Rock mass quality interpretation method based on multi-source geological forecast data fusion

Abstract

The invention discloses a rock mass quality interpretation method based on multi-source geological forecast data fusion, and relates to the technical field of geological survey. According to the interpretation method, a geological radar data interpretation result is regarded as soft data (with relatively low credibility), an advanced drilling data interpretation result is regarded as hard data (with relatively high credibility), and a corresponding prior probability distribution calculation method, a likelihood function calculation method and a global posteriori update mechanism are provided around a Bayesian probability update principle. The two geological forecast data are organically fused, and global, high-reliability and quantitative two-dimensional interpretation of the quality of the rock mass which is not excavated in the construction period is achieved; the method has the advantages of being high in accuracy, good in universality and high in operability, is suitable for being applied to interpretation of the quality of various unexcavated rock masses on a large scale, and plays a positive role in guaranteeing safe construction and application of excavation engineering.

Description

Technical Field

[0001] The present invention relates to the field of geological survey technology, in particular to a rock mass quality survey technology, and in particular to a rock mass quality interpretation method by fusing multi-source geological prediction data. Background Art

[0002] In order to ensure the safety of tunnel construction and to explore the structure and geological information of unexcavated rock mass, advanced geological forecasting is a necessary and key technical means.

[0003] At present, the commonly used advanced geological prediction is divided into two types: geophysical exploration and drilling. Among them, during the construction period, geophysical exploration is mainly based on geological radar detection, which is characterized by low cost, wide range (depth), and non-destructive detection. It can quickly obtain geological structure information under the two-dimensional profile, but is subject to the multi-solution nature of geophysical signals, and the interpretation results obtained are usually of general credibility; in contrast, drilling is a direct drilling test on the rock mass. Through the analysis of the drilling parameters of the core or the drilling process, the geological information of the rock mass can be interpreted more directly. The results obtained are generally more reliable, but its disadvantages are high cost, and the one-dimensional prediction data obtained are very limited and sparse (the number of boreholes is 2 to 3), making it difficult to accurately describe the geological conditions between the boreholes.

[0004] At present, in the survey technology of tunnel rock mass quality, usually one of the means is used to interpret geological information, and then another means is used to qualitatively correct the interpretation result. However, due to the influence of the dimensional difference at the forecast data level and other issues, the final survey results still have the problems of multiple solutions, low credibility, and incomplete interpretation range, which seriously affects the safe construction and application of tunnels. Therefore, providing a method that can fuse multi-source data of the two to organically form a quantitative, highly reliable information interpretation method that can interpret the quality of the unexcavated two-dimensional rock mass of the entire tunnel is of great significance to solving the above-mentioned technical problems in this field. Summary of the invention

[0005] The purpose of the present invention is to overcome the problems of low credibility and incomplete interpretation range in the existing rock mass quality survey and interpretation technology, and propose a rock mass quality interpretation method by fusing multi-source geological prediction data.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides a rock mass quality interpretation method by fusing multi-source geological prediction data, comprising the following steps:

[0007] S1. Obtaining the interpretation results of the geological radar data and the advance drilling data of the rock mass quality respectively;

[0008] S2, gridding and standardizing the interpretation results of the geological radar data and the advanced drilling data, respectively, to obtain a number of geological radar data grid units and advanced drilling data grid units with the same interpretation space and label;

[0009] S3. Based on the geological radar data grid unit and the advance drilling data grid unit, a two-dimensional prior probability distribution function of rock mass quality is constructed using a priori probability calculation formula; based on the grid length of the geological radar data grid unit, the advance drilling data grid unit is equidistantly divided and the secondary index is calculated and processed, and then the likelihood function of rock mass quality is constructed using a confusion matrix;

[0010] S4. Based on the two-dimensional prior probability distribution function and likelihood function of rock mass quality, the Bayesian probability update mechanism is used to perform fusion interpretation processing to obtain the fusion interpretation result of the rock mass quality at the drilling hole location;

[0011] S5. Based on the two-dimensional prior probability distribution function of rock mass quality and the fused interpretation results of the rock mass quality at the drilling hole location, a fused interpretation process is performed using a global update mechanism to obtain a fused interpretation result of the global rock mass quality.

[0012] The invention discloses a rock mass quality interpretation method by fusing multi-source geological prediction data, which regards the interpretation result of geological radar data as soft data (with low credibility) and the interpretation result of advance drilling data as hard data (with high credibility). Based on the Bayesian probability updating principle, a corresponding prior probability distribution calculation method, a likelihood function calculation method and a global posterior updating mechanism are proposed, and the two types of geological prediction data are organically integrated to realize a global, highly reliable and quantitative two-dimensional interpretation of the quality of unexcavated rock mass during the construction period. The method has the characteristics of high accuracy, good universality and strong operability, is suitable for large-scale application in the interpretation of various types of unexcavated rock mass quality, and plays a positive role in ensuring the safe construction and application of excavation projects.

[0013] Among them, in step S1, the electromagnetic wave propagation process is affected by various interference factors, and the original ground penetrating radar data usually contains various noises, so preprocessing is required to improve the signal-to-noise ratio to better highlight the characteristics of the disease; preferably, when obtaining the geological radar data interpretation results of the rock quality, preprocessing of the original ground penetrating radar data is also included; the preprocessing includes at least one of zero point reset, high-pass and low-pass filtering, and exponential gain.

[0014] Among them, in step S2, preferably, the gridding processing of the geological radar interpretation results includes: selecting an appropriate square grid length (the denser the grid, the higher the accuracy of the interpretation results obtained, but the larger the calculation scale, which is reasonably set according to the overall exploration area scale and exploration needs), gridding the two-dimensional geological radar interpretation results, and obtaining a number of radar data grid units; the gridding processing of the advance drilling data interpretation results includes: selecting the same grid length as the gridding processing of the geological radar interpretation results, gridding the advance drilling data interpretation results of the drilling position, and obtaining a number of drilling data grid units; through gridding processing, it is ensured that the grid units at the drilling position correspond to the grid unit positions of the geological radar interpretation results at the same position.

[0015] Preferably, the standardization process includes: unifying the radar data grid unit and the borehole data grid unit, and using the same interpretation space and label to represent them. For example, the radar data grid unit and the borehole data grid unit can be represented as follows:

[0016] S=(label1,label2...label n );

[0017] In the formula, label means label, S represents the grid unit interpretation space, the interpretation space is composed of interpretation labels, and there are n interpretation labels including label1, label2...labeln; the interpretation labels are generally numbers 0, 1, 2, etc., and each data represents a rock quality category. If the grid units have the same interpretation space and label, it means that the rock quality categories contained in the two data should be consistent.

[0018] Among them, in step S3, the prior probability calculation formula is:

[0019]

[0020] In the formula, represents the initial probability distribution of the interpreted labels in each radar data grid cell; represents the initial probability update value considering spatial correlation, where spatial correlation refers to the interpretation label situation within the neighborhood of the current radar data grid unit (eight-neighborhood system). The specific value is equal to the frequency of each interpretation label, and finally the result of the addition of the two is renormalized; f x (X) represents the prior probability distribution function of rock mass quality information X.

[0021] Preferably, the equidistant segmentation includes: segmenting the drilling data according to the grid length of the radar data grid unit; and the secondary index calculation includes: calculating the mean and variance of the drilling indicator parameters within the equidistant segmentation sections.

[0022] Preferably, the method of calculating the likelihood function of rock mass quality using a confusion matrix comprises the steps of:

[0023] (1) Using the optimized classifier model to train the grid cell data (drilling location) of the advanced drilling data after isometric segmentation and secondary index calculation;

[0024] (2) Using the trained classifier to predict the grid cell data of the advance drilling data (undrilled locations);

[0025] (3) The likelihood function is constructed based on the confusion matrix of the predicted output.

[0026] Preferably, in step (1), the classifier model is XGBoost; the model constructs multiple weak evaluators on the data one by one, and gradually accumulates and summarizes the modeling results of multiple weak evaluators through multiple iterations to obtain better regression or classification performance than a single model. This superposition strategy using a single decision tree as a weak evaluator can be expressed in an additive form, that is, the XGBoost classifier model can be expressed as:

[0027]

[0028] In the formula, represents the prediction result of the entire model on this sample, k represents the total number of weak evaluators (decision trees), and f k represents the kth decision tree, x i Represents the eigenvector corresponding to sample i.

[0029] Among them, the optimized XGBoost introduces model complexity to measure the computational efficiency of the algorithm. Therefore, the optimized XGBoost model consists of two parts: the traditional loss function and the model complexity. The specific formula is:

[0030]

[0031] In the formula, O bj represents the objective function of the model, n represents the total amount of data imported into the kth tree, and y i represents the true label, represents the predicted value, f k represents the kth decision tree, Ω represents the transformation value of the tree model, Represents the traditional loss function, measuring the true label y i and predicted values The difference between Represents the complexity of the model.

[0032] Among them, in step (2), the expression of constructing the likelihood function based on the confusion matrix is:

[0033]

[0034] Where TP and FN represent the number of true positive and false negative samples in the prediction confusion matrix, respectively.

[0035] Among them, in step S4, the Bayesian probability update mechanism includes: based on the two-dimensional prior probability distribution function and the likelihood function, using the Bayesian probability estimation formula to obtain the posterior probability distribution function, and then obtaining the fusion interpretation result of the rock mass quality at the drilling hole location through the posterior probability distribution function; the Bayesian probability estimation formula is:

[0036] f x ′(X)=aL(X)f x (X);

[0037] In the formula, f x ′(X) represents the posterior probability distribution function of rock mass quality information X; f x (X) represents the prior probability distribution function of rock mass quality information X; X is the original spatial random input vector, X = (X1, X2, ..., X n ) T , n is the number of random parameters; L(X) represents the likelihood function; a is a constant; T is the matrix transposition operator.

[0038] Among them, in step S5, the global update mechanism includes: based on the two-dimensional prior probability distribution function of rock mass quality and the fusion interpretation result of rock mass quality at the drilling hole position, using the global update formula, obtaining the fusion probability distribution function, and then obtaining the fusion interpretation result of the global rock mass quality by fusing the probability distribution function; the global update formula is:

[0039] f x ″(X)=f x (X)+d x (X);

[0040] In the formula, f x ″(X) represents the posterior probability distribution function of the rock mass quality information X that needs to be globally updated in addition to the Bayesian probability update position; d x (X) represents the probability update value calculated based on the distance from the drilling location; f x (X) represents the prior probability distribution function of rock mass quality information X.

[0041] Preferably, d x The calculation formula of (X) is:

[0042] dx(X)=f x ′(X)ω d ;

[0043] In the formula, ω d Represents the weight, which is related to the drilling distance. The closer the distance to the drilling location, the greater the impact of the drilling data, that is, the greater the weight; the farther the distance to the drilling location, the less impact of the drilling data, that is, the smaller the weight; f x ′(X) represents the posterior probability distribution function of rock mass quality information X.

[0044] Preferably, ω d The calculation formula is as follows: d =e -bd ; In the formula, d is the distance from the drill hole, b is the constant term, and e is a constant.

[0045] Furthermore, the present invention also proposes an interpretation device, comprising at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned rock quality interpretation method of multi-source geological prediction data fusion.

[0046] Furthermore, the present invention also proposes a computer-readable medium on which instructions executable by a processor are stored. When the instructions are executed by the processor, the processor executes the above-mentioned rock mass quality interpretation method by fusing multi-source geological prediction data.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The rock mass quality interpretation method of multi-source geological prediction data fusion of the present invention proposes a prior probability distribution calculation method considering spatial correlation for the geological radar data results at the two-dimensional level, which can effectively and reasonably convert the label interpretation results into the prior probability distribution required for the Bayesian probability update calculation, so that the data fusion update can be carried out through the Bayesian probability update principle, overcoming the shortcomings of traditional Kriging interpolation for data distribution form and the inability to consider the credibility of data from different sources, and significantly improving the accuracy of the interpretation results of rock mass quality geological prediction data.

[0049] 2. The rock quality interpretation method of multi-source geological prediction data fusion of the present invention, for the one-dimensional advanced drilling data results, proposes to train the model and obtain the confusion matrix for each interpretation label by machine learning, so as to reasonably calculate the likelihood function required for the Bayesian probability update calculation, and proposes a global update mechanism that takes the influence of drilling results into consideration, thereby realizing the final fusion of the two data by calculating the distance between the grid unit and the drilling grid unit and converting it into a weight, thereby significantly improving the accuracy and credibility of the global rock quality geological prediction data interpretation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the flow chart of the rock mass quality interpretation method embodiment 1 of the multi-source geological prediction data fusion of the present invention;

[0051] Figure 2 It is a schematic diagram of grid division of the interpretation result of geological radar in Example 1 of the rock mass quality interpretation method of multi-source geological prediction data fusion of the present invention;

[0052] Figure 3 This is a schematic diagram of grid division of an example of the rock mass quality interpretation result of the multi-source geological prediction data fusion in Example 1 of the present invention;

[0053] Figure 4 It is a graph showing grid standardization of geological radar interpretation results and prior probability distribution calculation results in Example 1 of the rock mass quality interpretation method for multi-source geological prediction data fusion of the present invention;

[0054] Figure 5 It is a grid-standardized result diagram of the advanced drilling interpretation result in Example 1 of the rock mass quality interpretation method for multi-source geological prediction data fusion of the present invention;

[0055] Figure 6 It is a diagram of the interpretation result of the rock mass quality fused with the drilling position after the Bayesian probability update in the embodiment 1 of the rock mass quality interpretation method for multi-source geological prediction data fusion of the present invention;

[0056] Figure 7 This is a diagram of the interpretation results of the rock mass by fusion of multi-source geological prediction data after global updating in Example 1 of the rock mass interpretation method by fusion of multi-source geological prediction data of the present invention. DETAILED DESCRIPTION

[0057] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. However, it should not be understood that the scope of the above subject matter of the present invention is limited to the following embodiments, and all technologies realized based on the content of the present invention belong to the scope of the present invention.

[0058] Embodiment 1:

[0059] A rock mass quality interpretation method based on multi-source geological prediction data fusion (flow chart as shown in Figure 1 As shown), comprising the following steps:

[0060] S1. Obtain the interpretation results of geological radar data at a certain tunnel construction stage and the interpretation results of advance drilling data under the same tunnel face (the two can basically be regarded as being on the same horizontal survey line, with a survey line width of 10m and a depth of 30m);

[0061] S2. Select an appropriate square grid length (grid length l is 0.5m, grid number is 20×60), grid the two-dimensional geological radar interpretation results, and obtain a number of radar data grid units; select the same grid length as the grid processing of the geological radar interpretation results (grid length l is 0.5m), grid the interpretation results of the advance drilling data at the borehole location, and obtain a number of borehole data grid units; then use the same interpretation space and label to represent the radar data grid unit and the borehole data grid unit ( Figure 5 , represented by the following formula: S = (label1, label2...label n ));

[0062] S3, based on the interpretation space and labels of radar data grid cells, using the prior probability calculation formula In the formula, Represents the initial probability distribution of each grid cell label in the geological radar interpretation result; represents the initial probability update value considering spatial correlation, where spatial correlation refers to the interpretation label situation in the neighborhood of the current grid unit (eight-neighborhood system). The specific value is equal to the frequency of each interpretation label, and finally the result of the addition of the two is renormalized; f x (X) represents the prior probability distribution function of rock mass quality information X) to construct a two-dimensional prior probability distribution function of rock mass quality ( Figure 4 );

[0063] For example, there are three types of labels in the interpretation space, represented by "1", "2", and "3". Assume that the initial probability distribution of each label in the radar interpretation result is 0.5 for the label itself and 0.25 for the other labels, that is, the initial probability distribution of label 1 is (0.5, 0.25, 025), the initial probability distribution of label 2 is (0.25, 0.5, 025), and the initial probability distribution of label 3 is (0.25, 0.25, 0.5);

[0064] Take the label 1 of the grid cell in the third row and third column as the reference point:

[0065] ① Its initial prior probability is (0.5, 0.25, 0.25);

[0066] ② Consider the labels of the surrounding adjacent grid cells and count the frequency of occurrence, such as Figure 2 There are 8 points in total around the middle, and the frequency statistics of labels 1, 2, and 3 are 5 / 8, 2 / 8, and 1 / 8 respectively;

[0067] ③ Correct the initial prior probability of the reference point, that is, add the statistical results and perform normalization, and calculate the prior probability distribution of the grid unit to be (0.56, 0.25, 0.19).

[0068] It should be noted that the neighborhood information statistics of the grid cells include three working conditions: middle grid cells (8 neighborhoods), edge grid cells (5 neighborhoods) and corner grid cells (3 neighborhoods).

[0069] Based on the grid length of the geological radar data grid unit, the interpretation space and labels of the advance drilling data grid unit are equidistantly segmented and the secondary indexes are calculated (the specific method can be referred to the relevant content in patent ZL202110674747.5), and then the likelihood function of the rock mass quality is constructed using the confusion matrix;

[0070] For example: Figure 3 The mesh division shown in the figure shows that the drilling data is Figure 2 The radar data has the same interpretation space and three types of labels. According to the confusion matrix, the likelihood functions of labels 1, 2, and 3 can be calculated as follows: Label1: (0.91, 0.09, 0); Label2: (0.08, 0.88, 0.04); Label3: (0.01, 0, 0.09);

[0071] Methods for calculating the likelihood function of rock mass quality using confusion matrix include:

[0072] (1) Using the optimized classifier model In the formula, O bj represents the objective function of the model, n represents the total amount of data imported into the kth tree, and y i represents the true label, represents the predicted value, f k represents the kth decision tree, Ω represents the transformation value of the tree model, Represents the traditional loss function, measuring the true label y i and predicted values The difference between The complexity of the representative model) is trained on the grid cell data (drilling location) of the advance drilling data that has been processed by equidistant segmentation and secondary index calculation;

[0073] (2) Using the trained classifier to predict the grid cell data of the advance drilling data (undrilled locations);

[0074] (3) According to the confusion matrix of the predicted output ( In the formula, TP and FN represent the number of true positive and false negative samples in the prediction confusion matrix respectively) to construct the likelihood function (① intact rock mass (0.98, 0.02, 0), ② broken rock mass (0.06, 0.92, 0.02), ③ void (0, 0, 1));

[0075] S4, based on the two-dimensional prior probability distribution function and likelihood function, using the Bayesian probability estimation formula (Bayesian probability estimation formula is: f x ′(X)=aL(X)f x (X); where f x ′(X) represents the posterior probability distribution function of rock mass quality information X; f x (X) represents the prior probability distribution function of rock mass quality information X; X is the original spatial random input vector, X = (X1, X2, ..., X n ) T , n is the number of random parameters; L(X) represents the likelihood function; a is a constant; T is the matrix transposition operator), and the posterior probability distribution function is obtained. Then, through the posterior probability distribution function (for example, there are three rock mass categories 0, 1, and 2, and the current grid unit overlaps with the borehole grid unit. First, when there is only radar data, the probability of calculating the prior probability distribution according to the radar interpretation result is (0.2, 0.3, 0.5), that is, the probability that the current grid unit belongs to label 0 is 0.2, the probability that it belongs to label 1 is 0.3, and the probability that it belongs to label 2 is is 0.5. After the borehole data is integrated and the global Bayesian update is performed, the probability distribution of the current grid unit becomes (0.1, 0.5, 0.4). This probability change process is also called the posterior update. (0.1, 0.5, 0.4) is the posterior probability distribution. According to the new probability distribution, we know that the maximum probability of the current grid belonging to label 1 is 0.5. Then, after the fusion update, it changes from belonging to label 2 to belonging to label 1, that is, a new interpretation result is obtained). The fusion interpretation result of the rock mass quality at the drilling hole location is obtained ( Figure 6 );

[0076] S5, based on the 2D prior probability distribution function of rock mass quality and the fusion interpretation results of rock mass quality at drilling hole location, using the global update formula (global update formula is: f x ″(X)=f x (X)+d x (X); where f x ″(X) represents the posterior probability distribution function of the rock mass quality information X that needs to be globally updated in addition to the Bayesian probability update position; d x (X) represents the probability update value calculated based on the distance from the drilling location; f x (X) represents the prior probability distribution function of rock mass quality information X; d x The calculation formula of (X) is: dx(X) = f x ′(X)ω d ; In the formula, ω dRepresents the weight, which is related to the drilling distance. The closer the distance to the drilling location, the greater the impact of the drilling data, that is, the greater the weight; the farther the distance to the drilling location, the less impact of the drilling data, that is, the smaller the weight; f x ′(X) represents the posterior probability distribution function of rock mass quality information X; ω d The calculation formula is as follows: d =e -bd ; where d is the distance from the borehole, b is a constant term, and e is a constant), the fusion probability distribution function is obtained, and then the fusion interpretation result of the global rock mass quality is obtained by fusion probability distribution function ( Figure 7 ).

[0077] Example 2

[0078] An interpretation device includes at least one processor and a memory connected to the at least one processor in communication; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a rock mass quality interpretation method for multi-source geological prediction data fusion in Example 1.

[0079] Example 3

[0080] A computer-readable medium stores instructions executable by a processor, and when the instructions are executed by the processor, the processor executes a rock mass quality interpretation method for multi-source geological prediction data fusion in Example 1.

[0081] Finally, it should be noted that the embodiments described in detail above are only the best practices of the invention and cannot be used to limit the scope of rights of the invention. Equivalent replacement of the technical solutions recorded in the aforementioned embodiments does not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the invention, and they should all be included in the scope of the claims and description of the invention.

Claims

1. A rock mass quality interpretation method based on multi-source geological prediction data fusion, characterized in that: The following steps are involved: S1. Obtaining the interpretation results of the geological radar data and the advance drilling data of the rock mass quality respectively; S2, gridding and standardizing the interpretation results of the geological radar data and the advanced drilling data, respectively, to obtain a number of geological radar data grid units and advanced drilling data grid units with the same interpretation space and label; S3. Based on the geological radar data grid unit and the advance drilling data grid unit, a two-dimensional prior probability distribution function of rock mass quality is constructed using a priori probability calculation formula; based on the grid length of the geological radar data grid unit, the advance drilling data grid unit is equidistantly divided and the secondary index is calculated and processed, and then the likelihood function of rock mass quality is constructed using a confusion matrix; S4. Based on the two-dimensional prior probability distribution function and likelihood function of rock mass quality, the Bayesian probability update mechanism is used to perform fusion interpretation processing to obtain the fusion interpretation result of the rock mass quality at the drilling hole location; S5. Based on the two-dimensional prior probability distribution function of rock mass quality and the fused interpretation results of the rock mass quality at the drilling hole location, a fused interpretation process is performed using a global update mechanism to obtain a fused interpretation result of the global rock mass quality.

2. The rock mass quality interpretation method according to claim 1, characterized in that: In step S2, the gridding processing of the geological radar interpretation result includes: selecting a suitable square grid length, gridding the two-dimensional geological radar interpretation result, and obtaining a plurality of radar data grid units; The gridding process of the advanced drilling data interpretation result comprises: selecting the same grid length as that of the geological radar interpretation result gridding process, gridding the advanced drilling data interpretation result of the borehole position, and obtaining a plurality of borehole data grid units; The standardization process includes: unifying radar data grid units and borehole data grid units, and using the same interpretation space and labels for representation.

3. The rock mass quality interpretation method according to claim 1, characterized in that: In step S3, the prior probability calculation formula is: In the formula, f x i (X) represents the initial probability distribution of the interpreted labels in each radar data grid cell; represents the initial probability update value considering spatial correlation; f x (X) represents the prior probability distribution function of rock mass quality information X.

4. The rock mass quality interpretation method according to claim 1, characterized in that: In step S3, the method of calculating the likelihood function of rock mass quality using the confusion matrix includes: (1) Using an optimized classifier model to train the grid cell data of the advanced drilling data after isometric segmentation and secondary index calculation; (2) Using the trained classifier to predict the grid cell data of the advance drilling data; (3) The likelihood function is constructed based on the confusion matrix of the predicted output.

5. The rock mass quality interpretation method according to claim 4, characterized in that: The optimized classifier model formula in step (1) is: In the formula, O bj represents the objective function of the model, n represents the total amount of data imported into the kth tree, and y i represents the true label, represents the predicted value, f k represents the kth decision tree, and Ω represents the transformation value of the tree model.

6. The rock mass quality interpretation method according to claim 5, characterized in that: In step (2), the expression of the likelihood function constructed based on the confusion matrix is: Where TP and FN represent the number of true positive and false negative samples in the prediction confusion matrix, respectively.

7. The rock mass quality interpretation method according to claim 1, characterized in that: In step S4, the Bayesian probability update mechanism includes: based on the two-dimensional prior probability distribution function and the likelihood function, using the Bayesian probability estimation formula to obtain the posterior probability distribution function, and then obtaining the fusion interpretation result of the rock mass quality at the drilling hole location through the posterior probability distribution function; the Bayesian probability estimation formula is: f x ′(X)=aL(X)f x (X); In the formula, f x ′(X) represents the posterior probability distribution function of rock mass quality information X; f x (X) represents the prior probability distribution function of rock mass quality information X; X is the original spatial random input vector, X = (X1, X2, ..., X n ) T , n is the number of random parameters; L(X) represents the likelihood function; a is a constant; T is the matrix transposition operator.

8. The rock mass quality interpretation method according to claim 1, characterized in that: In step S5, the global update mechanism includes: based on the two-dimensional prior probability distribution function of rock mass quality and the fused interpretation result of the rock mass quality at the drilling hole position, using the global update formula, obtaining the fused probability distribution function, and then obtaining the fused interpretation result of the global rock mass quality by fusing the probability distribution function; the global update formula is: f x ″(X)=f x (X)+d x (X); In the formula, f x ″(X) represents the posterior probability distribution function of the rock mass quality information X that needs to be globally updated in addition to the Bayesian probability update position; d x (X) represents the probability update value calculated based on the distance from the drilling location; f x (X) represents the prior probability distribution function of rock mass quality information X; The d x The calculation formula of (X) is: dx(X)=f x ′(X)ω d ; In the formula, ω d represents weight; f x ′(X) represents the posterior probability distribution function of rock mass quality information X; The ω d The calculation formula is as follows: d =e -bd ; In the formula, d is the distance from the drill hole, b is the constant term, and e is a constant.

9. A rock mass quality interpretation device, characterized in that: It includes at least one processor and a memory that is communicatively connected to the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the rock mass interpretation method described in any one of claims 1-8.

10. A computer-readable medium, characterized in that Instructions executable by a processor are stored thereon, and when the instructions are executed by the processor, the processor executes the rock mass quality interpretation method according to any one of claims 1 to 8.

Images