comparing the position adjustment value and the wavelet application value of the low elevation satellite combinations 22, 9, and 12.
7 is an exemplary view showing a GNSS quality control process according to an embodiment of the present invention.

Hereinafter, a GNSS-based bridge monitoring excessive error processing method based on a satellite navigation system (hereinafter referred to as "GNSS") according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flow chart showing the progress of the bridge monitoring excessive error processing method based on GNSS according to the present invention.

First, the position coordinate value of the receiver is obtained by adjusting the signal output from the satellite, and the multipath error included in the position coordinate value is corrected. In addition, a quality control process including a statistical verification process is performed on the position coordinate values deviating from the multipath influence, and finally, the position coordinate values from which the excessive errors are removed are output.

In order to correct the multipath error contained in the GNSS position adjustment value (GNSS position coordinate value), the present invention uses a wavelet transformation technique to de-nosing the noise, and the wavelet transformation Noise is selectively separated into high and low frequency bands.

The wavelet transform has an advantage that time and frequency information can be appropriately represented at the same time because various basis functions are generated by the process of scaling and moving the basic wavelet. Continuous wavelet transformation (CWT) is defined as

는 모웨이브릿 함수이다.Where s is a scale parameter and is a shift parameter. Also x (t) is the signal you want to interpret

Is the Mowavelet function.

Noise rejection using wavelets is commonly used with multi-resolution analysis (MRA), a fast, fast decrete wavelet algorithm.

Multiresolution analysis uses 2 ⁿ data manipulation discrete wavelet transform filters, which use the orthogonality of the filters to dramatically reduce the amount of computation.

The type of filter is determined by the basis function used. The scale function that acts as a low pass filter orthogonal to each other, and the wavelet function that functions as a high pass filter are defined as follows.

: 저역필터

: Low pass filter

: 고역필터

: High pass filter

Where h (m) is the low pass filter coefficient and g (m) is the high pass filter coefficient.

They go through a decomposition process and a reconstruction process according to the level, and a signal decomposed by the low pass filter in the decomposition process is called a scale component coefficient ( cA ), and a signal decomposed by the high pass filter is called a wavelet coefficient ( cD ).

In this case, each factor of decomposition is subjected to a downsampling process of reducing the number to half, and the factor decomposition factor is used to perform the next level. The reconstruction process combines the decomposed coefficients and restores the signal through an upsampling process that reduces the number half of the decomposition to its original state.

As shown in Figs. 2 and 3, a signal that is double-differentiated in the filtering process comes out as two signals after passing through two complementary filters, and thus is separated into a much lower resolution component part in succession. This process is called multiresolution analysis.

As with GNSS signals, whether they are code observations or carrier observations, they cause a lot of delay due to the process from the satellite to the receiver or for environmental reasons, and these delays are noises that directly affect accuracy.

In the GNSS noise removing process using the wavelet transform, a decomposition step of selecting a level and performing decomposition up to a selected level as shown in FIG. 4, setting a threshold value for each level, and calculating a shrinkage function Inverse transformation is performed by using the applied relationship and the modified detail and approximation coefficient.

Specifically, the purpose of the wavelet reduction function is to reduce or cancel the noise of the double differential signal, and three wavelet reduction functions are considered in the application of the GNSS. The critical wavelet coefficient processing is divided into two stages: the threshold function and the critical parameter determination. First, the threshold function is determined first, and then the critical parameter is determined. In general, three critical functions are considered in the application of GNSS.

Where λ is the critical parameter.

The first step, threshold function selection, selects one of the three functions listed above.

The second step, the selection of the parameters, is made by a universal threshold, in which, if the value of λ is too small, it will still contain a lot of noise. On the contrary, if the value of λ is too large, important information of the signal itself will be lost and large deviation May occur. In general, since GNSS is low in noise, a universal threshold is mainly used.

Where MAD is the absolute mean deviation and 0.6745 is the adjustment factor to have a Gaussian distribution.

In order to correct GNSS multipath error using the wavelet transform, selection of a threshold and a mowavelet is an important step in obtaining an optimal result value. Valid

[Table 1] and [Table 2] below show the threshold values in the GNSS data acquired from the bridge. The DB (the Daubechies), which was used for the GNSS data through the preliminary study due to the variety of types of the mower wavelet and the threshold function, The comparative study was limited to and Sym (Symlets) wavelets.

In addition, the threshold was limited to only the soft threshold of [Table 1] and the hard threshold of [Table 2].

The data was taken from the GNSS data installed on the cable-stayed bridge. Actually, the optimal wavelet was extracted by extracting the data from 07:00 to 09:00, which is expected to have the most multi-path among 24 hours of observation time. A comparative study was conducted. The table below summarizes these thresholds and the experimental details and results of the mowavelets.

Soft BLR Max Min ResidualAverage RMS Non-WT 0.0467 -0.0541 -0.0132 0.0163 DB2 0.0429 -0.0471 -0.0127 0.0158 DB4 0.0412 -0.0497 -0.0129 0.0159 DB8 0.0414 -0.049 -0.013 0.0159 DB12 0.0416 -0.0476 -0.0128 0.0159 Sym4 0.0401 -0.0461 -0.0119 0.0141 Sym8 0.0412 -0.0483 -0.0129 0.0159 Sym12 0.0409 -0.0466 -0.0109 0.0145 Soft BRR Max Min ResidualAverage RMS Non-WT 0.0397 -0.0372 0.0326 0.01158 DB2 0.0347 -0.03024 0.032 0.01105 DB4 0.0342 -0.0303 0.0321 0.011098 DB8 0.0341 -0.0293 0.0311 0.01111 DB12 0.0337 -0.0288 0.0309 0.01111 Sym4 0.0334 -0.03123 0.0255 0.011089 Sym8 0.0337 -0.030221 0.0322 0.01114 Sym12 0.0321 -0.0279 0.0244 0.01

Hard BLR Max Min ResidualAverage RMS Non-WT 0.0469 -0.0541 -0.0133 0.0163 DB2 0.0440 -0.0509 -0.0129 0.0160 DB4 0.0459 -0.0493 -0.0127 0.0160 DB8 0.0427 -0.0489 -0.0126 0.0160 DB12 0.0424 -0.0479 -0.0125 0.0160 Sym4 0.0391 -0.0459 -0.0109 0.0140 Sym8 0.0470 -0.0490 -0.0126 0.0160 Sym12 0.0381 -0.0445 -0.0100 0.0139 Hard BRR Max Min ResidualAverage RMS Non-WT 0.0397 -0.0373 0.0326 0.0116 DB2 0.0348 -0.0341 0.0320 0.0112 DB4 0.0343 -0.0340 0.0321 0.0112 DB8 0.0369 -0.0316 0.0311 0.0112 DB12 0.0337 -0.0309 0.0309 0.0112 Sym4 0.0334 -0.0293 0.0249 0.0101 Sym8 0.0338 -0.0367 0.0322 0.0112 Sym12 0.0317 -0.0246 0.0222 0.0080

As shown in [Table 1] and [Table 2], in the case of GNSS data acquired from the bridge, better results were obtained at the soft threshold. Therefore, in the case of the mowavelet, which is intended to be pursued in the present invention, the sym12 of the soft threshold value is considered first of all for the bridge according to the present experiment.

In order to confirm the applicability of wavelets to the correction of GNSS multipath error obtained in public bridges, about 5 (300 minutes) time data was extracted from the previously acquired data, and the effects of multipath reduction were investigated. A total of six satellites (PRN 9,12,14,18,22,30) were observed during the entire observation session, creating and adjusting satellite combinations with different altitude angles so that all conditions are identical and only multipath effects can be considered. Coordinate values were obtained and wavelet transform was applied.

Since the most stable elevation angle was maintained during the observation period of PRN 22, the satellite was selected as a reference satellite. The satellite combinations are high altitude satellite combinations 22, 14, and 18 as shown in Figs. Combinations (22, 9, 12) were used and Mosymbrits and Thresholds were Soft Sym12.

Two conditions divided by satellite altitude were found to show relatively high residuals in satellite pairs with low altitude, as expected.

FIG. 5 shows adjustment values and wavelet application values of the high elevation satellite combinations 22, 14 and 18, and FIG. 6 shows adjustment values and wavelet application values of the low elevation satellite combinations 22, 9 and 12. Shows. As shown in FIGS. 5 and 6, the difference was very large in the adjustment value according to the difference in the satellite altitude, but the difference was reduced in the result value of applying the wavelet. [Table 3] below is a statistical value for the result of reducing the multipath error through the WT.

Application of Wavelet Transformation to GNSS data Highest pair Max (m) Min (m) Residual Mean RMS Non-WT 0.048 -0.0498 0.0249 0.0238 Applied WT 0.0379 -0.0314 0.019 0.0195 Improvement - - - 22% Lowest pair Max (m) Min (m) Residual Mean RMS Non-WT 0.141 -0.061 0.022 0.0274 Applied WT 0.039 -0.0314 0.0137 0.016 Improvement - - - 71%

As can be seen from the above statistics and graphs, it can be seen that WT can not only reduce the fluctuation of the data but also significantly reduce the effect on multipath, thereby increasing the accuracy of the observations.

After the multipath error is corrected through the wavelet transform, other excessive errors are checked, detected, and corrected through the quality control algorithm.

Since the bridges are randomly generated according to the moving loads, it is necessary to check whether the displacements are the result of actual bridge movements or the excessive errors included in the GNSS positioning results.

Excessive errors that may be caused by sudden changes in ionospheric effects, rapid increases in multipath errors, or malfunctions in receivers in GNSS positioning are not considered in the function model, and if they are included in the observed data, they cause large displacements in the estimated coordinates. In order to estimate the final receiver position, these over-errors must be eliminated to provide users with the required accuracy.

To this end, a series of procedures for 'Detection', 'Identification' and 'Adaptation' of the excessive errors included in GNSS observation data processing is called 'Quality Control'. This theory of statistical quality control has already been used as a criterion for evaluating the quality and reliability of observations in the fields of surveying and topographical information.

However, the quality control process in the dynamic positioning field has rarely been carried out so far, and there is no case in the monitoring of bridge behavior based on GNSS. Therefore, this process is essential to provide users with more accurate data.

Excessive errors included in the GNSS location information extracted through the adjustment of the observation data are detected, identified and applied during the quality control process.

First, the detection process is to detect the presence of excessive error in the observed data, and is a process of confirming the excessive error calculated by the post-variance factor through hypothesis verification. In order to detect the presence of excessive error in the observed data, statistical hypothesis testing is performed on the post-variance factor, and the null and alternative hypotheses are expressed as follows

The rejection area of the null hypothesis is set as follows by the F ratio at the number of excess observation data, the significance level α, and the power of test 1-β.

When the null hypothesis is rejected according to the above equations, it is determined that the excessive error is included in the observed value, and at least one tracking satellite is required in the GNSS positioning because at least one redundant observation data is required for this purpose.

, 추정한 관측데이터를

, 정규 분포를 따르는 우연오차를

그리고 함수모형에 고려되지 않은 오차를

이라고 할 때 과대오차 규명을 위한 귀무가설과 대립가설은 다음 식과 같이 설정된다.After the detection process, it is necessary to identify observations that include excessive errors. I observation data from observation vector L

, The estimated observation data

, Coincidence error

And the error that is not considered in the function model

In this case, the null hypothesis and the alternative hypothesis for identifying the excessive error are set as follows.

Statistical values for hypothesis testing are computed through w testing, which is known to follow the standard normal distribution.

The adoption of the null hypothesis is determined by both tail tests on the standard normal distribution (N), and the rejection area for this is set as shown below and if it is satisfied, it is determined that an excessive error is included.

The application process is the process of removing contaminated observations identified after the detection and identification process. After removing the contaminated observations, it is repeated until the alternative hypothesis verification converges through the least-squares estimation.

Reliability is used as a reliability index indicating the performance of statistical verification used in the quality control procedure of GNSS observation data processing. Reliability is largely divided into internal reliability and external reliability, and internal reliability is simply referred to as MDB (Minimally Detectable Bias), which represents the minimum size of over-error found in the identification of over-errors present in the observed value. The external reliability represents the effect on the GNSS receiver coordinate estimation when the undetected overerror is used for the adjustment. If the estimated coordinate reliability is high, the internal and external reliability should have a small value.

The following is the internal reliability.

If there is an excessive error in the observed value, the type 1 error of the hypothesis verification will remain in the beta period of the significance interval α and the type 2 error. Generally, the verification power is set at 20%.

MDBs affect the covariance matrix of surplus observations and residuals, but are not affected by observations and residuals themselves.

If excessive errors that are not detected by the quality control procedure exist, their effects must be clearly identified and this process is defined as external reliability.

Where Q _x is the covariance matrix of the adjusted observation vector, A ^T is the coefficient matrix, and g _m is MDB.

Claims (8)

Correcting the multipath error, including de-nosing, using wavelet transformation in the multipath error correction included in the GNSS position coordinate value; And a quality control process for removing an excessive error included in the GNSS position coordinate value.
delete The method of claim 1,
The wavelet transformation is a GNSS-based bridge monitoring over-error processing method characterized by using a multi-resolution-analysis (MRA) method which is a fast discrete wavelet algorithm. . The method of claim 3,
The wavelet transformation is GNSS-based bridge monitoring over-error processing method characterized in that the noise is selectively separated into a high frequency band and a low frequency band. delete The method of claim 1,
The quality control process to remove the excessive error contained in the GNSS position coordinate value,
A detection step for detecting the presence of excessive error in the observation data;
Identifying excessive observations, identifying the observed values including the excessive errors; And
GNSS-based bridge monitoring over-error processing method characterized in that it comprises an adaptation step of removing the contaminated observations identified after the detection and identification step. The method according to claim 6,
The detection step is GNSS-based bridge monitoring over-error processing method characterized in that for performing the statistical hypothesis test for the post-dispersion factor. The method according to claim 6,
The quality control process,
Statistics using internal reliability, which is the smallest magnitude of the excess error that can be found in the identification of the excess error present in the observations, and an external reliability that indicates the effect on the GNSS receiver coordinate estimates when an undetected excess error is used for the estimation. Bridge monitoring over-error processing method based on GNSS characterized by indicating the performance of the verification.

Images