Micro-seismic / acoustic emission source positioning method for eliminating abnormal value based on non-parametric estimation
Through the combination of non-parametric estimation and damping linear correction, outliers are identified and eliminated, and the problem of outliers in micro-seismic/acoustic emission source positioning is solved, positioning accuracy and robustness are improved, and more efficient positioning results are achieved.
Patent Information
- Application Number
- CN202411091120.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art is difficult to effectively eliminate outliers in micro-seismic/acoustic emission source positioning, resulting in insufficient positioning accuracy and robustness, especially under the influence of low signal-to-noise ratio signal processing and environmental noise, propagation delay and multipath interference. Outliers are difficult to eliminate.
Using a non-parameter estimation method, through the steps of data preparation, preliminary positioning, non-parameter estimation, filtering abnormal data and repositioning, combining four-four combinations and damping linear correction, identifying and eliminating outliers to improve positioning accuracy and robustness. .
It effectively improves the accuracy and robustness of micro-seismic/acoustic emission source positioning, reduces the number of sensors, improves operating efficiency, and reduces positioning errors, and has high practical application value.
Smart Images

Figure CN119000883A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of microseismic / acoustic emission source positioning, and in particular to a microseismic / acoustic emission source positioning method based on non-parametric estimation to eliminate abnormal values. Background Art
[0002] Microseismic / acoustic emission source positioning has important application value in many fields such as mining engineering, geological disaster warning, hydraulic fracture monitoring, structural health monitoring, etc., and can help scientists and engineers evaluate geological stability and potential risks. However, in the process of positioning microseismic / acoustic emission sources, there are many obstacles that affect the accuracy of positioning. Among them, the most important is the accuracy of arrival data. Due to the complex deep mining environment, it has the characteristics of unpredictable initial signal arrival, poor signal-to-noise ratio, abnormal signal interference, and unclear propagation path, which inevitably leads to abnormal data with large errors in the recorded arrival data. On this basis, through the cumulative transmission and secondary amplification of the control equation (sensor geometric system), the error of the arrival data will eventually lead to serious positioning deviation.
[0003] In order to pick up accurate arrival data, researchers have proposed a variety of automatic arrival picking methods. Although they can avoid the subjectivity of manual arrival picking and improve the efficiency of arrival picking, it is still difficult to avoid the generation of abnormal arrivals. Especially when processing signals with low signal-to-noise ratios, even with advanced arrival detection algorithms, it is difficult to completely exclude all abnormal data. In addition, in addition to the influence of high environmental noise, multiple factors such as propagation delay, multipath interference, and sensor failure will also lead to significant arrival errors, and existing arrival picking algorithms cannot remove outliers caused by these factors. Therefore, it is crucial to identify and eliminate outliers from the perspective of source positioning.
[0004] In the prior art, the patent with publication number CN110376290A provides an acoustic emission source positioning method based on multidimensional kernel density estimation. This method does not use the prior knowledge of the data distribution of the preliminary positioning result, obtains a more accurate density estimate, avoids the error caused by the unreasonable assumed distribution, and thus improves the positioning accuracy to a certain extent. However, this patent only uses the quartile method to exclude abnormal positioning results in the preliminary positioning results, but this method does not identify and exclude abnormal sensors, nor does it use clean arrival data for repositioning. Therefore, the final positioning result has no optimization capability, so that the positioning accuracy and robustness of this method still need to be further improved. In addition, the preliminary positioning in this method requires the use of 5 sensors for positioning, the overall operating efficiency is not high, and the positioning accuracy is low.
[0005] In view of this, it is necessary to design an improved microseismic / acoustic emission source positioning method to solve the above problems. Summary of the invention
[0006] In view of the above-mentioned defects of the prior art, the object of the present invention is to provide a microseismic / acoustic emission source positioning method based on non-parametric estimation to eliminate outliers, so as to effectively eliminate outliers in arrival data and improve the accuracy and robustness of positioning.
[0007] To achieve the above object, the present invention provides a method for locating microseismic / acoustic emission sources based on non-parametric estimation to eliminate outliers, comprising the following steps:
[0008] S1. Data preparation: Collect the data collected by the monitoring system, and perform four-by-four combinations on multiple sensors of the monitoring system to obtain a combination of multiple sensors and arrival data;
[0009] S2, preliminary positioning: Based on the data prepared in step S1, an improved analytical method is used to perform preliminary positioning to obtain multiple groups of preliminary positioning results;
[0010] S3, non-parametric estimation: construct a non-parametric estimation function based on the multiple groups of preliminary positioning results obtained in step S2, calculate the probability density of the source coordinates, and determine the coordinates corresponding to the extreme value of the density;
[0011] S4, filtering abnormal data: according to the result obtained in step S3, the abnormal values in the arrival data are eliminated to obtain normal arrival data;
[0012] S5, repositioning: repositioning is performed based on the normal arrival time data obtained in step S4 in combination with the damped linear correction method.
[0013] As a further improvement of the present invention, in step S1, the collected data includes arrival time data, sensor position and wave speed.
[0014] As a further improvement of the present invention, in step S1, the four-four combination method is: randomly select four sensors from n sensors as a group, with a total of Sensor combinations and obtain Group arrival data.
[0015] As a further improvement of the present invention, in step S2, the preliminary positioning result obtained by each sensor combination is:
[0016]
[0017] in: M j and N j The determinant of (j=1,2,3) is obtained by replacing the jth column of D with (e 2 ,e 3 ,e 4 ) and (f2 ,f 3 ,f 4 );a i =x i -x 1 , b i =y i -y 1 , c i =z i -z 1 , f i =v 2 (t i -t 1 ), (x i ,y i ,z i )(i=1,2,3,4) are the spatial coordinates of the sensors in the four sensor combinations, v is the average wave velocity, t 0 is the triggering time of the microseismic / acoustic emission source, t i To reach sensor P i time.
[0018] As a further improvement of the present invention, in step S3, the expression of the non-parametric estimation function is:
[0019]
[0020] in: θ is the source coordinate of the microseismic / acoustic emission source, θ = [x, y, z] T , the superscript T indicates the matrix or vector transpose; is the kth preliminary positioning result, k = 1, 2, ..., m, m is the number of preliminary positioning results; H is the bandwidth matrix.
[0021] As a further improvement of the present invention, in step S4, the method for filtering abnormal data specifically includes the following steps:
[0022] S41. Calculate the coordinate θ corresponding to the density extreme value tri With the initial positioning result θ (k) The Euclidean distance D k , and then arrange these distances in ascending order to get D sort ;
[0023] S42, determine the distance threshold conf, and compare D sort,k The size of D and conf determines the quality of the positioning result; sort,k Indicates D sort The kth Euclidean distance in sort,k<conf, then the D sort,k The corresponding positioning result is judged as a "good" positioning result; otherwise, it is judged as a "bad" positioning result;
[0024] S43, calculating the usage frequency of the sensor / arrival data in the "good" positioning results, and eliminating abnormal values in the arrival data.
[0025] As a further improvement of the present invention, in step S42, the distance threshold conf is expressed as follows:
[0026]
[0027] Wherein, L, W, and H represent the three-dimensional dimensions of the monitoring area or rock sample; α represents the percentage relative to the size of the monitoring area or rock sample, which is a constant less than 1.
[0028] As a further improvement of the present invention, in step S43, the usage frequency of each sensor / arrival data in the "good" positioning result is calculated by the following formula:
[0029]
[0030] Where U represents the set of usage frequencies of each sensor / arrival data in the “good” positioning results; g represents the number of “good” positioning results, G represents the set of sensor coordinates in all g “good” positioning results; P 1 ,P 2 ,...,P n Represent the coordinates of the 1st to nth sensors respectively; I represents the indicator function, which returns 1 if the condition in the brackets is true, otherwise it returns 0; P i in G means P i In the set G, i=1,2,...,n.
[0031] As a further improvement of the present invention, in step S43, the abnormal value is determined according to the following formula:
[0032]
[0033] Among them, U i represents the i-th data in U; if the condition in the brackets is true, the arrival data corresponding to the i-th sensor is an abnormal value, recorded as outlier i .
[0034] As a further improvement of the present invention, in step S5, the damped linear correction method includes: adding a damping term to the Hessian matrix, expressed as:
[0035] [A TA+λdiag(A T A)]δh=A T r
[0036] Where A is a g×4 partial differential matrix, and the superscript T indicates the transpose of the matrix; λdiag(A T A) is the damping term, λ is the non-negative damping factor applied in each iteration; h is the vector of unknown parameters, h = [t 0 ,θ T ] T ; r = t′-t 0 IT(θ), t′ represents the normal arrival data, t′={t 1 ,t 2 ,…,t g} T ; I is a g-dimensional column vector, and T(θ) is the calculated travel time from the microseismic / acoustic emission source to the corresponding sensor.
[0037] The beneficial effects of the present invention are:
[0038] The microseismic / acoustic emission source positioning method based on non-parametric estimation to eliminate outliers provided by the present invention combines sensors into four-by-four combinations and uses an improved analytical method to perform preliminary positioning on the combined sensors. Compared with the existing method of using five sensors for preliminary positioning, the number of sensors required for each positioning is reduced, and the overall operating efficiency and positioning accuracy are improved. On this basis, the present invention performs non-parametric estimation on the preliminary positioning results, eliminates outliers in the arrival data according to the results of the non-parametric estimation, and then repositions based on the normal data after eliminating the outliers, effectively avoiding the influence of the existence of outliers on positioning, and improving the accuracy and robustness of positioning. In addition, the present invention can accelerate convergence and obtain the optimal positioning result by combining the repositioning process with damped linear correction. Compared with the prior art, the microseismic / acoustic emission source positioning method based on non-parametric estimation to eliminate outliers provided by the present invention has higher accuracy and smaller error, and has higher practical application value. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 A schematic flow chart of a microseismic / acoustic emission source localization method based on non-parametric estimation to eliminate outliers provided by the present invention.
[0040] Figure 2 The figure is a schematic diagram comparing a “good” positioning result and a “bad” positioning result in one embodiment of the present invention.
[0041] Figure 3 This is a comparison chart of the error distribution when using different methods to locate the acoustic emission source in the lead breaking experiment.
[0042] Figure 4 This is a comparison diagram of the error distribution when using different methods to locate the blasting source in the blasting experiment. DETAILED DESCRIPTION
[0043] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
[0044] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the scheme of the present invention are shown in the drawings, while other details that are not closely related to the present invention are omitted.
[0045] In addition, it should be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.
[0046] The present invention provides a method for locating microseismic / acoustic emission sources based on kernel density estimation to eliminate outliers, and the flow chart thereof is as follows: Figure 1 As shown, the specific steps include:
[0047] S1. Data preparation
[0048] Collects arrival time data, sensor positions, and wave speeds;
[0049] Data combination: Since at least four sensors are required to locate a single source in a positioning system, for a positioning system with n sensors, the data from n sensors P i (x i ,y i ,z i )(i=1,2,...,n) randomly select four sensors as a set of data, Sensor combinations can ultimately yield Group arrival data.
[0050] S2. Preliminary positioning
[0051] Take any four-sensor combination as an example, assuming that the coordinates of the sensors in the combination are P i (x i ,y i ,z i )(i=1,2,3,4), the source coordinates are S(x,y,z), and the average wave velocity is v, then the control equation they constitute can be expressed as:
[0052] (x i-x) 2 +(y i -y) 2 +(z i -z) 2 =v 2 (t i -t 0 ) 2 ,i=1,2,3,4 (1)
[0053] Among them, t 0 is the triggering time of the microseismic / acoustic emission source, t i To reach sensor P i time.
[0054] Subtracting the equation for i=1 from the equation for i=2,3,4 gives:
[0055] a i x+b i y+c i z=e i +f i t 0 (2)
[0056] in:
[0057] a i =x i -x 1
[0058] b i =y i -y 1
[0059] c i =z i -z 1
[0060]
[0061] f i =v 2 (t i -t 1 )
[0062]
[0063] Based on Cramer's law, we can calculate t 0 Find the solutions x, y, z of the three linear equations defined in equation (2), and the solutions are:
[0064]
[0065] in:
[0066]
[0067] M j and N j The determinant of (j=1,2,3) is similar to that of D, except that the j-th column of D is replaced by (e 2 ,e 3 ,e 4 ) and (f 2 ,f 3 ,f 4 ).
[0068] Substituting the three equations in (3) back into (1), we get t 0 The second-order equation is expressed as:
[0069] At 0 2 +Bt 0 +C=0 (4)
[0070] Among them, A, B, and C are constants, and the specific calculation method is as follows:
[0071]
[0072] By solving equation (4), we can get two solutions, which are expressed as t 01 and t 02 , where the real solution is t 0 . The obtained t 0 Substituting back into equation (3), the corresponding source coordinates are (x 01 ,y 01 ,z 01 ) or (x 02 ,y 02 ,z 02 ), which is the preliminary positioning result.
[0073] For a system with n sensors, the above method can be used to obtain m groups of The preliminary positioning result is expressed as (x k ,y k ,z k ),k=1,2,…,m.
[0074] Among these preliminary positioning results, there will be some "good" results and some "bad" results. Among them, the "good" results are usually obtained based on normal arrival data, and the "bad" results may use abnormal arrival data. Therefore, the present invention finds normal arrival data by identifying the "good" positioning results. The specific identification method will be described in detail below.
[0075] S3. Nonparametric Estimation
[0076] Nonparametric estimation infers the underlying probability distribution of a data set by directly examining the data, thereby avoiding any preconceived notions about the shape of the distribution model. This approach allows for flexible, data-driven estimation of the distribution, thereby capturing its inherent characteristics without the constraints of parametric assumptions. In the present invention, m groups of preliminary positioning results are processed by adopting nonparametric estimation techniques, and a nonparametric estimation function is generated, which is used as a benchmark for evaluating the quality of the positioning results. Among them, the source coordinate θ of the microseismic / acoustic emission source is regarded as a three-dimensional random vector with an unknown density function f. The preliminary positioning results of the microseismic / acoustic emission source can be expressed as:
[0077]
[0078] The superscript T represents the transpose of the matrix.
[0079] In some embodiments of the invention, non-parametric estimation specifically comprises the following steps:
[0080] S31. Selection of kernel function
[0081] The selection of kernel function is a key step in nonparametric estimation. Its characteristics are that it can weight the data points and make its integral in the entire space equal to 1 to ensure normalization. Any smooth and symmetrical peak function that meets the integral condition can be used as an effective kernel.
[0082] In the present invention, the source coordinate θ obtained by each preliminary positioning result is (k) The scale kernel function can be expressed as:
[0083]
[0084] Where θ is the source coordinate of the microseismic / acoustic emission source, θ = [x, y, z] T , the superscript T indicates the matrix or vector transpose; is the kth preliminary positioning result, k = 1, 2, ..., m, m is the number of preliminary positioning results; H is the bandwidth matrix.
[0085] The bandwidth matrix H plays a vital role in determining the shape, size and smoothness of the kernel function and has a significant impact on the estimation result. In the present invention, the bandwidth matrix H can be expressed as:
[0086]
[0087] Among them, s 1 、s 2 and 3 are the variances of the m preliminary positioning results in the x, y and z components respectively.
[0088] S32. Construction of non-parametric estimation function
[0089] The non-parametric estimation function is obtained by aggregating a single scale kernel function and then dividing it by the total number of source coordinates. It can be expressed as:
[0090]
[0091] S33. Calculate the probability density of the source coordinates and determine the density extreme value
[0092] After determining the kernel function and bandwidth matrix, the non-parametric estimation function can generate the probability density distribution model f of the source coordinates to display the kernel density corresponding to each group of preliminary positioning results and determine the coordinates θ corresponding to the density extreme value tri .
[0093] The above probability model is essential for objectively quantifying the spatial distribution of positioning results and identifying high-density areas corresponding to "good" results. In addition, non-parametric estimation does not only focus on "good" positioning results, but instead focuses on the distribution of all results to determine the clustering process. The non-parametric estimation function treats all results equally without prejudging their quality, and can mark results as "good" or "bad" based on their density.
[0094] The probability density distribution obtained based on the preliminary positioning results has obvious regularity. Among them, "good" positioning results are generally concentrated in high-density areas, while "bad" positioning results are concentrated in low-density areas. In the present invention, all positioning results within a specific range are determined as "good" results. The specific determination method will be described below.
[0095] S4. Filter abnormal data
[0096] S41. Calculate the coordinate θ corresponding to the density extreme value tri With the initial positioning result θ (k) The Euclidean distance D k , and then arrange these distances in ascending order to get D sort The specific expressions are as follows:
[0097] D k =|θ tri -θ (k) | (8)
[0098] D sort ={D sort,1 ,D sort,2 ,…D sort,k},D sort,1 <D sort,2 <···<D sort,k-1 <Dsort,k (9)
[0099] S42, determine the distance threshold conf, and compare D sort,k The size of conf determines the quality of the positioning result.
[0100] In general, the quality of the positioning result is closely related to the closeness between the result and the extreme value. Specifically, in the present invention, α is used to quantify the closeness, which is specifically expressed as follows:
[0101]
[0102] Among them, conf is the distance threshold, L, W, H represent the three-dimensional size of the monitoring area or rock sample, and α represents the percentage relative to the size of the monitoring area or rock sample, which is a constant less than 1. Based on a large number of simulation and experimental results, the value range of α is preferably 5% to 20%, within which a good positioning effect can be obtained, and 10% is more preferred.
[0103] On this basis, by comparing D sort,k The size of D and conf can be used to judge the quality of the positioning result. sort,k <conf, then the D sort,k The corresponding positioning result is judged as a "good" positioning result; otherwise, it is judged as a "bad" positioning result.
[0104] In one embodiment of the present invention, a comparison diagram of a "good" positioning result and a "bad" positioning result is shown in FIG. Figure 2 As shown. Figure 2 In the figure, the red five-pointed star represents the extreme value of the non-parametric estimation value, and each dot represents the preliminary positioning result. Among them, the purple dots within a certain distance threshold range close to the extreme value are identified as "good" positioning results, and the green dots outside the distance threshold range are identified as "bad" positioning results.
[0105] S43, calculating the usage frequency of the sensor / arrival data in the "good" positioning result, and determining an abnormal value; the abnormal value is the arrival data detected by the abnormal sensor.
[0106] First, count the number of "good" positioning results according to the following formula:
[0107]
[0108] Where g represents the number of “good” positioning results; m is the total number of preliminary positioning results; I is an indicator function that returns 1 if the condition in the brackets is true, otherwise it returns 0.
[0109] Then, count the sensors / arrival data used in the "good" positioning results, and calculate the usage frequency of each sensor / arrival data according to the following formula:
[0110]
[0111] Where U represents the set of usage frequencies of each sensor / arrival data in the “good” positioning results, G represents the set of sensor coordinates in all g “good” positioning results; P 1 ,P 2 ,...,P n Represent the coordinates of the 1st to nth sensors respectively; I represents the indicator function, which returns 1 if the condition in the brackets is true, otherwise it returns 0; P i in G means P i In the set G, i=1,2,...,n.
[0112] Finally, according to each frequency in the above set U, the "good" arrival data is identified and the "bad" arrival data (i.e., the arrival data corresponding to the abnormal sensor, also called outlier) is eliminated according to the following determination method:
[0113]
[0114] Among them, U i represents the i-th data in U; if the condition in the brackets is true, the arrival data corresponding to the i-th sensor is an abnormal value, recorded as outlier i .
[0115] Through the above method, outliers in the arrival data can be effectively identified. On this basis, outliers can be removed and then repositioned to avoid the impact of outliers on positioning accuracy and effectively improve positioning accuracy.
[0116] S5. Repositioning
[0117] After removing the outliers of the arrival data, the remaining "good" arrival data are all normal arrival data. Repositioning based on the normal arrival data can improve the positioning accuracy. The specific method is as follows:
[0118] The remaining normal arrival time data is expressed as a matrix t′, and the calculated travel time from the microseismic / acoustic emission source to the corresponding sensor is expressed as a matrix T(θ). The corresponding expressions are as follows:
[0119] t′={t 1 ,t 2 ,…,t g} T (14)
[0120] T(θ)={T 1 (θ),T 2 (θ),…,T g (θ)} T (15)
[0121] Trigger time t of microseismic / acoustic emission source 0 The source parameter h can be determined according to the following formula:
[0122] Φ(h)=r T r (16)
[0123] Where r = t′-t 0 IT(θ), I is a g-dimensional column vector; h = [t 0 ,θ T ] T , is the vector of unknown parameters.
[0124] Generally speaking, Equation (16) is an overdetermined system (n ≥ 4) that can be solved iteratively by the least squares method. Assume that the approximate value of the source parameter is known, which is expressed as:
[0125]
[0126] Observed arrival data t i This can be approximated by taking a first-order Taylor expansion of the travel time:
[0127]
[0128] Right now:
[0129]
[0130] Formula (19) can be expressed in the following matrix form:
[0131]
[0132] Among them, δθ is the parameter linear correction term of the microseismic / acoustic emission source, and A is the g×4 partial differential matrix, which is specifically expressed as follows:
[0133]
[0134] At the initial position h * , the partial derivative of formula (14) is calculated as:
[0135]
[0136] From formula (21), we can deduce:
[0137]
[0138] Where T represents the transpose of the matrix, A T A is the Hessian matrix.
[0139] In the above process, if there is a significant difference between the initial value and the actual solution, the first-order Taylor expansion is likely to produce a singular point in the Hessian matrix. The matrix inversion in equation (23) is difficult and the iteration will be stopped. In order to continue the iteration, the present invention adds a damping term λdiag(A T A) to perform damping linear correction. The specific modification results are as follows:
[0140] [A T A+λdiag(A T A)]δh=A T r (24)
[0141] where λ is a non-negative damping factor applied at each iteration.
[0142] In practical applications, if the residual decreases at a faster rate, a smaller value can be set for λ (λ = λ / β). The constant β is used as a step size control coefficient, and its value is β>1. On the contrary, if the residual decreases slowly, a larger value can be set for λ (λ = βλ) to obtain a larger step size in the direction of gradient descent.
[0143] When the calculation step size or the sum of squares of the residuals generated by the last parameter vector decreases to a predetermined threshold, the iteration ends. The final parameter vector h is the solution, and the diagonal matrix λdiag(A T A) represents the damping term, and the scale of each gradient component can be obtained by the diagonal component A T A is used to determine the displacement. Using a diagonal matrix ensures that the displacement is amplified in directions with smaller gradients to avoid slow convergence in these directions.
[0144] Through the above-mentioned method, the present invention combines kernel density estimation with damped linear correction, thereby effectively improving the accuracy of microseismic / acoustic emission source positioning.
[0145] In order to verify the technical effect of the microseismic / acoustic emission source location method based on kernel density estimation to eliminate outliers provided by the present invention, a lead breaking experiment was carried out using a granite cube sample with a side length of 100 mm. Among them, 16 acoustic emission sensors were set on the surface of the granite cube sample and arranged in an orderly geometric array, and their specific coordinates are shown in Table 1.
[0146] Table 1 Coordinates of 16 sensors in the lead breaking experiment
[0147]
[0148]
[0149] In the lead breaking experiment, an HB pencil with a lead diameter of 0.5 mm was used to break the lead at an angle of 30° relative to the surface of the rock sample, and the RS-2A piezoelectric sensor was used to measure the signal emitted by the signal source. The frequency range of the sensor is 50-400kHz, and the center frequency is 150kHz, which can better cover the main frequency band of the acoustic emission signal. Then a preamplifier with a gain of 40 dB is used to amplify the detected signal. The amplified signal is collected by an acquisition device of model DS5-16C, and the frequency is set to 3MHz. During the experiment, a total of 12 acoustic emission sources were selected, and their coordinates are shown in Table 2.
[0150] Table 2 Coordinates of 12 acoustic emission sources in the lead breaking experiment
[0151]
[0152]
[0153] The method provided by the present invention is used to locate each acoustic emission source, and it is compared with the four traditional positioning methods (PCFWE, LOG, CFS, LAG), and the positioning results are shown in Table 3.
[0154] Wherein, NEW represents the method provided in the present invention, PCFWE is the method disclosed in article 1 (Zhou ZL, Rui YC, Zhou J, Dong LJ, Chen LJ, Cai X, Chen RS. A new closed-form solution for acoustic emission source location in the presence of outliers. Appl Sci 2018; 8(6): 949), LOG is the method disclosed in article 2 (Dong LJ, Zou W, Li XB, Shu WW, Wang ZW. Collaborative localization method using analytical and iterative solutions for microseismic / acoustic emission sources in the rock mass structure for underground mining. Eng Fract Mech 2019; 210: 95-112), and CFS is the method disclosed in article 3 (Zhou ZL, Lan RY, Rui YC, Dong LJ, Cai XA new algebraic solution for acoustic emission source localization without premeasuring wave velocity.Sensors (Basel) 2021; 21(2):459), LAG is the method disclosed in article 4 (Zhou ZL, Rui YC, Cai XA novel linear-correction localization method of acoustic emission source for velocity-free system. Ultrasonics 2021; 115:106458), which will not be repeated here.
[0155] Table 3 Results of locating the acoustic emission source by different methods in the lead breaking experiment
[0156]
[0157]
[0158]
[0159]
[0160] It can be seen from Table 3 that among the 12 positioning results, the maximum absolute error of the method provided by the present invention is only 9.12 mm (relative error is 15.80%), while the maximum absolute error and relative error corresponding to the methods PCFWE, LOG, LAG, and CFS are 24.54 mm (relative error 42.50%), 36.28 mm (relative error 62.84%), 32.39 mm (relative error 56.10%), and 53.56 mm (relative error 92.77%), respectively, which are significantly higher than the method provided by the present invention. The smaller the absolute error, the higher the source positioning accuracy. Obviously, the method provided by the present invention has significantly smaller errors and higher accuracy than the above four traditional positioning methods.
[0161] In order to further analyze the error distribution of different methods when locating 12 acoustic emission sources, the following Figure 3 The violin plot shown. Figure 3 It can be seen intuitively that the method provided by the present invention has better positioning stability and significantly improved positioning accuracy, which is improved by 23.61%, 42.00%, 75.49% and 82.36% respectively compared with the methods PCFWE, LOG, LAG and CFS, achieving significant progress.
[0162] In order to further verify the effectiveness of the method provided by the present invention in practical engineering applications, the positioning method provided by the present invention and the above four traditional positioning methods are respectively applied to blasting experiments for comparison. The coordinates of the four explosion sources and 12 trigger sensors designed in the blasting experiment are shown in Table 4 and Table 5 respectively.
[0163] Table 4 Coordinates of the four explosion sources in the blasting experiment
[0164]
[0165] Table 5 Coordinates of 12 trigger sensors in the blasting experiment
[0166]
[0167] The positioning method provided by the present invention and the positioning results obtained by the above four traditional positioning methods are compared. Figure 4 shown. Figure 4 In the figure, S1, S2, S3, and S4 represent explosion source 1, explosion source 2, explosion source 3, and explosion source 4, respectively, and the ordinate represents the average absolute error of the positioning distance.
[0168] Depend on Figure 4It can be seen that the average absolute error of the positioning method provided by the present invention is only 11.71m, while the average absolute errors of methods PCFWE, LOG, LAG, and CFS are 23.56m, 66.09m, 16.95m, and 28.52m, respectively, indicating that the positioning method provided by the present invention also has smaller error and higher positioning accuracy in the blasting experiment. Compared with methods PCFWE, LOG, LAG, and CFS, the positioning accuracy of the present invention is improved by 50.30%, 82.28%, 30.91%, and 58.94%, respectively, achieving significant progress.
[0169] In summary, the present invention provides a method for locating microseismic / acoustic emission sources based on non-parametric estimation to eliminate outliers. The method first obtains multiple groups of preliminary positioning results through preliminary positioning, then performs non-parametric estimation based on the preliminary positioning results, and then eliminates outliers in the arrival data based on the non-parametric estimation results to obtain normal arrival data, and finally re-positions in combination with damped linear correction to obtain the optimal positioning result. Through the above-mentioned method, the present invention can effectively eliminate outliers, avoid the influence of the existence of outliers on positioning accuracy, and improve the accuracy and robustness of positioning; and the re-positioning process can further optimize the positioning results and accelerate convergence, so as to efficiently obtain the optimal positioning result. Compared with the existing methods, the method provided by the present invention has higher accuracy and smaller errors, and has higher practical application value.
[0170] The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.
Claims
1. A method for locating microseismic / acoustic emission sources based on non-parametric estimation to eliminate outliers, characterized in that: The steps include: S1. Data preparation: Collect the data collected by the monitoring system, and perform four-by-four combinations on multiple sensors of the monitoring system to obtain a combination of multiple sensors and arrival data; S2, preliminary positioning: Based on the data prepared in step S1, an improved analytical method is used to perform preliminary positioning to obtain multiple groups of preliminary positioning results; S3, non-parametric estimation: construct a non-parametric estimation function based on the multiple groups of preliminary positioning results obtained in step S2, calculate the probability density of the source coordinates, and determine the coordinates corresponding to the extreme value of the density; S4, filtering abnormal data: according to the result obtained in step S3, the abnormal values in the arrival data are eliminated to obtain normal arrival data; S5, repositioning: repositioning is performed based on the normal arrival time data obtained in step S4 in combination with the damped linear correction method.
2. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 1 is characterized in that: In step S1, the collected data include arrival time data, sensor position and wave speed.
3. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 1 is characterized in that: In step S1, the four-four combination method is: randomly select four sensors from n sensors as a group, with a total of C n 4 Sensor combinations, and get C n 4 Group arrival data.
4. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 3 is characterized by: In step S2, the preliminary positioning result obtained by each sensor combination is: in: M j and N j The determinant of (j=1,2,3) is obtained by replacing the j-th column of D with (e2,e3,e4) and (f2,f3,f4) respectively. i =x i -x1,b i =y i -y1,c i =z i -z1, f i =v 2 (t i -t1), is the spatial coordinate of the sensor in the four-sensor combination, v is the average wave velocity, t0 is the triggering time of the microseismic / acoustic emission source, t i To reach sensor P i time.
5. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 4 is characterized in that: In step S3, the expression of the non-parametric estimation function is: in: θ is the source coordinate of the microseismic / acoustic emission source, θ = [x, y, z] T , the superscript T indicates the matrix or vector transpose; is the kth preliminary positioning result, k = 1, 2, ..., m, m is the number of preliminary positioning results; H is the bandwidth matrix.
6. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 1, characterized in that: In step S4, the method for filtering abnormal data specifically includes the following steps: S41. Calculate the coordinate θ corresponding to the density extreme value tri Compared with the initial positioning result θ( k ) between k , and then arrange these distances in ascending order to get D sort ; S42, determine the distance threshold conf, and compare D sort,k The size of D and conf determines the quality of the positioning result; sort,k Indicates D sort The kth Euclidean distance in sort,k <conf, then the D sort,k The corresponding positioning result is judged as a "good" positioning result; otherwise, it is judged as a "bad" positioning result; S43, calculating the usage frequency of the sensor / arrival data in the "good" positioning results, and eliminating abnormal values in the arrival data.
7. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 6 is characterized by: In step S42, the distance threshold conf is expressed as follows: Wherein, L, W, and H represent the three-dimensional dimensions of the monitoring area or rock sample; α represents the percentage relative to the size of the monitoring area or rock sample, which is a constant less than 1.
8. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 6, characterized in that: In step S43, the usage frequency of each sensor / arrival data in the "good" positioning result is calculated by the following formula: Where U represents the set of usage frequencies of each sensor / arrival data in the "good" positioning results; g represents the number of "good" positioning results, G represents the set of sensor coordinates in all g "good" positioning results; P1, P2, ..., P n Represent the coordinates of the 1st to nth sensors respectively; I represents the indicator function, which returns 1 if the condition in the brackets is true, otherwise it returns 0; P i in G means P i In the set G, i=1,2,...,n.
9. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 8, characterized in that: In step S43, the abnormal value is determined according to the following formula: Among them, U i represents the i-th data in U; if the condition in the brackets is true, the arrival data corresponding to the i-th sensor is an abnormal value, recorded as outlier i .
10. The microseismic / acoustic emission source location method based on non-parametric estimation and elimination of outliers according to claim 5, characterized in that: In step S5, the damped linear correction method includes: adding a damping term to the Hessian matrix, expressed as: [A T A+λdiag(A T A)]δh=A T r Where A is a g×4 partial differential matrix, and the superscript T indicates the transpose of the matrix; λdiag(A T A) is the damping term, λ is the non-negative damping factor applied in each iteration; h is the vector of unknown parameters, h = [t0,θ T ] T ; r = t′-t0I-T(θ), t′ represents the normal arrival data, t′ = {t1, t2, …, t g } T ; I is a g-dimensional column vector, and T(θ) is the calculated travel time from the microseismic / acoustic emission source to the corresponding sensor.
Citation Information
Cited By
Two-dimensional sensor network layout method for optimizing acoustic emission positioning precision
CN119780243A
