Method for rapidly judging type of coal mine goaf for mine disasters

Abstract

The invention discloses a coal mine goaf type rapid discrimination method for mine disasters, and relates to the technical field of geophysical exploration. According to the rapid discrimination method, goaf type classification is carried out on measurement points in a known non-water-containing goaf, a water-containing goaf and a non-goaf range, a mapping relation between different parameters of each measurement point and goaf types is established, and a model training sample is formed. SPSS software is utilized to carry out Bayes modeling training on the established samples, and Bayes discrimination models of different types of goaf areas and three parameters are constructed. And performing Bayes model training by using SPSS software, and when the reestimation accuracy of a training sample exceeds 80%, inputting the optimal parameter combination calculated by the measuring points of the whole region into the Bayes model to realize rapid judgment of the goaf type of the measuring points of the whole region.

Description

Technical Field

[0001] The invention relates to the technical field of geophysical exploration, and in particular to a method for quickly distinguishing the type of coal mine goaf used for mine disasters. Background Art

[0002] After the early resource integration of coal, due to the lack of data, most coal mines have a large number of goafs left over from old kiln mining and goafs formed by illegal mining. In addition, with the continuous increase in the depth and intensity of coal mining, multi-layer goafs have also been formed. The cracks formed after coal seam mining and the cracks formed by goaf collapse are easy to conduct to the aquifer, which poses a huge threat of water disaster accidents to coal mine safety production and must be explored in advance.

[0003] The transient electromagnetic method is based on the principle of electromagnetic induction. It transmits a primary field through the ground transmitting loop power supply. After the current is turned off, the underground geological body is stimulated to produce a secondary eddy current field that decays continuously with time. The magnetic probe is used to receive it point by point in the loop, which is called the induced electromotive force. The resistivity of the water-containing goaf is low, and the induced electromotive force decays slowly over time; the resistivity of the water-free goaf is high, and the induced electromotive force decays quickly over time. When there is no goaf, the induced electromotive force decays at a medium rate over time. Usually, the induced electromotive force at each measuring point is inverted to obtain the resistivity value of the underground medium along the depth direction, and the type of goaf below the measuring point is analyzed according to the resistivity.

[0004] However, the amplitude of abnormal response caused by different types of goaf is slightly different, and the inversion results of transient electromagnetic data have multiple solutions. In addition, the scope of water-bearing goaf and water-free goaf is delineated based on the resistivity. Usually, relatively low resistance is water-bearing goaf and relatively high resistance is water-free goaf. The determination of the thresholds of relatively low resistance and relatively high resistance is more dependent on the experience of the interpreters and is highly subjective, resulting in the water-bearing goaf and water-free goaf delineated in the resistivity plane map of the detection target layer not being consistent with the actual situation. In view of this, we propose a method for rapid identification of coal mine goaf types for mine disasters. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for quickly distinguishing the type of coal mine goaf for mine disasters, which solves the problems mentioned in the above background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for quickly distinguishing the type of coal mine goaf for mine disasters, the method comprising the following steps:

[0007] S1. Calculate the transient electromagnetic induced electromotive force curves collected at different measuring points in the test area, including the "similar" resistivity and apparent depth, the "similar" resistivity along the apparent depth gradient, the logarithm of the induced electromotive force, and the logarithm of the induced electromotive force along the time window gradient;

[0008] S2. Perform normalization and principal component analysis on four parameters, and select the parameter combination with a cumulative contribution rate greater than 90%; and correct the apparent depth in combination with the electrical marker layer of the borehole logging resistivity curve;

[0009] S3. Calculate the buried depth of the target layer at each measuring point by combining the surface elevation of each measuring point and the bottom plate contour of the target layer. Take the optimal parameter values ​​corresponding to the depth positions of the target layer and the 5 layers above and below it after correction, and perform cubic spline interpolation respectively;

[0010] Then calculate the optimal parameter values ​​corresponding to five depth positions, namely, the detection target layer, the top plate and the bottom plate 5m, and the top plate and the bottom plate 10m;

[0011] S4. Obtain the contribution rate of each parameter at different depths by principal component analysis; multiply the sub-coefficient matrix obtained by principal component analysis at different depths by the corresponding contribution rate as the standard value of each parameter at the detection target layer position;

[0012] S5. Select the measuring points within the known water-free goaf, water-containing goaf and no goaf in the measuring area to classify the goaf types, establish the mapping relationship between the preferred parameters of each measuring point and the different goaf types, and construct the model training samples;

[0013] S6. Use SPSS software to perform Bayes sample training, obtain the discriminant function coefficients of different types of goaf areas, and construct Bayes discriminant models of different types of goaf areas;

[0014] S7. The optimal parameter values ​​corresponding to all the measuring points are combined and input into the established Bayesian discriminant model to realize the rapid discrimination of the goaf types at different measuring points in the whole area.

[0015] Optionally, the calculation method of the four parameters in S1 includes the following steps:

[0016] S11, firstly reduce the noise of the transient electromagnetic induction electromotive force attenuation curves at different measuring points in the measuring area;

[0017] S12, then using the induced electromotive force attenuation curve to calculate the induced electromotive force logarithm LV and the change gradient KLV of the induced electromotive force logarithm LV along the time window;

[0018] S13. Calculate the apparent resistivity and diffusion depth using the electromagnetic field propagation theory formula and the electromagnetic field diffusion depth formula under uniform half-space conditions;

[0019] S14. Calculate the “quasi” resistivity ρ using the smoke ring inversion formula r , visual depth H r , and then calculate the "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r .

[0020] Optionally, the normalization and principal component analysis and visual depth correction of the four parameters in S2 include:

[0021] S21, normalizing and principal component analysis the four parameters obtained by calculation, calculating the contribution rate and cumulative contribution rate of each parameter, and selecting a parameter combination with a cumulative contribution rate greater than 90%;

[0022] S22. Extract the resistivity curve of borehole logging in the test area along the depth, and draw the "quasi" resistivity ρ of the measuring point at the borehole position r and apparent depth curve;

[0023] S23. Compare and analyze the two groups of curves. According to the depth of the marker layer with obvious value change and large thickness in the borehole logging resistivity curve, the "similar" resistivity ρ calculated at the measuring point is calculated. r The depth position of the same marker layer in the apparent depth curve is corrected to be consistent with the borehole resistivity logging curve.

[0024] Optionally, the principal component analysis and dimensionality reduction processing of the parameters of the detection target layer and the five layers above and below it in S3 includes:

[0025] S31, taking the parameters corresponding to the five depths near the target layer after depth correction and performing cubic spline interpolation respectively, to obtain the different parameters corresponding to the detection target layer position, the top plate and the bottom plate 5m, and the top plate and the bottom plate 10m, and using the principal component analysis method to solve to obtain the correlation coefficient matrix and contribution rate;

[0026] S32. Identify the preferred layer number with cumulative contribution rate greater than 90% at different depths, select the sub-coefficient matrix with the same number of rows as the number of layers in the correlation coefficient according to the number of layers, multiply the sub-coefficient matrix with the contribution rate corresponding to the preferred number of layers, and obtain the standard value of the detection target layer position parameter combination respectively.

[0027] Optionally, in S7, the Bayesian discriminant model is first established using the training samples, and then the goaf type of all the measuring points in the whole area is quickly discriminated, including:

[0028] S71, dividing the known goaf type measuring points in the measuring area into three categories: goaf without water, goaf with water and no goaf, and marking them as category 1, category 2 and category 3 respectively;

[0029] S72. Use SPSS software to train and back-estimate the Bayesian discriminant model for the samples and calculate the back-estimation accuracy of the Bayesian discriminant model.

[0030] Optionally, the S71 further includes:

[0031] The same number of measuring points were selected within the scope of the three types of goaf, and the mapping relationship between the optimal parameters corresponding to each measuring point and the different types of goaf was established to construct training samples.

[0032] Optionally, the S72 further includes:

[0033] When the accuracy of back-estimation is lower than 80%, the samples with wrong back-estimation are replaced, and the training samples are re-established for Bayesian model training until the accuracy of sample back-estimation is greater than 80%. The final Bayesian model is used to distinguish the types of goafs at all measuring points in the whole area, and a plan view of the distribution of goaf types at the detection target layer position in the whole area is drawn based on the coordinates of each measuring point and the discrimination results.

[0034] The present invention provides a method for quickly distinguishing the type of coal mine goaf for mine disasters.

[0035] Beneficial effects:

[0036] The method for quickly distinguishing the type of coal mine goaf for mine disasters establishes training samples with different measuring point parameters and goaf types, uses SPSS software to train the Bayes model, retains the Bayes model with a training sample back-estimation accuracy rate of more than 80%, and then inputs the preferred parameter combination calculated by the remaining measuring points into the Bayes model to achieve rapid distinction of the goaf type of each measuring point in the entire measuring area. Compared with the existing method in which the interpreter distinguishes the type of goaf and delineates the range according to the resistivity distribution of the detected target layer, the present invention can avoid the adverse effects of the inversion multi-solution of the induced electromotive force data of each measuring point and the subjectivity of the interpreter on the distinction of the type of goaf and the delineation of the range, and achieves the purpose of quickly and accurately distinguishing the type and range of the goaf in the entire measuring area. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a schematic flow chart of an embodiment of the present invention;

[0038] Figure 2 It is the attenuation curve of the induced electromotive force actually measured at a certain measuring point of the present invention;

[0039] Figure 3 It is the curve of the change of the "similar" resistivity along the apparent depth calculated by the smoke ring inversion of the induced electromotive force at a certain measuring point of the present invention;

[0040] Figure 4 is the back-estimation discrimination result of the Bayesian model of the present invention;

[0041] Figure 5 It is a plane diagram of the results of the Bayes model rapid identification of the goaf type of the transient electromagnetic measuring points in the entire area of ​​the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] See also Figure 1-Figure 5 , the present invention provides a technical solution:

[0044] A method for quickly distinguishing the type of coal mine goaf for mine disasters comprises the following steps:

[0045] S1: For the transient electromagnetic induced electromotive force curves collected at different measuring points in the test area, four parameters are calculated respectively: "apparent" resistivity and apparent depth, "apparent" resistivity along the apparent depth gradient, induced electromotive force logarithm, and induced electromotive force logarithm along the time window gradient.

[0046] The apparent resistivity calculation is based on the theoretical formula of uniform half-space electromagnetic field. The apparent resistivity value corresponding to each time window is searched by binary search, and then the smoke ring inversion method is used to calculate the apparent resistivity and the "similar" resistivity and apparent depth corresponding to the time window.

[0047] The calculation of “approximate” resistivity and apparent depth using smoke ring inversion includes the following process:

[0048] t i,j =0.5·(t i +t j )

[0049]

[0050] In the formula, t i and t j are two adjacent time windows, t i,j is the “similar” time, ρ i and ρ j is the apparent resistivity corresponding to two adjacent time windows calculated based on the half-space electromagnetic field theory, ρ r is the “quasi” resistivity calculated by smoke ring inversion, μ0 is the magnetic permeability, which is 4π·10 7 H / m,d r,i is the electromagnetic field skin depth corresponding to each time window, H r is the apparent depth calculated by smoke ring inversion.

[0051] The “quasi” resistivity ρ calculated from smoke ring inversion r and visual depth H r , as well as the induced electromotive force V and each time window t, the "quasi" resistivity ρ corresponding to each transient electromagnetic measuring point is calculated r 、 "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r , the logarithm of the induced electromotive force LV, the gradient of the logarithm of the induced electromotive force LV along the time window KLV, the number of the four parameter values ​​calculated at each measuring point is the same as the length of the time window.

[0052] S2: normalize and perform principal component analysis on the four calculated parameters, and select the parameter combination with a cumulative contribution rate greater than 90%. Then, perform depth correction on the selected parameter combination in combination with the borehole logging resistivity curve.

[0053] The four parameters are normalized. Taking the logarithm of the induced electromotive force LV as an example, the normalization formula is:

[0054]

[0055] In the formula, Lv_norm i,j is the logarithmically normalized value of the induced electromotive force at the i-th measuring point in the j-th time window of a certain measuring line, Lv i,j is the logarithm of the induced electromotive force in the jth time window at the ith measuring point on a certain measuring line, is the mean value of the induced electromotive force in the jth time window of all measuring points on a certain measuring line, It is the variance of the induced electromotive force in the jth time window of all measuring points on a certain measuring line.

[0056] The principal component analysis of the four parameters was performed to calculate the contribution rate and cumulative contribution rate of each parameter, and the "quasi" resistivity ρ with a cumulative contribution rate greater than 90% was selected. r 、 "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r , and the logarithm LV of the induced electromotive force.

[0057] Extract the change curve of borehole resistivity logging along the depth in the mining area and draw the "quasi" resistivity ρ of the measuring point at the borehole location r The two sets of curves are compared and analyzed to find the marker layer with large thickness and obvious electrical property changes compared with the surrounding rock, and the position of the apparent depth marker layer is corrected to be consistent with the depth of the borehole resistivity logging curve.

[0058] like Figure 3 As shown, the "similar" resistivity ρ of the drilling position measurement point is plotted rThe apparent depth curve is compared with the resistivity curve of borehole logging in the test area to find the marker layer with larger thickness and more obvious electrical property changes relative to the surrounding rock. Figure 3 The position correction of the mid-view depth marker layer is consistent with the corresponding depth of the borehole resistivity logging curve.

[0059] S3: Combine the surface elevation of each measuring point and the bottom plate contour of the detection target layer to calculate the burial depth of the detection target layer at each measuring point. Take the optimal parameter values ​​corresponding to the 5-layer depth node positions near the coal seam burial depth, perform cubic spline interpolation, and obtain the optimal parameter values ​​corresponding to the 5 depth positions of the detection target layer, top and bottom plates 5m, and top and bottom plates 10m.

[0060] Using the elevation of each measuring point obtained by surface measurement, the elevation of the detection target layer of each measuring point is further extracted according to the contour lines of the detection target layer in the collected geological data, and the elevation of the detection target layer of each measuring point is subtracted from each other to obtain the buried depth of the detection target layer of each measuring point.

[0061] Take the buried depth position of the detection target layer, the depth positions of the two layers above and below, and the "similar" resistivity ρ corresponding to the five layers of depth node positions r 、 "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r The cubic spline interpolation of the induced electromotive force logarithm LV is performed along the depth direction to obtain the parameter values ​​of each parameter at each measuring point in the detection target layer, 5m above, 10m above, 5m below and 10m below, a total of 5 layers.

[0062] S4: Use principal component analysis to obtain the contribution rate of each parameter at each layer depth, and then use the product of the parameter values ​​at different depths and their corresponding contribution rates as the standard values ​​of each parameter at the detection target layer position.

[0063] The different parameters corresponding to the five depth layers extracted above are organized into arrays, and the principal component analysis method is used to first solve the correlation coefficient matrix and contribution rate.

[0064] Determine the preferred number of layers with cumulative contribution rates greater than 90% at different depths, select the sub-coefficient matrix with the same number of rows as the number of layers in the correlation coefficient according to the number of layers, multiply the sub-coefficient matrix by the contribution rate corresponding to the preferred number of layers, and obtain the final "similar" resistivity ρ of the detection target position. r 、 "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r , logarithm of induced electromotive force LV.

[0065] S5: Classify the goaf types of the measuring points within the known water-free goaf, water-containing goaf and no goaf, establish the mapping relationship between the different parameters of each measuring point and the goaf type, and form the model training samples. Combined with the known geological data of the mining area, the goaf types are divided into three categories: water-free goaf, water-containing goaf and no goaf, which are recorded as Class 1, Class 2 and Class 3 respectively.

[0066] Select 50 measuring points within different goaf types, each measuring point corresponds to the three parameters of the detection target layer depth position calculated in step 4, establish a mapping relationship between different parameters of each measuring point and the goaf type, and form 150 measuring point discrimination model training samples.

[0067] S6: Use SPSS software to conduct Bayesian modeling training on the established samples and construct Bayesian discriminant models for different types of goaf and three parameters.

[0068] The Bayes model was trained on 150 measuring point samples using SPSS software to obtain the classification function coefficients. The training samples of the established Bayes discriminant model were back-estimated, and the goaf type results of each training measuring point were discriminated. The results were compared with the actual goaf type of the corresponding measuring point to obtain the accuracy of the Bayes discriminant model.

[0069] The Bayes discriminant model formula is as follows:

[0070]

[0071] Among them, y1 corresponds to the water-filled goaf, y2 corresponds to the water-free goaf, y3 corresponds to the normal stratum, x, y, and z correspond to three parameters respectively, and a1~a4, b1~b4, and c1~c4 are classification function coefficients.

[0072] like Figure 4 As shown in the figure, when the accuracy of the back-estimation discrimination of the Bayes model is lower than 80%, the samples with back-estimation discrimination errors are replaced, and the training samples are re-established to train the Bayes model until the sample back-estimation discrimination accuracy is greater than 80%, and the Bayes discriminant model is saved.

[0073] S7: Using the Bayes discriminant model finally obtained above, the measuring points of unknown goaf types in the measuring area are discriminated to quickly obtain the goaf types of different measuring points in the whole area.

[0074] like Figure 5As shown in the figure, the unknown goaf type measuring points in the measuring area are counted, and the three optimal parameter values ​​calculated at each measuring point are input into the Bayes discriminant model to quickly discriminate the goaf types corresponding to different measuring points. The goaf types of the measuring points with known goaf types and the other measuring points quickly discriminated by the Bayes model are counted, and according to the coordinate position of each measuring point in the measuring area plane, a goaf type distribution plane diagram of the entire measuring area at the detection target layer position is drawn.

[0075] As an application of this embodiment:

[0076] The transient electromagnetic smoke ring inversion method is used to calculate the "similar" resistivity and apparent depth of the induced electromotive force at each point in the survey area, the "similar" resistivity along the apparent depth gradient, the logarithm of the induced electromotive force, and the logarithm of the induced electromotive force along the time window gradient. Secondly, the four parameters are normalized and principal component analysis is performed to select the parameter combination with a cumulative contribution rate greater than 90%. Then, the borehole resistivity logging curve is used to correct the depth of the marker layer. The buried depth of the target layer of each measuring point is calculated by combining the surface elevation of each measuring point and the bottom plate contour of the detection target layer. The preferred parameter values ​​corresponding to the target layer position and its upper and lower layers are calculated by the cubic spline interpolation method. The principal component analysis is used to obtain the contribution rate of each parameter at each layer depth, and then the product of the sub-coefficient matrix of different depths and its corresponding contribution rate is used as the standard value of each parameter at the detection target layer position. The goaf type is classified for the measuring points within the known water-free goaf, water-containing goaf and no goaf area, and the mapping relationship between different parameters of each measuring point and the goaf type is established to form a model training sample. The established samples were trained with Bayes modeling using SPSS software to construct Bayes discrimination models for different types of goaf and three parameters. When the back-estimation accuracy of the training samples exceeded 80%, the optimal parameter combination calculated at the measuring points in the whole area was input into the Bayes model to achieve rapid discrimination of the goaf type at the measuring points in the whole area.

[0077] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for quickly distinguishing the type of coal mine goaf for mine disasters, characterized by: The method comprises the following steps: S1. Calculate the transient electromagnetic induced electromotive force curves collected at different measuring points in the test area, including "similar" resistivity and apparent depth, "similar" resistivity along the apparent depth gradient, induced electromotive force logarithm, and induced electromotive force logarithm along the time window gradient; S2. Perform normalization and principal component analysis on four parameters, and select the parameter combination with a cumulative contribution rate greater than 90%; and correct the apparent depth in combination with the electrical marker layer of the borehole logging resistivity curve; S3. Calculate the buried depth of the target layer at each measuring point by combining the surface elevation of each measuring point and the bottom plate contour of the target layer; take the optimal parameter values ​​corresponding to the depth positions of the target layer and the 5 layers above and below it after correction, and perform cubic spline interpolation respectively; Then calculate the optimal parameter values ​​corresponding to five depth positions, namely, the detection target layer, the top plate and the bottom plate 5m, and the top plate and the bottom plate 10m; S4. Obtain the contribution rate of each parameter at different depths by principal component analysis; multiply the sub-coefficient matrix obtained by principal component analysis at different depths by the corresponding contribution rate as the standard value of each parameter at the detection target layer position; S5. Select the measuring points within the known water-free goaf, water-containing goaf and no goaf in the measuring area to classify the goaf types, establish the mapping relationship between the preferred parameters of each measuring point and the different goaf types, and construct the model training samples; S6. Use SPSS software to perform Bayes sample training, obtain the discriminant function coefficients of different types of goaf areas, and construct Bayes discriminant models of different types of goaf areas; S7. The optimal parameter values ​​corresponding to all the measuring points are combined and input into the established Bayesian discriminant model to realize the rapid discrimination of the goaf types at different measuring points in the whole area.

2. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 1, characterized in that: The calculation method of the four parameters in S1 includes the following steps: S11, firstly reduce the noise of the transient electromagnetic induction electromotive force attenuation curves at different measuring points in the measuring area; S12, then using the induced electromotive force attenuation curve to calculate the induced electromotive force logarithm LV and the change gradient KLV of the induced electromotive force logarithm LV along the time window; S13. Calculate the apparent resistivity and diffusion depth using the electromagnetic field propagation theory formula and the electromagnetic field diffusion depth formula under uniform half-space conditions; S14. Calculate the "quasi" resistivity ρ using the smoke ring inversion formula r , visual depth H r , and then calculate the "quasi" resistivity ρ r Along the visual depth H r The gradient of Kρ r .

3. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 1, characterized in that: The normalization and principal component analysis of the four parameters and the visual depth correction in S2 include: S21, normalizing and principal component analysis the four parameters obtained by calculation, calculating the contribution rate and cumulative contribution rate of each parameter, and selecting a parameter combination with a cumulative contribution rate greater than 90%; S22. Extract the resistivity curve of borehole logging in the test area along the depth, and draw the "quasi" resistivity ρ of the measuring point at the borehole position r and apparent depth curve; S23. Compare and analyze the two groups of curves. According to the depth of the marker layer with obvious value change and large thickness in the borehole logging resistivity curve, the "similar" resistivity ρ calculated at the measuring point is calculated. r The depth position of the same marker layer in the apparent depth curve is corrected to be consistent with the borehole resistivity logging curve.

4. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 1, characterized in that: The principal component analysis and dimensionality reduction processing of the parameters of the detection target layer and the five layers above and below it in S3 includes: S31, taking the parameters corresponding to the five depths near the target layer after depth correction and performing cubic spline interpolation respectively, to obtain the different parameters corresponding to the detection target layer position, the top plate and the bottom plate 5m, and the top plate and the bottom plate 10m, and using the principal component analysis method to solve to obtain the correlation coefficient matrix and contribution rate; S32. Identify the preferred layer number with cumulative contribution rate greater than 90% at different depths, select the sub-coefficient matrix with the same number of rows as the number of layers in the correlation coefficient according to the number of layers, multiply the sub-coefficient matrix with the contribution rate corresponding to the preferred number of layers, and obtain the standard value of the detection target layer position parameter combination respectively.

5. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 1, characterized in that: In S7, the training samples are first used to establish a Bayesian discriminant model, and then the goaf type of all the measuring points in the whole area is quickly discriminated, including: S71, dividing the known goaf type measuring points in the measuring area into three categories: goaf without water, goaf with water and no goaf, and marking them as category 1, category 2 and category 3 respectively; S72. Use SPSS software to train and back-estimate the Bayesian discriminant model for the samples and calculate the back-estimation accuracy of the Bayesian discriminant model.

6. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 5, characterized in that: The S71 further comprises: The same number of measuring points were selected within the scope of the three types of goaf, and the mapping relationship between the optimal parameters corresponding to each measuring point and the different types of goaf was established to construct training samples.

7. A method for quickly distinguishing the type of coal mine goaf for mine disasters according to claim 5, characterized in that: The S72 further includes: When the accuracy of back-estimation discrimination is lower than 80%, the samples with wrong back-estimation discrimination are replaced, and the training samples are re-established to train the Bayes model until the accuracy of sample back-estimation discrimination is greater than 80%; The final Bayes model is used to identify the goaf types at all measuring points in the entire area. Based on the coordinates of each measuring point and the identification results, a plan view of the goaf type distribution at the target layer location in the entire area is drawn.

Images