Intelligent risk early warning method and system based on multi-source data fusion

Abstract

The invention discloses an intelligent risk early warning method and system based on multi-source data fusion, and belongs to the technical field of construction risk assessment and early warning, and the method comprises the steps: obtaining multi-source information, and obtaining an early warning control quantity based on the multi-source information and a preset quantification processing mode; constructing an initial early warning cloud picture based on the early warning control quantity and a preset grading early warning standard; obtaining a three-dimensional early warning cloud picture based on the initial early warning cloud picture, the two-dimensional matrix distribution and a preset two-dimensional normal cloud model; optimizing and correcting the three-dimensional early warning cloud atlas by adopting a posterior probability support vector method to obtain an improved prediction model; and obtaining a risk early warning result based on the improved evidence fusion method and the improved prediction model. According to the intelligent risk early warning method and system based on multi-source data fusion provided by the invention, a corrected three-dimensional early warning model is constructed by combining historical data on the basis of multi-source data fusion and adopting a posterior probability support vector method and an improved evidence fusion method, and the accuracy, effectiveness and applicability of risk early warning are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of construction risk assessment and early warning, and in particular to an intelligent risk early warning method and system based on multi-source data fusion. Background Art

[0002] During subway foundation pit construction, many potential risks exist, as foundation pit projects typically involve deep excavation and complex support structures. Therefore, foundation pit construction risk warning is a crucial component of risk management, providing timely reminders to decision-makers to take appropriate safety measures, thereby effectively preventing accidents such as foundation pit collapse and support structure failure. Early warning mechanisms are the foundation of risk warning research, and early warning standards are a crucial basis for determining the safety status of foundation pit projects. By establishing reasonable early warning standards, key parameters such as pit deformation and displacement can be monitored in real time. Once these parameters approach or reach the early warning standards, timely alarms can be issued to ensure the safety of foundation pit construction. In engineering practice, the most common and widely used form of standard construction is based on national standards, local standards, or empirical knowledge. However, due to their high degree of subjectivity and regional specificity, early warning judgments are often affected, resulting in results that are inconsistent with reality.

[0003] Existing risk early warning methods for subway foundation pit construction suffer from the following challenges: Warning standards lack uniformity and adaptability, resulting in different warning levels for the same monitoring data across different projects. Early warning methods based on a single monitoring item fail to fully reflect the complex and ever-changing conditions at the construction site, easily leading to misjudgments. Traditional DS evidence theory can easily produce contradictory results when dealing with highly conflicting evidence. Risk early warning methods based on a single monitoring method, a single monitoring item, or a single monitoring point are common, but they often produce false alarms, missed alarms, and other anomalies in engineering practice, resulting in significant resource waste and potential safety hazards. These issues render existing early warning methods inaccurate and unreliable, making them ineffective in ensuring the safety of subway foundation pit construction. Therefore, the application of multi-source data fusion in foundation pit construction is crucial. A subway foundation pit construction risk early warning method based on multi-source data fusion is urgently needed to address these technical issues and provide effective and accurate early warning assessment data for actual construction. Summary of the Invention

[0004] The present invention provides an intelligent risk early warning method and system based on multi-source data fusion, which can solve the technical problems that the early warning method of a single monitoring project in the existing technology cannot fully reflect the complex and changeable conditions of the construction site, which is easy to cause misjudgment, and the traditional DS evidence theory is prone to produce results contrary to the facts when processing high-conflict evidence; it realizes the construction of a revised three-dimensional early warning model based on the fusion of multi-source data and the combination of historical data using the posterior probability support vector method and the improved evidence fusion method, which significantly improves the accuracy, effectiveness and applicability of risk early warning.

[0005] The present invention provides an intelligent risk early warning method based on multi-source data fusion, comprising:

[0006] Obtain multi-source information, and obtain early warning control quantities based on the multi-source information and preset quantitative processing methods;

[0007] Construct an initial warning cloud map based on the warning control quantity and preset graded warning standards;

[0008] Obtain a three-dimensional warning cloud map based on the initial warning cloud map, the two-dimensional matrix distribution and the preset two-dimensional normal cloud model;

[0009] The posterior probability support vector method is used to optimize and modify the three-dimensional warning cloud image to obtain an improved prediction model;

[0010] Obtain risk warning results based on improved evidence fusion methods and improved prediction models.

[0011] The present invention provides an intelligent risk warning method based on multi-source data fusion, which constructs a two-dimensional initial warning cloud map based on the fusion of multi-source data to obtain relevant risk warning data after fusion, and then constructs a three-dimensional warning cloud map. It is corrected by using the posterior probability support vector method and the improved evidence fusion method in combination with historical data to solve the ambiguity and randomness problems existing in the transition area of ​​two-dimensional adjacent warning levels. The improved integrated fusion method is used to obtain the risk change trend of foundation pit construction and its warning status, improve the accuracy and effectiveness of risk warning, establish a relatively unified warning prediction standard, and enhance the adaptability and uniformity of intelligent risk warning.

[0012] Furthermore, an initial warning cloud map is constructed based on the warning control quantity and the preset graded warning standard, including: obtaining the warning risk level based on the warning control quantity and the preset graded warning standard; obtaining the cumulative change value and the change rate value based on the warning risk level, and constructing a two-dimensional matrix distribution based on the cumulative change value and the change rate value; and constructing an initial warning cloud map based on the two-dimensional matrix distribution using traditional empirical methods.

[0013] The above scheme makes each monitoring item dimensionless by defining the early warning control quantity, thereby constructing a preliminary early warning distribution prediction model that can be applied to different application situations, and uses traditional empirical methods to visualize multi-source data and their corresponding early warning risk levels to prepare initial data for the subsequent three-dimensional model construction.

[0014] Furthermore, a three-dimensional warning cloud map is obtained based on the initial warning cloud map and the preset two-dimensional normal cloud model, including: obtaining a warning state interval number set based on the initial warning cloud map; obtaining cloud digital features and warning clouds based on the warning state interval number set and the preset two-dimensional normal cloud model; and constructing a three-dimensional warning cloud map based on the cloud digital features and the warning clouds according to a preset construction method.

[0015] The above scheme realizes the description of the ambiguity and randomness of the transition area of ​​the initial warning cloud map by constructing a three-dimensional warning cloud map. Each warning cloud has its corresponding membership degree, which can well express the duality reflected in the transition of warning levels. It realizes the idea of ​​cloud model-based solution to the problem of ambiguity and randomness in the transition area of ​​adjacent warning levels.

[0016] Furthermore, the posterior probability support vector method is used to optimize and correct the three-dimensional warning cloud map to obtain an improved prediction model, including: constructing a training set based on historical risk warning data, and constructing a target separation hyperplane based on the training set according to preset classification rules to obtain a positive sample set and a negative sample set; using the posterior probability support vector method to obtain a target decision surface based on the positive sample set and the negative sample set, so as to obtain an optimized posterior model based on a preset activation function and the target decision surface; obtaining historical cumulative change values ​​and historical change rates based on historical risk warning data; obtaining historical cloud membership based on historical cumulative change values, historical change rates and three-dimensional warning cloud maps; obtaining a target sample set based on historical cumulative change values, historical change rates and historical cloud membership, so as to obtain a posterior probability distribution based on the target sample set and the optimized posterior model; optimizing and correcting the three-dimensional warning cloud map based on the posterior probability distribution and the preset posterior probability principle to obtain an improved prediction model.

[0017] The above scheme adopts a highly adaptive improved support vector method to autonomously learn early warning cases of subway foundation pit projects in different regions, and corrects the results of empirical knowledge division to obtain a more objective early warning level division, thereby solving the problem that the warning levels corresponding to each block in the graded early warning are not universal, and establishing a more unified warning level division standard.

[0018] Furthermore, a training set is constructed based on historical risk warning data, and a target separation hyperplane is constructed based on the training set according to preset classification rules to obtain a positive sample set and a negative sample set, including: constructing a training set based on historical risk warning data; obtaining a minority feature set based on the training set, and obtaining neighbor data based on the minority feature set and a preset similarity calculation method; obtaining an optimized training set based on the neighbor data and a preset synthetic minority class oversampling method; and constructing a target separation hyperplane based on the preset classification rules of the optimized training set to obtain a positive sample set and a negative sample set.

[0019] In the above scheme, the optimal separation hyperplane for machine learning is constructed through the training set, so that the samples can be correctly classified into positive and negative categories according to the classification rules, and placed on both sides of the hyperplane respectively, to solve the problems of small sample data and nonlinear samples. At the same time, the SMOTE method is used to optimize the posterior probability support vector method, so that the number of samples of each warning level in the warning case reaches a balanced level.

[0020] Furthermore, the three-dimensional warning cloud map is optimized and corrected based on the posterior probability distribution and the preset posterior probability principle to obtain an improved prediction model, including: constructing an initial prediction model based on the three-dimensional warning cloud map; obtaining a training prediction model based on the initial prediction model using a preset stratified cross-validation method and a preset network search method; optimizing and correcting the training prediction model based on the posterior probability distribution, the preset posterior probability principle and the training prediction model to obtain an improved prediction model.

[0021] In the above scheme, the three-dimensional warning cloud map is corrected and updated based on the objective records of historical cases of foundation pit risk warning during the construction process. Through the support vector posterior probability output theory and based on engineering statistical cases, the cloud membership of each area to different warning levels is calculated, and the posterior probability support vector method is used to generate the partition correction result. This solves the problem that the warning levels corresponding to each block in the graded warning are not universal, and establishes a more unified warning level division standard to provide more accurate and effective risk warning results.

[0022] Furthermore, risk warning results are obtained based on the improved evidence fusion method and the improved prediction model, including: obtaining a membership matrix set based on the improved prediction model; introducing global uncertainty based on the membership matrix set to construct an initial trust; obtaining an initial trust two-dimensional distribution based on the dual-control index and the initial trust; obtaining the first type of high-conflict evidence and the second type of high-conflict evidence based on the initial trust two-dimensional distribution; obtaining the first trust based on the traditional evidence theory and the first type of high-conflict evidence; obtaining the target trust based on the first trust, the second type of high-conflict evidence and the improved evidence fusion method, so as to obtain risk warning results based on the target trust.

[0023] The above scheme is based on the ideas of model warning and quantitative analysis method. By constructing an intelligent warning model for construction risks based on multi-source monitoring projects, an improved evidence fusion method is adopted to effectively fuse monitoring data from different information sources to obtain more accurate warning results; global uncertainty is introduced to construct the initial trust BPA to form evidence, which objectively and effectively represents uncertainty information and realizes the universality of the construction method, reduces conflicts, reduces the impact of unreliable evidence on the synthesis results, and improves the rationality and accuracy of the final prediction results, providing strong support for the risk warning research of the complex multi-attribute decision-making problem of subway foundation pit construction collapse.

[0024] Furthermore, a first degree of trust is obtained based on traditional evidence theory and the first type of high-conflict evidence, including: obtaining the degree of conflict and the degree of difference based on traditional evidence theory and the first type of high-conflict evidence; obtaining an improved conflict factor according to a preset conflict improvement method based on the degree of conflict and the degree of difference, so as to obtain the first degree of trust based on the improved conflict factor and the initial degree of trust.

[0025] In the above scheme, by selecting two representative indicators, conflict degree α and difference degree β, for joint measurement, and combining them with two-dimensional coordinate mapping to construct a new comprehensive conflict factor, the first type of conflict problem is effectively solved, the conflict factor is considered more comprehensively, and the differences between evidence bodies are effectively weakened.

[0026] Furthermore, the target trust is obtained based on the first trust, the second type of high-conflict evidence and the improved evidence fusion method, so as to obtain the risk warning result based on the target trust, including: obtaining the comprehensive conflict factor based on the first trust; obtaining dynamic high conflict and dynamic low conflict based on the preset conflict perception critical value and the comprehensive conflict factor; performing weighted averaging based on the dynamic high conflict and the preset weighted averaging rule to obtain the first revised data; obtaining the second revised data based on the dynamic low conflict and the improved evidence fusion method; obtaining the target trust based on the first revised data and the second revised data, so as to obtain the risk warning result based on the target trust.

[0027] The above scheme solves the problem of invalidation of high-conflict evidence caused by defects in fusion rules by using a hybrid fusion rule based on improved DS and weighted average. Thresholds are set for case-by-case processing, and the synthesis rules are improved by introducing weighted average and discount factors, thereby improving the rationality and accuracy of the synthesis results.

[0028] In the above scheme, deep learning prediction data from multiple sources are used as evidence for risk warning judgment. The preliminary judgment of each evidence body is obtained by constructing a two-dimensional initial trust BPA. Then, the high conflict in the evidence body itself is adjusted through the first type of conflict, and the failure defects of the fusion rule are corrected through the second type of evidence. Finally, by perceiving and integrating the final trust of each information source at the four warning levels, the final warning level of the subway foundation pit risk is obtained through the principle of maximum trust. The two types of conflict adjustments greatly improve the effective fusion of multi-source data and the accuracy of risk warning.

[0029] The present invention provides an intelligent risk warning method based on multi-source data fusion, which constructs a two-dimensional initial warning cloud map based on the fusion of multi-source data to obtain relevant risk warning data after fusion, and then constructs a three-dimensional warning cloud map. The posterior probability support vector method and the improved evidence fusion method are used in combination with historical data to correct it to solve the ambiguity and randomness problems existing in the transition area of ​​two-dimensional adjacent warning levels. Based on the two-dimensional cloud BPA theory and DS evidence theory, a foundation pit construction risk warning method based on the improved two-dimensional BPA-DS multi-source information fusion method is proposed to effectively construct the initial trust and effectively integrate high-conflict multi-source information. The foundation pit construction risk change trend and its warning status are obtained through the improved evidence fusion method, which improves the accuracy and effectiveness of the risk warning. At the same time, a relatively unified warning prediction standard is established, and the adaptability and uniformity of the intelligent risk warning are enhanced, which provides strong support for the risk warning research of the complex multi-attribute decision-making problem of subway foundation pit construction collapse.

[0030] The present invention also provides an intelligent risk warning system based on multi-source data fusion, which is used to implement the above-mentioned intelligent risk warning method based on multi-source data fusion, including: a monitoring module, a prediction module and a correction module, wherein:

[0031] A monitoring module is used to obtain multi-source information and obtain early warning control quantities based on the multi-source information and a preset quantitative processing method;

[0032] The prediction module is used to construct an initial warning cloud map based on the warning control quantity and the preset graded warning standards;

[0033] The correction module is used to optimize and correct the three-dimensional warning cloud map using the posterior probability support vector method to obtain an improved prediction model, and obtain risk warning results based on the improved evidence fusion method and the improved prediction model.

[0034] The present invention provides an intelligent risk warning system based on multi-source data fusion. First, a monitoring module is used to obtain multi-source information and perform data preprocessing. The system is combined with a prediction module to construct a two-dimensional initial warning cloud map based on the fused multi-source data to obtain the relevant risk warning data after fusion, and then a three-dimensional warning cloud map is constructed. The correction module combines historical data with the posterior probability support vector method and the improved evidence fusion method to correct it to solve the ambiguity and randomness problems existing in the transition area of ​​two-dimensional adjacent warning levels. The improved integrated fusion method is used to obtain the risk change trend of foundation pit construction and its warning status, thereby improving the accuracy and effectiveness of risk warnings, and establishing a relatively unified warning prediction standard to enhance the adaptability and uniformity of intelligent risk warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0036] Figure 1 This is a flow chart of an intelligent risk warning method based on multi-source data fusion provided by this embodiment;

[0037] Figure 2 This is a schematic diagram of the causes of foundation pit collapse and their distribution ratios provided in this embodiment;

[0038] Figure 3 This is a schematic diagram of a two-dimensional matrix partition diagram of risk classification warning based on empirical knowledge provided by this embodiment;

[0039] Figure 4 This is a schematic diagram of the process of constructing a three-dimensional warning cloud map provided by this embodiment;

[0040] Figure 5 This is a schematic diagram of a three-dimensional warning cloud map provided by this embodiment;

[0041] Figure 6 Schematic diagram of the classification hyperplane of the support vector method provided in this embodiment;

[0042] Figure 7 This is a schematic diagram of a three-dimensional warning cloud image correction process provided by this embodiment;

[0043] Figure 8 This is a schematic diagram of a new sample synthesis process based on smote provided in this embodiment;

[0044] Figure 9 This is a schematic diagram of a data enhancement process based on SMOTE provided in this embodiment;

[0045] Figure 10 This is a schematic diagram of a data change process based on an improved cloud membership-support vector model provided by this embodiment;

[0046] Figure 11 is a schematic diagram of sample membership distribution provided in this embodiment;

[0047] Figure 12 Schematic diagram of the grid search parameter optimization process when the kernel is poly provided in this embodiment;

[0048] Figure 13 Schematic diagram of the grid search parameter optimization process when the kernel is rbf provided in this embodiment;

[0049] Figure 14 Schematic diagram of the grid search parameter optimization process when the kernel is sigmoid provided in this embodiment;

[0050] Figure 15 This is a schematic diagram of the two-dimensional initial trust BPA distribution provided by this embodiment;

[0051] Figure 16 is a schematic diagram of the improved conflict factors provided by this embodiment;

[0052] Figure 17 This is a schematic diagram of a risk warning process based on an improved two-dimensional BPA-DS provided in this embodiment;

[0053] Figure 18 This is a schematic diagram of the location of a research survey area provided in this embodiment;

[0054] Figure 19 This is a schematic diagram of an intelligent risk warning model based on multi-source data fusion provided in this embodiment. DETAILED DESCRIPTION

[0055] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0056] Example 1:

[0057] See Figure 1 This embodiment provides an intelligent risk warning method based on multi-source data fusion, including:

[0058] S1. Obtain multi-source information and obtain early warning control quantities based on the multi-source information and a preset quantitative processing method;

[0059] S2. Construct an initial warning cloud map based on the warning control quantity and the preset graded warning standards;

[0060] S3, obtaining a three-dimensional warning cloud map based on the initial warning cloud map, the two-dimensional matrix distribution and the preset two-dimensional normal cloud model;

[0061] S4. Use the posterior probability support vector method to optimize and modify the three-dimensional warning cloud image to obtain an improved prediction model;

[0062] S5. Obtain risk warning results based on improved evidence fusion methods and improved prediction models.

[0063] This embodiment provides an intelligent risk warning method based on multi-source data fusion, which constructs a two-dimensional initial warning cloud map based on the fusion of multi-source data to obtain fused relevant risk warning data, and then constructs a three-dimensional warning cloud map. It is corrected by using the posterior probability support vector method and the improved evidence fusion method in combination with historical data to solve the ambiguity and randomness problems in the transition area of ​​two-dimensional adjacent warning levels, enhance its adaptability and uniformity, establish a relatively unified warning level classification standard, and improve the accuracy and effectiveness of risk warnings.

[0064] During the specific implementation process, in order to analyze and summarize the risk factors of foundation pit collapse, reveal the relationship between the frequency of occurrence of different factors and collapse, and provide an objective basis for the subsequent research on the risk warning indicator system, the embodiment of the present invention first statistically analyzed 78 cases of subway foundation pit collapse. The collected collapse case data mainly came from the China Emergency Information Website, relevant literature records, news reports and engineering projects in which the author participated. Some typical cases are shown in Table 1.1 Typical Cases of Subway Foundation Pit Collapse Accidents.

[0065] Table 1.1 Typical cases of subway foundation pit collapse accidents

[0066]

[0067]

[0068] Case studies reveal that the causes of foundation pit collapse during actual construction are multifaceted and can be broadly categorized into two main categories, as shown in Table 1.2, Classification of Risk Factors for Subway Foundation Pit Construction. The first category is subjective factors, including design, construction, and management factors. For example, excessively deep foundation pits can increase lateral earth pressure, while excessively steep slopes can reduce soil stability. Insufficient drive depth of support piles and loose connections in the support structure can lead to support failure. Lax construction quality control by supervisors can lead to quality risks in key areas such as support structures and drainage facilities. The second category is objective factors, including geological and natural factors. For example, if a foundation pit is located in a high-aquifer area, long-term water accumulation and soil saturation can significantly reduce soil shear strength. Natural disasters such as earthquakes, typhoons, and heavy rains can damage soil structures and cause support structure failure.

[0069] Table 1.2 Classification of risk factors for subway foundation pits

[0070]

[0071]

[0072] In the specific implementation process, although it is difficult to obtain data on subway foundation pit collapse cases and its integrity cannot be guaranteed, the overall analysis and summary of a large amount of data can obtain real and reliable empirical data, which can reflect the basic laws of foundation pit collapse and has certain reference value. This embodiment will sort out and analyze several historical cases, and the distribution of the causes of foundation pit collapse, that is, the proportion of risk factors, is as follows: Figure 2 As shown in the figure, statistical results show that non-standard construction operations and lax construction process management are the most frequent risk factors, accounting for 62.8% and 57.6%, respectively. Complex geological structures, impacts of construction activities, inadequate drainage measures, high aquifer areas, lax construction quality control, and non-standard support structure construction are next, accounting for 35% or more. Finally, improper support structure selection, active fault zones, natural disasters, and improper foundation pit size design all account for less than 10%. Calculating the sum of the proportions of each factor reveals that foundation pit collapse is caused by a combination of factors, often reflected in on-site monitoring information. Therefore, during construction, on-site monitoring efforts and monitoring information processing capabilities should be strengthened. By mastering current monitoring information and future development trends, a more comprehensive understanding of foundation pit status changes can be achieved, thereby minimizing the likelihood of foundation pit collapse accidents during construction. Considering that a single monitoring data set cannot fully capture risk information, and given the diverse monitoring methods used by different monitoring data sources, the data exhibit multi-source heterogeneity, commonly used data monitoring methods include instrument monitoring, unmanned aerial photography, and video recognition. This example studies and calculates subway foundation pit construction risk warnings from the perspective of multi-source information fusion. The advantages of multi-source information fusion are: by integrating monitoring data obtained through multiple means, it can more comprehensively reflect the current status and changing trends of the foundation pit, thereby improving the accuracy of the warning; at the same time, multi-source information can verify and complement each other, reducing the errors and blind spots that may exist in a single information source and enhancing the reliability of the warning. This chapter will then focus on how to construct a subway foundation pit construction risk warning indicator system and standards based on multi-source information, providing a direction and foundation for further exploration of subsequent warning method research.

[0073] Optionally, an initial warning cloud map is constructed based on the warning control quantity and preset graded warning standards, including: obtaining the warning risk level based on the warning control quantity and the preset graded warning standards; obtaining the cumulative change value and the change rate value based on the warning risk level, and constructing a two-dimensional matrix distribution based on the cumulative change value and the change rate value; and constructing an initial warning cloud map based on the two-dimensional matrix distribution using traditional empirical methods.

[0074] During implementation, to achieve uniformity in warning levels across different information sources, and considering the trend uniformity between warning results and actual monitoring values ​​(the ratio of actual monitoring values ​​to control values), this paper defines the K value as the ratio of actual monitoring values ​​to control values, using K as the warning control variable. K1 and K2 are used to represent the cumulative change value and change rate of each risk factor, respectively. Each monitoring item is dimensionless, eliminating the need to consider the specific numerical value of the control value. This allows the use of the same two-dimensional cloud model regardless of the number of monitoring items selected, avoiding the inefficiency caused by the repeated construction of different cloud models. The foundation pit construction risk warning indicators and corresponding K values ​​are shown in Table 1.3.

[0075] Table 1.3 Foundation pit construction risk warning indicators and corresponding K values

[0076]

[0077]

[0078] Control values, determined based on monitoring specifications, references, and actual engineering projects, are crucial for assessing the safety of engineering structures and their surrounding environments. In this example, 70%, 85%, and 100% of the control values ​​are used as the classification criteria based on the preset graded warning standards. Furthermore, to facilitate subsequent modeling, the right boundary value is set to 1.7. Consequently, the warning risk level is divided into four levels: 1 (blue), 2 (yellow), 3 (orange), and 4 (red), from low to high, as shown in Table 1.3. Level 1 (blue) represents a safe state, requiring continued implementation of basic safety precautions; Level 2 (yellow) represents a tracking state, requiring enhanced routine inspections and maintenance to ensure the normal operation of equipment and facilities; Level 3 (orange) represents an alarm state, requiring adequate attention and appropriate control measures; and Level 4 (red) represents a dangerous state, indicating an extremely high risk, requiring immediate action and the activation of the highest-level emergency plan. To facilitate analysis, the graded warning standards described in Table 1.4 are visualized. The ratio of the cumulative change value to the corresponding control value is defined as K1 as the horizontal coordinate, and the ratio of the change rate value to the corresponding control value is defined as K2 as the vertical coordinate. After the graded warning standard analysis and judgment, the warning results are obtained into a two-dimensional matrix distribution, such as Figure 3 As shown in the figure, when the initial warning cloud map is constructed based on the two-dimensional matrix distribution using the traditional empirical method, the two-dimensional initial warning cloud map of the two-dimensional cloud model is mapped into the two-dimensional space according to the two-dimensional matrix distribution according to the traditional empirical method, laying the foundation for the subsequent three-dimensional distribution model construction.

[0079] Table 1.4 Risk classification and early warning standards for foundation pit construction based on empirical knowledge

[0080]

[0081] Optionally, a three-dimensional warning cloud map is obtained based on the initial warning cloud map and a preset two-dimensional normal cloud model, including: obtaining a warning state interval number set based on the initial warning cloud map; obtaining cloud digital features and a warning cloud based on the warning state interval number set and a preset two-dimensional normal cloud model; and constructing a three-dimensional warning cloud map based on the cloud digital features and the warning cloud according to a preset construction method.

[0082] In the specific implementation process, when constructing a three-dimensional warning cloud map, it is necessary to solve the problem of ambiguity and randomness in the transition areas of adjacent warning levels and the problem that the warning levels corresponding to each area in the map are not universal. The specific steps are as follows:

[0083] (1) According to the actual situation, the ratio K1 and the ratio K2 are divided into several warning levels, each level corresponds to an interval number [L min , L max ];

[0084] (2) For the warning levels of the ratio K1 and the ratio K2, an interval number is taken for each, and converted into cloud digital features (Ex, En, He) through the preset two-dimensional normal cloud model formula: the cloud model can be described according to the cloud digital features, which are expected Ex, entropy En and super entropy He. Among them, expected Ex is the mean value of cloud droplets, entropy En is the deviation about expected Ex, reflecting the fuzziness of information; He is the entropy of En, reflecting the randomness of information, affecting the degree of cloud droplet aggregation. The two-dimensional cloud model can be described by two sets of cloud digital features (Ex, En x , He x ) and (Ey, En y , He y ) to describe, which can simultaneously express the randomness and fuzziness generated by the cumulative value and the rate of change dual control indicators. Assume that F is a two-dimensional random function that obeys the normal distribution, with (Ex, Ey) as the expectation, (En x ,En y ) is the standard deviation, generate a random number (x i ,y i ); x ,En y ) is the expectation, (He x , He y ) is the standard deviation, generate a random number (Px i , Py i ), the cloud drop Drop(x i ,y i , m i ) is called a two-dimensional normal cloud model, and its expression is as follows:

[0085]

[0086] In the formula, (x i ,y i ) represents the cloud droplet coordinates, m i To obtain the target cloud membership m i , we need to first calculate the two-dimensional cloud numerical characteristics (Ex, En) corresponding to the cumulative value and the rate of change of the dual control indicators. x , He x ) and (Ey, En y , He y In the construction risk early warning indicator system, the monitoring indicators used to judge the early warning level are marked as Ti (i = 1, 2, 3, ..., 8), and T1 to T8 correspond to groundwater level, pile horizontal displacement, column settlement, surface settlement, support axial force, foundation pit convergence, wall top horizontal displacement and wall top vertical displacement. Each monitoring indicator Ti corresponds to a pair of dual control indicators, and the corresponding early warning state is recorded as T i (k j ), where j = 1, 2. Each warning state can correspond to a specific double limit interval, denoted as [t ij (L), t ij (R)], the double limit interval is converted into a two-dimensional cloud model (Ex, En x , He x ) and (Ey, En y , He y ):

[0087]

[0088] In the formula, the value range of the constant He is [0, En], which is used to describe the uncertainty of the current indicator; the cloud digital characteristics of the dual control indicators of this study calculated from Table 1.4 are shown in Table 1.5, which solves the uncertainty in the transition area.

[0089] Table 1.5 Cloud digital characteristics of dual control indicators

[0090]

[0091] (3) Cloud droplets are generated according to the two-dimensional matrix distribution using a preset two-dimensional normal cloud model formula to generate a warning cloud. To solve the problem of not considering the fuzziness and randomness of the transition area in the warning level zoning based on empirical knowledge, this embodiment uses a two-dimensional cloud model to represent the warning zone, such as Figure 4 (a) shows that the x-axis is the ratio of the cumulative change value to the corresponding control value K1, the y-axis is the ratio of the change rate value to the corresponding control value K2, and the z-axis is the membership degree. Each area in the matrix corresponds to a two-dimensional cloud, which is called a warning cloud. Figure 4 As shown in (b), when all areas correspond to warning clouds, a three-dimensional warning cloud map based on the cloud model can be obtained;

[0092] (4) Repeat steps (2) and (3) to represent all areas in the two-dimensional matrix diagram as warning clouds, thereby forming a three-dimensional warning cloud map. The vertical coordinate of the cloud droplet is the membership degree. The cloud droplets between different warning clouds overlap with each other, reflecting the fuzziness and randomness of the transition area of ​​the warning partition. With the ratio K1 as the x-axis, the ratio K2 as the y-axis, and the two-dimensional cloud membership as the z-axis, the three-dimensional warning cloud map of the foundation pit construction risk is finally obtained, as shown in Figure 5 shown.

[0093] Optionally, the posterior probability support vector method is used to optimize and correct the three-dimensional warning cloud map to obtain an improved prediction model, including: constructing a training set based on historical risk warning data, and constructing a target separation hyperplane based on the training set according to preset classification rules to obtain a positive sample set and a negative sample set; using the posterior probability support vector method to obtain a target decision surface based on the positive sample set and the negative sample set, so as to obtain an optimized posterior model based on a preset activation function and the target decision surface; obtaining historical cumulative change values ​​and historical change rates based on historical risk warning data; obtaining historical cloud membership based on historical cumulative change values, historical change rates and three-dimensional warning cloud maps; obtaining a target sample set based on historical cumulative change values, historical change rates and historical cloud membership, so as to obtain a posterior probability distribution based on the target sample set and the optimized posterior model; optimizing and correcting the three-dimensional warning cloud map based on the posterior probability distribution and the preset posterior probability principle to obtain an improved prediction model.

[0094] In the specific implementation process, Support Vector Machine (SVM) is a machine learning method jointly proposed by Vapnik and Corinna Cortes, which is good at solving small sample data and nonlinear sample problems. For linear binary classification tasks, the training sample is represented as (x i ,y i ), where x i ∈R n ,y n ∈{-1, 1}, i=1, 2, ..., n. The purpose of learning is to construct an optimal separating hyperplane, aiming to correctly classify the samples into positive and negative categories according to the classification rules, and place them on both sides of the hyperplane. For the same classification task, there are often multiple different decision boundaries, such as Figure 6 As shown, the squares and triangles represent two types of samples respectively, and the dotted lines represent different decision planes, namely H1:w T x+b=-1 and H2:w T x+b=+1, used to separate the two types of samples. Figure 6 (a) represents multiple classification hyperplanes, Figure 6(b) represents the maximum margin principle. When there exists a decision plane H whose distance to H1 and to H2 is the same, the optimal decision plane H can be obtained: where w=(w1,w2,...,w n ), which represents the normal vector of plane H; b is the displacement, which represents the distance between plane H and the origin (0, 0). The distance between H1 and H2 represents the classification interval d = 2 / ||w|| 2 , solving the target decision surface is to find a w that maximizes d. Based on the above analysis, the calculation process of finding the optimal hyperplane, that is, the target decision surface, is described as follows:

[0095] The function of the optimal decision surface H is:

[0096] f(x)=w T x+b;

[0097] Training samples to the optimal decision surface H: The distance is:

[0098]

[0099] Training sample set to the optimal decision surface H: The distance is:

[0100]

[0101] Where S is the sample space;

[0102] At this time, the optimization goal is:

[0103]

[0104] make Then the optimization objective becomes:

[0105]

[0106] In order to simplify the derivation process, the above formula is transformed into the problem of maximizing the classification interval, that is, The maximum value of Therefore, we need to find a set (w, b) that satisfies all training samples:

[0107]

[0108] At this time, the optimization objective becomes:

[0109]

[0110] Then, by constructing the Lagrangian function, the above problem is further transformed into:

[0111]

[0112] Where ai is the constrained Lagrange multiplier, and all are non-negative;

[0113] Calculate the partial derivatives of w and b for the above formula:

[0114]

[0115] Then we can derive the dual expression:

[0116]

[0117] According to the duality principle, the minimization process of the original objective function J(w, b, a) can be transformed into the maximization solution of the corresponding Lagrangian dual problem, namely:

[0118]

[0119] Therefore, the above optimization problem is equivalent to solving a convex quadratic programming optimization problem, that is, there is a global optimal solution Thus b* is calculated:

[0120]

[0121] Therefore, the optimal decision plane can be described as:

[0122]

[0123] In the formula, each component ai* of the dual variable a* has a strict one-to-one mapping with the training sample. The sparsity of its multipliers makes most ai* approach zero. According to the KKT complementary relaxation condition, the samples corresponding to non-zero multipliers are support vectors, which constitute the geometric characteristics of the decision boundary.

[0124] Then, the Sigmoid activation function is used to map the hard output f(x) of the support vector to [0, 1], and the posterior probability output of the support vector can be obtained:

[0125]

[0126] Where P(y=1|f) is the category where y=1 The posterior probability of , where A and B are parameters, can be obtained by solving the minimum negative log-likelihood of the parameter set as follows;

[0127]

[0128] Where pi represents p(yi=1|xi), N+ is the number of samples with yi=1, and N- is the number of samples with yi=-1.

[0129] At the same time, in the specific application process, according to the objective record of foundation pit risk warning cases during construction, the three-dimensional warning cloud map is revised and updated. Through the support vector posterior probability output theory, based on engineering statistical cases, the cloud membership of each area to different warning levels is calculated, and the posterior probability support vector method is used to generate the partition correction result. The engineering data processing process is as follows: Figure 7 As shown, the red font highlights the posterior probability distribution, which can be summarized into the following steps:

[0130] (1) The cumulative changes, change rates and corresponding warning levels of foundation pit construction monitoring items in historical risk warning are statistically analyzed. The ratio of cumulative changes K1 and the ratio of change rates K2 are converted into membership (u1, u2, ..., um) through the cloud model (2dcloud). The membership and the corresponding warning level form a sample data {(u1, u2, ..., um), (Ln, Level)}. After training and learning with a large amount of sample data, the optimal posterior probability support vector model for foundation pit risk warning zoning correction can be obtained; (2) The membership of the regional center point {(u1, 0, ..., 0), (u2, 0, ..., 0), ..., (um, 0, ..., 0)} is sent as input to the optimal model, and the posterior probability distribution {(p11, p12, ..., p1m), (p21, p22, ..., p2m), ... (pm1, pm2, ..., pmm)} can be obtained. The posterior probability distribution is used as the probability that each area belongs to the four warning levels, reflecting the membership of each area in the warning cloud map to the four warning levels; (3) According to the maximum posterior probability principle, SVM is used for correction processing to obtain the corrected warning cloud map, and the level corresponding to each area is (Ln, Level). Among them, Ln is the partition number in the warning cloud map, n = 1-1, 1-2, ..., 3-4, 4; Level is the warning level of the area Ln after correction. In this study, blue, yellow, orange and red are used.

[0131] Optionally, a training set is constructed based on historical risk warning data, and a target separation hyperplane is constructed based on the training set according to preset classification rules to obtain positive sample sets and negative sample sets, including: constructing a training set based on historical risk warning data; obtaining a minority feature set based on the training set, and obtaining neighbor data based on the minority feature set and a preset similarity calculation method; obtaining an optimized training set based on the neighbor data and a preset synthetic minority class oversampling method; and constructing a target separation hyperplane based on the preset classification rules of the optimized training set to obtain positive sample sets and negative sample sets.

[0132] In the specific implementation process, when the machine learning method handles the classification problem, the smaller the difference between the number of positive and negative samples in the data set, the better the classification effect will be. When the proportion of negative samples in the unbalanced data set is relatively small, it is easy to fail to fully learn the information of negative samples due to the small number of negative samples during the learning process. Although the model finally trained can learn the characteristics of positive samples well, the learning effect is not good for negative samples with a small number and low probability of occurrence, which greatly reduces the prediction accuracy of negative samples and even leads to extreme cases where all the prediction results are positive samples. Therefore, this embodiment adopts the synthetic minority over-sampling technique (Synthetic Minority Over-sampling Technique, SMOTE) to achieve inter-class balance optimization through intelligent analysis of the neighborhood relationship of minority class samples, realize data enhancement of the training set, and use the SMOTE method to optimize the posterior probability support vector method so that the number of samples of each warning level in the warning case reaches a balanced level. The SMOTE method takes each minority class sample as the center point, randomly searches for the remaining samples of the minority class, forms k nearest neighbors, connects them to the center sample, and performs interpolation at the same time to finally form a new minority class sample. This method can reduce the impact of data imbalance and inaccurate data classification. The process is described as follows: (1) Assume that there are M minority class samples in total, the eigenvalue dimension is N, and its eigenvalue is expressed as (x1, x2, ..., x N ). For each sample in the minority class sample set, calculate its similarity to all other samples in the minority class sample set. Assume that the feature values ​​of the i-th sample and the j-th sample in the minority class sample set are {x i :(x i 1,x i 2,......,x i N )} and {x j :(x j 1,x j 2,......,x j N )}, the similarity between the two is dist(x i ,x j ), select the k samples closest to i, and thus identify the k nearest neighbors in the feature space. The similarity calculation formula is as follows:

[0133]

[0134] (2) Based on the imbalance of the actual samples, a sampling ratio is set. For each sample xi of the minority class, several samples xt are randomly selected from its k nearest neighbors, and they are combined with the original sample xi to generate a new sample xnew according to the following formula:

[0135] xnew =x i +rand(0,1)·(x t -x i )

[0136] If the k value is too small, the synthesis effect will be close to the effect of data generated by random oversampling; if the k value is too large, it will increase the execution cost and the probability of introducing noise. The k value is generally set to 5; Figure 8 As shown, set k=5 in the k nearest neighbors, X1 is the minority class sample point, and the five nearest neighbor sample points found are X 11 ,X 12 ,X 13 ,X 14 and X 15 . In X1 and X 11 In the interpolation between, dist is the distance between the two sample points, and the newly generated sample point x new On the connecting line, rand_dist is from X1 to X 11 The random distance between them is described by rand(0,1). The target sample, i.e. the new sample, can be described as x new =X1+rand_dist*dist. Since the more balanced the data, the better the classification effect, the best effect is achieved by synthesizing samples in a 1:1 manner. This embodiment is based on actual engineering warning cases. The ratio of the number of warning cases collected is 66:42:31:19, corresponding to the number of cases of blue warning, yellow warning, orange warning and red warning respectively. The process is summarized as follows: The engineering data obtains cloud membership through the cloud model. This step only converts the membership of the original monitoring data and does not change the number of samples. When the data is enhanced by the SMOTE method, the ratio of the number of warning cases is adjusted to 66:66:66:66, so that the number of samples of each warning level reaches a balanced state, which can better adapt to subsequent classification predictions. The support vector posterior probability output model after adding the SMOTE method. The process is as follows Figure 9 As shown in the figure, the red highlighted font indicates the change in the amount of data at each warning level before and after data enhancement.

[0137] Optionally, the three-dimensional warning cloud map is optimized and corrected based on the posterior probability distribution and the preset posterior probability principle to obtain an improved prediction model, including: constructing an initial prediction model based on the three-dimensional warning cloud map; obtaining a training prediction model based on the initial prediction model using a preset stratified cross-validation method and a preset network search method; optimizing and correcting the training prediction model based on the posterior probability distribution, the preset posterior probability principle and the training prediction model to obtain an improved prediction model.

[0138] In order to solve the problem of non-universality of warning level partitioning during the specific implementation process, the posterior probability support vector method with self-learning ability is used to optimize and improve the traditional empirical partitioning based on actual engineering case data. In order to solve the sample imbalance problem existing in the data set itself, the SMOTE method is introduced to balance the current sample categories. Therefore, this embodiment proposes an improved cloud membership-support vector model based on the two-dimensional cloud model, the posterior probability support vector method and the SMOTE method. Its architecture is as follows: Figure 10 As shown in the figure, the model can optimize and improve the warning level partitioning obtained through traditional empirical knowledge, making the risk warning standard more adaptable and universal. The specific implementation process is described as follows: (1) Collect actual engineering cases and build a data set: This paper collects a total of 218 actual engineering cases, including warning data of subway foundation pit projects in multiple regions. In order to evaluate the performance of the model on unknown data and ensure that the model does not overfit during training, the data set is often divided into training set and test set in a ratio of 70%:30%. Therefore, 60 cases are selected as the test set (15 cases for each level), which do not participate in training and are only used for evaluating the generalization ability of the model. The remaining 158 cases are used as the training set. The initial input features include two dimensions (x, y) of cumulative value and change rate, which are converted into (K1, K2) by ratio. The output features are (Ln, Level), where Ln represents the area where the partition number n is located, and the value range is 1-1, 1-2, ..., 3-4, 4. Level is the warning level of the area where Ln is located, and the values ​​are blue, yellow, orange and red. (2) Solve the uncertainty problem of the transition area of ​​the warning level: Calculate the cloud membership. From process (1), we can know that the current input feature is (K1, K2). In the dataset of this paper, there are 218 such features. Now we need to uniformly convert the input features of these 218 samples into cloud membership and calculate the membership (u) of (K1, K2) relative to all partitions Ln. n 1,u n 2,......,u n m ), where u n m represents the cloud membership of the current monitoring data relative to the partition Ln; since there are 16 clouds in the 3D warning cloud map, the value of m is 16, and the membership of each monitoring data to the 16 warning clouds is calculated. The traditional warning partition is to directly determine the warning level by locating the location of (K1, K2), but due to the membership (u n 1,u n 2,......,u n m) is no longer a fixed value, but a probability distribution, which describes the degree of belonging of (K1, K2) in the entire warning zone. This description method can flexibly represent and process fuzzy information. (3) Solve the problem of sample imbalance in the engineering data set and perform data enhancement: For the 158 samples in the training set, the number of warning cases in the four categories of blue, yellow, orange and red are 66, 42, 31 and 19 respectively. Obviously, the number of samples in each category is not balanced. According to the distribution of the neighboring samples of the minority class samples, SMOTE data enhancement is used to generate new samples for improving the training and evaluation of the cloud membership-support vector model. Assume that the sample set with a large number of blue warnings is M, and the sample sets with relatively small numbers of yellow, orange and red warnings are N1, N2 and N3 respectively. For each sample in N1, N2 and N3, the nearest neighbor point when k=5 is obtained, forming three sample sets Q1, Q2 and Q3 respectively. For each sample in Q1, Q2 and Q3, a new sample is generated by the SMOTE algorithm until the data volume of these three warning levels is 66. Therefore, the number of warning samples in the four categories of blue, yellow, orange and red is the same. (4) Model training and evaluation to obtain the optimal model: This embodiment uses stratified cross-validation and grid search technology to optimize the model parameters. The stratified 5-fold cross-validation method can increase the diversity of model training samples when the number of cases is limited. At the same time, this method can make the proportion of the target variable in each fold the same as the proportion of the entire data set, so that each fold can better represent the whole. The grid search method is used to optimize the parameters of "kernel function, gamma parameter and penalty parameter C". The optimization ranges are [linear, poly, rbf], [0.01, 10] and [0.001, 10] respectively. The search step size of gamma and C is 1. In order to obtain a better warning partition correction effect, it is stipulated that the model test accuracy cannot be lower than 95%. (5) Correction of warning level zoning to solve the problem of non-universality of warning level zoning: In the three-dimensional warning cloud map, the center point position of each cloud represents the expected value of the warning level of the area. The 16 center point positions are converted into cloud membership through the cloud model, which is expressed as {(u1, 0, ..., 0), (0, u2, ..., 0), ..., (0, 0, ..., u 16 )}, and use it as the input sample to send it to the optimal model obtained in step (4) for posterior probability estimation, and the posterior probability distribution of each region for the four warning levels can be obtained, which is expressed as {(p 1 1, p 1 2,......,p 1 m ), (p 2 1, p 2 2,......,p 2m ),......,(p 16 1, p 16 2,......,p 16 m )}, which reflects the probability distribution corresponding to each warning cloud at different levels, and then the revised warning classification results (Ln, Level) of 16 cloud areas can be obtained, thus solving the problem of incompatibility in the construction risk warning level zoning of different subway foundation pit projects.

[0139] During the specific implementation process, the engineering case data collected by this embodiment has strong authenticity and reliability. The data sets are divided into three categories, and each category of data has input and output. The first category is the training set, which is used for model training; the second category is the test set, which is used to evaluate the generalization ability of the model; the third category is the input data set for early warning partition correction. The initial input is the cumulative value x and the change rate y of each monitoring item, and the final output is the partition number Ln and the early warning level Level corresponding to the maximum membership. In order to reduce the impact of uncertainty in the transition area, it is necessary to convert the ratio K into membership, and learn the fuzzy information and random information of the transition area through model training. For the above three categories of data sets, the input features of the samples need to be converted into membership, and the output features do not need to be converted, and the early warning level description is still used. In order to clearly describe the membership, a radar chart can be used to draw the membership distribution, where the cloud membership distribution diagram of some samples in the training set is as shown below. Figure 11 As shown in (a), the membership distribution diagram of some sample clouds in the test set is as follows Figure 11 As shown in (b), the membership distribution of some sample clouds in the warning partition correction sample set is as follows Figure 11 As shown in (c), the figure shows the membership distribution of 30 groups of samples. The outer peripheral coordinates 1, 2, 3, ..., 16 represent the 16 warning zones 1 to 16. The radius coordinates 0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2 represent the size of the membership in the current zone. The values ​​of the 16 zones on the radius together constitute the membership of a certain sample. All points of the same color represent the membership (u1, u2, ..., u2) generated by a group of samples (K1, K2). 16 ), different colors represent the membership generated by different samples.

[0140] In the specific implementation process, this embodiment compares the traditional support vector and improved cloud membership-support vector models, allows C and gamma to operate within a certain range (the search range of C and gamma in this study is between 0.01 and 10, and the search step is 1), and then finds the C and gamma combination with the highest accuracy through cross-validation. Considering that different combinations of C and gamma may correspond to the highest accuracy, and that too high a penalty parameter C will cause a transitional learning state, which will in turn make the model's generalization ability worse, the C and gamma corresponding to the minimum C value are considered to be the best combination. There are 3 combinations that can achieve the highest accuracy of 82.74%, namely ①kernel=poly; C=0.001; gamma=8.001; (such as Figure 12 ②kernel=rbf;C=1.001;gamma=1.001(as shown) Figure 13 ③ kernel=sigmoid;C=1.001;gamma=1.001 (as shown) and ③ kernel=sigmoid;C=1.001;gamma=1.001 (as shown) Figure 14 As shown). Figure 12 、 13 As shown in Figure 14, the kernel functions selected include poly, rbf, and sigmoid. To clearly see the change process, a contour map is used to describe it. In the three-dimensional graph, the X-axis is C, the Y-axis is gamma, and the Z-axis is accuracy. The color change in the graph describes the change in accuracy. The closer the color is to red, the higher the accuracy, and the closer it is to blue, the lower the accuracy. Among them: Figure 12 (a) shows the improved cloud membership-support vector optimization process when the kernel is poly, where the optimal point is; C = 1.001, gamma = 2.001, accuracy = 94.94%; Figure 12 (b) The traditional support vector optimization process when the kernel is poly, where the optimal point is; C = 0.001, gamma = 8.001, accuracy = 82.74%; Figure 13 (a) shows the improved cloud membership-support vector optimization process when the kernel is rbf, where the optimal point is; C = 4.001, gamma = 5.001, accuracy = 95.84%; Figure 13 (b) shows the traditional support vector optimization process when the kernel is rbf, where the optimal point is; C = 1.001, gamma = 1.001, accuracy = 82.74%; Figure 14 (a) shows the improved cloud membership-support vector optimization process when the kernel is sigmoid, where the optimal point is; C = 2.001, gamma = 21.001, accuracy = 96.46%; Figure 14(b) shows the traditional support vector optimization process when the kernel is sigmoid, where the optimal point is; C = 1.001, gamma = 1.001, accuracy = 92.74%. The poly kernel function can map data from low-dimensional space to high-dimensional space and can better handle nonlinear problems. However, RBF and sigmoid will oscillate when searching for parameters, such as Figure 13 (b) and 13(b), the color depth alternates, indicating that as C and gamma increase, the accuracy sometimes decreases and sometimes increases, indicating that the model training process is unstable and not easy to converge. Figure 12 、 13 ,14The following conclusions are drawn: the accuracy of the kernel function poly is higher than that of rbf and sigmoid; compared with the traditional support vector model, the improved cloud membership-support vector model has higher accuracy.

[0141] In the specific implementation process, this embodiment proposes an improved cloud membership-support vector warning level partition correction model based on the risk characteristics of soft soil foundation pits and the empirical graded warning zoning method. By converting the two-dimensional warning partition into a three-dimensional warning cloud map, it not only characterizes the fuzziness and randomness of the transition area between adjacent warning levels, but also solves the problem that the traditional warning level partition is not universal. Based on the engineering case data set, cloud membership is used to express uncertain information, and the balanced category data set is obtained through the SMOTE data enhancement technology. The best model is obtained by cross-validation and grid search. Based on the maximum a posteriori probability principle, the traditional warning partition is corrected, and the actual sample distribution is compared to demonstrate the feasibility of the corrected model. It has the following technical effects: (1) A relatively practical and complete foundation pit construction risk warning mechanism is constructed, covering monitoring function, prediction function, alarm function and correction function. The cause of foundation pit construction collapse is analyzed, and the relationship between multiple factors, multiple sources of information and multiple monitoring data of collapse is discussed. A variety of warning information processing methods are analyzed, and based on the ideas of model warning and quantitative analysis method, a risk warning information processing model framework based on multiple sources of information is proposed. (2) Based on the types of underground engineering safety monitoring indicators, the construction principles and standards of the early warning indicator system are proposed, which is mainly based on strain indicators and supplemented by stress and other indicators. According to the risk characteristics of soft soil metropolitan pits and the principle of indicator selection, eight typical early warning indicators covering strain, stress and other categories are selected based on three common monitoring methods. Based on the cumulative change and change rate, the early warning indicator system for soft soil metropolitan pit construction risk is constructed. (3) Based on empirical knowledge such as specifications, literature, and engineering records, the early warning indicator level is unified through actual monitoring values ​​and control values, the right boundary value of the early warning level is set, and the four-level early warning classification standard is divided. The dual control indicators are spatially mapped, and the monitoring items of different information sources are uniformly visualized. A warning level zoning method based on empirical knowledge is further proposed. (4) Based on the two-dimensional cloud model, the two-dimensional early warning zoning is converted into a three-dimensional early warning cloud map, which expresses the fuzziness and randomness of the transition area between adjacent warning levels. In order to address the sample imbalance problem in the dataset itself, the SMOTE technology was introduced to balance the engineering case data. At the same time, based on the support vector method with autonomous learning ability, an improved cloud membership-support vector warning level zoning correction model was proposed, and an adaptive and unified soft soil subway foundation pit construction risk warning level zoning correction method was established.

[0142] Optionally, risk warning results are obtained based on an improved evidence fusion method and an improved prediction model, including: obtaining a membership matrix set based on the improved prediction model; introducing global uncertainty based on the membership matrix set to construct an initial trust; obtaining an initial trust two-dimensional distribution based on dual-control indicators and the initial trust; obtaining a first type of high-conflict evidence and a second type of high-conflict evidence based on the initial trust two-dimensional distribution; obtaining a first trust based on traditional evidence theory and the first type of high-conflict evidence; obtaining a target trust based on the first trust, the second type of high-conflict evidence and the improved evidence fusion method, so as to obtain risk warning results based on the target trust.

[0143] In its implementation, this embodiment, based on the aforementioned early warning standards, combines numerical simulation with deep learning models, and uses actual engineering projects as a backdrop to predict information from multiple monitoring sources. It then employs a multi-source information fusion approach to achieve both theoretical research and practical application of intelligent early warning for foundation pit construction risks. The application of multi-source information fusion is crucial in foundation pit engineering. To ensure safety during subway pit construction, a large number of monitoring points are typically deployed at the construction site to collect various monitoring information in real time. This embodiment processes multi-source data from multiple aspects, including type, fusion method, and application scenario, specifically including: (1) Types of multi-source information: Monitoring data: By arranging monitoring points in and around the foundation pit, data such as displacement, settlement, stress, and water level can be obtained, which can reflect the actual status of the foundation pit in real time; Geological survey data: Covering soil layer distribution, physical and mechanical properties, groundwater conditions, etc., it is the basic data for foundation pit design and construction, and is used to evaluate stratum stability, etc.; Design drawings and parameters: Including detailed design information such as the shape, size, and support form of the foundation pit, clarifying the goals and requirements of the project; Construction process data: Data such as construction progress, construction process parameters, and material properties can be used to monitor whether the construction quality and progress meet the requirements. (2) Fusion methods: Data layer fusion: directly fuse the raw data from different sensors or data sources, such as unifying the analysis of the raw signals of displacement sensors and stress sensors; Feature layer fusion: first extract features from each source data, and then fuse these features, such as extracting the trend characteristics of the monitoring data and the soil layer characteristics in the geological data for comprehensive analysis; Decision layer fusion: each data source is processed independently and a preliminary decision is made, and then these decisions are integrated, such as combining the early warning results of the monitoring data analysis with the judgment of the construction experts to decide whether to take emergency measures. (3) Application scenarios: Real-time monitoring and early warning: the integration of multiple monitoring data and geological information can more accurately judge the safety status of the foundation pit and issue early warnings in a timely manner; Optimization design: based on the integrated analysis of actual data and geological conditions during the construction process, the original design can be optimized and adjusted, such as adjusting the support structure parameters; Construction process control: by integrating construction data and monitoring data, the impact of construction on the foundation pit and the surrounding environment can be grasped in real time, and the construction process and progress can be adjusted in a timely manner; Risk assessment: integrating multi-source information can comprehensively assess the risks of foundation pit projects at different construction stages, providing a basis for risk management. Improved accuracy: Avoids the one-sidedness of single information, providing a more comprehensive and accurate understanding of foundation pit projects. Enhanced reliability: Multi-source information verifies and complements each other, improving the reliability of decision-making and judgment. Intelligent: Supports intelligent management of foundation pit projects, facilitating automated monitoring, analysis, and decision-making.

[0144] Optionally, obtaining a first degree of trust based on traditional evidence theory and a first type of high-conflict evidence includes: obtaining a degree of conflict and a degree of difference based on traditional evidence theory and a first type of high-conflict evidence; obtaining an improved conflict factor according to a preset conflict improvement method based on the degree of conflict and the degree of difference, so as to obtain a first degree of trust based on the improved conflict factor and the initial degree of trust.

[0145] In the specific implementation process, the membership matrix ui of the i-th information source is constructed from the cloud membership through the improved prediction model mentioned above as follows:

[0146]

[0147] In the formula, n = 3, indicating that there are three information sources for this foundation pit project; m = 4, indicating that the early warning standard has four risk warning levels, with A1, A2, A3, and A4 representing blue warning, yellow warning, orange warning, and red warning, respectively. uij is the cloud membership of the i-th information source at the j-th warning level, where i = 1, 2, 3, and j = 1, 2, 3, 4. The cloud membership is further defined as follows:

[0148] θ i =1-max(u ij )

[0149] m i (Θ)=θ i

[0150]

[0151] Where θi is the degree of uncertainty that the i-th information source belongs to the j-th warning level; mi(Θ) is the global uncertainty, and mi(Aj) is the initial trust level BPA of the i-th information source at the j-th warning level. The initial trust level BPA is constructed in two dimensions based on the dual control indicators. The two-dimensional BPA matrix is ​​constructed as shown in the following formula:

[0152]

[0153] And map it to the two-dimensional cloud warning, such as Figure 15 As shown in the figure, the sum of the values ​​of all areas is 1, and m(Ajs) represents the BPA score of the sth area of ​​a certain information source in the jth level warning. Then the BPA of a certain information source in the jth level warning is equal to the sum of all BPA components under this level of warning.

[0154] When adjusting the first type of high-conflict evidence, in the traditional DS evidence theory, the conflict factor only considers the conflict degree as a measurement indicator. At present, scholars have conducted a lot of research on conflict factors, such as conflict coefficient, Pignistic probability distance and Josselme distance. These indicators describe the conflict caused by the evidence body itself from different angles. However, a single indicator can only reflect one side of the conflict. In order to more comprehensively characterize its contradiction, two representative indicators, conflict degree α and difference degree β, are selected for joint measurement. A new comprehensive conflict factor is constructed through two-dimensional coordinate mapping, thereby effectively solving the first type of conflict problem. Specifically including: (1) Calculating the conflict degree α: Based on the traditional DS evidence theory, the conflict degree is expressed by the conflict coefficient K, which describes the overall conflict degree between the evidence bodies. The smaller the conflict degree, the smaller the conflict. In the construction risk warning, for the evidence body composed of different information sources, let αij be the conflict degree between the i-th information source and the j-th information source, expressed as: α ij =∑ i≠ j m i (A)m j (A), where mi(A) and mj(A) represent the initial trust BPA corresponding to the i-th information source and the j-th information source when the warning level is A; (2) Calculate the difference β: The difference can be quantified by the distance in the vector space, which can effectively reflect the similarity of different evidence. The smaller the distance, the greater the similarity of the evidence and the smaller the conflict. This paper adopts the Euclidean distance representation, and let βij be the difference between the i-th information source and the j-th information source, expressed as follows: (3) Improved conflict factor: According to the calculation method of conflict degree α and difference degree β, the two have the meaning of vector and distance in mathematics, and both show the same monotonic change trend. The smaller the value, the smaller the contradiction between the evidence. For further optimization, the properties of two-dimensional space are introduced, and the conflict degree α is mapped to the x-axis to reflect the overall conflict intensity between the evidence bodies; the difference degree β is mapped to the y-axis to reflect the local difference between the evidence bodies; the comprehensive state of the evidence conflict is intuitively represented by the point (α, β) in the coordinate system, that is, the distance from (α, β) to (0, 0) is expressed as the comprehensive conflict factor. The improved conflict factor is as follows Figure 16 As shown. Let δ represent the credibility of the i-th information source, which is used to adjust the initial trust of the evidence body. The calculation formula is as follows: (4) First-type conflict adjustment: Applying the improved conflict factor to the initial trust level BPA corresponding to the corresponding information source can obtain the trust level after first-type conflict adjustment. The calculation formula is as follows:

[0155] m adjust,i (A)=(1-δ)·m i (A)

[0156] madjust,j (A)=(1-δ)·m j (A)

[0157] where madjust,i(A) and madjust,i(A) represent the temporary trust of the i-th and j-th information sources on the warning level A, respectively.

[0158] Optionally, the target trust is obtained based on the first trust, the second type of high-conflict evidence and the improved evidence fusion method, so as to obtain a risk warning result based on the target trust, including: obtaining a comprehensive conflict factor based on the first trust; obtaining dynamic high conflict and dynamic low conflict based on a preset conflict perception critical value and the comprehensive conflict factor; performing weighted averaging based on dynamic high conflict and a preset weighted averaging rule to obtain first revised data; obtaining second revised data based on dynamic low conflict and the improved evidence fusion method; obtaining the target trust based on the first revised data and the second revised data, so as to obtain a risk warning result based on the target trust.

[0159] In practice, the first type of high-conflict evidence adjustment takes more comprehensive account of the conflict factor, effectively reducing the differences between evidence bodies. However, it does not address the issue of high-conflict evidence failing due to flaws in the fusion rule. Traditional DS fusion rules can easily produce inconsistent results when faced with high-conflict evidence. This is because when synthesizing conflicting evidence, the conflicting component is distributed proportionally to each evidence body, which can lead to unjustified increases or decreases in the trustworthiness of some evidence bodies. The conflict factor should be less than 1. When it is very close to 1, the traditional DS fusion rule will produce the opposite result. To address the failure of the traditional DS fusion rule in high-conflict situations, this section proposes a hybrid fusion rule based on an improved DS and weighted averaging to address this deficiency. A threshold ζ is set as the critical point for conflict perception, dynamically distinguishing between high and low conflict. When the comprehensive conflict factor K>ζ, the situation is high conflict, and the weighted averaging rule is used, which performs a weighted average based on the similarity between the current evidence body and other evidence bodies. Otherwise, the improved DS rule is automatically switched to, which first fuses and then normalizes. The threshold value is related to the actual project. This study focuses on intelligent early warning of subway foundation pit construction risks. Its multi-source information comes from the prediction results of monitoring data by machine learning. This paper sets the value to 0.9. When δ<0.9, the improved DS rule is used, as shown in the following formula:

[0160]

[0161] Where, Based on the improved DS rule, the target trust of the i-th and j-th information sources for warning level A after the second type of conflict correction is calculated. When A = Θ, mi(Θ) ≤ 0.1 and max((mrevised(Ak)) - max(mrevised(Ak)) ≥ 0.2 must be satisfied, where k ≠ p. That is, the difference between the maximum and the second largest values ​​of the mass function must exceed 0.2. The fusion result is considered valid; otherwise, the fusion fails. When δ ≥ 0.9, the weighted average rule is adopted, as shown in the following formula:

[0162]

[0163] Where ωi is the weighted average coefficient, which is also the weight of the i-th information source. di is the similarity between the i-th information source and other information sources. Based on the weighted average rule, the target trust of the i-th and j-th information sources for the warning level A after the second type of conflict correction is calculated. Similarly, when A = Θ, mi(Θ) ≤ 0.1 and max((mrevised(Ak)) - max(mrevised(Ak)) ≥ 0.2 must be satisfied, where k ≠ p. That is, the difference between the maximum and the second largest values ​​of the mass function must exceed 0.2. Otherwise, the fusion result is considered valid.

[0164] During implementation, regarding the construction of an initial trustworthiness prediction (BPA), since BPA is a preliminary judgment of evidence and a description of uncertain information, and cloud membership has the advantage of relatively accurately expressing information uncertainty, this embodiment adopts the concept of cloud membership to construct BPA. However, when the sum of cloud memberships is not 1, it does not meet the BPA definition and cannot be used directly as a BPA. Therefore, this embodiment, based on cloud membership, introduces global uncertainty to construct an initial trustworthiness prediction (BPA) to form evidence, thereby objectively and effectively representing uncertain information while achieving the universality of the construction method. When addressing the issue of high-conflict evidence fusion failure, the conflict is divided into two categories based on the source. The first category is conflict between evidence bodies, which is caused by inconsistencies or contradictions in the evidence itself, such as significantly different trust assignments for the same event from different evidence sources. The second category is defects in the evidence fusion rules themselves and their limitations in high-conflict scenarios, such as the DS rule, which amplifies conflicts due to normalization. Regarding the first category of high conflict. Before evidence synthesis, the original evidence is preprocessed to reduce conflict. By discounting the evidence—that is, weighting it according to its reliability and accuracy—the impact of unreliable evidence on the synthesis results is reduced. Regarding the second type of high-conflict problem, traditional DS synthesis rules may produce unreasonable results when processing highly conflicting evidence. Therefore, this embodiment improves the synthesis rules by introducing weighted averaging and discount factors, thereby improving the rationality and accuracy of the synthesis results.

[0165] In the specific implementation process, based on the two-dimensional cloud BPA theory and DS evidence theory, this embodiment proposes a foundation pit construction risk warning method based on the improved two-dimensional BPA-DS (2dBPA-DS) multi-source information fusion method to effectively build initial trust and effectively integrate high-conflict multi-source information, providing strong support for the risk warning research of subway foundation pit construction collapse, a complex multi-attribute decision-making problem. The process is as follows: Figure 17 As shown in the figure, the specific steps include: based on the monitoring data obtained by three monitoring methods, namely instrument monitoring, computer vision and drone aerial photography, firstly, the spatiotemporal features of the support axial force, pile horizontal displacement, surface settlement, column settlement, groundwater level, foundation pit convergence, wall horizontal displacement and vertical displacement are extracted to obtain the corresponding SelfC-BiGRU prediction value, which is used as the input evidence of multi-source information fusion to perceive the risk status in advance and issue an alarm; after the cloud membership is converted into the initial trust degree BPA, the basic probability distribution BPA is constructed based on the dual-control indicator two-dimensional cloud generator constructed by cumulative deformation and deformation rate to obtain The initial state of the evidence is described; the new conflict coefficient is obtained by mapping the conflict degree and difference degree to the two-dimensional space, and the BPA is adjusted to the intermediate trust mass. According to the two-dimensional coordinate system projection principle and the related properties of the two-dimensional space, a new conflict coefficient is generated, thereby completing the adjustment of the first type of high-conflict evidence; by setting a threshold, the new conflict coefficient and the threshold are judged to perceive the degree of conflict. If it is greater than the threshold, it is regarded as high-conflict evidence and integrated according to the weighted average rule to avoid fusion failure; if it is less than the threshold, it is integrated according to the improved DS fusion rule; thereby completing the correction of the second type of conflict evidence, and then obtaining the final warning result according to the maximum trust principle.

[0166] This embodiment takes into account the prediction data of multiple information sources, and obtains early warning results by fusing multi-source prediction data. It not only solves the deviation caused by the early warning of a single monitoring project, but also can analyze the warning situation in advance and give an alarm prompt. The overall warning level calculation process is mainly divided into three steps: ① Collect the prediction value of multi-source information of the SelfC-BiGRU (GicAdam) model; ② Use the improved two-dimensional BPA-DS to fuse the multi-source prediction value, including two-dimensional BPA conversion and intermediate trust calculation; ③ Calculate the final trust and obtain the risk warning result of the multi-source monitoring project according to the maximum trust principle. The improved early warning method of this embodiment obtains a more comprehensive representation of multiple future change trends of construction risks by collecting prediction values ​​from multiple information sources, and uses multiple trends as multiple input evidences of multi-source information fusion, and fuses them through the improved two-dimensional BPA-DS method to obtain a comprehensive representation of the risk status, thereby realizing intelligent early warning of subway foundation pit construction risks with multi-source information fusion.

[0167] Example 2:

[0168] This embodiment provides an intelligent risk warning system based on multi-source data fusion, which is used to implement the above-mentioned intelligent risk warning method based on multi-source data fusion, including: a monitoring module, a prediction module and a correction module, wherein:

[0169] A monitoring module is used to obtain multi-source information and obtain early warning control quantities based on the multi-source information and a preset quantitative processing method;

[0170] The prediction module is used to construct an initial warning cloud map based on the warning control quantity and the preset graded warning standards;

[0171] The correction module is used to optimize and correct the three-dimensional warning cloud map using the posterior probability support vector method to obtain an improved prediction model, and obtain risk warning results based on the improved evidence fusion method and the improved prediction model.

[0172] During the specific implementation process, the monitoring module realizes its monitoring function. During the construction of the subway foundation pit, the safety status of the construction site is monitored through multiple sources such as advanced geological forecasting, on-site inspections, instrument monitoring, drone aerial photography, computer vision, etc., and monitoring results are formed. The monitoring implementation plan for the foundation pit is formulated based on geological survey data and construction plans. The main purpose is to monitor the safety risks in various links such as retaining structures, drainage, earth excavation, and support erection, and fully grasp the dynamic changes in the foundation pit construction process to guide the entire process of foundation pit construction. The prediction module realizes its prediction function and alarm function. The prediction function refers to the use of the intelligent risk warning based on multi-source data fusion provided in this embodiment during the construction of the subway foundation pit to analyze and predict the deformation of the foundation pit and the surrounding environment, including analysis of the current state and prediction of the future state. The alarm function monitors various parameters of the subway excavation in real time during construction. Once a measured value reaches or exceeds a preset warning limit, the system automatically triggers an alarm and generates a warning notification to the responsible person. Based on engineering geology and other conditions, the system defines warning standards, indicator systems, and warning limits, and uses direct comparisons and other methods to assess warning status at each construction stage. Through the implementation of this warning system, various channels, such as text messages and emails, constantly alert all parties involved in the construction process, mobilizing their emergency response efforts and reducing the likelihood of accidents. The correction module implements this correction function. During subway excavation construction, after monitoring and predictions trigger alarms, it promptly adjusts to unsafe and unbalanced conditions, analyzes the cause of the alarm, and initiates emergency response plans. If the early warning control measures are effectively implemented and the monitored values ​​gradually stabilize, the monitoring density can be appropriately reduced. If there is no convergence within a certain period of time and the values ​​continue to increase, the alert level should be upgraded until the conditions converge. The significance of the correction function is to reduce the risk and losses of accidents through effective response measures before they occur. This constitutes a complete risk early warning mechanism for subway foundation pit construction.

[0173] During the specific implementation process, when the monitoring module obtains multi-source information, the monitoring objects mainly include the support system and the surrounding environment. The monitoring items involved cover stress, strain, and other categories. Risk monitoring is mainly carried out through on-site inspections, instrument monitoring, computer vision, and drone aerial photography. Since traditional instrument monitoring methods are easily affected by factors such as the environment, personnel, and weather, the monitoring efficiency is low, data acquisition is delayed, and the safety status of the foundation pit project cannot be reflected in a timely manner. Therefore, this embodiment uses computer vision and drone aerial photography to supplement and strengthen monitoring of points with higher risks to ensure normal progress during the foundation pit construction. Among them, computer vision is simple, convenient, and highly accurate. This embodiment is used to supplement the horizontal and vertical displacements of the top of the wall that have not been measured in time due to damage to the instrument monitoring equipment. In addition, since the foundation pit is irregular and prone to construction safety accidents, drone aerial photography, which is more flexible and capable of global three-dimensional imaging, is used to strengthen monitoring of the overall convergence of the foundation pit. During drone aerial photography, the drone is equipped with optical photography equipment (DJI MAVIC PRO) to obtain high-resolution images of specific sub-areas of the foundation pit. Image processing technology is also used to obtain real-time data on the deformation of the foundation pit project. This real-time data is then used to implement dynamic monitoring and early warning of the foundation pit construction. The drone is used to perform a full-link process of image acquisition, point cloud construction, data simplification, and detection and early warning. First, based on drone image acquisition methods and point cloud modeling theory, a three-dimensional point cloud model of the foundation pit is constructed using Pix4Dmapper software. To speed up the processing of the point cloud data, mainstream algorithms such as the bounding box method, random sampling, and curvature sampling are used to simplify the foundation pit point cloud data while ensuring the accuracy of the point cloud model. Finally, based on the fitting results of the foundation pit edge point cloud data, in-depth mining is performed to obtain the foundation pit convergence deformation value, which is used to analyze the safety status of the foundation pit. Using drone imagery for foundation pit safety monitoring provides an effective way to improve project management efficiency, optimize resource allocation, reduce operating costs, and prevent safety accidents.

[0174] In the specific implementation process, since information fusion must be targeted at the same research object, each research object is composed of multiple individuals. For example, the support system refers to the enclosure structure, concrete support, steel support and other individuals. Each individual is equipped with a large number of monitoring points. The data of each monitoring point is different, and the stress and strain conditions at different locations are also different. Therefore, it is inaccurate to perform information fusion on the entire support system to judge the warning status. For another example, the surrounding environment is composed of the surface around the foundation pit. A large number of monitoring points are arranged on the surface. The settlement and water level of each monitoring point are different. The damage conditions such as collapse and cracks at different locations are also different. Therefore, it is inaccurate to perform information fusion on the entire surrounding environment to judge the warning status. On the other hand, if each monitoring point is treated as a research object for information fusion according to the layout of the monitoring points, the data volume will be extremely large and the timeliness will be very low. Therefore, the fusion measurement area division criteria in this paper are: (1) information fusion must be for the same research object; (2) according to the principles of location proximity and mergibility, the warning fusion measurement area should be reasonably planned and determined. According to the above principles, the survey area is used as the fusion object and warning object, and the constructed improved two-dimensional BPA-DS evidence fusion model is used to carry out multi-source information fusion construction risk warning. The warning survey area is used as the research object to obtain the risk warning status of the survey area at different times. On the basis of the on-site monitoring point layout map, according to the principles of location proximity and mergibility, the warning fusion survey area is re-planned and reasonably determined, and the survey area is used as the fusion object and warning object, in order to achieve efficient warning fusion and alarm indication. In the actual application process, the final warning fusion survey area is divided as follows Figure 18 As shown in the figure, the black highlights indicate the early warning survey area division results of the support system, and the gray highlights indicate the early warning survey area division results of the surrounding environment. This area includes 8 early warning survey areas, namely SS-MK1, SS-MK2, SS-MK3, SS-MK4 and SE-MK1, SE-MK2, SE-MK3, and SE-MK4.

[0175] Example 3:

[0176] This embodiment provides an intelligent risk warning platform based on multi-source data fusion, which is implemented using the above-mentioned intelligent risk warning system based on multi-source data fusion. The improved intelligent risk warning method based on multi-source data fusion provided in the above embodiment is integrated, encapsulated and packaged to form a lightweight, convenient application that can be run and installed independently, including: a prediction interface, an early warning interface and a details interface, wherein: the optimized early warning standards are written into the background program as the default setting, no additional operations are required by the operator; the numerical simulation data and measured data obtained from the test and numerical model are loaded into the system through Excel, and the optimal prediction model under the current data set is obtained by adjusting the parameters; the predicted multi-source data is input into the early warning module, and a comprehensive description of the future foundation pit risk status from multiple information sources can be obtained. The interface mainly includes the user manual, software version and developer information.

[0177] During the specific implementation process, in response to the problems of lack of uniformity in the initial trust construction and failure of high-conflict evidence fusion in the traditional DS fusion method, this embodiment proposes an intelligent early warning model for subway foundation pit construction risks based on the improved 2dBPA-DS. Based on actual engineering cases, three types of information source data, namely instrument monitoring, computer vision and drone aerial photography, are collected. The entire foundation pit is divided into several areas based on the fusion measurement area. Through historical measured data and future prediction data, and combined with the actual situation of the construction site, the accuracy and effectiveness of the improved method in this chapter are verified from both theoretical and practical aspects. Based on computer programming language and combined with the improved early warning method, an intelligent early warning information system platform for subway foundation pit construction risks is constructed. The main technical effects are as follows: (1) An improved 2dBPA-DS intelligent early warning method for foundation pit construction risks based on multi-source information fusion is proposed, which not only constructs the initial trust, but also adjusts and corrects the high-conflict failure problem; a two-dimensional cloud BPA is used to express the initial trust, and a new conflict factor is generated by introducing the conflict degree and difference degree and performing two-dimensional projection, and unreliable evidence is discounted to achieve the adjustment of the first type of conflict; a new synthesis rule is established by improving the DS fusion strategy and weighted average strategy to achieve the correction of the second type of conflict. (2) The limitations of the warning level calculation method for a single monitoring project are analyzed, and a warning level calculation method suitable for multi-source information fusion is proposed; based on the identity of the information fusion research object, with actual engineering cases as the background, the standard end and the special-shaped expansion end are re-planned according to the principles of location proximity and mergibility, and their respective warning fusion measurement areas are determined; the measurement area is used as the basic unit of data fusion and warning monitoring, realizing the efficient integration and accurate prompting of warning information. (3) Based on the actual engineering case and the historical measured data of the previous month, the accuracy of the improved early warning method was verified by comparing the single information source, the traditional DS fusion method and the improved DS fusion method; based on the predicted data of the next 5 months, the improved DS fusion method was used to obtain the risk change trend of foundation pit construction and its early warning status, and the effectiveness of the improved early warning method was verified in combination with the construction site conditions. (4) Based on the proposed optimized graded early warning standard, constitutive model and formation seepage law, construction deformation dynamic prediction model and multi-source information risk intelligent early warning model, an intelligent early warning information system platform for subway foundation pit construction risk was constructed, which integrates the main functions such as standards, prediction, and early warning. Complex calculations can be completed through simple parameter settings and mouse clicks, forming a set of lightweight, convenient, independently run and installable applications, which greatly reduces the workload of on-site operators.

[0178] During implementation, the platform's user interface primarily consists of three panels: input, output, and operation. The input panel is primarily used for parameter setting, while the operation panel controls data import, calculation execution, result export, clearing current results, and exiting the system. The prediction interface only requires setting simple model parameters and loading training data. The calculation function yields loss changes, error changes, prediction curves, and evaluation results. The export function also allows the prediction results to be exported to a local Excel file. Based on the aforementioned early warning method, a SelfC-BIGRU foundation pit construction deformation prediction program based on improved gradient optimization was developed. The prediction interface typically includes a parameter setting window, a control panel, a prediction result graphical window, an evaluation indicator window, and a prediction result window. Based on the proposed optimized hierarchical early warning standards, constitutive models and formation seepage patterns, a construction deformation prediction model, and a multi-source information fusion risk early warning model, an intelligent early warning information system for subway foundation pit construction risks was constructed. The early warning interface integrates information from three sources. By inputting the prediction interface's predicted values ​​for the next three days, the corresponding target confidence and its early warning status can be calculated using the calculation function. The early warning results can also be exported to a local Excel file using the export function. The program was developed based on the improved two-dimensional BPA-DS risk warning model proposed above. The intelligent warning interface typically includes a data input window, a result display window, and a control panel. It typically displays forecast data for the next three days, with warning results for the first, second, and third days displayed in different colors.

[0179] Example 4:

[0180] This embodiment provides an intelligent risk warning model based on multi-source data fusion, which is implemented by using an intelligent risk warning method based on multi-source data fusion in the above embodiment. Figure 19 As shown, the following steps are included:

[0181] Establish a deformation prediction model based on multi-source information, Y (n) pre =F pre (X (n) act ), where X (n) act is the actual value of the nth information source, which is obtained through the prediction model F pre Then get the predicted value Y of the nth information source (n) pre ;

[0182] Establish a risk warning model based on multi-source information prediction, Y warn =F warn (Y (1) pre ,Y (2) pre,......,Y (n) pre ), where Y (n) pre is the prediction result of the nth information source, which is obtained through the early warning model F warn Then get the final warning result Y warn .

[0183] In the process of establishing a risk warning model based on multi-source information prediction, the posterior probability support vector method and improved evidence fusion method are used to optimize and correct the model acquisition, so as to establish an emerging warning information processing method based on statistics and artificial intelligence.

[0184] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An intelligent risk warning method based on multi-source data fusion, characterized in that: include: Acquire multi-source information, and obtain an early warning control quantity based on the multi-source information and a preset quantitative processing method; Constructing an initial warning cloud map based on the warning control quantity and the preset graded warning standard; Acquire a three-dimensional warning cloud image based on the initial warning cloud image, the two-dimensional matrix distribution and the preset two-dimensional normal cloud model; Using a posterior probability support vector method to optimize and correct the three-dimensional warning cloud image to obtain an improved prediction model; Risk warning results are obtained based on the improved evidence fusion method and the improved prediction model.

2. The intelligent risk warning method based on multi-source data fusion according to claim 1, characterized in that: The constructing of the initial warning cloud map based on the warning control quantity and the preset graded warning standard includes: Obtaining a warning risk level based on the warning control amount and a preset graded warning standard; Acquire a cumulative change value and a change rate value based on the warning risk level, and construct a two-dimensional matrix distribution based on the cumulative change value and the change rate value; An initial warning cloud map is constructed based on the two-dimensional matrix distribution using traditional empirical methods.

3. The intelligent risk warning method based on multi-source data fusion according to claim 1, characterized in that: The obtaining of a three-dimensional warning cloud image based on the initial warning cloud image and a preset two-dimensional normal cloud model includes: Acquire a set of warning status intervals based on the initial warning cloud map; Obtain cloud digital features and warning clouds based on the warning state interval number set and a preset two-dimensional normal cloud model; A three-dimensional warning cloud map is constructed based on the cloud digital features and the warning cloud according to a preset construction method.

4. The intelligent risk early warning method based on multi-source data fusion according to claim 1, characterized in that: The method of optimizing and correcting the three-dimensional warning cloud image by using the posterior probability support vector method to obtain an improved prediction model includes: Constructing a training set based on historical risk warning data, and constructing a target separation hyperplane based on the training set according to preset classification rules to obtain a positive sample set and a negative sample set; A posterior probability support vector method is used to obtain a target decision surface based on the positive sample set and the negative sample set, so as to obtain an optimized posterior model based on a preset activation function and the target decision surface; Obtain historical cumulative change values ​​and historical change rates based on historical risk warning data; Obtaining a historical cloud membership based on the historical cumulative change value, the historical change rate, and the three-dimensional warning cloud map; Acquire a target sample set based on the historical cumulative change value, the historical change rate, and the historical cloud membership, and acquire a posterior probability distribution based on the target sample set and the optimized posterior model; Based on the posterior probability distribution and the preset posterior probability principle, the three-dimensional warning cloud map is optimized and corrected to obtain an improved prediction model.

5. The intelligent risk warning method based on multi-source data fusion according to claim 4 is characterized in that: The method of constructing a training set based on historical risk warning data, and constructing a target separation hyperplane based on the training set according to preset classification rules to obtain a positive sample set and a negative sample set includes: Build a training set based on historical risk warning data; Obtaining a minority feature set based on the training set, and obtaining neighbor data based on the minority feature set and a preset similarity calculation method; Obtaining an optimized training set based on the neighbor data and a preset synthetic minority class oversampling method; A target separation hyperplane is constructed based on the preset classification rules of the optimized training set to obtain a positive sample set and a negative sample set.

6. The intelligent risk warning method based on multi-source data fusion according to claim 4, characterized in that: The optimizing and correcting the three-dimensional warning cloud image based on the posterior probability distribution and the preset posterior probability principle to obtain an improved prediction model includes: Building an initial prediction model based on the three-dimensional warning cloud map; Based on the initial prediction model, a preset stratified cross-validation method and a preset network search method are used to obtain a training prediction model; Based on the posterior probability distribution, the preset posterior probability principle and the trained prediction model, optimization and correction are performed to obtain an improved prediction model.

7. The intelligent risk warning method based on multi-source data fusion according to claim 1, characterized in that: The obtaining of risk warning results based on the improved evidence fusion method and the improved prediction model includes: Acquire a membership matrix set based on the improved prediction model; Introducing global uncertainty based on the membership matrix set to construct an initial trust; Obtaining an initial trust two-dimensional distribution based on the dual control indicators and the initial trust; Acquire a first type of high-conflict evidence and a second type of high-conflict evidence based on the initial two-dimensional distribution of trust; Obtaining a first degree of confidence based on traditional evidence theory and the first type of high-conflict evidence; A target trust level is obtained based on the first trust level, the second type of high-conflict evidence, and the improved evidence fusion method, so as to obtain a risk warning result based on the target trust level.

8. The intelligent risk warning method based on multi-source data fusion according to claim 7, characterized in that: The obtaining of the first confidence level based on traditional evidence theory and the first type of high-conflict evidence includes: Obtain the conflict degree and difference degree based on traditional evidence theory and the first type of high-conflict evidence; An improved conflict factor is obtained based on the conflict degree and the difference degree in a preset conflict improvement manner, so as to obtain a first trust degree based on the improved conflict factor and the initial trust degree.

9. The intelligent risk warning method based on multi-source data fusion according to claim 7, characterized in that: The step of obtaining a target trust level based on the first trust level, the second type of high-conflict evidence, and the improved evidence fusion method, and obtaining a risk warning result based on the target trust level, includes: Obtaining a comprehensive conflict factor based on the first trust level; Obtaining dynamic high conflict and dynamic low conflict based on a preset conflict perception threshold and the comprehensive conflict factor; Performing weighted averaging based on the dynamic high conflict and a preset weighted averaging rule to obtain first revised data; Acquire second revised data based on the dynamic low-conflict and improved evidence fusion method; A target confidence level is obtained based on the first revised data and the second revised data, so as to obtain a risk warning result based on the target confidence level.

10. An intelligent risk warning system based on multi-source data fusion, characterized in that: The method is used to implement an intelligent risk warning method based on multi-source data fusion as claimed in any one of claims 1 to 9, comprising: a monitoring module, a prediction module and a correction module, wherein: The monitoring module is used to obtain multi-source information and obtain early warning control quantities based on the multi-source information and a preset quantitative processing method; The prediction module is used to construct an initial warning cloud map based on the warning control quantity and the preset graded warning standard; The correction module is used to optimize and correct the three-dimensional warning cloud map using the posterior probability support vector method to obtain an improved prediction model, and obtain risk warning results based on the improved evidence fusion method and the improved prediction model.

Images