A data set partitioning method for foundation pit monitoring neural network prediction

Abstract

The present invention provides a data set division method for foundation pit monitoring neural network prediction, the method comprising: obtaining the number of stratum types, groundwater height, project monitoring data during foundation pit construction, and subsequent monitoring data after the foundation pit construction is completed; classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit; dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted, and obtaining a training set and a test set in the data set. The present invention classifies the complex geological environment of the foundation pit into four categories according to the number of strata and the groundwater height at the foundation pit, and determines different data set division methods for each category of the foundation pit according to the construction progress of the foundation pit, so that the data in the training set and the test set are more representative, so that the performance of the model on the actual data can be more accurately evaluated, and the generalization ability of the model is improved.

Description

Technical Field

[0001] The invention relates to the technical field of foundation pit engineering, and in particular to a data set partitioning method for foundation pit monitoring neural network prediction. Background Art

[0002] The excavation of foundation pits will have a certain impact on the surrounding surface settlement, and then affect the displacement and deformation of nearby buildings. Among the monitoring projects of foundation pit engineering, there is a certain correlation between many projects. With the excavation of the foundation pit, the horizontal and vertical displacements of the retaining piles will increase, resulting in increased surface settlement. Therefore, a neural network model is used to predict the settlement of the surface around the foundation pit based on the monitoring data of the foundation pit.

[0003] When using a neural network to predict the displacement of a monitoring project, after determining the input and output, the data set needs to be divided into a training set and a test set. Traditionally, 70% of the data set is randomly selected for the training set and 30% for the test set.

[0004] However, due to the different geological environments of different foundation pits, the training set and test set are constructed based on the monitoring data in a fixed proportion by random selection, which makes it difficult for the constructed data set to reflect the impact of the complex geological conditions at the foundation pit on the foundation pit monitoring project, thereby affecting the generalization ability of the neural network trained using the data set. Summary of the invention

[0005] The present invention provides a data set division method for foundation pit monitoring neural network prediction, which is used to solve the defect that the data set randomly divided at a fixed ratio in the prior art is difficult to reflect the impact of the complex geological environment on the foundation pit monitoring project, and realizes a data set division method with a more scientific division ratio.

[0006] The present invention provides a data set partitioning method for foundation pit monitoring neural network prediction, comprising:

[0007] Obtain the number of strata types, groundwater height, project monitoring data during the foundation pit construction process, and subsequent monitoring data after the foundation pit construction is completed;

[0008] Classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit;

[0009] The project monitoring data and the subsequent monitoring data are divided according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set.

[0010] According to a data set division method for foundation pit monitoring neural network prediction provided by the present invention, the step of classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the position relationship between the groundwater height and the target bottom of the foundation pit specifically includes:

[0011] When the number of stratum types of the foundation pit to be predicted is one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a first type of foundation pit;

[0012] When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is not below the target bottom of the foundation pit, classifying the foundation pit to be predicted as a second type of foundation pit;

[0013] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a third type of foundation pit;

[0014] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater height is not below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a fourth type of foundation pit.

[0015] According to a data set division method for foundation pit monitoring neural network prediction provided by the present invention, when the foundation pit to be predicted is a first type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0016] When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is below the target bottom of the foundation pit, the project monitoring data during the foundation pit construction process is divided into the training set, and the subsequent monitoring data after the foundation pit construction is completed is divided into the test set.

[0017] According to a data set division method for foundation pit monitoring neural network prediction provided by the present invention, when the foundation pit to be predicted is a second type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0018] When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is not below the target bottom of the foundation pit, the project monitoring data before the foundation pit is constructed to the same height below the groundwater level is divided into the training set, and the remaining project monitoring data and the subsequent monitoring data are divided into the test set.

[0019] According to a data set division method for foundation pit monitoring neural network prediction provided by the present invention, when the foundation pit to be predicted is a third type foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0020] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater height is below the target bottom of the foundation pit, the project monitoring data when the foundation pit is constructed until the stratum above the target bottom of the foundation pit is completely excavated is divided into the training set, and the subsequent monitoring data is divided into the test set.

[0021] According to a data set division method for foundation pit monitoring neural network prediction provided by the present invention, when the foundation pit to be predicted is a fourth type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0022] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater level is not below the target bottom of the foundation pit, the project monitoring data before the foundation pit is constructed until the bottom of the foundation pit is not lower than the groundwater level is divided into the training set;

[0023] Among the project monitoring data generated during the construction of each stratum below the groundwater level, a preset number of the project monitoring data are divided into the training set;

[0024] The remaining project monitoring data and the subsequent monitoring data are divided into the test set.

[0025] The present invention also provides a training method for a foundation pit monitoring neural network model, comprising: using any of the data set division methods for foundation pit monitoring neural network prediction as described above, using project monitoring data during foundation pit construction and subsequent monitoring data after foundation pit construction is completed as training data to construct a data set, and using the data set to train the foundation pit monitoring neural network model.

[0026] The present invention also provides a data set partitioning device for foundation pit monitoring neural network prediction, comprising:

[0027] An acquisition module is used to obtain the number of stratum types of the foundation pit to be predicted, the groundwater height, the project monitoring data during the foundation pit construction process, and the subsequent monitoring data after the foundation pit construction is completed;

[0028] A classification module, used for classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit;

[0029] The division module is used to divide the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted, so as to obtain a training set and a test set in the data set.

[0030] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for dividing a data set for foundation pit monitoring neural network prediction as described in any one of the above is implemented.

[0031] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the data set partitioning methods described above for foundation pit monitoring neural network prediction.

[0032] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned data set partitioning methods for foundation pit monitoring neural network prediction.

[0033] The present invention provides a data set division method for foundation pit monitoring neural network prediction, which classifies the complex geological environment of the foundation pit according to the number of strata and groundwater height at the foundation pit, determines different data set division methods for each category according to the construction progress of the foundation pit, and divides the project monitoring data of different construction stages and subsequent monitoring data into training sets and / or test sets, so that the data in the training sets and test sets are more representative, thereby being able to more accurately evaluate the performance of the model on real data and improve the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 It is one of the flow diagrams of a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0036] Figure 2 (a) is one of the schematic diagrams for showing the first type of foundation pit in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0037] Figure 2(b) is a second schematic diagram for showing the first type of foundation pit in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0038] Figure 3 It is a schematic diagram for displaying the second type of foundation pit in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0039] Figure 4 It is a schematic diagram for displaying the third type of foundation pit in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0040] Figure 5 It is a schematic diagram for displaying the fourth type of foundation pit in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0041] Figure 6 (a) is a schematic diagram of the geological environment of a foundation pit for testing case 1 in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0042] Figure 6 (b) is a schematic diagram of the geological environment of the foundation pit of Case 2 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0043] Figure 6 (c) is a schematic diagram of the geological environment of the foundation pit of Case 3 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0044] Figure 6 (d) is a schematic diagram of the geological environment of the foundation pit of Case 4 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0045] Figure 7 (a) is one of the comparison diagrams of the prediction results of case 1 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0046] Figure 7 (b) is one of the comparison diagrams of the prediction results of Case 2 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0047] Figure 7 (c) is one of the comparison diagrams of the prediction results of case 3 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0048] Figure 7(d) is one of the comparison diagrams of the prediction results of case 4 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0049] Figure 8 (a) is the second comparison chart of evaluation indicators of case 1 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0050] Figure 8 (b) is a second comparison chart of evaluation indicators of case 2 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0051] Figure 8 (c) is the second comparison chart of evaluation indicators of case 3 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0052] Figure 8 (d) is the second comparison chart of evaluation indicators of case 4 used for testing in a data set partitioning method for foundation pit monitoring neural network prediction provided by the present invention;

[0053] Fig. 9 It is a structural schematic diagram of a data set partitioning device for foundation pit monitoring neural network prediction provided by the present invention;

[0054] Fig.10 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0056] First, the following contents are introduced:

[0057] There are many factors affecting foundation pit monitoring, including geological conditions, engineering design, construction technology, etc., and geological conditions directly affect the stability and deformation characteristics of foundation pits, such as the nature, distribution, and hydrogeological conditions of the strata. Deep foundation pits often contain many strata, such as fill, sand, clay, etc. When the foundation pit is excavated to different strata, the monitoring data is affected to different degrees, so the traditional division method of foundation pit data sets is inaccurate, which in turn leads to insufficient generalization ability of neural network models.

[0058] It is precisely because of the complex geological conditions in the foundation pit that it is difficult to determine the division ratio when dividing the neural network data set. Different division ratios have different effects on the results of model prediction. Random division ratios are unscientific when facing complex foundation pit strata, which will lead to inaccurate prediction results.

[0059] The nonlinearity of some influencing factors cannot be considered intuitively when analyzing foundation pit monitoring data. Machine learning models have great flexibility in dealing with these complex nonlinear relationships. In machine learning research, the division of data sets is often a random selection of proportions. If the proportion of the training set is too small, it will affect the accuracy of the model; if the training set is too large, a large amount of data needs to be collected and the monitoring cycle is long. When the monitoring data cannot be obtained, the development trend of the monitoring deformation cannot be judged. Therefore, it is necessary to propose a method for dividing the foundation pit monitoring data set according to the excavation progress, soil layer distribution, and groundwater distribution to overcome the limitations of the traditional data set division method in foundation pit data.

[0060] Combine the following Figures 1 to 8 The present invention introduces a data set partitioning method for foundation pit monitoring neural network prediction, such as Figure 1 As shown, including:

[0061] Step 101, obtaining the number of stratum types, groundwater height, project monitoring data during the foundation pit construction process, and subsequent monitoring data after the foundation pit construction is completed;

[0062] The number of stratum types of the foundation pit to be predicted is obtained. The number of stratum types reflects whether the geological environment of the foundation pit to be predicted is a single stratum or a complex stratum. The monitoring data under different stratum environments can reflect the impact of different strata on the settlement of the foundation pit or other displacement changes.

[0063] The groundwater height at the foundation pit to be predicted is obtained, so as to determine the impact of groundwater on the foundation pit settlement or other displacement changes based on the groundwater height and the target construction depth of the foundation pit.

[0064] Obtain project monitoring data during foundation pit construction and follow-up monitoring data after foundation pit construction is completed. Use the two types of monitoring data as the specific content of the data set of the foundation pit monitoring neural network, select data from the project monitoring data and follow-up monitoring data and divide them into training sets and / or test sets to construct a data set.

[0065] It should be noted that the specific content of the project monitoring data and the subsequent monitoring data is determined according to the actual practical foundation pit monitoring neural network, and the data type required for the input of the foundation pit monitoring neural network can be selected from the project monitoring data and the subsequent monitoring data.

[0066] In a feasible implementation, a piece of data in the data set is constructed by taking the horizontal displacement and vertical displacement of the retaining piles as input data and taking the vertical displacement data of the ground surface around the foundation pit as output data.

[0067] In other feasible implementations, the displacement or deformation monitoring data of different elements in the foundation pit can be selected as input according to the input requirements of the neural network, and the deformation monitoring data of the surface in various directions around the foundation pit can be used as output.

[0068] Optionally, the present application does not limit the method for obtaining specific monitoring data in the project monitoring data and subsequent monitoring data, and the monitoring data can be data that can be obtained at the foundation pit construction site.

[0069] Taking the monitoring data of the horizontal displacement and vertical displacement of the retaining piles as an example, the horizontal displacement and vertical displacement data at any position within a preset radius around the retaining piles can be used as the displacement monitoring data of the retaining piles; multiple retaining piles in the foundation pit can be selected, and the horizontal displacement and vertical displacement data of the multiple retaining piles can be used as the horizontal displacement and vertical displacement of the retaining piles used in the data set.

[0070] Optionally, this application does not limit the time length corresponding to each piece of data in the project monitoring data and subsequent monitoring data. Taking the horizontal displacement data of the retaining piles as an example, the horizontal displacement data of the retaining piles within 12 hours can be regarded as one piece of data, or the horizontal displacement data of the retaining piles within 48 hours can be regarded as one piece of data. The specific time length division can be determined according to the actual construction period of the foundation pit.

[0071] Step 102, classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit;

[0072] The number of stratum types at the foundation pit to be predicted represents the strata encountered during the foundation pit construction process, and different strata have different effects on the deformation of the foundation pit.

[0073] The groundwater height is the height of the groundwater level at the foundation pit; the target bottom of the foundation pit is the expected construction depth of the foundation pit. By comparing the positional relationship between the groundwater height and the target bottom of the foundation pit, the impact of groundwater on the foundation pit construction can be determined.

[0074] The project monitoring data and the subsequent monitoring data are divided according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set.

[0075] The foundation pits are classified according to the number of stratum types to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit, so that different data set division strategies can be adopted for each type of foundation pit, so that the divided data sets can better consider the influence of the stratum environment and hydrological environment of the foundation pit on the deformation of the foundation pit.

[0076] Specifically, if the groundwater height is below the target bottom of the foundation pit, it is considered that the impact of groundwater on the settlement of the foundation pit is small; if the groundwater height is above the target bottom of the foundation pit, the impact of groundwater on the foundation pit also needs to be considered during the construction phase of the foundation pit.

[0077] The impact of the number of stratum types to be predicted on the foundation pit is similar to that of groundwater. It is also necessary to consider the strata at each stage of the foundation pit construction, so as to select data from the project monitoring data when the construction reaches each stratum and put it into the training set.

[0078] The present invention classifies foundation pits according to the complex geological environment of the foundation pits according to the number of strata and groundwater height at the foundation pits, determines different data set division methods for each category according to the construction progress of the foundation pits, and divides the project monitoring data of different construction stages and subsequent monitoring data into training sets and / or test sets, so that the data in the training sets and test sets are more representative, thereby being able to more accurately evaluate the performance of the model on real data and improve the generalization ability of the model.

[0079] In a data set division method for foundation pit monitoring neural network prediction of the present invention, the step of classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit specifically includes:

[0080] When the number of stratum types of the foundation pit to be predicted is one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a first type of foundation pit;

[0081] The number of stratum types is one, which means that the foundation pit to be predicted is in a single stratum environment. The stratum environment in which the foundation pit is located does not change when the construction is completed, and there is no situation in which different strata affect the monitoring data during the construction process.

[0082] It should be noted that the type of stratum refers to the type of stratum passed through when the foundation pit construction reaches the target bottom. If there are multiple strata at the foundation pit, but the foundation pit is only in one type of stratum during the construction to the bottom of the foundation pit, it is also considered that the foundation pit to be predicted is in a single stratum environment.

[0083] The groundwater height refers to the groundwater level height, which is below the target bottom of the foundation pit, that is, it is not affected by groundwater during the process of foundation pit construction to completion.

[0084] It should be noted that if there is no groundwater in the foundation pit, the groundwater height can also be understood as being below the target bottom of the foundation pit, such as Figure 2 As shown in Figure (a), Figure 2 Figure (b) shows that there is groundwater in the foundation pit but the groundwater level is below the target bottom of the foundation pit. The dotted line in the figure represents the groundwater height.

[0085] Foundation pits located in a single geological environment and not affected by groundwater are identified as Type I foundation pits.

[0086] When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is not below the target bottom of the foundation pit, classifying the foundation pit to be predicted as a second type of foundation pit;

[0087] The groundwater height is not below the target bottom of the foundation pit, which means that the foundation pit construction process will be affected by groundwater. That is, during one part of the construction process, there is no groundwater in the construction environment, and during the other part of the construction process, the foundation pit will be constructed in a groundwater environment.

[0088] like Figure 3 As shown, Figure 3 The gap indicates the completed foundation pit, and the dotted line indicates the groundwater height. That is, after the construction of the foundation pit with a single stratum is completed, the target bottom of the foundation pit is lower than the groundwater height.

[0089] Foundation pits located in a single geological environment and affected by groundwater are identified as Type II foundation pits.

[0090] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a third type of foundation pit;

[0091] The number of stratum types of the foundation pit to be predicted is greater than one, that is, the foundation pit passes through different strata during the construction to its target bottom. Therefore, it is necessary to consider the impact of different strata on the project monitoring data.

[0092] The foundation pits in complex geological environments and not affected by groundwater are identified as the third type of foundation pits, such as Figure 4 As shown, Figure 4 Different fill patterns in the represent different strata.

[0093] When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater height is not below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a fourth type of foundation pit.

[0094] The fourth type of foundation pit has more than one type of strata and the groundwater level is not below the target bottom of the foundation pit. In other words, the foundation pit passes through different strata during the construction to its target bottom and is affected by groundwater during the construction process. Figure 5 shown.

[0095] In a data set division method for foundation pit monitoring neural network prediction of the present invention, when the foundation pit to be predicted is a first type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0096] The project monitoring data during the foundation pit construction process is divided into the training set, and the subsequent monitoring data after the foundation pit construction is completed is divided into the test set.

[0097] Since the first type of foundation pit is in a single stratum and is not affected by groundwater, the data during the foundation pit excavation process, that is, the project monitoring data of the foundation pit, can be directly divided into the training set, and the data after the foundation pit excavation is completed, that is, the subsequent monitoring data, can be divided into the test set.

[0098] In a data set division method for foundation pit monitoring neural network prediction of the present invention, when the foundation pit to be predicted is a second type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0099] The project monitoring data before the foundation pit is constructed to a height where the bottom of the foundation pit is lower than the groundwater level is the same as the groundwater level is divided into the training set, and the remaining project monitoring data and the subsequent monitoring data are divided into the test set.

[0100] Since the presence of groundwater will affect foundation pit monitoring projects, such as the displacement of retaining piles, and affect the stability of the foundation pit, therefore, when the foundation pit to be predicted is the second type of foundation pit, according to the construction progress of the foundation pit, the project monitoring data when the foundation pit bottom is at the same height as the groundwater level is divided into the training set, and the remaining project monitoring data and subsequent monitoring data are divided into the test set.

[0101] Among them, the bottom of the foundation pit refers to the bottom height during the foundation pit construction process. When the foundation pit is constructed until the bottom of the foundation pit is the same height as the groundwater level, the project monitoring data generated will not be affected by groundwater.

[0102] The remaining project monitoring data are the project monitoring data when the foundation pit is constructed from the groundwater height to the height of the target bottom of the foundation pit. The project monitoring data generated in this process are affected by groundwater.

[0103] In a specific implementation, the project monitoring data is stored in days. When the height of the bottom of the foundation pit is the same as the groundwater height in the project monitoring data on the 130th day, the project monitoring data of the first 130 days are divided into a training set, and the project monitoring data of subsequent days and subsequent monitoring data are divided into a test set.

[0104] Through the above division, the data in the test set not only includes the subsequent monitoring data after the completion of the foundation pit construction, but also includes the project monitoring data when the foundation pit construction is affected by groundwater, so that the impact of groundwater on the foundation pit is taken into consideration during the data set division process.

[0105] In a data set division method for foundation pit monitoring neural network prediction of the present invention, when the foundation pit to be predicted is a third type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0106] The project monitoring data when the foundation pit is constructed until the stratum above the target bottom of the foundation pit is completely excavated is divided into the training set, and the subsequent monitoring data is divided into the test set.

[0107] When the foundation pit to be predicted is the third type, considering the diverse distribution of strata and the uneven impact of monitoring data, it is impossible to randomly select the data set ratio, nor can the excavation process monitoring data be directly divided into a training set.

[0108] In order to reduce the impact of uneven factors on the data, when there are multiple strata and the foundation pit is not affected by groundwater, the project monitoring data when the foundation pit is constructed to the strata above its target bottom are completely excavated are divided into the training set.

[0109] Since incomplete excavation of the strata may result in uneven stratum conditions for the supporting structure, which may affect the project monitoring data of this construction phase and cause errors, the remaining project monitoring data are neither included in the training set nor in the test set.

[0110] At the same time, the subsequent monitoring data is divided into the test set.

[0111] In a specific embodiment, during the excavation of the foundation pit to the target bottom of the foundation pit, three different strata are passed through, wherein the target bottom of the foundation pit is still located in the third stratum, then the first two strata passed through are considered to be completely excavated strata, and the third stratum is considered to be an incompletely excavated stratum.

[0112] On this basis, the project monitoring data generated during the foundation pit construction until the first two strata were completely excavated were divided into the training set, and the subsequent monitoring data generated after the foundation pit construction was completed were divided into the test set.

[0113] In a data set division method for foundation pit monitoring neural network prediction of the present invention, when the foundation pit to be predicted is a fourth type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes:

[0114] The monitoring data of the project before the foundation pit is constructed until the bottom of the foundation pit is not lower than the groundwater level is divided into the training set;

[0115] When the foundation pit to be predicted is the fourth type of foundation pit, the impact of groundwater on foundation pit construction is given priority. Therefore, all project monitoring data before the foundation pit is excavated to a level not lower than the groundwater level are first divided into the training set.

[0116] Among the project monitoring data and the subsequent monitoring data generated during the construction of each stratum below the groundwater level, a preset number of monitoring data are divided into the training set;

[0117] The remaining monitoring data are divided into the test set.

[0118] Furthermore, in order to consider the impact of different strata on the foundation pit, a preset number of monitoring data are divided into the training set from the project monitoring data generated during the construction process of the bottom of the foundation pit from a position not lower than the groundwater level to the target bottom position and the subsequent monitoring data after the construction is completed.

[0119] It should be noted that the number of subsequent monitoring data will be different depending on the preset number of copies. The preset number of monitoring data may only include project monitoring data, or may include project monitoring data and part of the subsequent monitoring data, which is determined by the project monitoring data and subsequent monitoring data actually obtained from the foundation pit.

[0120] Optionally, the preset number of copies is determined according to the total number of subsequent monitoring data. In a feasible implementation, the first one-third of the monitoring data is divided into the training set.

[0121] On this basis, the remaining monitoring data are divided into the test set. At this time, the test set also includes the monitoring data of each soil layer project constructed below the groundwater level, so as to consider the impact of complex strata and groundwater on the foundation pit at the same time.

[0122] The present invention summarizes four common types of foundation pits according to the complex stratigraphic characteristics of the foundation pits and the relationship between the groundwater height and the target bottom of the foundation pits, and determines the corresponding data set division method for each type of foundation pit, so that the training set and the verification set in the divided data set can both reflect the influence of the geological environment of each type of foundation pit on the deformation or displacement of the foundation pit, thereby making the data in the training set and the test set more representative. The more common method of random division according to a fixed ratio is accurate, so that a more accurate foundation pit monitoring model can be trained. Compared with the traditional method of manually randomly dividing the data set, it can not only reduce the time of model training, improve the model speed, but also reduce the risk of model overfitting.

[0123] In order to further verify the data set division method for foundation pit monitoring neural network prediction provided in this application, verification was carried out based on project monitoring data and follow-up monitoring data obtained from a foundation pit project in a county, a water purification plant foundation pit project, a high-speed railway station foundation pit project, and a subway foundation pit project.

[0124] Specifically, relevant parameters of the foundation pit project are obtained, including the excavation depth of the foundation pit, the soil type of the foundation pit site, the thickness of the site soil layer, the depth of the groundwater level, the foundation pit monitoring project data, as well as the progress of the foundation pit excavation, and the start and end time.

[0125] Among them, the excavation depth of the foundation pit represents the target bottom of the foundation pit, the site soil type of the foundation pit represents the number of stratum types of the foundation pit, the excavation progress, start and end time of the foundation pit represent the construction progress of the foundation pit, the groundwater level burial depth of the foundation pit represents the groundwater height, and the foundation pit monitoring project data includes project monitoring data and follow-up monitoring data.

[0126] Case 1 is a foundation pit project in a certain county. The excavation depth of the foundation pit is 7.5m, the groundwater level is 9.2m deep, and the foundation pit monitoring time is from July 5, 2018 to July 5, 2019, which lasts for 365 days. A total of 305 sets of data were collected. The foundation pit excavation time is from July 5, 2018 to January 1, 2019, which lasts for 180 days.

[0127] like Figure 6 As shown in Figure (a), there are four engineering geological layers distributed from top to bottom in the foundation pit site, namely: plain fill layer, pebble layer, strongly weathered granite, and moderately weathered granite. The foundation pit excavation ends at the strongly weathered granite, and the groundwater is located below the bottom of the foundation pit.

[0128] Furthermore, the type of the foundation pit is determined according to the number of types of strata in the foundation pit, the positional relationship between the groundwater height and the target bottom of the foundation pit, and the progress of the foundation pit construction.

[0129] Based on the above-mentioned foundation pit related parameters, it can be determined that the foundation pit belongs to the third type of foundation pit, that is, the geological environment is complex but not affected by groundwater.

[0130] In this application, the horizontal displacement data of the retaining piles at a certain monitoring point and the vertical displacement data of the retaining piles at the monitoring point are used as the input of the foundation pit monitoring neural network, and the surface settlement data of another monitoring point are selected as the output of the foundation pit monitoring neural network. Among them, since the monitoring data of most monitoring projects are time series data, the foundation pit monitoring neural network of this application uses the LSTM neural network (Long Short Term Memory) model during verification.

[0131] On this basis, a data set is constructed based on the project monitoring data and subsequent monitoring data of the foundation pit. Specifically, the project monitoring data generated during the first two strata of the foundation pit are used as the training set, and the monitoring data of the third stratum is not trained, and the subsequent monitoring data is used as the test set. According to the reference data, the monitoring data of the first 150 days is used as the training set, and the group data of the last 180 days is used as the test set, that is, the training set ratio is 45% and the test set ratio is 55%.

[0132] The constructed training set is used to train the LSTM model, and the test set is used to evaluate the LSTM model. The output prediction results are as follows: Figure 7 At the same time, 75% and 80% were selected as the training set ratio in the randomly divided data set, and the data set was constructed based on the project monitoring data and subsequent monitoring data, and the LSTM model was trained and predicted. The prediction results were compared with the 45% training set ratio determined in this application based on the mean square error (MSE), mean absolute error (MAE), determination coefficient (R2), etc. to evaluate the model accuracy. Figure 8 As shown in Figure (a), the closer R2 is to 1, the better the model fitting effect is; the lower the error indicators MSE and MAE are, the smaller the model error is.

[0133] Figure 8 In Figure (a), the R2 of the model trained with the training set divided at a ratio of 45% is higher than that of the models trained with other conventional division ratios, and the error indicators MSE and MAE are lower than those of the models trained with other conventional division ratios. Therefore, through comparison, it is believed that the model trained with the data set division method provided in this application has better performance.

[0134] Case 2 is a water treatment plant foundation pit project with a foundation pit excavation depth of 7.8m and a groundwater level of 5.2m. The foundation pit monitoring period was from June 20, 2019 to August 13, 2020, for 420 days, and a total of 189 sets of data were collected. The foundation pit excavation period was from June 20, 2019 to June 20, 2020, for 365 days.

[0135] like Figure 6 As shown in Figure (b), the foundation pit site has five engineering geological layers from top to bottom, namely: miscellaneous fill, silty clay, round gravel, peat soil, and silty soil. The foundation pit excavation ends at the round gravel layer, and the groundwater is above the target bottom of the foundation pit. It belongs to the fourth type of foundation pit.

[0136] In this application, the horizontal displacement data of the retaining piles at a certain monitoring point and the vertical displacement data of the retaining piles at the monitoring point are used as the input of the foundation pit monitoring neural network, and the surface settlement data of another monitoring point is selected as the output of the foundation pit monitoring neural network to train the LSTM model.

[0137] On this basis, a data set is constructed based on the project monitoring data and subsequent monitoring data of the foundation pit. Specifically, all project monitoring data of the excavation progress above the groundwater level of the foundation pit are taken as the training set, and 1 / 2 of the project monitoring data of each soil layer below the groundwater level, including subsequent monitoring data, are additionally taken as the training set, and the remaining project monitoring data are taken as the test set. According to the referenced information, the monitoring data above the groundwater level in the first 330 days are taken as the training set, 1 / 2 of the subsequent data, 45 days of project monitoring data and subsequent monitoring data are taken as the training set, and the remaining 45 days of 21 groups of data are taken as the test set, that is, the training set division ratio is 89%, and the test set ratio is 11%.

[0138] The constructed training set is used to train the LSTM model, and the test set is used to test the LSTM model. The output prediction results are as follows: Figure 7 At the same time, 50% and 70% were selected as the training set ratio in the randomly divided data set, and a data set was constructed based on the project monitoring data and subsequent monitoring data. The LSTM model was trained and predicted. The prediction results were compared with the 89% training set ratio determined in this application. The results of evaluating the model accuracy based on mean square error (MSE), mean absolute error (MAE), determination coefficient (R2), etc. are shown in the figure below. Figure 8 As shown in Figure (b).

[0139] Figure 8 In Figure (b), the R2 of the model trained with the training set divided at a ratio of 89% is higher than that of the models trained with other conventional division ratios, and the error indicators MSE and MAE are lower than those of the models trained with other conventional division ratios. Therefore, through comparison, it is believed that the model trained with the data set division method provided in this application has better performance.

[0140] Case 3 is a foundation pit project in a high-speed railway station. The excavation depth of the foundation pit is 12m, the groundwater level is 13.65m deep, the foundation pit monitoring time lasts for 100 days, a total of 170 sets of data are collected, and the foundation pit excavation time lasts for 70 days.

[0141] The foundation pit site has five engineering geological layers from top to bottom, such as Figure 6 As shown in (c), they are: miscellaneous fill, clay, silt, silt sand, and clay. The excavation of the foundation pit ends at the silt layer, and the groundwater is below the bottom of the foundation pit, which belongs to the third type of foundation pit.

[0142] In this application, the horizontal displacement data of the retaining piles at a certain monitoring point and the vertical displacement data of the retaining piles at the monitoring point are used as the input of the foundation pit monitoring neural network, and the surface settlement data of another monitoring point is selected as the output of the foundation pit monitoring neural network to train the LSTM model.

[0143] On this basis, a data set is constructed based on the project monitoring data and subsequent monitoring data of the foundation pit. Specifically, the project monitoring data of the first two strata of the foundation pit excavation are taken as the training set, the project monitoring data of the third stratum is not trained, and the subsequent monitoring data is taken as the test set. According to the referenced data, the project monitoring data of the first 60 days is taken as the training set, and the 56 groups of subsequent monitoring data of the last 30 days are taken as the test set, that is, the training set division ratio is 67%, and the test set ratio is 33%.

[0144] The constructed training set is used to train the LSTM model, and the test set is used to test the LSTM model. The output prediction results are as follows: Figure 7 At the same time, 50% and 80% were selected as the training set ratio in the randomly divided data set, and a data set was constructed based on the project monitoring data and subsequent monitoring data. The LSTM model was trained and predicted. The prediction results were compared with the 67% training set ratio determined in this application. The results of evaluating the model accuracy based on mean square error (MSE), mean absolute error (MAE), determination coefficient (R2), etc. are shown in the figure below. Figure 8 As shown in Figure (c).

[0145] Figure 8 In Figure (c), the R2 of the model trained with the training set divided at a ratio of 67% is higher than that of the models trained with other conventional division ratios, and the error indicators MSE and MAE are lower than those of the models trained with other conventional division ratios. Therefore, through comparison, it is believed that the model trained with the data set division method provided in this application has better performance.

[0146] Case 4 is a foundation pit project in a subway station. The excavation depth of the foundation pit is 28.52m, the groundwater level is 5.8m deep, and the foundation pit monitoring time is from September 13, 2017 to February 10, 2018, which lasts for 150 days. A total of 156 sets of data were collected; the foundation pit excavation time was from September 13, 2017 to December 25, 2017, which lasted for 103 days.

[0147] The foundation pit site has four engineering geological layers from top to bottom, such as Figure 6 As shown in (d) in the figure, they are: fill soil, clay, silt, pebble soil 1, dense sand, and pebble soil 2. The excavation of the foundation pit ends at the pebble soil layer, and the groundwater is located in the clay layer. It belongs to the fourth type of foundation pit.

[0148] In this application, the horizontal displacement data of the retaining piles at a certain monitoring point and the vertical displacement data of the retaining piles at the monitoring point are used as the input of the foundation pit monitoring neural network, and the surface settlement data of another monitoring point is selected as the output of the foundation pit monitoring neural network to train the LSTM model.

[0149] On this basis, a data set is constructed based on the project monitoring data and subsequent monitoring data of the foundation pit. Specifically, the project monitoring data of the excavation height of the groundwater is taken as the training set, 1 / 3 of the subsequent data is used as the training set, and the remaining data is used as the test set. According to the reference data, the data of the 60 days before monitoring is taken as the training set, and 1 / 3 of the data of the 90 days after monitoring is taken, that is, the data of the first 30 days is taken as the training set, and the 60 groups of data of the 60 days after monitoring are taken as the test set. The data set is divided into 60% for training set and 40% for test set.

[0150] The constructed training set is used to train the LSTM model, and the test set is used to test the LSTM model. The output prediction results are as follows: Figure 7 At the same time, 50% and 80% were selected as the training set ratio in the randomly divided data set, and a data set was constructed based on the project monitoring data and subsequent monitoring data. The LSTM model was trained and predicted. The prediction results were compared with the 60% training set ratio determined in this application. The results of evaluating the model accuracy based on mean square error (MSE), mean absolute error (MAE), determination coefficient (R2), etc. are shown in the figure below. Figure 8 As shown in Figure (d).

[0151] Figure 8 In Figure (d), the R2 of the model trained with the training set divided at a ratio of 60% is higher than that of the models trained with other conventional division ratios, and the error indicators MSE and MAE are lower than those of the models trained with other conventional division ratios. Therefore, through comparison, it is believed that the model trained with the data set division method provided in this application has better performance.

[0152] The following describes a training method for a foundation pit monitoring neural network model provided by the present invention. The training method for a foundation pit monitoring neural network model described below and the data set division method for foundation pit monitoring neural network prediction described above can correspond to each other.

[0153] The present application also provides a training method for a foundation pit monitoring neural network model, including: obtaining the number of stratum types of the foundation pit to be predicted, the groundwater height, the project monitoring data during the foundation pit construction, and the subsequent monitoring data after the foundation pit construction is completed;

[0154] The number of stratum types of the foundation pit to be predicted is obtained. The number of stratum types reflects whether the geological environment of the foundation pit to be predicted is a single stratum or a complex stratum. The monitoring data under different stratum environments can reflect the impact of different strata on the settlement of the foundation pit or other displacement changes.

[0155] The groundwater height at the foundation pit to be predicted is obtained, so as to determine the impact of groundwater on the foundation pit settlement or other displacement changes based on the groundwater height and the target construction depth of the foundation pit.

[0156] Obtain project monitoring data during foundation pit construction and follow-up monitoring data after foundation pit construction is completed. Use the two types of monitoring data as the specific content of the data set of the foundation pit monitoring neural network, select data from the project monitoring data and follow-up monitoring data and divide them into training sets and / or test sets to construct a data set.

[0157] The project monitoring data and the subsequent monitoring data are divided according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit, and the progress of the foundation pit construction to obtain a training set and a test set in the data set;

[0158] In order to consider the influence of strata and groundwater at the predicted foundation pit on the settlement of the foundation pit, the project monitoring data at different foundation pit construction stages and the subsequent monitoring data after the construction are completed are divided according to the number of strata types at the predicted foundation pit, the positional relationship between the groundwater height and the target bottom of the foundation pit.

[0159] The target bottom of the foundation pit is the estimated construction depth of the foundation pit. If the groundwater level is below the target bottom of the foundation pit, it is considered that the impact of groundwater on the settlement of the foundation pit is small; if the groundwater level is above the target bottom of the foundation pit, the impact of groundwater on the foundation pit also needs to be considered during the construction stage of the foundation pit.

[0160] The impact of the number of stratum types to be predicted on the foundation pit is similar to that of groundwater. It is also necessary to consider the strata at each stage of the foundation pit construction, so as to select data from the project monitoring data when the construction reaches each stratum and put it into the training set.

[0161] Therefore, according to the number of strata types and groundwater height of the foundation pit to be predicted, the geological environment of the foundation pit is divided into four types: a single stratum not affected by groundwater, a single stratum affected by groundwater, a complex stratum not affected by groundwater, and a complex stratum affected by groundwater. For each type, data are selected from the project monitoring data and subsequent monitoring data corresponding to the progress of different foundation pit construction and divided into training sets and / or test sets.

[0162] The present invention classifies the complex geological environment of the foundation pit into four categories according to the number of strata and groundwater height at the foundation pit, determines different data set division methods for each category according to the construction progress of the foundation pit, and divides the project monitoring data of different construction stages and subsequent monitoring data into training sets and / or test sets, so that the data in the training sets and test sets are more representative, thereby being able to more accurately evaluate the performance of the model on real data and improve the generalization ability of the model.

[0163] The dataset is used to train a neural network model for foundation pit monitoring.

[0164] The training set and test set in the constructed dataset are used to train and test the foundation pit monitoring neural network model.

[0165] The following is a description of a data set division device for foundation pit monitoring neural network prediction provided by the present invention. The data set division device for foundation pit monitoring neural network prediction described below and the data set division method for foundation pit monitoring neural network prediction described above can correspond to each other.

[0166] like Fig. 9 As shown, a data set partitioning device for foundation pit monitoring neural network prediction includes an acquisition module 901, a classification module 902 and a partitioning module 903:

[0167] The acquisition module 901 is used to obtain the number of stratum types of the foundation pit to be predicted, the groundwater height, the project monitoring data during the foundation pit construction, and the subsequent monitoring data after the foundation pit construction is completed;

[0168] The number of stratum types of the foundation pit to be predicted is obtained. The number of stratum types reflects whether the geological environment of the foundation pit to be predicted is a single stratum or a complex stratum. The monitoring data under different stratum environments can reflect the impact of different strata on the settlement of the foundation pit or other displacement changes.

[0169] The groundwater height at the foundation pit to be predicted is obtained, so as to determine the impact of groundwater on the foundation pit settlement or other displacement changes based on the groundwater height and the target construction depth of the foundation pit.

[0170] Obtain project monitoring data during foundation pit construction and follow-up monitoring data after foundation pit construction is completed. Use the two types of monitoring data as the specific content of the data set of the foundation pit monitoring neural network, select data from the project monitoring data and follow-up monitoring data and divide them into training sets and / or test sets to construct a data set.

[0171] It should be noted that the specific content of the project monitoring data and the subsequent monitoring data is determined according to the actual practical foundation pit monitoring neural network, and the data type required for the input of the foundation pit monitoring neural network can be selected from the project monitoring data and the subsequent monitoring data.

[0172] In a feasible implementation, a piece of data in the data set is constructed by taking the horizontal displacement and vertical displacement of the retaining piles as input data and taking the vertical displacement data of the ground surface around the foundation pit as output data.

[0173] In other feasible implementations, the displacement or deformation monitoring data of different elements in the foundation pit can be selected as input according to the input requirements of the neural network, and the deformation monitoring data of the surface in various directions around the foundation pit can be used as output.

[0174] Optionally, the present application does not limit the method for obtaining specific monitoring data in the project monitoring data and subsequent monitoring data, and the monitoring data can be data that can be obtained at the foundation pit construction site.

[0175] Taking the monitoring data of the horizontal displacement and vertical displacement of the retaining piles as an example, the horizontal displacement and vertical displacement data at any position within a preset radius around the retaining piles can be used as the displacement monitoring data of the retaining piles; multiple retaining piles in the foundation pit can be selected, and the horizontal displacement and vertical displacement data of the multiple retaining piles can be used as the horizontal displacement and vertical displacement of the retaining piles used in the data set.

[0176] Optionally, this application does not limit the time length corresponding to each piece of data in the project monitoring data and subsequent monitoring data. Taking the horizontal displacement data of the retaining piles as an example, the horizontal displacement data of the retaining piles within 12 hours can be regarded as one piece of data, or the horizontal displacement data of the retaining piles within 48 hours can be regarded as one piece of data. The specific time length division can be determined according to the actual construction period of the foundation pit.

[0177] A classification module 902 is used to classify the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the position relationship between the groundwater height and the target bottom of the foundation pit;

[0178] The number of stratum types at the foundation pit to be predicted represents the strata encountered during the foundation pit construction process, and different strata have different effects on the deformation of the foundation pit.

[0179] The groundwater height is the height of the groundwater level at the foundation pit; the target bottom of the foundation pit is the expected construction depth of the foundation pit. By comparing the positional relationship between the groundwater height and the target bottom of the foundation pit, the impact of groundwater on the foundation pit construction can be determined.

[0180] The division module 903 is used to divide the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted, so as to obtain a training set and a test set in the data set.

[0181] The foundation pits are classified according to the number of stratum types to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit, so that different data set division strategies can be adopted for each type of foundation pit, so that the divided data sets can better consider the influence of the stratum environment and hydrological environment of the foundation pit on the deformation of the foundation pit.

[0182] Specifically, if the groundwater height is below the target bottom of the foundation pit, it is considered that the impact of groundwater on the settlement of the foundation pit is small; if the groundwater height is above the target bottom of the foundation pit, the impact of groundwater on the foundation pit also needs to be considered during the construction phase of the foundation pit.

[0183] The impact of the number of stratum types to be predicted on the foundation pit is similar to that of groundwater. It is also necessary to consider the strata at each stage of the foundation pit construction, so as to select data from the project monitoring data when the construction reaches each stratum and put it into the training set.

[0184] The present invention classifies foundation pits according to the complex geological environment of the foundation pits according to the number of strata and groundwater height at the foundation pits, determines different data set division methods for each category according to the construction progress of the foundation pits, and divides the project monitoring data of different construction stages and subsequent monitoring data into training sets and / or test sets, so that the data in the training sets and test sets are more representative, thereby being able to more accurately evaluate the performance of the model on real data and improve the generalization ability of the model.

[0185] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030 and a communication bus 1040, wherein the processor 1010, the communication interface 1020 and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 may call the logic instructions in the memory 1030 to execute a data set division method for foundation pit monitoring neural network prediction, the method comprising: obtaining the number of stratum types, groundwater height, project monitoring data during the foundation pit construction process and subsequent monitoring data after the foundation pit construction is completed; dividing the project monitoring data and the subsequent monitoring data according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit and the progress of the foundation pit construction, and obtaining a training set and a test set in the data set.

[0186] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0187] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute a data set partitioning method for foundation pit monitoring neural network prediction provided by the above-mentioned methods, the method comprising: obtaining the number of stratum types, groundwater height, project monitoring data during foundation pit construction, and subsequent monitoring data after the foundation pit construction is completed; dividing the project monitoring data and the subsequent monitoring data according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit, and the progress of the foundation pit construction, to obtain a training set and a test set in the data set.

[0188] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0189] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data set partitioning method for foundation pit monitoring neural network prediction, characterized in that: include: Obtain the number of strata types, groundwater height, project monitoring data during the foundation pit construction process, and subsequent monitoring data after the foundation pit construction is completed; Classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit; Dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted to obtain a training set and a test set in the data set; The step of classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit specifically includes: When the number of stratum types of the foundation pit to be predicted is one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a first type of foundation pit; When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is not below the target bottom of the foundation pit, classifying the foundation pit to be predicted as a second type of foundation pit; When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a third type of foundation pit; When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater height is not below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a fourth type of foundation pit; In the case where the foundation pit to be predicted is a first type foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the type of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes: The project monitoring data during the foundation pit construction process is divided into the training set, and the subsequent monitoring data after the foundation pit construction is completed is divided into the test set; In the case where the foundation pit to be predicted is a second type foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the type of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes: The project monitoring data before the foundation pit is constructed to the same height below the groundwater level is divided into the training set, and the remaining project monitoring data and the subsequent monitoring data are divided into the test set; In the case where the foundation pit to be predicted is a third type foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the type of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes: The project monitoring data when the foundation pit is constructed until the stratum above the target bottom of the foundation pit is completely excavated is divided into the training set, and the subsequent monitoring data is divided into the test set; In the case where the foundation pit to be predicted is a fourth type of foundation pit, the step of dividing the project monitoring data and the subsequent monitoring data according to the type of the foundation pit to be predicted to obtain a training set and a test set in the data set specifically includes: The monitoring data of the project before the foundation pit is constructed until the bottom of the foundation pit is not lower than the groundwater level is divided into the training set; Among the project monitoring data and the subsequent monitoring data generated during the construction of each stratum below the groundwater level, a preset number of monitoring data are divided into the training set; The remaining monitoring data are divided into the test set.

2. A training method for a foundation pit monitoring neural network model, characterized in that: Using the data set partitioning method for foundation pit monitoring neural network prediction as described in claim 1, the project monitoring data during the foundation pit construction process and the subsequent monitoring data after the foundation pit construction is completed are used as training data to construct a data set, and the data set is used to train the foundation pit monitoring neural network model.

3. A data set partitioning device for foundation pit monitoring neural network prediction, characterized in that: include: An acquisition module is used to obtain the number of stratum types of the foundation pit to be predicted, the groundwater height, the project monitoring data during the foundation pit construction process, and the subsequent monitoring data after the foundation pit construction is completed; A classification module, used for classifying the foundation pit to be predicted according to the number of stratum types of the foundation pit to be predicted, the positional relationship between the groundwater height and the target bottom of the foundation pit; A division module, used for dividing the project monitoring data and the subsequent monitoring data according to the category of the foundation pit to be predicted, to obtain a training set and a test set in the data set; The classification module is further used for classifying the foundation pit to be predicted as a first type of foundation pit when the number of stratum types of the foundation pit to be predicted is one and the groundwater level is below the target bottom of the foundation pit; When the number of stratum types of the foundation pit to be predicted is one and the groundwater height is not below the target bottom of the foundation pit, classifying the foundation pit to be predicted as a second type of foundation pit; When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater level is below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a third type of foundation pit; When the number of stratum types of the foundation pit to be predicted is greater than one and the groundwater height is not below the target bottom of the foundation pit, the foundation pit to be predicted is classified as a fourth type of foundation pit; The division module is also used for, when the foundation pit to be predicted is a first type of foundation pit, dividing the project monitoring data during the foundation pit construction process into the training set, and dividing the subsequent monitoring data after the foundation pit construction is completed into the test set; The division module is also used for, when the foundation pit to be predicted is a second type of foundation pit, dividing the project monitoring data before the foundation pit is constructed to the same height below the groundwater level at the bottom of the foundation pit into the training set, and dividing the remaining project monitoring data and the subsequent monitoring data into the test set; The division module is further used for, when the foundation pit to be predicted is a third type foundation pit, dividing the project monitoring data when the foundation pit is constructed until the stratum above the target bottom of the foundation pit is completely excavated into the training set, and dividing the subsequent monitoring data into the test set; The division module is also used for, when the foundation pit to be predicted is a fourth type of foundation pit, dividing the project monitoring data before the foundation pit is constructed to a level not lower than the groundwater level at the bottom of the foundation pit into the training set; Among the project monitoring data and the subsequent monitoring data generated during the construction of each stratum below the groundwater level, a preset number of monitoring data are divided into the training set; The remaining monitoring data are divided into the test set.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements a data set partitioning method for foundation pit monitoring neural network prediction as described in claim 1.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements a data set partitioning method for foundation pit monitoring neural network prediction as described in claim 1.

Images