Drilling hole arrangement optimization method for geological exploration of channel dredging engineering

Abstract

The invention relates to a channel dredging engineering geological exploration drill hole arrangement optimization method, which comprises the following steps of: firstly, on the basis of a geostatistics principle, analyzing and constructing a direct sampling method geological section construction model capable of being used for optimizing a subsequent drill hole position by adopting a Bayesian theory; and proposing a specific simulation calculation method for the soil texture type of each unknown unit in the simulation grid, and then carrying out statistical analysis on a stochastic simulation result to realize uncertainty quantification so as to optimize a drilling position. The method provides a theoretical basis for optimizing the drilling position of channel dredging geological exploration, is beneficial to optimizing the drilling arrangement position of geological exploration, provides decision support for geological exploration of channel dredging engineering, provides reliable soil information for subsequent construction, and is suitable for popularization and application. And the method is of great significance to cost budgeting, construction period arrangement and dredger type selection of dredging engineering.

Description

Technical Field

[0001] The invention relates to the technical field of waterway dredging and geological engineering, and in particular to a method for optimizing the arrangement of geological exploration drilling holes for waterway dredging engineering. Background Art

[0002] The shipping industry accounts for more than 90% of the world's trade transportation services and plays a vital role in the global trade system. If shipping is disrupted due to factors such as channel blockage, global trade stability may be significantly affected. Therefore, dredging projects are the foundation for promoting the stability and prosperity of the shipping industry. Dredging is a technology that uses professional machinery to excavate underwater rock and soil and transport the excavated materials to designated areas. The soil quality information of the dredged area plays a key role in the difficulty of construction. It not only affects the setting of the dredger construction parameters, but also involves the subsequent filling efficiency, yard design, etc. Therefore, an accurate understanding of the soil conditions in the dredged area is crucial for the smooth implementation of dredging projects.

[0003] Borehole sampling is the main method for obtaining soil information through geological exploration in waterway dredging projects. However, offshore exploration and construction conditions are more difficult and costly than on land, so samples are usually limited and sparse. In this process, in order to measure the soil information of the entire dredging area as much as possible, the equal spacing method is often used for drilling sampling. This sampling strategy does not fully consider the actual distribution of underground soil layers, such as the spatial variability and uncertainty of the soil body, and lacks scientific basis. Sparse borehole data and unreasonable sampling strategies make it difficult to accurately construct geological profiles, and thus cannot provide accurate soil layer distribution information, affecting the design of dredging project implementation plans. Therefore, it is necessary to propose a method for optimizing the arrangement of boreholes for geological exploration of waterway dredging projects, which provides a new way to solve the above problems. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a method for optimizing the arrangement of geological exploration drilling holes for waterway dredging projects, which is beneficial to optimizing the arrangement of geological exploration drilling holes, providing decision support for geological surveys of waterway dredging projects, and providing reliable soil information for subsequent construction.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for optimizing drilling arrangement in geological exploration of waterway dredging engineering comprises the following steps:

[0007] S1. Select appropriate simulation grids, training images and standard simulation images from geological profile data, divide the grids, and assign soil types to the grids at specified locations to form three input data set files;

[0008] S2. assigning the prior attributes in the standard simulation image to all the cells to be simulated in the simulation grid to construct a prior model;

[0009] S3. Setting a random path in the simulation grid and the training image respectively;

[0010] S4. Determine the search radius r according to the number of cells input, and extract the data template d in the simulation grid g , set the search ratio f in the training image;

[0011] S5. According to the random path, use the data template d g Perform mobile scanning on the training image to extract data events d t ;

[0012] S6. Calculation data template d g With data event d t The similarity distance d{d g ,d t}, update the simulation grid;

[0013] S7. Obtaining units with geological property probabilities, simulating the current unknown units multiple times in the training image, and outputting geological profiles after all units are simulated;

[0014] S8. Use discretization to quantify the uncertainty of simulation results and predict the optimal drilling location.

[0015] According to the above technical solution, preferably, in step S4, in any one of the data templates d g Under the condition of occurrence, the probability of any soil type appearing at an unknown unit in the simulation grid is expressed as P[G(c i )=g|d g ],in:

[0016]

[0017] G(c i ) = g

[0018] Among them, c i represents an unknown unit in the simulation grid; G(c i ) = g represents a unit attribute, such as G(c i ), i∈{1,2,3,…,n}, is the soil type; P[d g and G(c i )=g] represents s and d g Both appear at the same time in c i The probability of position; P(d g ) means dg The probability of appearing in the simulation grid is called the prior probability. According to the Bayesian formula, when the prior probability occurs, P[G(c i )=g|d g ] is called the posterior probability, which refers to i The soil type at is s.

[0019] According to the above technical solution, preferably, in step S5, considering that the borehole samples in the waterway dredging project are sparse, the available d g Very rarely, since the training images and the simulation grid have similar geological structures and have enough data events d t , can be indirectly derived from the training images,

[0020] P(d g )=P(d t )

[0021] Assume that the prior probabilities of the two are equal, where T(c i ) represents a unit in TI,

[0022] P[d g andG(c i )=g]=P[d t and T(c i )=t]

[0023] After merging,

[0024]

[0025] Prior probability P(d t ) can be calculated by the following equation:

[0026]

[0027] Among them, I(.) is an indicator function. When d is given, t When the value is equal to T(c i ), exists in the training image; N l ×N h is the total number of cells in the training image, l and h represent the horizontal and vertical depth directions respectively; P(d g ) can be interpreted in Bayesian terms as the evidence factor, P[d g and G(c i )=g] can be interpreted as the product of the prior probability and the likelihood probability, which are directly derived from the different d t and different soil types.

[0028] According to the above technical solution, preferably, in step S6, the d of the simulation grid is g Each position g(c i ) is transformed into an indicator variable. If the soil type of the training image is equal to that of the simulation grid, the indicator variable a = 0, otherwise it is equal to 1. Then the sum is calculated and the average value is taken to obtain d g and d t d, that is:

[0029]

[0030] Among them, d g and d t are the data templates in the simulation grid and the data events in the training image respectively; g(c i ) and t(c i ) are the measured data of the corresponding positions; n is the number of surrounding units.

[0031] According to the above technical solution, preferably, in step S6, in order to emphasize the relative spatial position between the drilling data, a specific weight of the conditional data is defined, and its value is inversely proportional to the vector distance from the central estimation point.

[0032]

[0033] h i =||c i -c0||

[0034] Among them, h i is d g Each condition point c i The vector distance from the central estimated point x0, δ is h i The weight adjustment coefficient of d g and d t The difference between them; when d>0, a corresponding penalty will be imposed on the mismatched units near g(c), and the larger the value, the greater the penalty; when d<0, more weights will be imposed on the mismatched units far away from g(c).

[0035] Here, a specific weight of the conditional data is defined, and d=1 is set, which further highlights the importance of nearby units.

[0036] According to the above technical solution, preferably, in step S6, in order to solve the spatial variability of soil layer distribution, the Mahalanobis distance is introduced.

[0037]

[0038] Among them, S -1It can exclude the interference of the correlation between variables and eliminate the influence of the measurement dimension; the value of d satisfies 0 £ d £ 1, and this distance represents d g and d t the similarity between; the value 0 indicates that the two are exactly the same, on the contrary, 1 indicates that the two are completely different, and the larger the value, the worse the similarity;

[0039] In the general simulation process, it is not that d g and d t are exactly equal to be accepted, but when d{d g ,d t} is less than the similarity threshold t, that is, d{d g ,d t} < t, then the two are considered similar.

[0040] According to the above technical solution, preferably, in step S7, the dispersion can be used to quantify the uncertainty of the simulation result, that is, the probability distribution of different soil types is obtained at the same position. The greater the dispersion, the higher the uncertainty of the simulation result.

[0041]

[0042] Among them, Dd represents the ratio of the number of times of non - matching correct categories at a certain position x in the simulation image to the total number of successful matches; G(c i ) represents the r - th non - matching correct category at position x; S(c i ) represents the true type value at position x; N max represents the maximum number of matches.

[0043] According to the above technical solution, preferably, in step S7, a statistical analysis is performed on the dispersion. It is considered that the greater the vertical average dispersion, the higher the uncertainty of this position. Based on this, the drilling position can be optimized.

[0044]

[0045] Among them, Ad represents the vertical average dispersion at a certain horizontal position, Dd represents the dispersion of a certain unit, and N h represents the total number of vertical units.

[0046] The beneficial effects of the present invention are:

[0047] The present invention proposes a method for optimizing the arrangement of drilling holes for geological exploration of waterway dredging projects based on the direct sampling method. The method is divided into three stages. The first stage is based on the principle of geostatistics and adopts Bayesian theory analysis to construct a direct sampling method geological profile construction model that can be used for subsequent optimization of drilling positions. The second stage analyzes the principle of direct sampling method and proposes a specific simulation calculation method for the soil type of each unknown unit in the simulation grid. The third stage performs statistical analysis on the random simulation results to realize uncertainty quantification to optimize the drilling position, in order to provide decision support for geological exploration of dredging projects. The present invention provides a theoretical basis for optimizing the location of drilling holes for geological exploration of waterway dredging, and is also of great significance for the cost budget, schedule arrangement and dredger selection of dredging projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a technical route for optimizing drilling arrangement in geological exploration for waterway dredging projects.

[0049] Figure 2 This is a schematic diagram of the principle of direct sampling method.

[0050] Figure 3 These are standard simulation images and training images selected from geological sections.

[0051] Figure 4 These are the random simulation results: (a) simulation image; (b) soil layer depth; (c) soil type frequency.

[0052] Figure 5 It is the Ad distribution of two sampling strategies: equal spacing and direct sampling.

[0053] Figure 6 These are the simulated images and Dd distribution diagrams of the two sampling strategies: equal spacing and direct sampling: (a) simulated image; (b) Dd distribution diagram.

[0054] Figure 7 It is a quantitative analysis of the uncertainty of two sampling strategies: equal spacing and direct sampling. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and the best embodiment. Based on the embodiments in the invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the invention.

[0056] Embodiment 1: The present invention comprises the following steps:

[0057] Step A: According to the first aspect of the present invention, a direct sampling geological profile construction model is proposed through geostatistical learning model analysis, which can be used for subsequent drilling location optimization, including:

[0058] Step A1: Select appropriate simulation grids, training images and standard simulation images from the geological profile data and convert them into digital form. According to the actual engineering needs, divide the grids according to the set resolution size and assign the soil type to the grid at the specified location to form three input data set files.

[0059] Step A2: Assign the prior attributes in the standard simulation image to all the cells to be simulated in the simulation grid to construct a prior model.

[0060] Step A3: Randomly assign a number between 0 and 1 to each unknown cell of the simulation grid and each cell of the training image, ensuring that each cell has equal priority. Then sort them in descending order, and set a random path in the simulation grid and training image according to this process.

[0061] Step A4: Generally, in any data template (d g ) occurs, the probability of any soil type appearing at an unknown unit in the simulation grid is expressed as P[G(c i )=g|d g ]. According to the definition of Bayes’ formula, P[G(c i )=g|d g ] can be expressed as:

[0062]

[0063] Among them, c i represents an unknown unit in the simulation grid; G(c i ) = g represents a unit attribute, such as G(c i ), i∈{1,2,3,…,n}, is the soil type; P[d g and G(c i )=g] represents s and d g Both appear at the same time in c i The probability of position; P(d g ) means d g The probability of appearing in the simulation grid is called the prior probability. According to the Bayesian formula, when the prior probability occurs, P[G(c i )=g|d g ] is called the posterior probability, which refers to i The soil type at is s.

[0064] Step A5: Considering that the borehole samples in the channel dredging project are sparse, the available d g Very few. Because the training images and simulation grids have similar geological structures and have enough data events (d t ), which can be indirectly derived from the training image, as shown in formula (2). Assuming that the prior probabilities of the two are equal, T(c i ) represents a unit in TI.

[0065] P(d g )=P(d t ) (2)

[0066] P[d g andG(c i )=g]=P[d t and T(c i )=t] (3)

[0067] Substituting formula (2) and (3) into (1), we get a new formula as follows:

[0068]

[0069] However, the prior probability P(d t ) can be calculated by the following equation:

[0070]

[0071] Among them, I(.) is an indicator function. When d is given, t When the value is equal to T(c i ), exists in the training image; N l ×N h is the total number of cells in the training image. l and h represent the horizontal and vertical depth directions, respectively. g ) can be interpreted in Bayesian terms as the evidence factor, P[d g and G(c i ) = g] can be interpreted as the product of the prior probability and the likelihood probability. They are directly derived from the different d t and different soil types. Therefore, the performance of the direct sampling method is affected by the similarity between the training images and the standard simulation images.

[0072] Step A6: To compare d g and d t To measure the similarity of , we need to define a similarity measure. The direct sampling method uses the similarity distance (d) to measure d g and d tFirst, the similarity between each position g(c i ) is transformed into an indicator of the soil type. If the soil type of the training image is equal to that of the simulation grid, the indicator variable a = 0, otherwise it is equal to 1. Then the sum is taken and the average is taken to obtain d g and d t d, that is:

[0073]

[0074]

[0075] Among them, d g and d t are the data templates in the simulation grid and the data events in the training image respectively; g(c i ) and t(c i ) are the measured data of the corresponding positions; n is the number of surrounding units.

[0076] Step A7: Formula (6) only considers the similarity of soil types and assumes that all units have the same importance. In fact, for the geological data of the waterway dredging project, d g and d t The unit is the conditional data, and the adjacent units should be given more information about the central proxy point in order to better utilize the geological information of the adjacent units. In order to emphasize the relative spatial position between the borehole data, a specific weight of the conditional data is defined, and its value is inversely proportional to the vector distance from the central proxy point. As shown in formula (8):

[0077]

[0078] h i =||c i -c0|| (9)

[0079] Among them, h i is d g Each condition point c i The vector distance from the central estimated point x0, δ is h i The calculated distance reflects the weight adjustment coefficient of d g and d t When d>0, a corresponding penalty will be imposed on the mismatched units near g(c), and the larger the value, the greater the penalty. On the contrary, when d<0, more weight will be imposed on the mismatched units far away from g(c). For the research object of this paper, a specific weight of conditional data is defined, and d=1 is set, which highlights the importance of nearby units.

[0080] Step A8: The d calculated by formula (9) is independent of direction. In order to solve the problem of spatial variability of soil layer distribution, the Mahalanobis distance is introduced.

[0081]

[0082] Among them, S -1 The interference of correlation between variables can be eliminated, and the influence of measurement dimension can be eliminated. The value of d satisfies 0￡d￡1. This distance indicates that d g and d t The value 0 means that the two are exactly the same, and 1 means that the two are completely different. The larger the value, the worse the similarity. g and d t Completely equal to accept, but when d{d g ,d t When} is less than the similarity threshold (t) (Formula (11)), the two are considered similar. This avoids multiple invalid cycles and improves the simulation speed.

[0083] d{d g , d t} <t (11)

[0084] Step B: Based on the geostatistical model in step A, the soil type of each unknown unit in the simulation grid is simulated and calculated one by one, including:

[0085] Step B1: Determine the search radius (r) based on the number of cells entered and extract the d g . Set the search ratio f in the training image, generally f = 0.7.

[0086] Step B2: Follow the random path, using d g Perform mobile scanning on the training image and extract d t The scanning range is determined according to the set search box. The center of the scan is the position of g(c).

[0087] Step B3: Calculate d g With d t The similarity distance d{d g ,d t}, using formulas (8)-(10), where formula (10) takes into account the spatial variability of the soil.

[0088] Step B4: The d obtained from the training image t With the simulation grid d g Compare, if d{d g ,d tIf it is less than t, the soil type in the training image is directly assigned to the current unknown cell; if it is greater than the set t, the current cell is discarded and the next cell is scanned. The attributes of previously simulated successful cells are used in subsequent simulation processes.

[0089] Step B5: For the current unknown cell, perform multiple simulations on the training image. When the maximum number of matching times N max is reached, the scanning stops and the next g(c) is transferred in the simulation grid. If the number of successful matches does not reach N max , the cell being simulated is temporarily frozen. After simulating all the remaining cells, the frozen cells are simulated again. When all cells are simulated, the geological profile is output.

[0090] Step C: According to the second aspect of the present invention, in combination with the geological profile construction model of the direct sampling method in Step A, an optimization method for the layout of exploration boreholes in waterway dredging engineering geological exploration is proposed, including:

[0091] Step C1: Since the same location will be matched multiple times, different soil types may be successfully matched. To study the proportion of other error types, it is necessary to introduce the concept of dispersion (Formula (12)). Dispersion can be used to quantify the uncertainty of simulation results, that is, the probability distribution of obtaining different soil types at the same location. The greater the dispersion, the higher the uncertainty of the simulation results.

[0092]

[0093] Among them, Dd represents the ratio of the number of times of non-matching correct categories to the total number of successful matches at a certain position x in the simulation image. G(c i ) represents the r-th non-matching correct category at position x; S(c i ) represents the true type value at position x. Nmax represents the maximum number of matching times. The above calculation results can be displayed by a dispersion diagram to show the dispersion of each position, so as to clearly show the interpolation uncertainty of different positions. Compared with traditional categorical variable statistics, this method has been normalized, with theoretical maximum and minimum values of 0 and 1 respectively, and is not affected by the size of the classification values.

[0094] Step C2: In actual dredging exploration, it is usually necessary to rely on the experience of geotechnical engineers to evaluate areas with high uncertainty and increase borehole sampling accordingly. However, this method has certain subjectivity and limitations, resulting in the accuracy and efficiency of borehole location selection being affected. To solve this problem, a statistical analysis of dispersion is carried out. It is considered that the greater the vertical average dispersion (see Formula (13)), the higher the uncertainty of this position, and the drilling location can be optimized based on this.

[0095]

[0096] Where Ad represents the vertical average dispersion at a certain horizontal position, Dd represents the dispersion of a certain unit, and its value can be calculated by formula (12). h Represents the total number of vertical units.

[0097] Embodiment 2:

[0098] (1) Project Overview:

[0099] In order to overcome the shortcomings of the existing equally spaced borehole sampling method, a method for optimizing the arrangement of boreholes for geological exploration of waterway dredging engineering based on direct sampling method was provided. Figure 1 For the technical route of the present invention, Figure 2 This is a schematic diagram of the calculation flow of the direct sampling method. For a detailed introduction to the principle, see step B in the invention content. A 100,000-ton waterway (southern extension) was selected as an example project. The new length of the waterway is 3135m, the navigation width is 190m, and the design bottom elevation of the waterway is -13.3m. The total dredging volume of the waterway is 6.7551 million m3. The waterway exploration line is arranged along the center line of the long axis of the dredging area, and a total of 1 exploration line is set up. The drilling density of the geological survey design is relatively low. A total of 8 boreholes are set up, and the boreholes are evenly spaced at 400m. From top to bottom, the soil layers are distributed in the following order: silt, silty clay, clay, and gravel. Two parts were selected from the geological profile as training images and standard simulation images (such as Figure 3 This covers all soil types in this project and is representative. The training images and standard simulation images have the same size, with a total horizontal length of 1200m and a vertical depth of 12m.

[0100] (2) Data preprocessing:

[0101] According to the actual engineering situation, combined with the training image and the standard simulation image size, the unit is divided into 10m long and 0.1m wide, that is, the resolution in the horizontal and vertical directions is 10m and 0.1m respectively. This is ideal for the actual geological survey process, especially considering that the horizontal drilling spacing is 400m. When dividing the soil layer, it is generally carried out in dm units, so the resolution of 0.1m meets the engineering technical requirements.

[0102] In this study, the performance of the direct sampling method is mainly affected by N, t, and l. First, in group #1, t is set to 0 and l is set to 400, and the optimal values ​​when n = 5, 10, 20, 30, 50, and 100 are explored. Then, in group #2, the best n obtained in group #1 is used, and l is also set to 400, and the optimal values ​​when t = 0, 0.1, 0.2, 0.3, 0.5, and 0.7 are explored. Finally, in group #3, the optimal n and t are set, and the impact of l = 400, 200, 100, and 50 on the performance of the proposed method is studied. The boreholes are evenly spaced, l is 400, which is the actual exploration distance, and the rest are added virtual boreholes. Preliminary experiments found that when n = 30 and t = 0 is the best, so n = 30, t = 0, and l = 400 are set in this study. In addition, before evaluating the impact of simulation parameters, it is necessary to clarify the Nmax of a single random simulation. This study conducted a 100-cycle pilot experiment and found that the curve in the initial stage decreased exponentially and rapidly, and stabilized after the 50th time. In order to ensure the accuracy of subsequent simulation parameters, the number of cycles was set to 80 to meet the pre-set requirements.

[0103] (3) Geological profile construction:

[0104] In the case of the present invention, the boreholes are arranged at equal intervals along the dredging center area during geological survey. At this time, the borehole spacing is sparse, and it is unreasonable to sample at equal intervals in areas with complex terrain conditions. The geological profile is randomly simulated by the optimal simulation parameters set above. Figure 4 (a) shows a cross-section MN and lists four representative positions. This simulation image is the result of simulation based on the optimal simulation parameters n=30, t=0, l=400. Figure 4 (b) shows the soil distribution at different depths under this section line. The simulated image basically reproduces the soil spatial distribution of the standard simulated image. The difference is that there are some differences in the depth of soil distribution, especially at the boundary of the soil layers. The predictions of the four locations are very different, especially at the boundary of the two soil layers, even if the prediction is accurate, the uncertainty is still high. For example, at location A, the frequencies of silt and silty clay are 46 and 34 times respectively. For location B, although the final result is predicted incorrectly, there are still 28 accurate predictions. At location C, far away from the stratification, the uncertainty is 0, and the prediction results are all the same soil type because the surrounding cells have the same properties. At location D, three possible soil types appear. Although the accuracy is high, the uncertainty is also high. This location is in the interlayer, corresponding to the gravel of silty clay. In the random matching process of the training image, there is even a matching clay situation, because clay is adjacent to silty clay and gravel.

[0105] (4) Random simulation results:

[0106] Figure 5 The Ad under the equally spaced sampling method and the sampling method we proposed is shown. The Ad values ​​at the known borehole data locations are set to 0 by default because they are known. When l = 400m, it corresponds to the actual exploration situation in the dredging area. The farther away from the borehole location, the higher the Ad. From left to right, Ad rises in a fluctuating manner and reaches a maximum value at about 220m. Then it begins to decline in a fluctuating manner, with another peak at 530m, and then continues to decline in a fluctuating manner, with the last larger peak at 950m, and then continues to decline in a fluctuating manner to 1200m. There are three larger peaks in the figure, decreasing from left to right. It can be seen from the standard simulation image that the positions of the A and B peaks are located on the left and right of the boundary of the gravel interlayer of the silty clay, respectively, and these two positions are also far away from the initial drilling position, so Ad is larger. According to the conclusions in Section 3.2.5, the Dd in this vicinity is very high and Pa is also very low. However, although the positions of 220m and 530m are both located at the boundary of this interlayer, the Pa of the other three soils at 220m is lower than that at 530m, so their Ad is higher than that at 530m. These fluctuations are affected by the wave-like changes at the stratification, which reflects the randomness and scientificity of the method we proposed in dredging survey application.

[0107] Figure 6 The Ad under the equally spaced sampling method and the sampling method we proposed is shown. The Ad values ​​at the known borehole data locations are set to 0 by default because they are known. When l = 400m, it corresponds to the actual exploration situation in the dredging area. The farther away from the borehole location, the higher the Ad. From left to right, Ad rises in a fluctuating manner and reaches a maximum value at about 220m. Then it begins to decline in a fluctuating manner, with another peak at 530m, and then continues to decline in a fluctuating manner, with the last larger peak at 950m, and then it continues to decline in a fluctuating manner to 1200m. There are three larger peaks in the figure, decreasing from left to right. It can be seen from the standard simulation image that the positions of the A and B peaks are located on the left and right of the boundary of the gravel interlayer of the silty clay, and these two positions are also far away from the initial drilling position, so Ad is larger. However, although the positions of 220m and 530m are both located at the boundary of this interlayer, the Dd of the other three soils at 220m is higher than that at 530m, so their Ad is higher than that at 530m. These fluctuations are affected by the wave-like changes at the stratification, which reflects the randomness and scientificity of the method we proposed in dredging survey application.

[0108] In order to compare the advantages and disadvantages of the two sampling strategies, we selected Figure 5The three higher peak points A, B and C (Ad are 0.153, 0.134, 0.126 respectively) are used to set virtual boreholes, and 7 drilling positions are formed with the 4 existing boreholes. It can be seen that when the number of boreholes increases from 4 to 7 (corresponding to l changing from 400m to 200m), the number of peaks increases from 3 to 6, and the peak value decreases significantly. In addition, compared with the equidistant sampling method, the peak value of the direct sampling method is lower, indicating that the uncertainty of the soil layer is affected by different sampling strategies. Increasing the number of boreholes can reduce the uncertainty of the soil layer. At the same time, under the same number of boreholes, the rationality of the borehole location will also affect the uncertainty of the soil layer. It should be noted that only an ideal method for optimizing the drilling position is provided here. In actual engineering, corresponding judgments need to be made based on factors such as underwater geological conditions and construction costs of dredging projects.

[0109] (5) Uncertainty analysis and quantification:

[0110] Furthermore, the effects of the equal spacing method and direct sampling method on soil layer distribution prediction under the same number of drilling holes are analyzed in detail. Figure 6 The effects of the two sampling strategies on soil layer distribution and Dd are shown, and a wave-shaped distribution is presented. However, their morphologies are slightly different, which is caused by the differences in the locations of their respective boreholes. The results show that, overall, the area where Dd>0 for the direct sampling method is smaller than that for the equal-interval method, especially in the interlayer location with complex geological conditions. The direct sampling method selects the location with the largest vertical difference as the new drilling location, so it is more suitable for dredging projects with complex geological conditions. It infers the location with the highest soil uncertainty based on the previously available exploration data and determines the most reasonable drilling location. In contrast, the equal-interval method only relies on distance for sampling strategy. Although soil properties are easier to predict in areas close to the borehole location, the direct sampling method is more scientific, reasonable and applicable under the premise of a limited and equal number of boreholes. It takes into account the spatial distribution of soil uncertainty, can more accurately predict soil layer distribution, and reduce the discreteness of the prediction results. This provides a new method for dredging surveys to explore the distribution of underwater soil layers.

[0111] In order to quantify the uncertainty of the direct sampling method, the uncertainty is divided into 6 different regions according to the range of Dd and displayed by fan diagrams (such as Figure 7As shown). The entire simulated image area is divided into low uncertainty areas (0-0.1), medium uncertainty areas (0.1-0.5) and high uncertainty areas (0.5-1). In the simulated images of the equal spacing method and the direct sampling method, most of the areas belong to the low uncertainty area, accounting for 88.4% and 89.5% respectively. In contrast, the high uncertainty area only accounts for 3.5% and 3.1% respectively. The medium uncertainty area occupies the remaining proportion. The entire simulated image area has a low uncertainty level as a whole. This verifies the advanced nature of the method for constructing geological profiles in this study. It is worth noting that the direct sampling method reduces uncertainty compared to the equal spacing method. Specifically, the proportion of low uncertainty areas increased by 1.24% compared to the equal spacing method, while the medium uncertainty area and high uncertainty area decreased by up to 8.53% and 11.43% respectively. This result further verifies the effectiveness of the direct sampling method in soil layer prediction. It can accurately determine the sampling location and reduce uncertainty to a certain extent, especially at the boundary of soil layers. Therefore, the direct sampling method is more scientific, reasonable and applicable than the equal-interval method. It is of great significance for constructing geological profiles and optimizing drilling locations, and can provide assistance for dredging engineering exploration decisions.

[0112] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for optimizing drilling arrangement for geological exploration of waterway dredging engineering, characterized in that: The steps include: S1. Select appropriate simulation grids, training images and standard simulation images from geological profile data, divide the grids, and assign soil types to the grids at specified locations to form three input data set files; S2. assigning the prior attributes in the standard simulation image to all the cells to be simulated in the simulation grid to construct a prior model; S3. Setting a random path in the simulation grid and the training image respectively; S4. Determine the search radius r according to the number of cells input, and extract the data template d in the simulation grid g , set the search ratio f in the training image; S5. According to the random path, use the data template d g Perform mobile scanning on the training image to extract data events d t ; S6. Calculation data template d g With data event d t The similarity distance d{d g ,d t }, update the simulation grid; S7. Obtaining units with geological property probabilities, simulating the current unknown units multiple times in the training image, and outputting geological profiles after all units are simulated; S8. Use discretization to quantify the uncertainty of simulation results and predict the optimal drilling location.

2. The method for optimizing the arrangement of drilling holes for geological exploration of waterway dredging engineering according to claim 1, characterized in that: In step S4, in any of the data templates d g Under the condition of occurrence, the probability of any soil type appearing at an unknown unit in the simulation grid is expressed as P[G(c i )=g|d g ],in: G(c i )=g Among them, c i represents an unknown unit in the simulation grid; G(c i ) = g represents a unit attribute, such as G(c i ), i∈{1,2,3,…,n}, is the soil type; P[d g and G(c i )=g] represents s and d g Both appear at the same time in c i The probability of position; P(d g ) means d g The probability of appearing in the simulation grid is called the prior probability. According to the Bayesian formula, when the prior probability occurs, P[G(c i )=g|d g ] is called the posterior probability, which refers to i The soil type at is s.

3. The method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 2, characterized in that: In step S5, considering that the borehole samples in the channel dredging project are sparse, the available d g Very rarely, since the training images and the simulation grid have similar geological structures and have enough data events d t , can be indirectly derived from the training images, P(d g )=P(d t ) Assume that the prior probabilities of the two are equal, where T(c i ) represents a unit in TI, P[d g andG(c i )=g]=P[d t and T(c i )=t] After merging, Prior probability P(d t ) can be calculated by the following equation: Among them, I(.) is an indicator function. When d is given, t When the value is equal to T(c i ), exists in the training images; N l ×N h is the total number of cells in the training image, l and h represent the horizontal and vertical depth directions respectively; P(d g ) can be interpreted in Bayesian terms as the evidence factor, P[d g and G(c i )=g] can be interpreted as the product of the prior probability and the likelihood probability, which are directly derived from the different d t and different soil types.

4. A method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 3, characterized in that: In step S6, each position g(c i ) is transformed into an indicator variable. If the soil type of the training image is equal to that of the simulation grid, the indicator variable a = 0, otherwise it is equal to 1. Then the sum is calculated and the average value is taken to obtain d g and d t d, that is: Among them, d g and d t are the data templates in the simulation grid and the data events in the training image respectively; g(c i ) and t(c i ) are the measured data of the corresponding positions; n is the number of surrounding units.

5. A method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 4, characterized in that: In step S6, in order to emphasize the relative spatial position between the drilling data, a specific weight of the conditional data is defined, and its value is inversely proportional to the vector distance from the central estimation point. h i =||c i -c0|| Among them, h i is d g Each condition point c i The vector distance from the central estimated point x0, δ is h i The weight adjustment coefficient of d g and d t The difference between them; when d>0, a corresponding penalty will be imposed on the mismatched units near g(c), and the larger the value, the greater the penalty; when d<0, more weights will be imposed on the mismatched units far away from g(c). Here, a specific weight of the conditional data is defined, and d=1 is set, which further highlights the importance of nearby units.

6. A method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 5, characterized in that: In step S6, the Mahalanobis distance is introduced to solve the spatial variability of soil layer distribution. Among them, S -1 The interference of correlation between variables can be eliminated, and the influence of measurement dimension can be eliminated; the value of d satisfies 0￡d￡1, and this distance represents d g and d t The similarity between them; a value of 0 means that the two are exactly the same, on the contrary 1 means that the two are completely different, and the larger the value, the worse the similarity; In a general simulation process, it is not that d g and d t are exactly equal to be accepted. Instead, when d{d g ,d t} is less than the similarity threshold t, that is, d{d g ,d t} < t, it is considered that the two are similar.

7. A method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 6, characterized in that: In step S7, the discreteness can be used to quantify the uncertainty of the simulation results, that is, the probability distribution of different soil types at the same location. The larger the discreteness, the higher the uncertainty of the simulation results. Where Dd represents the ratio of the number of non-matching correct categories to the total number of successful matches at a certain position x in the simulated image; G(c i ) represents the rth non-matching correct category at position x; S(c i ) represents the real type value at position x; N max Indicates the maximum number of matches.

8. The method for optimizing drilling arrangement for geological exploration of waterway dredging engineering according to claim 7, characterized in that: In step S7, a statistical analysis is performed on the dispersion, and it is considered that the greater the vertical average dispersion, the higher the uncertainty of the position, and the drilling position can be optimized based on this. Among them, Ad represents the vertical average dispersion at a certain horizontal position, Dd represents the dispersion of a certain unit, and N h Represents the total number of vertical units.

Images