Foundation pit deformation monitoring method, device and equipment under density alternating iteration and medium
Through the method of alternating and density iteration, the problem of discretization of observation data in deep foundation pit deformation monitoring is solved, and the efficiency and accuracy of foundation pit deformation monitoring is improved. By training the model in small intervals in high information time periods and low information time periods The large interval training model achieves efficient and accurate prediction.
Patent Information
- Application Number
- CN202411193701.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2024-12-03
Smart Images

Figure CN119063685A_ABST
Abstract
Description
Technical Field
[0001] The present invention is applicable to the technical field of foundation pit deformation monitoring, and in particular relates to a foundation pit deformation monitoring method, device, equipment and medium under sparse and dense alternating iteration. Background Art
[0002] With the rapid development of urban construction, urban subways and super high-rise building projects are increasing day by day, and with them comes an increase in deep foundation pit projects. Deep foundation pit projects are relatively risky projects, so deformation monitoring of deep foundation pits is particularly important.
[0003] At present, deep foundation pit deformation monitoring mainly relies on instruments with high measurement accuracy such as total stations and leveling instruments to monitor foundation pit parameters. However, due to factors such as limited construction surface of the foundation pit, limited space, and the concentration of equipment and personnel, it is difficult to find suitable observation points for long-term measurement of foundation pit deformation, which makes the observation data discrete in time and space. Since the input interval of the foundation pit deformation prediction model needs to be kept fixed, the discretization or unevenness of this kind of observation data will directly affect the utilization rate of the foundation pit deformation prediction model for the observation data, and directly affect the prediction efficiency and accuracy of the foundation pit deformation, that is, the efficiency and accuracy of foundation pit deformation monitoring are low.
[0004] Therefore, how to improve the efficiency and accuracy of foundation pit deformation monitoring has become a current problem that needs to be solved. Summary of the invention
[0005] In view of this, an embodiment of the present invention provides a method, device, equipment and medium for monitoring foundation pit deformation under alternating density and sparseness iteration to solve the technical problem of low efficiency and accuracy of current foundation pit deformation monitoring.
[0006] In a first aspect, a method for monitoring foundation pit deformation under sparse and dense alternating iteration is provided, the method comprising:
[0007] Continuously observing the foundation pit parameters for a preset number of times and recording the corresponding observation time period as a target interval, dividing the target interval into split points at a first interval, wherein the first interval makes each sampling time point of the foundation pit parameter be on the split point and the number of the split points is the least;
[0008] Alternately divide the target interval into a first-class subinterval and a second-class subinterval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class subinterval;
[0009] All the segmentation points in the first type of subintervals are used as iteration points for training the first prediction model, and iteration points are selected from the segmentation points of the second type of subintervals at intervals of a preset multiple of the first interval for training the second prediction model, and when the prediction model corresponding to the latter subinterval in the adjacent subintervals is first iterated, the output of the prediction model corresponding to the previous subinterval in the adjacent subintervals after the last iteration is input;
[0010] The trained prediction model is used to obtain the predicted foundation pit parameters and complete the foundation pit deformation monitoring.
[0011] In a second aspect, a foundation pit deformation monitoring device under sparse and dense alternating iteration is provided, the device comprising:
[0012] A basic interval determination module, used for continuously observing the foundation pit parameters for a preset number of times and recording the corresponding observation time period as a target interval, dividing the segmentation points in the target interval with a first interval, wherein the first interval makes the sampling time point of each of the foundation pit parameters be on the segmentation point and the number of the segmentation points is the minimum;
[0013] An alternating sub-interval determination module, configured to alternately divide the target interval into a first-class sub-interval and a second-class sub-interval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class sub-interval;
[0014] A prediction model training module, used to use all the segmentation points in the first type of sub-interval as iteration points for training the first prediction model, select iteration points from the segmentation points of the second type of sub-interval at intervals of a preset multiple of the first interval for training the second prediction model, and input the output of the last iteration of the prediction model corresponding to the previous sub-interval in the adjacent sub-interval when the prediction model corresponding to the next sub-interval in the adjacent sub-interval is first iterated;
[0015] The foundation pit deformation monitoring module is used to obtain the predicted foundation pit parameters with the trained prediction model and complete the foundation pit deformation monitoring.
[0016] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for monitoring foundation pit deformation under alternating sparse and dense iterations as described in the first aspect is implemented.
[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for monitoring foundation pit deformation under alternating sparse and dense iterations as described in the first aspect is implemented.
[0018] Compared with the prior art, the present invention has the following beneficial effects:
[0019] During the time period when the actually observed foundation pit parameter information accounts for a relatively high proportion, the first prediction model is trained at a relatively small interval so that the first prediction model can learn useful information to the greatest extent to improve the training accuracy. At the same time, during the time period when the actually observed foundation pit parameter information accounts for a relatively low proportion, the second prediction model is trained at a relatively large interval so that the second prediction model training process can reduce the learning of useless information, thereby reducing the amount of calculation and improving the training efficiency. At the same time, by transferring the learned memory information between the first and second prediction models, it is ensured that the learned information is not lost when switching the prediction model. Different from completing the training with the high-precision first prediction model alone or the low-precision second prediction model alone, it can achieve the training effect of high training efficiency and accuracy at the same time, thereby ultimately improving the efficiency and accuracy of foundation pit deformation monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, 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 only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0021] Figure 1 It is a schematic diagram of an application environment of a foundation pit deformation monitoring method under sparse-dense alternating iteration provided by the first embodiment of the present invention;
[0022] Figure 2 It is a flow chart of a method for monitoring foundation pit deformation under sparse-dense alternating iterations provided in the first embodiment of the present invention;
[0023] Figure 3 It is a flowchart of a method for training a prediction model when the prediction model is a long short-term memory network provided by the second embodiment of the present invention;
[0024] Figure 4 It is a structural schematic diagram of a foundation pit deformation monitoring device under sparse and dense alternating iterations provided in the third embodiment of the present invention;
[0025] Figure 5 It is a structural diagram of a computer device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION
[0026] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.
[0027] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.
[0028] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
[0029] As used in the present specification and the appended claims, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to monitoring, depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is monitored" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is monitored" or "in response to monitoring [described condition or event]", depending on the context.
[0030] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.
[0031] References to "one embodiment" or "some embodiments" etc. described in the specification of the present invention mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.
[0032] The embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0033] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0034] It should be understood that the order of execution of the steps in the following embodiments does not imply a precedence of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0035] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.
[0036] A foundation pit deformation monitoring method under sparse and dense alternating iteration provided in the first embodiment of the present invention can be applied to Figure 1 In the application environment, the client communicates with the server, and the client is used to input the observed foundation pit parameter information. The client includes but is not limited to PDA, desktop computer, notebook computer, ultra-mobile personal computer (UMPC), netbook, cloud terminal device, personal digital assistant (PDA) and other terminal devices. The server can be implemented by an independent server or a server cluster composed of multiple servers.
[0037] See also Figure 2 , is a flow chart of a method for monitoring foundation pit deformation under alternating density iterations provided by the first embodiment of the present invention. The above-mentioned method for monitoring foundation pit deformation under alternating density iterations can be applied to Figure 1 The client in the client, the terminal device corresponding to the client connects to the target database through a preset application programming interface (Application Programming Interface, API). When the target data is driven to execute the corresponding task, a corresponding task log will be generated, and the above task log can be collected through the API.
[0038] The above-mentioned foundation pit deformation monitoring method under the alternating iteration of density and sparseness can also be applied to Figure 1 The server in the server, the computer device corresponding to the server is connected to the corresponding database, rule base, etc. to obtain the corresponding data in the database. The above-mentioned computer device can also be connected to the client to collect the foundation pit parameter information and control instructions sent by the client user.
[0039] like Figure 2 As shown, a method for monitoring foundation pit deformation under sparse-dense alternating iteration provided by the first embodiment of the present invention may include the following steps:
[0040] Step S101, continuously observe the foundation pit parameters for a preset number of times and record the corresponding observation time period as a target interval, divide the target interval into segmentation points with a first interval, and the first interval makes each sampling time point of the foundation pit parameters fall on the segmentation point and the number of the segmentation points is the minimum.
[0041] Due to the limited construction area at the foundation pit site, the frequent movement and gathering of equipment and personnel, it is difficult for the equipment that observes the foundation pit parameters to remain observable for a long time, which results in the observed foundation pit parameters being relatively scattered in time, and the time intervals between any two observations of the foundation pit parameters are different. This discretization characteristic makes it difficult for the prediction model based on the foundation pit parameters to estimate the future parameters of the foundation pit and complete the foundation pit deformation monitoring to fully and efficiently utilize the acquired foundation pit parameters. The reason is that the input interval of the prediction model needs to be fixed, that is, the foundation pit parameters at continuous sampling moments with a fixed time interval are required as training samples to complete the training better. In reality, the intervals between the sampling time points corresponding to adjacent foundation pit parameters are not fixed, which makes the prediction model unable to utilize all the foundation pit parameter information, affecting the efficiency and accuracy of foundation pit deformation detection.
[0042] In response to the above problems, this embodiment considers using two prediction models with different input intervals to complete the learning of the acquired foundation pit parameters, so that while fully utilizing the information of all foundation pit parameters, it can also more quickly complete the transition of the time period in which the foundation pit parameters have not been successfully acquired, thereby achieving rapid and accurate learning of the foundation pit parameter information. The corresponding prediction model training is also more efficient and accurate, which can ultimately improve the efficiency and accuracy of foundation pit deformation monitoring.
[0043] Then, in order to make full use of the foundation pit parameters collected at each sampling time point, first of all, the sampling time point corresponding to each foundation pit parameter should be as close as possible to the iteration point of the subsequent prediction model, that is, the input and output point.
[0044] To this end, this embodiment first selects the foundation pit parameters observed for a preset number of times as the training data of the prediction model, and records the observation time corresponding to the foundation pit parameters observed for a preset number of times as the target interval. The preset number of times can be determined by the implementer according to the situation, such as two hundred times, etc.; the foundation pit parameters can refer to the three-dimensional coordinate information at key structural points, such as the three-dimensional coordinate information at certain positions on the foundation pit where stress is more concentrated.
[0045] In particular, the foundation pit parameters of the preset number of times indicate the amount of data. Therefore, under the premise of a certain amount of data, in order to further improve the training efficiency, it is preferred to continuously observe the foundation pit parameters of the preset number of times and the shortest time period corresponding to the time period is used as the target interval.
[0046] After selecting the training data, since the interval lengths between adjacent sampling time points in the process of observing and acquiring the foundation pit parameters are different, as mentioned above, in order to make full use of all the foundation pit parameters as much as possible, first select the first interval to divide the target interval to determine each segmentation point. The size of the first interval should be selected so that the sampling time point corresponding to each foundation pit parameter falls on a segmentation point, and at the same time, the target interval is divided into as few segmentation points as possible. That is, after determining the observation interval between any two adjacent observation moments, or after determining the observation interval between any two adjacent sampling time points, select the greatest common divisor of all observation intervals, and use the greatest common divisor as the first interval. This ensures that the foundation pit parameters obtained from each observation are utilized to the greatest extent during the subsequent prediction model training, and minimizes the number of training iterations of the prediction model.
[0047] For example, assuming that there are five foundation pit parameters obtained by observation, and their observation times are the 1st second, 3rd second, 6th second, 7th second and 9th second respectively, then the observation interval between any two adjacent observation times, or the observation interval between any two adjacent sampling time points (i.e., interval time), has four observation intervals of 2 seconds, 3 seconds, 1 second and 2 seconds in length. There are three types of observation intervals, namely, 1 second, 2 seconds and 3 seconds in length. Since the greatest common divisor of the three observation intervals is 1 second, the length of the first interval at this time is 1 second, and the corresponding segmentation points are the 1st second, 2 seconds, 3 seconds, 4 seconds, 5 seconds, 6 seconds, 7 seconds, 8 seconds and 9 seconds, a total of 9.
[0048] Step S102, alternately dividing the target interval into a first-class sub-interval and a second-class sub-interval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class sub-interval.
[0049] As described in step S102, the foundation pit parameters have discrete characteristics in time. Therefore, among the segmentation points obtained after step S101, those segmentation points corresponding to sampling time points also have the characteristics of appearing more frequently in certain time periods and less frequently in other time periods.
[0050] To this end, this embodiment continues to alternately divide the target interval into Class I sub-intervals and Class II sub-intervals, each of which has a length of the second interval, after obtaining the segmentation point. That is, in chronological order, the first sub-interval on the target interval is a Class I sub-interval, the second sub-interval is a Class II sub-interval, the third sub-interval is a Class I sub-interval, the fourth sub-interval is a Class II sub-interval, the fifth sub-interval is a Class I sub-interval, and so on.
[0051] The reason for dividing into two types of sub-intervals is that, as described in the above step S101, two prediction models with different input intervals are prepared to learn the information of each segmentation point. Then, in order to learn as much information as possible about the segmentation points corresponding to the sampling time points to complete effective learning, and at the same time learn as little information as possible about the segmentation points corresponding to the non-sampling time points to avoid invalid learning, it is necessary to classify the segmentation points corresponding to the sampling time points into the first type of sub-intervals to the greatest extent, and classify the segmentation points corresponding to the non-sampling time points into the second type of sub-intervals to the greatest extent, and determine the value of the second interval according to the maximum number of sampling time points that can fall into the first type of sub-intervals as desired.
[0052] Preferably, in one embodiment, the method for determining the value of the second interval, or the method for determining the second interval so that the number of the sampling time points falling into the first type of subinterval is maximized, may be:
[0053] Set the corresponding weight of the first type of sub-interval to a non-zero value and the corresponding weight of the second type of sub-interval to zero, multiply each of the foundation pit parameters by the corresponding weight of the sub-interval in which it is located and record the product as a temporary value, and select the sub-interval length when the number of non-zero values in the temporary value is the largest as the second interval.
[0054] Step S103, using all the segmentation points in the one type of sub-interval as iteration points for training the first prediction model, selecting iteration points from the segmentation points of the two types of sub-intervals at intervals of a preset multiple of the first interval for training the second prediction model, and when the prediction model corresponding to the latter sub-interval in the adjacent sub-intervals is iterated for the first time, inputting the output after the last iteration of the prediction model corresponding to the previous sub-interval in the adjacent sub-intervals.
[0055] Since a class of sub-intervals contains a large number of segmentation points corresponding to sampling time points, in order to fully learn the observed foundation pit information corresponding to the sampling time points, all segmentation points in a class of sub-intervals are selected as iteration points for training the first prediction model.
[0056] The second-class sub-interval contains a large number of segmentation points corresponding to non-sampling time points. In order to avoid invalid learning, the iterative points are selected from the segmentation points of the second-class sub-interval at intervals of the first interval of the preset multiple for the training of the second prediction model. It is easy to understand that in order to successfully avoid invalid information during the learning process, that is, the prediction model training process, the iterative points selected in the second-class sub-interval should be less than the iterative points selected in the first-class sub-interval. At the same time, in order to ensure that the selected iterative points correspond to certain segmentation points, the corresponding preset multiple should be an integer greater than 1. And because the selection interval of the iterative points in the second-class sub-interval should be less than the length of the second-class sub-interval, that is, the second interval, it should also be ensured that the product of the preset multiple and the first interval is less than the second interval.
[0057] The key concept of this embodiment is, first of all, to divide the sub-intervals into one type and the sub-intervals into two types, so as to realize the training of different prediction models with different input intervals, so that one of the prediction models, namely the first prediction model, can learn the vast majority of the actually observed foundation pit parameters, ensuring the accuracy of the prediction model training, and the other prediction model, namely the second prediction model, can use a larger input interval to iterate more quickly to complete the useless iteration points, namely the time period where there are no actually observed foundation pit parameters, thereby improving the prediction model training speed to improve efficiency.
[0058] Secondly, it should be emphasized that simply subdividing the training data into two learning categories corresponding to two types of sub-intervals and training them separately with different prediction models will not actually successfully achieve the above-mentioned efficient and high-accuracy prediction model training effect. The reason is that the learning of the prediction model needs to inherit the results of previous learning or learn and remember information, so if only the first prediction model is trained and learned in discontinuous first-class sub-intervals, or the second prediction model is trained and learned in discontinuous second-class sub-intervals, the training of the two models will be discontinuous and the learned information will be lost. To this end, the present embodiment is also specially arranged that when the prediction model corresponding to the latter subinterval in the adjacent subintervals is iterated for the first time, the output of the prediction model corresponding to the previous subinterval in the adjacent subinterval after the last iteration is input, so that in the process of alternating training of the first prediction model and the second prediction model, the learned memory information is transferred between the first prediction model and the second prediction model, thereby ensuring that the learning memory learned in the iteration of the first prediction model, which can more accurately predict the foundation pit parameters, can also be successfully inherited in the extensive and rapid iterative learning process of the second prediction model, thereby improving the training efficiency of the prediction model and at the same time improving the prediction accuracy of the prediction model obtained after the training is completed in a manner in which the first and second prediction models alternately inherit the learning memory.
[0059] Step S104, obtaining predicted foundation pit parameters with the trained prediction model, and completing foundation pit deformation monitoring.
[0060] After training the first and second prediction models by observing the foundation pit parameters obtained by a preset number of times to obtain the trained first and second prediction models, for the foundation pit status at a certain moment in the future after the target interval, the predicted foundation pit parameters at that moment can be obtained with the help of the trained prediction model, and the foundation pit status at that moment can be characterized by the obtained predicted foundation pit parameters.
[0061] To this end, the number of complete sub-intervals required to reach the future moment can be determined based on the ratio of the time interval length between the future moment to be predicted and the end time of the target interval to the second interval, and the number of times and the order of use of the first and second prediction models can be selected for each complete sub-interval that must be experienced before reaching the future moment to complete the prediction iteration process on each of the complete sub-intervals. In this process, it is only necessary to ensure that the sum of the number of times the first and second prediction models are used is equal to the number of complete sub-intervals that must be experienced before reaching the future moment; thereafter, after experiencing the prediction iteration of the above number of complete sub-intervals, the first prediction model is used to complete the prediction iteration from the end of the last complete sub-interval iteration to the future moment, so as to use the first prediction model with a smaller prediction interval to ensure the prediction accuracy of the predicted foundation pit parameters at the future moment.
[0062] It is easy to understand that, in each complete sub-interval that must be experienced before reaching the future moment, the more times the second prediction model is used, or the higher the proportion of using the second prediction model, the faster the prediction speed. However, correspondingly, since the number of training iterations of the second prediction model during the training process is less than that of the first prediction model, the more times the second prediction model is used, the lower the prediction accuracy.
[0063] That is, in each complete sub-interval that must be experienced before reaching the future moment, the higher the proportion of using the second prediction model, the higher the prediction efficiency, and the higher the proportion of using the first prediction model, the higher the prediction accuracy.
[0064] Preferably, in order to ensure the prediction accuracy as much as possible, this embodiment selects to obtain the prediction parameters corresponding to the time to be predicted after the target interval, that is, the above-mentioned future time, using only the first prediction model after training as the prediction pit parameters.
[0065] After obtaining the predicted foundation pit parameters at the time to be predicted, that is, the future time to be predicted, the predicted foundation pit parameters can be input into the pre-trained deformation detection model to determine the degree of foundation pit deformation, thereby completing the foundation pit deformation monitoring. The pre-trained deformation detection model used is an existing technology, and completing the foundation pit deformation monitoring according to the degree of deformation is also an existing technology. For example, the degree of foundation pit deformation can be monitored by comparing the deformation level with a preset threshold, so it will not be repeated here.
[0066] In this embodiment, the method trains the first prediction model at a smaller interval during a time period when the actually observed foundation pit parameter information accounts for a higher proportion, so that the first prediction model can learn useful information to the greatest extent to improve the training accuracy. At the same time, the second prediction model is trained at a larger interval during a time period when the actually observed foundation pit parameter information accounts for a lower proportion, so that the second prediction model training process can reduce the learning of useless information, thereby reducing the amount of calculation and improving the training efficiency. At the same time, by transferring the learned memory information between the first and second prediction models, it is ensured that the learned information is not lost when switching the prediction model. Different from completing the training with the high-precision first prediction model alone or the low-precision second prediction model alone, it can achieve the training effect of high training efficiency and accuracy at the same time, and ultimately improve the efficiency and accuracy of foundation pit deformation monitoring.
[0067] See also Figure 3 , is a flow chart of a method for training a prediction model when the prediction model is a long short-term memory network provided in the second embodiment of the present invention, such as Figure 3 As shown, the training method may include the following steps:
[0068] Step S201, in the first class-one sub-interval, the preset memory information corresponding to the first iteration point in the first class-one sub-interval, the preset prediction parameters corresponding to the first iteration point in the first class-one sub-interval, and the foundation pit parameters corresponding to the first iteration point in the first class-one sub-interval are input into the first prediction model to obtain the memory information corresponding to the second iteration point in the first class-one sub-interval and the prediction parameters corresponding to the second iteration point in the first class-one sub-interval.
[0069] First, in order to overcome the long-term dependency problem that cannot be solved by recurrent neural networks, that is, the long-distance dependency problem, in this embodiment, the first and second prediction models are preferably long short-term memory networks.
[0070] Since the first sampling time point corresponds to the starting moment of the target interval, and the sub-interval division process follows the principle that a class of sub-intervals includes as many sampling time points as possible, the first sub-interval must be a class of sub-intervals that includes the first sampling time point; at the same time, the sampling time point of each foundation pit parameter in a class of sub-intervals is at the division point, and each division point of a class of sub-intervals is an iteration point, so the first iteration point in the first class of sub-intervals must be the sampling time point corresponding to the foundation pit parameters actually observed.
[0071] However, before the first iteration point in the first class sub-interval, there is no available historical information for comparison with the actually observed foundation pit parameters corresponding to the first iteration point in the first class sub-interval. Therefore, although there are corresponding foundation pit parameters obtained through actual observations for the first iteration point in the first class sub-interval, it is impossible to correct the relevant parameters in the first prediction model, that is, it is impossible to effectively train the first prediction model.
[0072] Then, we can only preset the memory information c1 corresponding to the first iteration point in the first class subinterval and the prediction parameter y1 corresponding to the first iteration point in the first class subinterval. Among them, the memory information c1 corresponding to the first iteration point in the first class subinterval is the historical experience information used for prediction learned by the prediction model before the assumed or estimated iteration point, and the prediction parameter y1 corresponding to the first iteration point in the first class subinterval is the prediction of the foundation pit parameters corresponding to the first iteration point in the first class subinterval based on the memory information c1.
[0073] After completing the preset of memory information c1 and prediction parameter y1, input memory information c1, prediction parameter y1 and foundation pit parameter x1 corresponding to the first iteration point in the first class subinterval into the first prediction model, and then memory information c2 corresponding to the second iteration point in the first class subinterval and prediction parameter y2 corresponding to the second iteration point in the first class subinterval can be obtained. Among them, memory information c2 is corrected based on the difference between foundation pit parameter x1 corresponding to the first iteration point in the first class subinterval and the preset prediction parameter y1 corresponding to the first iteration point in the first class subinterval on the basis of the previous iteration point, that is, the memory information c1 corresponding to the first iteration point in the first class subinterval. After obtaining memory information c2, the predicted value of foundation pit parameter corresponding to the second iteration point in the first class subinterval, that is, prediction parameter y2, can be obtained based on memory information c2.
[0074] Step S202, among all iteration points excluding the first iteration point in the first class of sub-intervals, if the current iteration point is the iteration point corresponding to the sampling time point, a loss function is constructed according to the difference between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point, a prediction model for generating the prediction parameters corresponding to the current iteration point is trained until the loss function converges to obtain a converged prediction model, and the memory information corresponding to the previous iteration point, the foundation pit parameters corresponding to the previous iteration point, and the prediction parameters corresponding to the previous iteration point are re-input into the converged prediction model to obtain the updated prediction parameters corresponding to the current iteration point and the updated memory information corresponding to the current iteration point.
[0075] For all iteration points except the first iteration point in the first class of sub-intervals, since there is at least one iteration point as a sampling time point before any of these iteration points, the prediction model has learned the foundation pit parameter information corresponding to at least one sampling time point, that is, the real parameter information has been learned at least once.
[0076] Then, after selecting any one of all the iteration points excluding the first iteration point in the first class of sub-intervals as the current iteration point, if the current iteration point is also the iteration point corresponding to the sampling time point, then the foundation pit parameter corresponding to the current iteration point is the actual observed value, and the prediction model predicts the foundation pit parameter of the current iteration point based on the previous actual observed foundation pit parameter information, that is, the prediction parameter corresponding to the current iteration point. Then, by comparing the prediction parameter corresponding to the current iteration point with the foundation pit parameter, the accuracy of the prediction model's prediction at this time can be verified, and the prediction model can be further corrected and trained.
[0077] Specifically, the mean square error between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point is calculated, and the loss function is constructed with the mean square error, which is used to adjust the parameters of the prediction model that generates the prediction parameters corresponding to the current iteration point to achieve training of the model, until the loss function converges, then the training is considered to be completed and the prediction model at this time is recorded as a converged prediction model. Among them, constructing the loss function with the mean square error between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point is only preferred in this embodiment. In other embodiments, the loss function can also be directly constructed with the absolute value of the difference between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point.
[0078] After obtaining the convergent prediction model, since it has not combined the real information of the foundation pit parameters corresponding to the current iteration point, the prediction parameters and memory information corresponding to the current iteration point may have inaccurate results. Therefore, the memory information corresponding to the previous iteration point, the foundation pit parameters corresponding to the previous iteration point and the prediction parameters corresponding to the previous iteration point are re-input into the convergent prediction model to obtain the updated prediction parameters corresponding to the current iteration point and the updated memory information corresponding to the current iteration point.
[0079] The above process can be, for example:
[0080] Assuming that the current iteration point is the second iteration point in the first class subinterval and the current iteration point is a sampling time point, then the foundation pit parameter corresponding to the current iteration point is x2, its corresponding prediction parameter is y2, and the corresponding memory information is c2.
[0081] The mean square error between the foundation pit parameter x2 and the prediction parameter y2 is calculated as the loss function loss2. Since the prediction parameter y2 corresponding to the current iteration point is predicted by the first prediction model, the loss function loss2 is used to train the first prediction model, and the first prediction model after training is recorded as the convergence prediction model. The memory information c1, foundation pit parameter x1 and prediction parameter y1 corresponding to the previous iteration point, that is, the first iteration point in the first class sub-interval, are re-input into the convergence prediction model to obtain the updated prediction parameter y'2 and memory information c'2 corresponding to the current iteration point, that is, the second iteration point in the first class sub-interval.
[0082] It should be emphasized that when the current iteration point is the first iteration point in a certain second-category sub-interval and is a sampling time point, what is trained after the loss function is constructed at this time is not the second prediction model corresponding to the second-category sub-interval to which the current iteration point belongs, but the first prediction model corresponding to the first-category sub-interval to which the previous iteration point that generates the prediction parameters corresponding to the current iteration point belongs; similarly, when the current iteration point is the first iteration point in a certain first-category sub-interval excluding the first first-category sub-interval and is a sampling time point, what is trained after the loss function is constructed at this time is the second prediction model corresponding to the second-category sub-interval to which the previous iteration point belongs.
[0083] Step S203, among all iteration points excluding the first iteration point in the first class of sub-intervals, if the current iteration point is not the iteration point corresponding to the sampling time point, the prediction parameter corresponding to the current iteration point is used as the foundation pit parameter corresponding to the current iteration point, and the foundation pit parameter corresponding to the current iteration point, the prediction parameter corresponding to the current iteration point and the memory information corresponding to the current iteration point are input into the prediction model corresponding to the current iteration point to obtain the prediction parameter corresponding to the next iteration point and the memory information corresponding to the next iteration point.
[0084] When an iteration point is not the iteration point corresponding to the sampling time point among all the iteration points except the first iteration point in the first type of subinterval, it means that there is no actual observed foundation pit parameter at the iteration point, that is, there is no real useful information that can be used for model iteration training. Therefore, the improvement of the prediction model cannot be completed in the iterative training process corresponding to the iteration point. Then, the prediction parameter corresponding to the iteration point can only be used as the corresponding foundation pit parameter to continue the iteration. The so-called continued iteration is to input the foundation pit parameter, prediction parameter and memory information corresponding to the iteration point into the prediction model corresponding to the iteration point to obtain the prediction parameter and memory information corresponding to the next iteration point, so as to complete the next iteration.
[0085] From step S201 to step S203, the complete iterative method from the first iteration point to the last iteration point is clarified, thereby clarifying the training process in the preferred case where the prediction model is a long short-term memory network.
[0086] Corresponding to the method of the above embodiment, Figure 4 The structural block diagram of the foundation pit deformation monitoring device under the alternating iteration of density provided in the third embodiment of the present invention is shown. The foundation pit deformation monitoring device under the alternating iteration of density is applied to a computer device, and the computer device is connected to the target database through a preset application program interface. When the target database is driven to execute the corresponding task, a corresponding task log will be generated, and the task log can be collected through the API. For the convenience of explanation, only the part related to the embodiment of the present invention is shown.
[0087] See also Figure 4 The foundation pit deformation monitoring device under the sparse-dense alternating iteration includes:
[0088] A basic interval determination module 31 is used to continuously observe the foundation pit parameters for a preset number of times and record the corresponding observation time period as a target interval, and divide the segmentation points in the target interval by a first interval, wherein the first interval makes each sampling time point of the foundation pit parameter be on the segmentation point and the number of the segmentation points is the minimum;
[0089] An alternating sub-interval determining module 32, configured to alternately divide the target interval into a first-class sub-interval and a second-class sub-interval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class sub-interval;
[0090] The prediction model training module 33 is used to use all the segmentation points in the first type of subinterval as iteration points for training the first prediction model, select iteration points from the segmentation points of the second type of subinterval at intervals of a preset multiple of the first interval for training the second prediction model, and input the output of the last iteration of the prediction model corresponding to the previous subinterval in the adjacent subinterval when the prediction model corresponding to the next subinterval in the adjacent subinterval is first iterated;
[0091] The foundation pit deformation monitoring module 34 is used to obtain predicted foundation pit parameters with the trained prediction model and complete the foundation pit deformation monitoring.
[0092] Optionally, the basic interval determination module 31 includes:
[0093] The target interval determination unit is used to select the shortest time period corresponding to the continuous observation of the foundation pit parameters for the preset number of times as the target interval.
[0094] Optionally, the basic interval determination module 31 further includes:
[0095] The first interval determination unit is used to determine the observation interval between any two adjacent observation moments, select the greatest common divisor of all the observation interval durations, and use the greatest common divisor as the first interval.
[0096] Optionally, the alternating sub-interval determining module 32 includes:
[0097] The second interval determination unit is used to set the corresponding weight of the first type of sub-interval to a non-zero value and the corresponding weight of the second type of sub-interval to zero, multiply each of the foundation pit parameters by the corresponding weight of the sub-interval in which it is located and record the product as a temporary value, and select the sub-interval length when the number of non-zero values in the temporary value is the largest as the second interval.
[0098] Optionally, the prediction model training module 33 includes:
[0099] The preset multiple limiting unit is used to limit the preset multiple to an integer greater than 1 and the product of the preset multiple and the first interval is smaller than the second interval.
[0100] Optionally, the prediction model training module 33 further includes:
[0101] A prediction model limiting unit, used to limit the prediction model to a long short-term memory network;
[0102] A first iteration unit is used to input the preset memory information corresponding to the first iteration point in the first class-one subinterval, the preset prediction parameters corresponding to the first iteration point in the first class-one subinterval, and the foundation pit parameters corresponding to the first iteration point in the first class-one subinterval into the first prediction model in the first prediction model to obtain the memory information corresponding to the second iteration point in the first class-one subinterval and the prediction parameters corresponding to the second iteration point in the first class-one subinterval;
[0103] A measured data iteration unit, for constructing a loss function according to the difference between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point, training a prediction model for generating the prediction parameters corresponding to the current iteration point until the loss function converges to obtain a converged prediction model, re-inputting the memory information corresponding to the previous iteration point, the foundation pit parameters corresponding to the previous iteration point, and the prediction parameters corresponding to the previous iteration point into the converged prediction model, and obtaining the updated prediction parameters corresponding to the current iteration point and the updated memory information corresponding to the current iteration point, if the current iteration point is the iteration point corresponding to the sampling time point among all the iteration points excluding the first iteration point in the first class of subintervals;
[0104] The non-measured data iteration unit is used to, among all iteration points excluding the first iteration point in the first type of sub-interval, if the current iteration point is not the iteration point corresponding to the sampling time point, use the prediction parameters corresponding to the current iteration point as the foundation pit parameters corresponding to the current iteration point, input the foundation pit parameters corresponding to the current iteration point, the prediction parameters corresponding to the current iteration point and the memory information corresponding to the current iteration point into the prediction model corresponding to the current iteration point to obtain the prediction parameters corresponding to the next iteration point and the memory information corresponding to the next iteration point.
[0105] It should be noted that the information interaction, execution process and other contents between the above modules are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.
[0106] Figure 5 This is a schematic diagram of the structure of a computer device provided in Embodiment 4 of the present invention. Figure 5 As shown, the computer device of this embodiment includes: at least one processor ( Figure 5 Only one is shown), a memory, and a computer program stored in the memory and executable on at least one processor, and when the processor executes the computer program, the steps in any of the above-mentioned embodiments of the method for monitoring foundation pit deformation under alternating sparse and dense iterations are implemented.
[0107] The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that Figure 5 This is merely an example of a computer device and does not constitute a limitation on the computer device. The computer device may include more or fewer components than shown in the figure, or a combination of certain components, or different components.
[0108] The processor may be a CPU, or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0109] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory may be the memory of a computer device, and the internal memory provides an environment for the operation of an operating system and computer-readable instructions in the readable storage medium. The readable storage medium may be a hard disk of a computer device, and in other embodiments may also be an external storage device of a computer device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on a computer device. Further, the memory may also include both an internal storage unit of a computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as program codes of computer programs, etc. The memory may also be used to temporarily store data that has been output or is to be output.
[0110] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device capable of carrying computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.
[0111] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by a computer program product. When the computer program product runs on a computer device, the computer device can implement the steps in the above-mentioned method embodiment when executing.
[0112] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0113] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0114] In the embodiments provided by the present invention, it should be understood that the disclosed devices / computer equipment and methods can be implemented in other ways. For example, the device / computer equipment embodiments described above are only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0115] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0116] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such 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, and should be included in the protection scope of the present invention.
Claims
1. A foundation pit deformation monitoring method under sparse and dense alternating iteration, It is characterized in that The method comprises: Continuously observing the foundation pit parameters for a preset number of times and recording the corresponding observation time period as a target interval, dividing the target interval into split points at a first interval, wherein the first interval makes each sampling time point of the foundation pit parameter be on the split point and the number of the split points is the least; Alternately divide the target interval into a first-class subinterval and a second-class subinterval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class subinterval; All the segmentation points in the first type of subintervals are used as iteration points for training the first prediction model, and iteration points are selected from the segmentation points of the second type of subintervals at intervals of a preset multiple of the first interval for training the second prediction model, and when the prediction model corresponding to the latter subinterval in the adjacent subintervals is first iterated, the output of the prediction model corresponding to the previous subinterval in the adjacent subintervals after the last iteration is input; The trained prediction model is used to obtain the predicted foundation pit parameters and complete the foundation pit deformation monitoring.
2. The foundation pit deformation monitoring method under sparse and dense alternating iteration according to claim 1, It is characterized in that The method of continuously observing the foundation pit parameters for a preset number of times and recording the corresponding observation time period as the target interval is: The shortest time period corresponding to the continuous observation of the foundation pit parameters for the preset number of times is selected as the target interval.
3. The foundation pit deformation monitoring method under sparse and dense alternating iteration according to claim 1, It is characterized in that The method for making the first interval so that the sampling time point of each of the foundation pit parameters is on the segmentation point and the number of the segmentation points is the minimum is: An observation interval between any two adjacent observation moments is determined, a greatest common divisor of all the observation intervals is selected, and the greatest common divisor is used as the first interval.
4. The method for monitoring foundation pit deformation under sparse and dense alternating iteration according to claim 1, It is characterized in that The method for the second interval to maximize the number of the sampling time points falling into the first type of subinterval is: Set the corresponding weight of the first type of sub-interval to a non-zero value and the corresponding weight of the second type of sub-interval to zero, multiply each of the foundation pit parameters by the corresponding weight of the sub-interval in which it is located and record the product as a temporary value, and select the sub-interval length when the number of non-zero values in the temporary value is the largest as the second interval.
5. The method for monitoring foundation pit deformation under sparse and dense alternating iteration according to claim 1, It is characterized in that The preset multiple is an integer greater than 1, and a product of the preset multiple and the first interval is smaller than the second interval.
6. The method for monitoring foundation pit deformation under sparse and dense alternating iteration according to claim 1, It is characterized in that The prediction model is a long short-term memory network, and the training method of the prediction model is: In the first class-one subinterval, the preset memory information corresponding to the first iteration point in the first class-one subinterval, the preset prediction parameters corresponding to the first iteration point in the first class-one subinterval, and the foundation pit parameters corresponding to the first iteration point in the first class-one subinterval are input into the first prediction model to obtain the memory information corresponding to the second iteration point in the first class-one subinterval and the prediction parameters corresponding to the second iteration point in the first class-one subinterval; Among all iteration points excluding the first iteration point in the first class of subintervals, if the current iteration point is the iteration point corresponding to the sampling time point, a loss function is constructed according to the difference between the foundation pit parameters corresponding to the current iteration point and the prediction parameters corresponding to the current iteration point, a prediction model for generating the prediction parameters corresponding to the current iteration point is trained until the loss function converges to obtain a converged prediction model, and memory information corresponding to the previous iteration point, the foundation pit parameters corresponding to the previous iteration point, and the prediction parameters corresponding to the previous iteration point are re-input into the converged prediction model to obtain updated prediction parameters corresponding to the current iteration point and updated memory information corresponding to the current iteration point; Among all iteration points excluding the first iteration point in the first class of subintervals, if the current iteration point is not the iteration point corresponding to the sampling time point, the prediction parameter corresponding to the current iteration point is used as the foundation pit parameter corresponding to the current iteration point, and the foundation pit parameter corresponding to the current iteration point, the prediction parameter corresponding to the current iteration point and the memory information corresponding to the current iteration point are input into the prediction model corresponding to the current iteration point to obtain the prediction parameter corresponding to the next iteration point and the memory information corresponding to the next iteration point.
7. The method for monitoring foundation pit deformation under sparse and dense alternating iteration according to claim 1, It is characterized in that The method for obtaining the predicted foundation pit parameters by using the trained prediction model is as follows: The prediction parameters corresponding to the time to be predicted after the target interval are obtained by using the first prediction model after training as the prediction pit parameters corresponding to the time to be predicted.
8. A foundation pit deformation monitoring device under sparse and dense alternating iteration, It is characterized in that The device comprises: A basic interval determination module, used for continuously observing the foundation pit parameters for a preset number of times and recording the corresponding observation time period as a target interval, dividing the segmentation points in the target interval with a first interval, wherein the first interval makes the sampling time point of each of the foundation pit parameters be on the segmentation point and the number of the segmentation points is the minimum; An alternating sub-interval determination module, configured to alternately divide the target interval into a first-class sub-interval and a second-class sub-interval of a second interval length, wherein the second interval maximizes the number of sampling time points falling into the first-class sub-interval; A prediction model training module, used to use all the segmentation points in the first type of sub-interval as iteration points for training the first prediction model, select iteration points from the segmentation points of the second type of sub-interval at intervals of a preset multiple of the first interval for training the second prediction model, and input the output of the last iteration of the prediction model corresponding to the previous sub-interval in the adjacent sub-interval when the prediction model corresponding to the next sub-interval in the adjacent sub-interval is first iterated; The foundation pit deformation monitoring module is used to obtain the predicted foundation pit parameters with the trained prediction model and complete the foundation pit deformation monitoring.
9. A computer device, It is characterized in that The computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for monitoring foundation pit deformation under alternating density and sparseness iterations as described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, the method for monitoring foundation pit deformation under alternating sparse and dense iterations as described in any one of claims 1 to 7 is implemented.
