WO2023193304A1 - Crack detection method and detection device based on data fusion and storage medium - Google Patents

Abstract

A crack detection method based on data fusion, comprising: acquiring multiple crack response signals, and performing continuous wavelet transform on the crack response signals, each crack response signal comprising at least one channel (101); for multiple translation quantity factors in each channel, respectively calculating a maximum crack test statistic within a scale factor range according to a wavelet transform coefficient, and determining a continuous first translation quantity factor range of each channel, a first translation quantity factor satisfying the condition that the maximum crack test statistic corresponding to at least one detection principle is greater than a test threshold (102); and selecting an optimal crack test statistic, and determining a corresponding crack position (103). According to the method, the advantages of various detection technologies can be complemented, and the crack detection probability is greatly improved under the condition of a relatively small constant false alarm rate. Suitable objects of the crack detection method and detection device comprise but are not limited to the field of defect detection such as oil and gas pipeline girth welding crack detection, storage tank bottom plate crack detection, and steel rail crack detection.

Description

Crack detection method, detection device and storage medium based on data fusion

This application requests the priority of the Chinese patent application submitted to the China Patent Office on April 8, 2022, with the application number 202210370023.6 and the invention title "A crack detection method and detection device and storage medium based on data fusion", and its content shall be understood to be incorporated by reference into this application.

Technical field

Embodiments of the present disclosure relate to, but are not limited to, the technical field of defect detection, and in particular, to a crack detection method, detection device, and storage medium based on data fusion. The applicable detection objects include, but are not limited to, detection of girth weld cracks in oil and gas pipelines, and tank bottom plate cracks. Detection, rail crack detection, etc.

Background technique

Ferromagnetic materials such as oil and gas pipelines and petroleum storage tanks have been operating in complex natural environments for a long time, and metal loss or crack defects will occur on the inner and outer wall surfaces. Compared with cracks, metal loss is easier to detect due to its larger size. Currently, there are a variety of relatively mature technologies that can detect metal loss defects. However, crack defects are usually smaller in size and are more difficult to detect, especially under high-speed inspection conditions. Therefore, accurate detection of crack defects is of great significance to the safe operation of oil and gas pipelines and petroleum storage tanks.

Contents of the invention

The following is an overview of the topics described in detail in this article. This summary is not intended to limit the scope of the claims.

The embodiment of the present disclosure provides a crack detection method based on data fusion, including:

Acquire a variety of crack response signals obtained based on different detection principles, perform continuous wavelet transformation on the acquired crack response signals, obtain wavelet transformation coefficients, determine the scale factor and translation factor range, each of the crack response signals includes at least one aisle;

For multiple translation factors in each channel, calculate the maximum crack test statistic within the scale factor range according to the wavelet transform coefficient, and determine the first continuous translation factor range of each channel based on the calculation results, in The maximum crack inspection statistic calculated from the first translation factor position satisfies that the maximum crack inspection statistic corresponding to at least one detection principle is greater than the inspection threshold;

Select an optimal crack inspection statistic from the maximum crack inspection statistic calculated from the continuous first translation factor range of each channel, and determine the crack position corresponding to the optimal crack inspection statistic.

An embodiment of the present disclosure also provides a crack detection device, including a memory; and a processor connected to the memory, the processor being configured to execute any embodiment of the present disclosure based on instructions stored in the memory. The steps of the crack detection method based on data fusion.

An embodiment of the present disclosure also provides a storage medium in a crack detection device, on which a computer program is stored. When the program is executed by a processor, the crack detection method based on data fusion described in any embodiment of the present disclosure is implemented.

The crack detection method, detection device, and storage medium based on data fusion of the embodiments of the present disclosure can calculate the optimal crack inspection statistics based on crack response signals based on multiple detection principles, and can combine the advantages of multiple detection technologies in the same detector. Complementary, crack judgment and positioning are carried out through optimal crack inspection statistics, which improves the detection performance and positioning accuracy of cracks, ensuring that the probability of crack detection is increased under the condition that the false alarm probability remains unchanged; by using the wavelet basis function as the basis for crack detection The reference signal can make full use of the time-frequency analysis advantages of wavelet transform, thereby improving the sensitivity of crack detection. The applicable objects of this crack detection method and detection device include but are not limited to defect detection fields such as oil and gas pipeline girth weld crack detection, storage tank bottom plate crack detection, and rail crack detection.

After reading and understanding the drawings and detailed description, other aspects can be understood.

Description of drawings

The drawings are used to provide an understanding of the technical solution of the present disclosure and constitute a part of the specification. They are used to explain the technical solution of the present disclosure together with the embodiments of the present disclosure and do not constitute a limitation of the technical solution of the present disclosure.

Figure 1 is a schematic flow chart of a crack detection method based on data fusion according to an exemplary embodiment of the present disclosure;

Figure 2 is a graph of the radial component of magnetic flux leakage according to an exemplary embodiment of the present disclosure;

Figure 3 is a graph of a moving magnetic signal according to an exemplary embodiment of the present disclosure;

Figure 4 is a schematic flow chart of another crack detection method based on data fusion according to an exemplary embodiment of the present disclosure;

Figure 5a is the crack test statistic distribution diagram of the channel with the largest peak value in Figure 2;

Figure 5b is the crack test statistic distribution diagram of the channel with the largest peak value in Figure 3;

Figure 6 is a schematic structural diagram of a crack detection device according to an exemplary embodiment of the present disclosure.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. It should be noted that, as long as there is no conflict, the embodiments and features in the embodiments can be arbitrarily combined with each other.

Unless otherwise defined, the technical terms or scientific terms used in the disclosure of the embodiments of the present disclosure shall have the usual meanings understood by those with ordinary skill in the art to which the disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. Words such as "include" or "include" mean that the elements or things preceding the word include the elements or things listed after the word and their equivalents, without excluding other elements or things.

Currently, in scenarios such as oil and gas pipelines and petroleum storage tanks, crack detection usually uses a single detection technology such as magnetic flux leakage, eddy current or ultrasonic, electromagnetic ultrasonic, etc., and the detection effect has limitations. For example, magnetic flux leakage testing has poor detection effect on cracks in the parallel excitation direction, and the signal-to-noise ratio of the response signal is low; ordinary eddy current testing and pulsed eddy current testing can only detect surface defects; ultrasonic testing is only suitable for oil pipelines and not for natural gas. pipelines, and the speed is lower than 2m/s; electromagnetic ultrasound is suitable for oil and gas pipelines, but the energy conversion efficiency is low and the detection speed is usually lower than 2m/s. If multiple detection technologies are implemented in the same detector and the acquired measurement data are fused, complementary advantages can be achieved, which in turn helps improve the performance of crack detection.

As shown in Figure 1, the embodiment of the present disclosure provides a crack detection method based on data fusion, which includes the following steps:

Step 101: Acquire a variety of crack response signals obtained based on different detection principles, perform continuous wavelet transformation on the acquired crack response signals, obtain wavelet transformation coefficients, determine the scale factor and translation factor range, each crack response signal includes at least one channel ;

Step 102: For multiple translation factors in each channel, calculate the maximum crack test statistic within the scale factor range according to the wavelet transform coefficient, and determine the continuous first translation factor range of each channel based on the calculation results, The maximum crack inspection statistic calculated at the first translation factor position satisfies that the maximum crack inspection statistic corresponding to at least one detection principle is greater than the inspection threshold;

Step 103: Select an optimal crack inspection statistic from the maximum crack inspection statistic calculated from the continuous first translation factor range of each channel, and determine the crack position corresponding to the optimal crack inspection statistic.

The crack detection method based on data fusion in the embodiment of the present disclosure can complement the advantages of multiple detection technologies in the same detector by calculating optimal crack inspection statistics based on crack response signals based on multiple detection principles. The test statistics are used for crack judgment and positioning, which improves the detection performance and positioning accuracy of cracks, ensuring that the probability of crack detection is increased while the probability of false alarm remains unchanged; by using the wavelet basis function as a reference signal for crack detection, it can be fully utilized The advantages of time-frequency analysis of wavelet transform can further improve the sensitivity of crack detection.

The crack detection method based on data fusion in the embodiment of the present disclosure can perform data fusion processing on measurement data obtained by multiple detection technologies, including but not limited to magnetic flux leakage detection technology, moving magnetic detection technology, and ordinary eddy current detection technology. , pulsed eddy current detection technology, etc., the measurement data obtained by multiple detection technologies can be obtained by one detector or multiple detectors, and the embodiment of the present disclosure does not limit this. For example, for the same measured area, the same detector contains at least two array signals of detection principles. These array signals are not coupled to each other for calculation, but will be jointly considered in the process of detecting cracks. By combining multiple Set up an array signal matrix to jointly complete the detection of crack defects.

In some exemplary embodiments, in step 101, the method may further include: selecting a wavelet basis function similar in shape to the crack response signal according to different detection principles.

In the embodiments of the present disclosure, a wavelet basis function with a shape similar to the crack response signal can be manually selected, or a wavelet basis function with a shape similar to the crack response signal can be selected through a computer program. The embodiments of the present disclosure do not limit this.

For example, when manually selecting a wavelet basis function that is similar in shape to the crack response signal, we know through theoretical analysis that the waveform of the crack response signal is a waveform similar to a sine wave. Then, when selecting a wavelet basis function, select a shape more like Wavelet basis functions for sine waveforms.

For example, when a wavelet basis function with a similar shape to the crack response signal is selected by a computer program, a standard crack defect can be manually processed in advance and its crack response signal is obtained, and then the crack response signal is combined with the wavelet basis function set. Each wavelet basis function is used to calculate the crack test statistic. After polling all a and b in the definition domain [a ₁ , a _max1 ] and [b ₁ , b _max2 ], each wavelet basis function obtains a maximum crack Test statistic, select the wavelet basis function that maximizes the calculated value of the maximum crack test statistic as the wavelet basis function most similar to the shape of the crack response signal.

Assume that there are N detection principles for crack detection in the same measured area. According to different detection principles, a wavelet basis function similar in shape to the crack response signal is selected. The wavelet basis functions corresponding to these detection principles are respectively

Among them, a and b are both

The independent variable of , a is called the scale factor, a is a real number, and a∈[a ₁ ,a _max1 ], which represents the scale identification range of the detection object; b is called the translation factor, b is a real number, and b∈[b ₁ ,b _max2 ], which characterizes the axial length of the detection area.

The scale factor a usually depends on the period (or frequency) of the crack signal. For example, a can range from 0.5 to 2.0. In actual use, the minimum value a ₁ usually has a certain value for different wavelet basis functions. For example, assume that the wavelet basis function corresponding to a detection principle is selected as Gaussian wavelet No. 1, and the wavelet basis function corresponding to a detection principle is selected as No. 5. Gaussian wavelet, since the minimum value a 1 corresponding to Gaussian wavelet No. ₁ is 0.2, and the minimum value a _{1 corresponding to Gaussian wavelet No. 5} is 0.5, then when choosing the same definition domain [a ₁ , a _max1 ] for different wavelet basis functions When , a ₁ can be taken as 0.5. The value of a _max1 can usually be determined based on engineering experience. For example, a _max1 of 2 means that when a _max1 is less than or equal to 2, the corresponding defect is a crack. When a _max1 is greater than 2, the corresponding defect may be wider than the crack. Metal loss, such as corrosion defects, therefore, the value of a _max1 limits the type of defect under study.

The translation factor b is equivalent to the time domain interval. For example, b can be 0 to 100, 70 to 80, etc., and the embodiment of the present disclosure does not limit this. For example, when the target signal appears in the range of 70 to 80, b ₁ to b _max2 can range from 65 to 85. For another example, for an oscilloscope, if the display length of a single screen is 0 to 10, b ₁ to b _max2 can range from 0 to 10.

In some exemplary implementations, in step 101, the method may further include:

One or more crack response signals are interpolated so that the number of channels of the crack response signals obtained based on different detection principles is equal.

Figure 2 is a graph of the radial component of magnetic flux leakage according to an exemplary embodiment of the present disclosure; Figure 3 is a graph of a moving magnetic signal according to an exemplary embodiment of the present disclosure; each transverse curve in Figures 2 and 3 can be Represents the signal of one channel, the horizontal axis represents the translation factor, and the vertical axis represents the magnetic field distribution. The number of channels in Figure 2 is twice that of Figure 3. Therefore, before performing continuous wavelet transformation on the acquired crack response signal, The moving magnetic signals in Figure 3 are interpolated to make the number of channels of the crack response signals of the two detection principles the same, further simplifying subsequent calculations. However, the embodiments of the present disclosure do not limit this. For the crack response signals in Figures 2 and 3, the moving magnetic signal in Figure 3 does not need to be interpolated. In this case, one channel in Figure 3 can be connected to the crack response signal in Figure 2. corresponding to the two channels.

In some exemplary embodiments, in step 101, the continuous wavelet transform can be a discrete continuous wavelet transform, and the obtained crack response signal can be subjected to a discrete continuous wavelet transform according to the following formula:

Among them, x _i,k [n] is the observation signal of the kth channel of the i-th detection principle, i∈[1,N], N is a natural number greater than or equal to 2, k∈[1,m], m is greater than or a natural number equal to 1,

is the wavelet basis function,

is the wavelet transform coefficient, ψ ^* [n] represents the conjugate operation of ψ[n], when ψ is a real wavelet, since the conjugate of the real number is itself, therefore,

When ψ is a complex wavelet, it needs to be conjugated. The scale factor a is a real number, and a∈[a ₁ ,a _max1 ], and the translation factor b is a real number, and b∈[b ₁ ,b _max2 ]; * in Δ*a is the multiplication sign, and Δ is the basic wavelet ψ The half-width of [n], Δ is a real number, dj is the sampling step size, dj can be an integer or a decimal, for example, when the sampling points of the observed signal x _i,k [n] are 1, 2, 3... like this sequence, dj=1.

The continuous wavelet transform in the embodiment of the present disclosure may be a discrete continuous wavelet transform, or may not be a discrete continuous wavelet transform, and the embodiment of the present disclosure does not limit this.

In actual use, the value of Δ can be determined based on the selected wavelet basis function. For example, still taking the aforementioned Gaussian wavelet No. 1 and Gaussian wavelet No. 5 as an example, Gaussian wavelet No. 1 is narrower, only

The integration result within is not 0. Gaussian wavelet No. 5 is wider, basically

The integral result within is not 0, therefore, the complement of the two can be taken, that is

That is Δ=5.

In some exemplary embodiments, after step 101, the method may further include: establishing a crack test statistic matrix corresponding to the i-th detection principle according to the following formula:

in,

expresses right

Taking the absolute value, T _i is a three-dimensional matrix of m×max2×max1, and a sub-two-dimensional matrix represents the max1×max2 crack test statistic matrix of one channel.

In the crack detection method of the embodiment of the present disclosure, it is assumed that the background noise standard deviation of the i-th detection principle in the k-th channel is σ _i,k , and the inverse function of the complementary cumulative distribution function of the standard normal distribution is Q ^-1 (·) , the false alarm probability is P _FA , then the crack inspection statistics and the detection criterion for the existence of cracks can be defined as:

in,

expresses right

Take the absolute value, i∈[1,N]; k∈[1,m];

is the detection threshold; if _Ti (a, b, k) exceeds the detection threshold, then there is a crack at the (a, b, k) position, otherwise there is no crack at the (a, b, k) position.

In this embodiment, the method for obtaining the background noise standard deviation σ _i,k can be as follows: taking a section of noisy crack response signal that is known to have only a background signal and no crack signal, subtracting the mean and then calculating the standard deviation, that is, σ _i, can be obtained _k . The false alarm probability P _FA may be a pre-specified value. For example, the false alarm probability P _FA may be pre-specified as 0.05. Generally speaking, set a lower constant false alarm probability P _FA and optimize the detector parameters so that the higher the detection probability, the better.

In some exemplary embodiments, in step 102, determining the continuous first translation factor range of each channel according to the calculation results may include:

For the currently calculated translation factor, determine whether the maximum crack inspection statistic corresponding to each detection principle is greater than the inspection threshold;

When there is a maximum crack inspection statistic corresponding to any detection principle that is greater than the inspection threshold, the maximum crack inspection statistic corresponding to the detection principle, and the translation factor and scale factor corresponding to the maximum crack inspection statistic are saved in the cache, and Check whether the translation factor of the current channel has been calculated;

When the translation factor of the current channel has not been calculated, the currently calculated translation factor is incremented by the step size, and the maximum crack test statistic within the scale factor range is calculated cyclically for the auto-increased translation factor until the current After the calculation of the translation factor of the channel is completed or the maximum crack test statistic corresponding to all detection principles is less than or equal to the test threshold, the range of the translation factor in the cache constitutes the continuous first translation factor range of each channel.

In the crack detection method of the embodiment of the present disclosure, in the three-dimensional matrix of each detection principle, the search is performed row by row and column by row in the direction of the data channel (m-dimensional direction) and the direction of the translation factor (b-dimensional direction). For example, The translation factor in Figure 2 and Figure 3 is searched from left to right, and the channel is searched from top to bottom, row by row, column by column. However, the embodiment of the present disclosure does not limit this. For example, the direction of the translation factor can also be from right to left. Searches can be performed, and channel directions can also be searched in any order.

In the crack response signal corresponding to each detection principle, in the first channel data, find the maximum crack test statistic value in the variation range of the scale factor a for each position of the translation factor b; in the second channel data, for each position of the translation factor b, Find the maximum crack test statistic value in the variation range of scale factor a for each position of translation factor b,..., that is, for each (i, k, b), there is a maximum crack test statistic value corresponding to it. The maximum crack test statistic value corresponds to the maximum detection probability.

In the embodiment of the present disclosure, a golden section algorithm based _on a dynamic search domain _can be used to calculate the maximum crack test statistic maxT _{i (} b _q ,k), where [a _i,L ,a _i,R ] is the search range of the scale factor of the i-th detection principle in this step, i∈[1,N], b _q∈ [b ₁ , b _max2 ], k∈[1,m]. The crack detection method of the embodiment of the present disclosure can speed up the overall speed of detecting cracks in a three-dimensional matrix through the golden section algorithm based on the dynamic search domain, which is beneficial to the application of the crack detection method of the embodiment of the disclosure in actual engineering.

In some exemplary embodiments, in step 102, for multiple translation factors in each channel, calculating the maximum crack test statistic within the scale factor range according to the wavelet transform coefficients may include:

For the currently calculated translation factor b _q , set the search boundary initial values a _i,L = a ₁ , a _i,R = a _max1 ;

Calculate the crack test statistics y ₁ and y ₂ of the golden section points x ₁ and x ₂ respectively according to the following formula: x ₁ =a _i,L +0.382(a _i,R -a _i,L ),

x ₂ =a _i,L +0.618(a _i,R -a _i,L ),

Update the golden section point according to the following formula, and calculate the value of the crack test statistic of the updated golden section point in a loop until a _i,R -a _i,L ≤ε, ε is the preset value, for example, ε =0.01: When y ₁ ≥ y ₂ , a _i,R =x ₂ ; x ₂ =x ₁ ; y ₂ =y ₁ ; x ₁ =a _i,L +0.382(a _i,R -a _i,L );

When y ₁ <y ₂ , set a _i,L =x ₁ ; x ₁ =x ₂ ; y ₁ =y ₂ ; x ₂ =a _i,L +0.618(a _i,R -a _i,L );

It is determined that the optimal scale factor corresponding to the currently calculated translation factor b _q is x ₁ or x ₂ , and the maximum crack test statistic corresponding to the currently calculated translation factor b _q is y ₁ or y ₂ .

In some exemplary embodiments, in step 102, for multiple translation factors in each channel, calculate the maximum crack test statistic within the scale factor range according to the wavelet transform coefficient, which may also include:

When the maximum crack test statistic corresponding to the currently calculated translation factor b _q is greater than the test threshold, for the next translation factor b _q+1 of the current channel, set the search boundary initial value a _i,L =max{a ₁ , a _q -δ}; a _i,R =min{a _max1 ,a _q +δ}, where δ is the adjustment amount of the search domain;

When the maximum crack test statistic corresponding to the currently calculated translation factor b _q is less than or equal to the test threshold, set the search boundary initial value a _i,L = a ₁ for the next translation factor b _q+1 of the current channel; a _i,R = a _max1 .

For example, δ=0.15(a _max1 -a _min ), δ can also be set to other values, and the embodiment of the present disclosure does not limit this.

In some exemplary embodiments, in step 103, selecting an optimal crack test statistic from the maximum crack test statistic calculated from the continuous first translation factor range of each channel may include:

Determine the local optimal crack inspection statistic corresponding to each detection principle. The local optimal crack inspection statistic is one maximum crack inspection statistic corresponding to each detection principle, or multiple maximum crack inspection statistics corresponding to each detection principle. The largest crack test statistic in

Detect the number of local optimal crack test statistics;

When the number of local optimal crack test statistics is 1, the local optimal crack test statistic is used as the optimal crack test statistic;

When the number of local optimal crack test statistics is greater than 1, calculate the signal-to-noise ratio corresponding to each local optimal crack test statistic, and select the local optimal crack test statistic with the highest signal-to-noise ratio as the optimal crack test statistic. quantity.

In the embodiment of the present disclosure, multiple detection principles are synchronously detected point by point. Every time a position point (b, k) is calculated, each detection principle may find a maximum crack detection statistic that is greater than the detection threshold or multiple detection statistics that are greater than the detection threshold. The maximum crack test statistic value for the threshold. The local optimal crack test statistic T _{i_max} refers to the largest maximum crack test statistic in the cache corresponding to each detection principle. For example: when reaching a certain position point (b, k), the cache ROM ₁ corresponding to the first detection principle stores 10 maxT ₁ (b, k) and the corresponding 10 a and 10 b; The cache ROM ₂ corresponding to the second detection principle stores 5 maxT ₂ (b, k) and the corresponding 5 a and 5 b. Then, in this step, find a local optimum in ROM ₁ respectively. The crack test statistic T _{1_max} and the corresponding a _{1_best} and b _{1_best} are found in ROM _2. A local optimal crack test statistic T _{2_max} and the corresponding a _{2_best} and b _{2_best} are found. Then one of the multiple local optimal crack test statistics is selected as the final optimal crack test statistic based on the signal-to-noise ratio. When there is only one local optimal crack test statistic, the local optimal crack test statistic is is the final optimal crack test statistic.

In some exemplary embodiments, the signal-to-noise ratio SNR _i corresponding to each local optimal crack test statistic can be calculated according to the following formula:

in,

In some exemplary embodiments, in step 103, the crack position corresponding to the determined optimal crack test statistic is [b _best -Δ*a _best , b _best +Δ*a _best ], and b _best is the optimal crack The translation factor corresponding to the test statistic, a _best is the scale factor corresponding to the optimal crack test statistic.

Each channel k may have one or more optimal crack test statistics (corresponding to one or more local area cracks). Different channels k include the spatial location of the respective local area cracks, and the local area cracks of adjacent channels k The spatial position simultaneously draws the overall two-dimensional range of the crack.

The two-dimensional space of the overall crack drawn simultaneously with the spatial position of the crack in the local area of adjacent channel k may be an irregular figure. Draw a rectangle around the outer edge of the irregular figure, and use this rectangle to identify the two-dimensional space of the overall crack. The scope of the space is shown in the rectangular boxes in Figures 2 and 3.

In some exemplary embodiments, as shown in Figure 4, embodiments of the present disclosure provide a crack detection method, including the following steps:

1) According to different detection principles, select a wavelet basis function that is similar in shape to the crack response signal; the crack response signal here includes but is not limited to magnetic flux leakage detection signal, moving magnet detection signal, ordinary eddy current detection signal, pulse eddy current detection signal, etc.;

Assume that there are N detection principles for crack detection in the same measured area. The wavelet basis functions corresponding to these detection principles are:

Among them, a and b are the independent variables of ψ, a is a real number, and a∈[a ₁ ,a _max1 ], which is called the scale factor, which represents the scale identification range of the detection object; b is a real number, and b∈[b ₁ , b _max2 ], called the translation factor, which characterizes the axial length of the detection area; for different wavelet basis functions, a or b respectively select their respective domains [a ₁ , a _max1 ] or [b ₁ , b _max2 ].

2) Discretize the continuous wavelet transform formula;

Assume that different detection principles have m data channels in the detection area; the continuous wavelet transform discretization formula of the observation signal x _i,k [n] of the i-th detection principle in the k-th channel is defined as:

Among them, ψ[n] is the basic wavelet or mother wavelet, ψ ^* [n] represents the conjugate operation of ψ[n], Δ is the half-width of ψ[n], Δ is a real number, and dj is the sampling step size.

3) Define crack inspection statistics and detection criteria;

Assume that the background noise standard deviation of the i-th detection principle in the k-th channel is σ _i,k , the inverse function of the complementary cumulative distribution function of the standard normal distribution is Q ^-1 (·), and the false alarm probability is P _FA , then the crack The test statistic and the detection criterion for the presence of cracks are defined as:

in,

expresses right

Take the absolute value, i∈[1,N]; k∈[1,m];

is the detection threshold; if _Ti (a, b, k) exceeds the detection threshold, then there is a crack at the (a, b, k) position, otherwise there is no crack at the (a, b, k) position.

The test statistic matrix defined for the i-th detection principle in the detection area is:

T _i is a three-dimensional matrix of m×max2×max1. Each sub-two-dimensional matrix represents the test statistic matrix of a channel. In the a-dimensional direction, it includes a total of max1 rows from a ₁ to a _max1 . In the b-dimensional direction, it includes from b ₁ to b _max2 total max2 columns.

4) Calculate the maximum test statistic of the crack at each location point. In the three-dimensional matrix of each detection principle, the search is traversed row by row and column by row in the direction of the data channel (m-dimensional direction) and the direction of the translation factor (b-dimensional direction). If the row and column search is not ended, then under the condition of determining the row and column, Calculate the maximum test statistic maxT _i (b,k),i∈[1,N] for the position point (b,k ₎ in the range of [a 1 ,a _max1 ]; if all row and column searches are completed, jump to the step 10);

5) Determine whether the maximum test statistic of the crack at each location point exceeds the detection threshold. If for any i∈[1,N], it satisfies

Then jump to step 6); otherwise if i∈[1,N] exists, satisfy

Then save the corresponding a, b, maxT _i (b, k) in the cache ROM _i ; if the search for this column (b-dimensional direction) has not ended, jump to step 4), otherwise jump to step 6);

6) Calculate the maximum test statistic for cracks in the non-empty cache. If all ROM _i is empty, jump to step 4), otherwise for all non-empty ROM _i , calculate the local maximum test statistic _{Ti_max} and the corresponding a _{i_best} , b _{i_best} and signal-to-noise ratio

in

7) Screen the optimal scale factor and translation factor in the local area. If there is only one non-empty cache ROM _i in the local area, then the optimal scale factor a _best = a _{i_best} in the local area, and the optimal translation factor b _best = b _{i_best} ; if there are multiple non-empty caches in the local area, then take The scale factor and translation factor corresponding to the non-empty buffer with the highest signal-to-noise ratio in the area are a _best and b _best ;

8) Determine the spatial location of cracks in local areas. The spatial position of the crack in the local area is [b _best -Δ*a _best , b _best +Δ*a _best ];

9) Clear all temporary cache ROM _i . For any i∈[1,N], clear ROM _i and jump to step 4);

10) The testing process is officially ended.

In step 4), in order to improve the speed of calculating the maximum test statistic maxT _i (b _q ,k) of a certain position point (b _q ,k), a golden section algorithm based on the dynamic search domain can be used, which calculates The body process is as follows:

S1. Whenever a data channel is newly switched (k is updated), the initial values a _i,L = a ₁ , a _i,R = a _max , are set, and the temporary cache ROM _i is cleared;

S2) Calculate x ₁ =a _i,L +0.382(a _i,R -a _i,L ),

x ₂ =a _i,L +0.618(a _i,R -a _i,L ),

S3) Repeat step S4 until a _i,R -a _i,L ≤ε, where for example ε=0.01; then jump to step S5;

S4) If y ₁ ≥ y ₂ , set a _i,R = x ₂ ; x ₂ = x ₁ ; y ₂ = y ₁ ; x ₁ = a _i,L +0.382(a _i,R -a _i,L ) ;

Otherwise, set a _i,L =x ₁ ; x ₁ =x ₂ ; y ₁ =y ₂ ; x ₂ =a _i,L +0.618(a _i,R -a _i,L );

S5) The current optimal scale factor is a _q =x ₁ , and the maximum test statistic maxT _i (b _q ,k)=y ₁ of the location point (b _q ,k);

S6) If

Set a _i,L =max{a ₁ ,a _q -δ}; a _i,R =min{a _max ,a _q +δ} as the search domain boundary of the next point (b _q+1 ,k), where δ is the adjustment amount of the search domain; for example, δ=0.15(a _max -a _min ) can be set;

Otherwise, set a _i,L =a ₁ ; a _i,R =a _max as the search domain boundary of the next position point (b _q+1 ,k).

The technical solutions of the embodiments of the present disclosure will be further described below in conjunction with non-destructive testing of oil and gas pipelines.

Non-destructive testing of oil and gas pipelines usually adopts the principle of magnetic flux leakage internal testing, because compared with other electromagnetic non-destructive testing technologies, magnetic flux leakage testing technology has many advantages such as simple principle, easy engineering implementation, and high detection efficiency. However, the magnetic flux leakage detection principle is not sensitive enough to small defects such as cracks, especially the detection ability of small defects in the parallel excitation direction is limited. Therefore, it is hoped that the shortcomings of single magnetic flux leakage detection can be made up by integrating other detection technologies. This embodiment lists a crack detection method based on the fusion of magnetic flux leakage and moving magnetic data. The moving magnetic detection technology has high sensitivity for detecting crack defects in any direction, which can make up for the shortcomings of the magnetic flux leakage detection technology. The implementation steps of the method are introduced in detail below.

First, we use an in-oil and gas pipeline detection sensor that combines magnetic flux leakage and moving magnetism to simultaneously implement magnetic flux leakage detection and moving magnetism detection for the same measured area, and obtain magnetic flux leakage detection data and moving magnetism detection data at the same time.

As an example, we choose the magnetic flux leakage radial component signal (MFLY) and the moving magnet signal (DM) for data fusion. During a certain pipeline inspection process, the MFLY signal obtained for the same measured area is shown in Figure 2, and the DM signal is shown in Figure 3. In this embodiment, the number of magnetic leakage channels is twice the number of moving magnetic channels. Before performing the following steps, the number of moving magnetic channels can be expanded to be equal to the number of magnetic leakage channels through the cubic spline interpolation method.

According to step 1), select No. 1 Gaussian real wavelet gaus1 as the wavelet basis function of the MFLY signal, and select No. 5 Gaussian real wavelet gaus5 as the wavelet basis function of the DM signal.

According to step 2), the continuous wavelet transform of the MFLY signal is discretized according to the following formula, where

Assume that the sampling step size is equal to 1;

According to step 2), the continuous wavelet transform of the DM signal is discretized according to the following formula, where

Assume that the sampling step size is equal to 1;

According to step 3), assume that the background noise standard deviation of the MFLY signal in the k-th channel is σ _MFLY,k , the inverse function of the complementary cumulative distribution function of the standard normal distribution is Q ^-1 (·), and the false alarm probability is P _FA , Then the magnetic flux leakage test statistics of cracks and the detection criteria for the existence of cracks are defined as:

According to step 3), assume that the background noise standard deviation of the DM signal in the k-th channel is σ _DM,k , the inverse function of the complementary cumulative distribution function of the standard normal distribution is Q ^-1 (·), and the false alarm probability is P _FA , Then the dynamic magnetic test statistics of cracks and the detection criteria for the existence of cracks are defined as:

According to step 4), calculate the maximum crack test statistic at each position point of the magnetic flux leakage signal in Figure 2 and the moving magnetic signal in Figure 3. In the three-dimensional matrix of the magnetic flux leakage signal and the moving magnetic signal, the search is performed row by row and column by row in the direction of the data channel (m-dimensional direction) and the direction of the translation factor (b-dimensional direction). Under the condition, calculate the maximum test statistics maxT _MFLY (b,k) and maxT _DM (b,k) of the position point (b,k) in the range of [a ₁ , a _max1 ]; if all row and row searches are completed, then Jump to step 10);

Follow step 5) if

and

Then jump to step 6); otherwise if

or

Then save the corresponding a _MFLY , b _MFLY , maxT _MFLY (b,k) in the cache ROM _MFLY , or save the corresponding a _DM , b _DM , maxT _DM (b,k) in the cache ROM _DM ; if this If the column (b-dimensional direction) search has not ended, jump to step 4), otherwise jump to step 6);

Follow step 6), if ROM _MFLY and ROM _DM are empty, jump to step 4); otherwise, if ROM _MFLY is not empty, calculate the local maximum test statistic T _{MFLY_max} and the corresponding a _{MFLY_best} , b _{MFLY_best} and SNR. _MFLY ; if ROM _DM is not empty, calculate the local maximum test statistic T _{DM_max} and the corresponding a _{DM_best} , b _{DM_best} and SNR _DM ;

According to step 7), if the local area ROM _MFLY is not empty and ROM _DM is empty, then a _best =a _{MFLY_best} , b _best =b _{MFLY_best} ; if the local area ROM _DM is not empty and ROM _MFLY is empty, then a _best =a _{DM_best} , b _best = b _{DM_best} ; if the local area ROM _MFLY is not empty and the ROM _DM is not empty, then further determine if SNR _MFLY ≥ SNR _DM , then a _best = a _{MFLY_best} , b _best = b _{MFLY_best} ; and if SNR _DM >SNR _MFLY , then a _best =a _{DM_best} , b _best =b _{DM_best} ;

According to step 8), the spatial location of the crack in the local area is [b _best -5a _best , b _best +5a _best ];

Follow step 9), clear ROM _MFLY and ROM _DM , and jump to step 4);

Follow step 10) to officially end the detection process.

The rectangular frame areas in Figures 2 and 3 are the crack detection results achieved through the above method. A total of 6 cracks were detected in Figures 2 and 3. The specific locations of the cracks are marked with boxes, numbered 1-6 respectively. It can be seen that cracks No. 1 and 2 are almost impossible to identify in Figure 2, but they have a relatively high signal-to-noise ratio in Figure 3; crack No. 6 is almost impossible to identify in Figure 3, but they are in Figure 2 It has a relatively high signal-to-noise ratio; and the crack detection method based on the fusion of magnetic flux leakage and moving magnetic data can combine the advantages of the two detection technologies to detect all cracks in the area.

Figure 5a and Figure 5b are respectively the three-dimensional distribution diagram of the crack test statistic for the channel with the largest peak value of the above-mentioned No. 4 crack. The vertical axis is the test statistic, the horizontal axis scale represents a, and the translation represents b. The maximum value of the MFLY test statistic appears at the 16 channels, the corresponding coordinate position (a, b, T) is (26, 7.80, 1340.12), the maximum DM test statistic appears in the 8th channel, the corresponding coordinate position (a, b, T) is (36, 7.60, 214.33). Since crack No. 4 has a relatively high signal-to-noise ratio in Figures 2 and 3, the MFLY test statistic and DM test statistic of crack No. 4 have relatively obvious peaks in Figure 5a and Figure 5b, respectively. It's 1340.12 and 214.33. By observing the distribution of crack inspection statistics in Figure 5a and Figure 5b, we can help determine whether the detected defect is a crack, thereby reducing the probability of false alarms for cracks. The front view of the crack test statistics of the MFLY signal and the DM signal in Figure 5a and Figure 5b all approximately obey the Gaussian distribution, and the right view approximately obeys the Rayleigh distribution, so it can be confirmed that the defect is a crack defect.

Embodiments of the present disclosure also provide a crack detection device, including a memory; and a processor connected to the memory. The processor executes instructions based on the memory, and executes the above-mentioned method based on the previous item. Steps of data fusion crack detection method.

In one example, as shown in FIG. 6 , the crack detection device may include: a processor 610 , a memory 620 , a bus system 630 and a sensor 640 , wherein the processor 610 , the memory 620 and the sensor 640 pass through the bus system 630 Connected, the memory 620 is used to store instructions and the optimal crack test statistics, optimal scale factors, optimal translation factors, etc., and the processor 610 is used to execute the instructions stored in the memory 620. On the one hand, it is used to The sensor 640 is controlled to receive signals, and on the other hand, execute the crack detection procedure. Specifically, the sensor 640 can acquire a variety of crack response signals obtained based on different detection principles under the control of the processor 610. The processor 610 performs continuous wavelet transformation on the acquired crack response signals to obtain wavelet transformation coefficients and determine the scale factor. and translation factor range, each crack response signal includes at least one channel; for multiple translation factors in each channel, the maximum crack test statistic within the scale factor range is calculated respectively according to the wavelet transform coefficient, And determine the continuous first translation factor range of each channel according to the calculation results. The maximum crack inspection statistic calculated at the first translation factor position satisfies that the maximum crack inspection statistic corresponding to at least one detection principle is greater than the inspection threshold. ; Select an optimal crack inspection statistic from the maximum crack inspection statistic calculated from the continuous first translation factor range of each channel, and determine the crack position corresponding to the optimal crack inspection statistic.

It should be understood that the processor 610 can be a central processing unit (Central Processing Unit, CPU). The processor 610 can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), and off-the-shelf programmable gate arrays. (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor 610 may be any conventional processor or the like.

The memory 620 may include a read-only memory and a random access memory, and provide instructions and data to the processor 610, including the optimal crack test statistic, optimal scaling factor, optimal translation factor, etc. A portion of memory 620 may also include non-volatile random access memory. For example, memory 620 may also store device type information.

In addition to the data bus, the bus system 630 may also include a power bus, a control bus, a status signal bus, etc.

During the implementation process, the processing performed by the crack detection device can be completed by instructions in the form of hardware integrated logic circuits or software in the processor 610 . That is, the steps of the crack detection method in the embodiment of the present disclosure can be executed by a hardware processor, or by a combination of hardware and software modules in the processor 610 . Software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media. The storage medium is located in the memory 620. The processor 610 reads the information in the memory 620 and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

Embodiments of the present disclosure also provide a storage medium in a crack detection device. The storage medium in the crack detection device stores executable instructions. When executed by a processor, the executable instructions can implement the provisions of any of the above embodiments of the present disclosure. A crack detection method based on data fusion. This crack detection method can obtain a variety of crack response signals obtained based on different detection principles, perform continuous wavelet transformation on the obtained crack response signals, obtain wavelet transformation coefficients, and determine scale factors and translations. Scale factor range, each crack response signal includes at least one channel; for multiple translation factors in each channel, the maximum crack test statistic within the scale factor range is calculated according to the wavelet transform coefficient, and based on The calculation results determine the continuous first translation factor range of each channel, and the maximum crack inspection statistic calculated at the first translation factor position satisfies that the maximum crack inspection statistic corresponding to at least one detection principle is greater than the inspection threshold; from Select an optimal crack inspection statistic from the maximum crack inspection statistic calculated in the continuous first translation factor range of each channel, and determine the crack position corresponding to the optimal crack inspection statistic. The method of realizing crack detection by executing executable instructions is basically the same as the crack detection method based on data fusion provided by the above embodiments of the present disclosure, and will not be described again here.

Those of ordinary skill in the art can understand that all or some steps, systems, and functional modules/units in the devices disclosed above can be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may consist of several physical components. Components execute cooperatively. Some or all of the components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer-readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As is known to those of ordinary skill in the art, the term computer storage media includes volatile and nonvolatile media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. removable, removable and non-removable media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disk (DVD) or other optical disk storage, magnetic cassettes, tapes, disk storage or other magnetic storage devices, or may Any other medium used to store the desired information and that can be accessed by a computer. Additionally, it is known to those of ordinary skill in the art that communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media .

Although the embodiments disclosed in the present disclosure are as above, the described contents are only used to facilitate the understanding of the present disclosure and are not intended to limit the present disclosure. Any person skilled in the field to which this disclosure belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope of this disclosure. However, the protection scope of this disclosure must still be The scope defined by the appended claims shall prevail.

Claims (12)

1. A crack detection method based on data fusion, including:

   Acquire a variety of crack response signals obtained based on different detection principles, perform continuous wavelet transformation on the acquired crack response signals, obtain wavelet transformation coefficients, determine the scale factor and translation factor range, each of the crack response signals includes at least one aisle;

   For multiple translation factors in each channel, calculate the maximum crack test statistic within the scale factor range according to the wavelet transform coefficient, and determine the first continuous translation factor range of each channel based on the calculation results, in The maximum crack inspection statistic calculated from the first translation factor position satisfies that the maximum crack inspection statistic corresponding to at least one detection principle is greater than the inspection threshold;

   Select an optimal crack inspection statistic from the maximum crack inspection statistic calculated from the continuous first translation factor range of each channel, and determine the crack position corresponding to the optimal crack inspection statistic.
2. The crack detection method according to claim 1, wherein determining the continuous first translation factor range of each channel according to the calculation results includes:

   For the currently calculated translation factor, determine whether the maximum crack inspection statistic corresponding to each detection principle is greater than the inspection threshold;

   When there is a maximum crack inspection statistic corresponding to any detection principle that is greater than the inspection threshold, the maximum crack inspection statistic corresponding to the detection principle, as well as the scale factor and translation factor corresponding to the maximum crack inspection statistic are saved in the cache, and Check whether the translation factor of the current channel has been calculated;

   When the translation factor of the current channel has not been calculated, the currently calculated translation factor is incremented by the step size, and the maximum crack test statistic within the scale factor range is calculated cyclically for the auto-increased translation factor. Until the calculation of the translation amount factor of the current channel is completed or the maximum crack test statistic corresponding to all detection principles is less than or equal to the test threshold, the range of the translation amount factor in the buffer constitutes the continuous first translation amount factor of each channel scope.
3. The crack detection method according to claim 1, wherein selecting an optimal crack inspection statistic from the maximum crack inspection statistic calculated from the continuous first translation factor range of each channel includes:

   Determine the local optimal crack inspection statistic corresponding to each detection principle. The local optimal crack inspection statistic is a maximum crack inspection statistic corresponding to each detection principle, or multiple crack inspection statistics corresponding to each detection principle. The largest maximum crack test statistic among the maximum crack test statistics;

   Detect the number of local optimal crack test statistics;

   When the number of the local optimal crack test statistics is 1, the local optimal crack test statistic is used as the optimal crack test statistic;

   When the number of local optimal crack test statistics is greater than 1, the signal-to-noise ratio corresponding to each local optimal crack test statistic is calculated, and the local optimal crack test statistic with the highest signal-to-noise ratio is selected. as the optimal crack test statistic.
4. The crack detection method according to claim 3, wherein the signal-to-noise ratio SNR _i corresponding to each local optimal crack inspection statistic is calculated according to the following formula:

   

   in,

   

   T _{i_max} is the local optimal crack test statistic, a _{i_best} is the scale factor corresponding to the local optimal test statistic, dj is the sampling step, σ _i,k is the background noise standard deviation of the i-th detection principle in the k-th channel , i∈[1, N], N is a natural number greater than or equal to 2, and Δ is the half-width of the basic wavelet.
5. The crack detection method according to claim 1, wherein the continuous wavelet transform is a discrete continuous wavelet transform, and the obtained crack response signal is subjected to a discrete continuous wavelet transform according to the following formula:

   

   Among them, x _{i, k} [n] is the observation signal of the k-th channel of the i-th detection principle, i∈[1, N], N is a natural number greater than or equal to 2, k∈[1, m], m is greater than or is a natural number equal to 1, ψ[n] is the basic wavelet,

   

   is the wavelet transform coefficient, ψ ^* [n] represents the conjugate operation of ψ[n], the scale factor a is a real number, and a∈[a ₁ , a _max1 ], the translation factor b is a real number, and b∈[b ₁ , b _max2 ], max1 and max2 are both natural numbers greater than 1, Δ is the half-width of ψ[n], Δ is a real number, dj is the sampling step size, and dj is a real number.
6. The crack detection method according to claim 5, further comprising: establishing a crack inspection statistic matrix corresponding to the i-th detection principle according to the following formula:

   

   in,

   

   expresses right

   

   Taking the absolute value, T _i is a three-dimensional matrix of m×max2×max1, and a sub-two-dimensional matrix represents the max1×max2 crack test statistic matrix of one channel.
7. The crack detection method according to claim 5, wherein for the plurality of translation factors in each channel, calculating the maximum crack test statistic within the scale factor range according to the wavelet transform coefficients includes:

   For the currently calculated translation factor b _q , set the search boundary initial value a _{i, L} = a ₁ , a _{i, R} = a _max1 ;

   Calculate the crack test statistics y ₁ and y ₂ of the golden section points x ₁ and x ₂ respectively according to the following formula: x ₁ =a _{i, L} +0.382 (a _{i, R} -a _{i, L} ),

   

   x ₂ =a _i,L +0.618(a _i,R -a _i,L ),

   

   Update the golden section point according to the following formula, and calculate the value of the crack test statistic of the updated golden section point in a loop until a _{i, R} -a _{i, L} ≤ ε, ε is the default value: when y ₁ ≥ y When ₂ , a _{i, R} = x ₂ ; x ₂ = x ₁ ; y ₂ = y ₁ ; x ₁ = a _{i, L} +0.382 (a _{i, R} - _{a i, L} );

   

   When y ₁ < y ₂ , set a _{i, L} = x ₁ ; x ₁ = x ₂ ; y ₁ = y ₂ ; x ₂ = a _{i, L} +0.618 (a _{i, R} - _{a i, L} );

   

   It is determined that the optimal scale factor corresponding to the currently calculated translation factor b _q is x ₁ or x ₂ , and the maximum crack test statistic corresponding to the currently calculated translation factor b _q is y ₁ or y ₂ .
8. The crack detection method according to claim 7, wherein the maximum crack test statistic within the scale factor range is calculated respectively for the plurality of translation factors in each channel according to the wavelet transform coefficient, and further includes:

   When the maximum crack test statistic corresponding to the currently calculated translation factor b _q is greater than the test threshold, for the next translation factor b _q+1 of the current channel, set the search boundary initial value a _{i, L} = max{a ₁ , a _q -δ}; a _{i, R} =min{a _max1 , a _q +δ}, where δ is the adjustment amount of the search domain;

   When the maximum crack test statistic corresponding to the currently calculated translation factor b _q is less than or equal to the test threshold, set the search boundary initial value a _{i, L} = a ₁ for the next translation factor b _q+1 of the current channel; a _{i, R} = a _max1 .
9. The crack detection method according to claim 5, wherein the crack position corresponding to the determined optimal crack inspection statistic is [b _best -Δ*a _best , b _best +Δ*a _best ], and b _best is the a is the translation factor corresponding to the optimal crack test statistic, and a _best is the scale factor corresponding to the optimal crack test statistic.
10. The crack detection method according to claim 1, wherein the inspection threshold is:

    

    Among them, dj is the sampling step size, σ _{i, k} is the background noise standard deviation of the i-th detection principle in the k-th channel, i∈[1, N], N is a natural number greater than or equal to 2, k∈[1, m], m is a natural number greater than or equal to 1, Q ^-1 (·) is the inverse function of the complementary cumulative distribution function of the standard normal distribution, and P _FA is the false alarm probability.
11. A crack detection device, comprising a sensor for detecting a crack response signal, a memory storing instructions; and a processor connected to the memory, the processor can execute the instructions of claim 1 based on the instructions stored in the memory The steps of the crack detection method based on data fusion according to any one of to 10.
12. A storage medium in a crack detection device, on which a crack detection program is stored. When the program is executed by a processor, the crack detection method based on data fusion as described in any one of claims 1 to 10 is implemented.

Images