Method and system for predicting micp intensity based on machine learning and multi-feature fusion

Abstract

The application discloses a MICP intensity prediction method and system based on machine learning and multi-feature fusion, relates to the field of curing data analysis, and constructs a precise quality prediction mapping model and an abnormal compensation parameter mapping model based on historical data, so that the system has the ability to predict the curing quality from process parameters and intelligently recommend an optimal treatment scheme from an abnormal state; through setting a double-threshold quality control mechanism and a real-time abnormal monitoring and early warning system, various abnormal conditions in the curing process can be actively identified and responded in time, from parameter deviation detection to automatic compensation measure execution to complete closed-loop processing of hardware backup switching, so that the stability of the curing process and the consistency of the engineering quality are effectively guaranteed, the demand for manual intervention and the risk of misjudgment are greatly reduced, strong technical support is provided for the standardization, normalization and industrialization development of the sand soil curing technology, and engineering efficiency and economic benefits are significantly improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of solidification data analysis, and particularly relates to a MICP strength prediction method and system based on machine learning and multi-feature fusion. BACKGROUND

[0002] The conventional solidification process lacks real-time quantitative evaluation means for solidification quality, and cannot realize accurate prediction from process parameters to solidification effect, so that engineering quality control relies on experience judgment, and lacks scientificity and consistency; the existing technology has insufficient identification and response ability for abnormal state in the solidification process, and it is difficult to timely find and handle sudden process problems such as bacteria solution activity decline, pipeline blockage, and environmental temperature abnormality, which often leads to substandard solidification quality or process interruption, and parameter adjustment under abnormal conditions mainly relies on manual experience, with slow response speed, unstable treatment effect, and lack of effective fault tolerance and redundancy design, which cannot guarantee the continuity and engineering reliability of the solidification process. SUMMARY

[0003] In view of the problems in the related art, the application provides a MICP strength prediction method and system based on machine learning and multi-feature fusion to overcome the above technical problems existing in the prior art.

[0004] To solve the above technical problems, the application is realized by the following technical scheme:

[0005] The application is a MICP strength prediction method based on machine learning and multi-feature fusion, comprising the following steps:

[0006] S1, setting average unconfined compressive strength and calcium carbonate distribution difference coefficient quality indexes, and setting a plurality of influence factor parameters to collect historical sand column solidification data;

[0007] S2, constructing a final sand column solidification quality index mapping model based on the data collected in S1;

[0008] S3, identifying a plurality of abnormal parameters including bacteria solution activity, grouting pressure, and environmental temperature, setting corresponding measures and parameters, and collecting historical data based thereon to construct a final sand column solidification abnormal measure parameter mapping model;

[0009] S4, collecting current sand column solidification parameter data in real time and inputting the same to the mapping model in S2 for mapping;

[0010] S5, setting double thresholds of compressive strength and distribution difference coefficient, if the mapping result in S4 meets the double thresholds, repeatedly adjusting the solidification parameters and mapping again until the result meets the requirements; then collecting and calculating the deviation of the abnormal solidification parameters in real time, if the deviation meets the corresponding threshold, no measures need to be taken; otherwise, S6 is executed;

[0011] S6. Input the real-time parameters that are prone to anomalies and the preset parameters collected in S5 into the mapping model in S3 for mapping. Perform compensation operation based on the mapping result and then provide feedback to verify the recovery status of the parameters prone to anomalies. If the expected results are not achieved, start hardware backup switching; otherwise, the compensation is completed.

[0012] Preferably, step S1 includes the following steps:

[0013] S11. Define several index types for measuring the quality of sand columns after curing using the MICP method, and obtain a set of sand column curing quality index types; the set of sand column curing quality index types includes the average unconfined compressive strength of the cured sand column and the calcium carbonate distribution difference coefficient.

[0014] Then, several parameter types that affect the quality after curing are defined during the solidification process of sand columns using the MICP method, resulting in a set of sand column solidification parameter types. The set of sand column solidification parameter types includes bacterial solution characteristics, cementing solution characteristics, environmental parameters, grouting method, grouting interval time, number of grouting cycles, soil properties, and geometric parameters of the sand column before solidification.

[0015] S12. Based on the sand column solidification quality index type set and sand column solidification parameter type set, collect the sand column solidification quality index data and sand column solidification parameter data corresponding to multiple historical MIP sand column solidification operations to obtain historical sand column solidification quality index dataset and historical sand column solidification parameter dataset.

[0016] By introducing key quality indicators such as average unconfined compressive strength and calcium carbonate distribution difference coefficient, the curing effect can be comprehensively evaluated from two dimensions: mechanical properties and curing uniformity, thus providing a scientific basis for subsequent process parameter adjustments.

[0017] Preferably, step S2 includes the following steps:

[0018] S21. Based on the historical sand column solidification quality index dataset and the historical sand column solidification parameter dataset, construct a mapping model with various sand column solidification parameter data as input and various sand column solidification quality index data as output, and obtain the final sand column solidification quality index mapping model.

[0019] This mapping model can transform complex multi-parameter input relationships into quantifiable output results, so that the prediction of solidification quality no longer relies on simple empirical judgment of a single parameter, but is based on comprehensive parameter combination analysis, thereby improving the prediction accuracy and reliability.

[0020] Preferably, step S3 includes the following steps:

[0021] S31. Based on the operational steps of several historical sand column solidification processes, the types of parameters prone to anomalies during the sand column solidification process are defined, resulting in a set of easily abnormal solidification parameter types. This set includes bacterial activity, grouting pipeline pressure, and ambient temperature. Then, based on this set, corresponding countermeasures and parameter types are defined for each easily abnormal solidification parameter, resulting in a set of solidification parameter anomaly countermeasures and a set of solidification parameter anomaly countermeasures. These countermeasures include replenishing bacterial solution, executing backflushing procedures, and activating the cooling device. The set of solidification parameter anomaly countermeasures includes the flow rate of the bacterial solution, replenishment time, total bacterial solution replenishment volume, backflushing direction, flushing medium type, flushing pressure, single flushing duration, number of flushing cycles, flushing flow rate, cooling power, target cooling temperature, and cooling duration.

[0022] S32. Based on the set of easily abnormal curing parameter types, the set of measures to deal with abnormal curing parameters, and the set of measures to deal with abnormal curing parameters, collect the values ​​of various easily abnormal curing parameters before and after the abnormality and the parameter data of the corresponding measures taken in the several historical sand column curing operation steps in S31, and obtain the historical pre-abnormal curing parameter dataset, the historical post-abnormal curing parameter dataset, and the historical abnormal measure parameter dataset.

[0023] S33. Based on the historical pre-abnormal solidification parameter dataset, the historical post-abnormal solidification parameter dataset, and the historical abnormal measure parameter dataset, construct a mapping model with the input of the solidification parameter data before and after the abnormality and the output of the abnormal measure parameter data, and obtain the final sand column solidification abnormal measure parameter mapping model.

[0024] By constructing a mapping model based on historical anomaly data, the system is able to predict the optimal handling solution from anomaly states. By inputting the parameter states before and after the anomaly, it can generate specific parameter settings for corresponding countermeasures, which greatly improves the efficiency and accuracy of anomaly handling.

[0025] Preferably, step S4 includes the following steps:

[0026] S41. Based on the set of sand column curing parameter types, collect the corresponding sand column curing parameter data in real time during the current sand column curing process to obtain the current sand column curing parameter dataset.

[0027] S42. Input the current sand column solidification parameter dataset into the final sand column solidification quality index mapping model for mapping to obtain the current sand column solidification quality index dataset.

[0028] By inputting real-time parameter data into a pre-built mapping model for instant calculation, online prediction of curing quality is achieved, avoiding the lag problem of traditional post-curing testing, greatly improving engineering efficiency, and enabling the curing process to be adjusted and optimized in real time based on the prediction results, ensuring that the curing quality is always maintained at the expected level.

[0029] Preferably, step S5 includes the following steps:

[0030] S51. Set the corresponding unconfined compressive strength threshold and calcium carbonate distribution difference coefficient threshold according to the current requirements for sand column solidification.

[0031] S52. If the unconfined compressive strength data in the current sand column solidification quality index dataset is less than the unconfined compressive strength threshold, or the calcium carbonate distribution difference coefficient data in the current sand column solidification quality index dataset is greater than or equal to the calcium carbonate distribution difference coefficient threshold, the current sand column solidification parameter dataset is repeatedly adjusted. In each repetition, the adjusted current sand column solidification parameter dataset is input again into the final sand column solidification quality index mapping model for mapping, until the unconfined compressive strength data in the mapping result is greater than or equal to the unconfined compressive strength threshold and the calcium carbonate distribution difference coefficient data in the mapping result is less than the calcium carbonate distribution difference coefficient threshold, thus obtaining the current final sand column solidification parameter dataset; otherwise, no adjustment is required, and the current sand column solidification parameter dataset is used as the current final sand column solidification parameter dataset.

[0032] S53. Based on the current actual requirements for sand column solidification, set the deviation threshold between each type of easily abnormal solidification parameter and the corresponding preset value to obtain the current set of deviation thresholds for easily abnormal solidification parameters.

[0033] S54. Set the parameters for the current sand column curing based on the current final sand column curing parameter dataset; after setting, start the current sand column curing operation; during the operation, collect the corresponding abnormal curing parameter data in real time according to the abnormal curing parameter type set to obtain the current real-time abnormal curing parameter set; then obtain the preset values ​​corresponding to the current various abnormal curing parameters based on the current final sand column curing parameter dataset to obtain the current initial abnormal curing parameter set.

[0034] Based on the current set of deviation thresholds for easily abnormal solidified parameters, if the absolute value of the difference between the corresponding parameters in the current real-time set of easily abnormal solidified parameters and the current initial set of easily abnormal solidified parameters is greater than or equal to the corresponding deviation threshold, execute S6; otherwise, no action is required.

[0035] An iterative optimization mechanism is adopted to dynamically adjust the curing parameters. Through repeated verification and correction, it is ensured that the final curing parameters can accurately meet the quality requirements, avoiding the blindness and uncertainty of parameter setting in traditional methods. Furthermore, by establishing a set of deviation thresholds for easily abnormal curing parameters, real-time monitoring and early warning of key process parameters are realized. Potential risks can be detected in the early stages when parameters deviate from the normal range, significantly improving the system's proactive prevention capabilities.

[0036] Preferably, step S6 includes the following steps:

[0037] S61. Input the current initial set of easily abnormal solidification parameters and the current real-time set of easily abnormal solidification parameters into the final sand column solidification abnormality measure parameter mapping model for mapping, and obtain the current abnormality measure parameter dataset.

[0038] S62. Set the execution feedback cycle; set and execute the corresponding parameters during the replenishment of bacterial solution, execution of backwashing procedure and start-up of cooling device according to the current abnormal measure parameter dataset; after the execution feedback cycle, collect the corresponding easily abnormal solidification parameter data in real time again to obtain the current compensated easily abnormal solidification parameter set;

[0039] If the absolute value of the difference between the corresponding parameters in the current compensated set of parameters prone to anomalies and the current initial set of parameters prone to anomalies is less than the corresponding deviation threshold, the anomaly compensation is completed; otherwise, execute S63.

[0040] S63. Switch from the currently used bacterial solution tank to the standby bacterial solution tank and from the currently used grouting pipeline to the standby pipeline;

[0041] By establishing a dynamic feedback control loop and setting an adaptively adjustable execution feedback cycle, the system ensures that anomaly compensation measures can be fully verified and adjusted, greatly enhancing the stability and reliability of the system. In addition, a comprehensive fault-tolerance and redundancy design is prepared. When the parameter deviation cannot be controlled within the allowable range after the compensation measures are implemented, the hardware-level switching mechanism can be automatically triggered, significantly improving the continuous operation capability and engineering quality stability of the sand and soil solidification project.

[0042] The MICP intensity prediction system based on machine learning and multi-feature fusion includes a historical sand column solidification process data acquisition module, a sand column solidification quality index mapping model construction module, a sand column solidification abnormal measure parameter mapping model construction module, a current sand column solidification parameter mapping module, a current solidification parameter adjustment and easily abnormal solidification parameter judgment module, and a current abnormal measure parameter setting and measure execution feedback module.

[0043] The present invention has the following beneficial effects:

[0044] 1. This invention constructs a precise quality prediction mapping model and anomaly compensation parameter mapping model based on historical data, enabling the system to predict solidification quality from process parameters and intelligently recommend optimal handling solutions from abnormal states, significantly improving the accuracy and intelligence of process control. By setting a dual-threshold quality control mechanism and a real-time anomaly monitoring and early warning system, it can proactively identify and promptly respond to various abnormal situations in the solidification process. The complete closed-loop processing from parameter deviation detection to automatic compensation measures and hardware backup switching effectively ensures the stability of the solidification process and the consistency of engineering quality, greatly reducing the need for manual intervention and the risk of misjudgment. It provides strong technical support for the standardization, normalization, and industrialization of sand and soil solidification technology, significantly improving engineering efficiency and economic benefits.

[0045] 2. In this invention, by inputting real-time parameter data into a pre-built mapping model for instant calculation, online prediction of curing quality is achieved, avoiding the lag problem of traditional post-curing detection, greatly improving engineering efficiency, and enabling the curing process to be adjusted and optimized in real time according to the prediction results, ensuring that the curing quality is always maintained at the expected level.

[0046] 3. In this invention, by inputting the initial parameter state and real-time monitoring state into the anomaly response mapping model, intelligent decision-making from anomaly identification to the optimal handling plan is realized, avoiding the lag and subjectivity of traditional manual judgment and significantly improving the timeliness and accuracy of anomaly handling. Secondly, a dynamic feedback control loop is established. By setting an adaptively adjustable execution feedback cycle, it is ensured that anomaly compensation measures can be fully verified and adjusted, forming a closed-loop control process of "detection-decision-execution-verification-readjustment". Furthermore, a sound fault tolerance and redundancy design is prepared. When the parameter deviation still cannot be controlled within the allowable range after the compensation measures are executed, a hardware-level switching mechanism can be automatically triggered.

[0047] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart illustrating the MICP intensity prediction method based on machine learning and multi-feature fusion of the present invention.

[0050] Figure 2A schematic diagram illustrating the process of constructing the final sand column solidification quality index mapping model for this invention;

[0051] Figure 3 This is a schematic diagram of the process for constructing the parameter mapping model for the final sand column solidification anomaly measures in this invention;

[0052] Figure 4 This is a schematic diagram illustrating the process of setting parameters for current abnormal measures and providing feedback on the execution of abnormal measures according to the present invention.

[0053] Figure 5 This is a flowchart illustrating the process of adjusting current curing parameters and determining easily abnormal curing parameters according to the present invention.

[0054] Figure 6 This is a schematic diagram of the process of mapping the current sand column solidification parameters according to the present invention;

[0055] Figure 7 This is a schematic diagram of the MIP intensity prediction system based on machine learning and multi-feature fusion according to the present invention.

[0056] Figure 8 This is a line graph showing the real-time monitoring data of the sand column solidification quality of this invention. Detailed Implementation

[0057] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0058] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0059] Example 1

[0060] Please see Figure 1 This embodiment describes a MIP intensity prediction method based on machine learning and multi-feature fusion, including the following steps:

[0061] S1. Set the quality indicators of average unconfined compressive strength and calcium carbonate distribution difference coefficient, and set several influencing factor parameters to collect historical sand column solidification data.

[0062] Please see Figure 2 , Figure 8 S1 includes the following steps:

[0063] S11. Define several index types to measure the quality of sand columns after solidification using the MIP (Microbially Induced Calcite Precipitation) method, resulting in a set of sand column solidification quality index types. The set of sand column solidification quality index types includes the average unconfined compressive strength of the solidified sand column and the calcium carbonate distribution difference coefficient (the difference coefficient can be used to quantitatively evaluate the dispersion of calcium carbonate distribution, and its calculation formula is the ratio of the standard deviation of the sample to the mean; where, the larger the difference coefficient, the greater the difference in calcium carbonate content deposited in different parts of the sand column, and the more uneven the solidification of the overall sand column).

[0064] Several parameter types that affect the post-curing quality during the MICP method for solidifying sand columns are then defined, resulting in a set of sand column solidification parameter types. This set includes bacterial solution characteristics (such as injection times, bacterial density, etc.), cementing solution characteristics (such as calcium chloride and urea concentrations, etc.), environmental parameters (such as pH value, temperature, etc.), grouting method, grouting interval, grouting cycle count, soil properties (such as particle size, gradation, density, void ratio, etc.), and geometric parameters of the sand column before solidification (such as height, diameter, etc.).

[0065] S12. Based on the sand column solidification quality index type set and sand column solidification parameter type set, collect the sand column solidification quality index data and sand column solidification parameter data corresponding to multiple historical MIP sand column solidification operations to obtain historical sand column solidification quality index dataset and historical sand column solidification parameter dataset.

[0066] By systematically constructing a set of quality index types and curing parameter types for sand column curing, comprehensive quantitative analysis and optimized control of the MIP curing process were achieved. Specifically, by introducing key quality indicators such as average unconfined compressive strength and calcium carbonate distribution difference coefficient, the curing effect can be comprehensively evaluated from two dimensions: mechanical properties and curing uniformity, thus providing a scientific basis for subsequent process parameter adjustments. Secondly, incorporating various parameters affecting curing (including bacterial solution characteristics, cementitious liquid composition, environmental conditions, grouting methods, etc.) into a unified framework helps identify the degree of influence of each parameter on curing quality, thereby achieving precise control of key parameters. In addition, the dataset established based on historical data provides data support for the subsequent construction of corresponding mapping models, thereby improving the controllability and predictability of the MIP curing process.

[0067] S2. Construct a final sand column solidification quality index mapping model based on the data collected in S1;

[0068] S2 includes the following steps:

[0069] S21. Based on the historical sand column solidification quality index dataset and the historical sand column solidification parameter dataset, construct a mapping model with various sand column solidification parameter data as input and various sand column solidification quality index data as output, and obtain the final sand column solidification quality index mapping model.

[0070] S21 includes the following steps:

[0071] S211. Construct an initial sand column solidification quality index mapping model and set a first training data ratio (e.g., 8:2 or 7:3, which can be adjusted adaptively according to the actual training situation); divide the historical sand column solidification parameter dataset and the historical sand column solidification quality index dataset according to the first training data ratio to obtain the first training dataset and the first test dataset.

[0072] S212. Set a first training error threshold (10%~15%, which can be adjusted adaptively according to the actual training situation); input the first training dataset into the initial sand column solidification quality index mapping model for training; during the training process, if the training error is less than the first training error threshold, stop training and obtain the trained sand column solidification quality index mapping model; otherwise, continue training until the training error is less than the first training error threshold.

[0073] S213. Set a first test accuracy threshold (90%~95%, which can be adjusted adaptively according to the actual test situation); input the first test dataset into the trained sand column solidification quality index mapping model for testing; after the test is completed, obtain the first test accuracy data; if the first test accuracy data is greater than or equal to the first test accuracy threshold, use the trained sand column solidification quality index mapping model as the final sand column solidification quality index mapping model; otherwise, return to S212 to continue training the trained sand column solidification quality index mapping model and repeat S213 until the first test accuracy data is greater than or equal to the first test accuracy threshold.

[0074] The initial sand column solidification quality index mapping model adopts XGBoost, consisting of 100 regression trees, with a maximum depth of 6 layers per tree to ensure sufficient expressive power while avoiding overfitting. A greedy algorithm is used for feature selection during node splitting, determining the optimal split point by minimizing the mean squared error. The learning rate (shrinkage rate) is set to 0.1 to control the contribution weight of each tree and improve model stability. The model incorporates an L1 regularization coefficient of 0.01 and an L2 regularization coefficient of 1.0 to constrain model complexity and enhance generalization performance. In terms of feature processing, the model automatically processes the input solidification parameter features, including the number of bacterial infusions, bacterial density, and chlorination. Multidimensional data such as calcium concentration, urea concentration, pH value, temperature, and grouting method are encoded and nonlinear mapping relationships are achieved through hierarchical judgment in a tree structure. The model uses the weighted quantile Sketch algorithm for quantile compression to improve training efficiency on large-scale data, while supporting an automatic missing value handling mechanism, demonstrating good robustness to incomplete solidification parameter data. In the prediction stage, the model obtains the final predicted values ​​of solidification quality indicators by weighted summation of the outputs of all regression trees, including key performance parameters such as average unconfined compressive strength and calcium carbonate distribution difference coefficient. The entire model structure is optimized for hyperparameters through cross-validation to ensure the prediction accuracy and engineering practicality of sand column solidification quality.

[0075] By constructing a mapping model between sand column curing parameters and quality indicators, intelligent prediction and optimized control of the MICP curing process were achieved. Specifically, this mapping model can transform complex multi-parameter input relationships into quantifiable output results, so that the prediction of curing quality no longer relies on simple empirical judgment of a single parameter, but is based on comprehensive parameter combination analysis, thereby improving prediction accuracy and reliability. Secondly, by establishing the mathematical relationship between parameters and quality indicators, this model provides a scientific basis for parameter optimization of the curing process, and can quickly identify the key parameter combinations that have the greatest impact on curing quality, significantly improving the efficiency of process design. In addition, this mapping model has good generalization ability and can be applied to sand column curing processes under different conditions, providing a universal prediction tool for engineering practice and reducing experimental costs and time consumption. At the same time, the model also supports real-time monitoring and adjustment of the curing process. By inputting the current curing parameters, it can predict the quality status after curing in real time, providing technical support for the dynamic optimization of process parameters.

[0076] S3. Identify several parameters that are prone to anomalies, including bacterial activity, grouting pressure, and ambient temperature, then set corresponding countermeasures and their parameters, and collect historical data based on this to construct the final sand column solidification anomaly response parameter mapping model.

[0077] Please see Figure 3 S3 includes the following steps:

[0078] S31. Based on the operational steps of several historical sand column solidification processes, the types of parameters prone to anomalies (i.e., sudden deviations from preset values) during sand column solidification are defined, resulting in a set of easily abnormal solidification parameter types. This set includes bacterial activity, grouting pipeline pressure, and ambient temperature. Then, based on this set, corresponding countermeasures and parameter types are defined for each easily abnormal solidification parameter, resulting in a set of solidification parameter anomaly response measures and a set of solidification parameter anomaly response parameters. These countermeasures include replenishing bacterial solution and performing a backwashing procedure (referring to short-term, controlled modification of fluid transport or filtration systems). The automated operation process of changing the flow direction or state of the medium to remove blockages, deposits or restore system performance, and starting the cooling device, etc.; the parameters for abnormal solidification parameters include the flow rate of the bacterial solution when replenishing the bacterial solution, the replenishment time of the bacterial solution, the total amount of bacterial solution replenished, the backwash direction (usually opposite to the normal working flow direction: reverse flushing), the type of flushing medium (clean water, special cleaning fluid, or high-flow-rate flushing of the original medium (such as cementing liquid / bacterial solution) can be used), the flushing pressure (to enhance the flushing force), the duration of a single flush, the number of flushing cycles, the flushing flow rate (to ensure sufficient carrying capacity), the cooling power, the target cooling temperature, and the duration of cooling, etc.

[0079] S32. Based on the set of easily abnormal curing parameter types, the set of measures to deal with abnormal curing parameters, and the set of measures to deal with abnormal curing parameters, collect the values ​​of various easily abnormal curing parameters before and after the abnormality and the parameter data of the corresponding measures taken in the several historical sand column curing operation steps in S31, and obtain the historical pre-abnormal curing parameter dataset, the historical post-abnormal curing parameter dataset, and the historical abnormal measure parameter dataset.

[0080] S33. Based on the historical pre-abnormal solidification parameter dataset, the historical post-abnormal solidification parameter dataset, and the historical abnormal measure parameter dataset, construct a mapping model with the input of the solidification parameter data before and after the abnormality and the output of the abnormal measure parameter data, and obtain the final sand column solidification abnormal measure parameter mapping model.

[0081] S33 includes the following steps:

[0082] S331. Construct an initial sand column solidification anomaly measure parameter mapping model and set a second training data ratio (e.g., 8:2 or 7:3, which can be adjusted adaptively according to the actual training situation); divide the historical pre-anomaly solidification parameter dataset, historical post-anomaly solidification parameter dataset, and historical anomaly measure parameter dataset according to the second training data ratio to obtain the second training dataset and the second test dataset.

[0083] S332. Set a second training error threshold (10%~15%, which can be adjusted adaptively according to the actual training situation); input the second training dataset into the initial sand column solidification anomaly measure parameter mapping model for training; during the training process, if the training error is less than the second training error threshold, stop training and obtain the trained sand column solidification anomaly measure parameter mapping model; otherwise, continue training until the training error is less than the second training error threshold.

[0084] S333. Set a second test accuracy threshold (90%~95%, which can be adjusted adaptively according to the actual test situation); input the second test dataset into the trained sand column solidification anomaly measure parameter mapping model for testing; after the test is completed, obtain the second test accuracy data; if the second test accuracy data is greater than or equal to the second test accuracy threshold, use the trained sand column solidification anomaly measure parameter mapping model as the final sand column solidification anomaly measure parameter mapping model; otherwise, return to S332 to continue training the trained sand column solidification anomaly measure parameter mapping model and repeat S333 until the second test accuracy data is greater than or equal to the second test accuracy threshold;

[0085] The parameter mapping model for the initial sand column solidification anomaly measures adopts a multilayer perceptron regressor model.

[0086] The number of neurons in the input layer is equal to the number of input features. Assuming we select N parameters that are prone to abnormal solidification and extract their values ​​before and after the abnormality as features, the input dimension is approximately 2N (we also need to add possible manually constructed features, such as differences, ratios, etc.). For example, if there are three original parameters: bacterial activity, grouting pressure, and ambient temperature, there are at least 6 basic input features (3 before the abnormality + 3 after the abnormality). With the addition of derived features, the total input dimension may be around 10 to 20.

[0087] Its hidden layers include a first fully connected layer containing 50 neurons, with ReLU (Rectified Linear Unit) as the activation function, which can effectively alleviate the gradient vanishing problem and speed up training; and a second fully connected layer containing 30 neurons, also using the ReLU activation function (more hidden layers can be added depending on the data complexity).

[0088] The number of neurons in its output layer is equal to the total number of abnormal measure parameters that need to be predicted; for example, if we want to predict three continuous values, namely, the flow rate of the replenishing bacterial solution, the backwash pressure, and the cooling power, then the output layer has 3 neurons; the activation function is generally not strictly restricted here (the default is linear activation 'linear'), since this is a regression task, it needs to be able to output arbitrary real numbers;

[0089] Loss function: Mean squared error or mean absolute error is used, both of which are commonly used regression loss functions; Optimizer: Adam optimizer is used, which can converge quickly in the early stages of training and finely adjust the weights in the later stages; Learning rate is set to 0.001 and batch size is 64 or 128 during training;

[0090] By setting key parameters prone to anomalies during the curing process (such as bacterial activity, grouting pressure, and ambient temperature), a comprehensive anomaly early warning system was established, enabling the system to identify abnormal signs of process conditions in advance, shifting from passive response to proactive prevention. Secondly, the solution combines anomaly handling measures with specific parameters to form a structured response strategy library, including measures such as replenishing bacterial solution, backwashing procedures, and activating cooling devices, and sets detailed execution parameters for each measure to ensure the standardization and operability of anomaly handling. In addition, a mapping model built based on historical anomaly data enables the system to predict the optimal handling plan from anomaly states. By inputting the parameter states before and after the anomaly, the system can generate specific parameter settings for corresponding response measures, significantly improving the efficiency and accuracy of anomaly handling, thereby enhancing the intelligence level and engineering reliability of the entire curing process system, and providing a safer, more efficient, and stable construction guarantee for sand and soil reinforcement projects under complex geological conditions.

[0091] S4. Real-time acquisition of current sand column solidification parameter data and input into the mapping model in S2 for mapping;

[0092] Please see Figure 6 S4 includes the following steps:

[0093] S41. Based on the set of sand column curing parameter types, collect the corresponding sand column curing parameter data in real time during the current sand column curing process to obtain the current sand column curing parameter dataset.

[0094] S42. Input the current sand column solidification parameter dataset into the final sand column solidification quality index mapping model for mapping to obtain the current sand column solidification quality index dataset.

[0095] By collecting various parameter data in real time during the current curing process, the system can comprehensively grasp the real-time status of the curing process, including key information such as bacterial solution injection parameters, changes in cementitious liquid composition, and fluctuations in environmental conditions, providing timely data support for quality control. Secondly, by inputting real-time parameter data into a pre-built mapping model for instant calculation, the system enables online prediction of curing quality, avoiding the lag problem of traditional post-curing testing, greatly improving engineering efficiency, and allowing the curing process to be adjusted and optimized in real time based on the prediction results, ensuring that the curing quality is always maintained at the expected level. At the same time, the system has good adaptability and scalability, and can adjust parameters and optimize models for different types of sand and soil and engineering needs, reducing the subjectivity of human experience judgment, thereby effectively ensuring engineering quality and construction safety, and reducing the risk of rework and economic losses caused by substandard curing quality.

[0096] S5. Set dual thresholds for compressive strength and distribution difference coefficient. If the mapping result in S4 meets the dual thresholds, repeatedly adjust the curing parameters and remap until the result meets the requirements. Then collect and calculate the deviation of easily abnormal curing parameters in real time. If the deviation meets the corresponding threshold, no measures need to be taken. Otherwise, proceed to S6.

[0097] Please see Figure 5 S5 includes the following steps:

[0098] S51. Set the corresponding unconfined compressive strength threshold and calcium carbonate distribution difference coefficient threshold according to the current requirements for sand column solidification.

[0099] S52. If the unconfined compressive strength data in the current sand column solidification quality index dataset is less than the unconfined compressive strength threshold, or the calcium carbonate distribution difference coefficient data in the current sand column solidification quality index dataset is greater than or equal to the calcium carbonate distribution difference coefficient threshold, the current sand column solidification parameter dataset is repeatedly adjusted. In each repetition, the adjusted current sand column solidification parameter dataset is input again into the final sand column solidification quality index mapping model for mapping, until the unconfined compressive strength data in the mapping result is greater than or equal to the unconfined compressive strength threshold and the calcium carbonate distribution difference coefficient data in the mapping result is less than the calcium carbonate distribution difference coefficient threshold, thus obtaining the current final sand column solidification parameter dataset; otherwise, no adjustment is required, and the current sand column solidification parameter dataset is used as the current final sand column solidification parameter dataset.

[0100] S53. Based on the current actual requirements for sand column solidification, set the deviation threshold between each type of easily abnormal solidification parameter and the corresponding preset value to obtain the current set of deviation thresholds for easily abnormal solidification parameters.

[0101] S54. Set the parameters for the current sand column curing based on the current final sand column curing parameter dataset; after setting, start the current sand column curing operation; during the operation, collect the corresponding abnormal curing parameter data in real time according to the abnormal curing parameter type set to obtain the current real-time abnormal curing parameter set; then obtain the preset values ​​corresponding to the current various abnormal curing parameters based on the current final sand column curing parameter dataset to obtain the current initial abnormal curing parameter set.

[0102] Based on the current set of deviation thresholds for easily abnormal solidified parameters, if the absolute value of the difference between the corresponding parameters in the current real-time set of easily abnormal solidified parameters and the current initial set of easily abnormal solidified parameters is greater than or equal to the corresponding deviation threshold, execute S6; otherwise, no action is required.

[0103] By constructing a complete closed-loop control and anomaly monitoring system for sand column curing quality, intelligent management of the entire chain—from quality prediction to parameter optimization and real-time monitoring—is achieved. Specifically, by setting dual quality thresholds for unconfined compressive strength and calcium carbonate distribution difference coefficient, a multi-dimensional quality evaluation standard is established to ensure that the curing effect meets both mechanical performance requirements and has good uniformity, providing comprehensive quantitative assurance for engineering quality. Secondly, the scheme uses an iterative optimization mechanism to dynamically adjust curing parameters. Through repeated verification and correction, it ensures that the final curing parameters accurately meet quality requirements, avoiding the blindness and uncertainty of parameter settings in traditional methods. Thirdly, by establishing a set of deviation thresholds for easily abnormal curing parameters, real-time monitoring and early warning of key process parameters are achieved, enabling the timely detection of potential risks in the early stages when parameters deviate from the normal range, significantly improving the system's proactive prevention capabilities. In addition, this technical solution constructs a complete feedback loop from parameter setting, real-time monitoring to anomaly handling, forming a closed-loop control process of "setting-execution-monitoring-adjustment-re-execution," greatly improving the stability and repeatability of the curing process and providing a flexible and reliable solution for complex and ever-changing engineering environments.

[0104] S6. Input the real-time parameters that are prone to anomalies and the preset parameters collected in S5 into the mapping model in S3 for mapping. Perform compensation operation based on the mapping result and then provide feedback to verify the recovery status of the parameters prone to anomalies. If the expected results are not achieved, start hardware backup switching; otherwise, the compensation is completed.

[0105] Please see Figure 4 S6 includes the following steps:

[0106] S61. Input the current initial set of easily abnormal solidification parameters and the current real-time set of easily abnormal solidification parameters into the final sand column solidification abnormality measure parameter mapping model for mapping, and obtain the current abnormality measure parameter dataset.

[0107] S62. Set the execution feedback cycle (which can be adaptively set according to the actual sand column solidification requirements); set and execute the corresponding parameters during the replenishment of bacterial solution, execution of backwashing procedure and start-up of cooling device according to the current abnormal measure parameter dataset; after the execution feedback cycle, collect the corresponding easily abnormal solidification parameter data again in real time to obtain the current compensated easily abnormal solidification parameter set;

[0108] If the absolute value of the difference between the corresponding parameters in the current compensated set of parameters prone to anomalies and the current initial set of parameters prone to anomalies is less than the corresponding deviation threshold, the anomaly compensation is completed; otherwise, execute S63.

[0109] S63. Switch from the currently used bacterial solution tank to the standby bacterial solution tank and from the currently used grouting pipeline to the standby pipeline;

[0110] For example, taking the MIP solidification process control of a sand and soil reinforcement project as an example, the following is an example:

[0111] 1. Real-time parameter acquisition and quality prediction: In a sand and soil reinforcement project, key parameter data during the current solidification process were acquired in real time, including the number of bacterial infusions (3 times) and the bacterial density (8.5 × 10⁻⁶). 8 The parameters included CFU / mL, calcium chloride concentration of 0.25 mol / L, urea concentration of 0.15 mol / L, pH value of 7.8, temperature of 22.5℃, grouting pressure of 0.8 MPa, grouting flow rate of 120 mL / min, sand column height of 300 mm, and diameter of 100 mm. These parameters were input into the final sand column solidification quality index mapping model, predicting that the average unconfined compressive strength after solidification was 185 kPa, and the calcium carbonate distribution difference coefficient was 0.28.

[0112] 2. Quality Indicator Comparison and Parameter Optimization: Based on project requirements, quality thresholds were set: unconfined compressive strength threshold was 200 kPa, and calcium carbonate distribution difference coefficient threshold was 0.25. Since the predicted compressive strength of 185 kPa was less than the threshold of 200 kPa, and the distribution difference coefficient of 0.28 was greater than or equal to the threshold of 0.25, the curing quality was deemed substandard. Therefore, the parameter optimization process was initiated to adjust the current parameter set, increasing the bacterial density to 9.2 × 10⁻⁶. 8 The concentrations of CFU / mL, calcium chloride, and urea were increased to 0.30 mol / L, urea concentration to 0.18 mol / L, and grouting pressure to 1.0 MPa. The final sand column solidification quality index mapping model was input again, and the predicted result was a compressive strength of 215 kPa and a distribution difference coefficient of 0.23, which met the quality requirements. The final solidification parameter set was obtained.

[0113] 3. Anomaly Monitoring Threshold Setting: Based on actual engineering needs, a set of deviation thresholds for parameters prone to anomalies is set, including a bacterial activity deviation threshold of ±10%, a grouting pipeline pressure deviation threshold of ±0.1MPa, and an ambient temperature deviation threshold of ±2℃. Before the curing operation begins, the system obtains initial preset values ​​based on the final curing parameter set, with the bacterial activity set to 9.2×10⁻⁶. 8 CFU / mL, grouting pipeline pressure 1.0MPa, ambient temperature 22.5℃;

[0114] 4. Real-time monitoring and anomaly detection: During the curing process, various parameters are continuously monitored, and real-time data is collected: the bacterial activity is 8.9 × 10⁻⁶. 8 The parameters are: CFU / mL, grouting pipeline pressure, and ambient temperature, 24.0℃. Compared with the initial preset values, the bacterial activity deviation is -3.26%, the grouting pipeline pressure deviation is -0.05MPa, and the ambient temperature deviation is +1.5℃. Among these, the absolute value of the bacterial activity deviation is 3.26%, which is less than the set 10% threshold; the absolute value of the grouting pressure deviation is 0.05MPa, which is less than the set 0.1MPa threshold; and the absolute value of the ambient temperature deviation is 1.5℃, which is less than the set 2℃ threshold. Therefore, the system determines that the current state is normal and no action is required.

[0115] 5. Abnormal Response and Compensation: During subsequent continuous monitoring, the system collected new real-time data: the bacterial activity was 7.8 × 10⁻⁶. 8 The parameters are: CFU / mL, grouting pipeline pressure: 0.75 MPa, ambient temperature: 25.8℃. Compared with the initial preset values, the bacterial activity deviation is -15.22%, the grouting pipeline pressure deviation is -0.25 MPa, and the ambient temperature deviation is +3.3℃. Among these, the absolute value of the bacterial activity deviation is 15.22%, which is greater than the set 10% threshold; the absolute value of the grouting pressure deviation is 0.25 MPa, which is greater than the set 0.1 MPa threshold; and the absolute value of the ambient temperature deviation is 3.3℃, ​​which is greater than the set 2℃ threshold. The system determines that an anomaly has occurred and proceeds to the next step.

[0116] 6. Abnormal measure parameter mapping and execution: The system will map and execute the current initial parameter set (bacterial activity 9.2 × 10⁻⁶). 8 CFU / mL, injection pressure 1.0 MPa, ambient temperature 22.5℃) and real-time parameter set (bacterial activity 7.8 × 10⁻⁶). 8The parameters (CFU / mL, grouting pressure 0.75MPa, ambient temperature 25.8℃) are input into the final sand column solidification anomaly response parameter mapping model to obtain the anomaly response parameter set: replenishment bacterial solution flow rate 150mL / min, replenishment time 120 seconds, total replenishment volume 300mL, backwash pressure 1.2MPa, backwash duration 30 seconds, cooling power 800W, and cooling target temperature 20.0℃. The system executes corresponding anomaly compensation measures based on these parameters, including replenishing bacterial solution, executing the backwash procedure, and activating the cooling device.

[0117] 7. Feedback Verification and Final Processing: After implementing the anomaly compensation measures, the system is set to a feedback cycle of 30 minutes. Real-time parameter data is collected again after 30 minutes: the bacterial activity is 9.0 × 10⁻⁶. 8 The parameters were: CFU / mL, grouting pipeline pressure: 0.98 MPa, ambient temperature: 22.8℃. Compared with the initial preset values, the bacterial activity deviation was -2.17%, the grouting pipeline pressure deviation was -0.02 MPa, and the ambient temperature deviation was +0.3℃. Among these, the absolute value of the bacterial activity deviation was 2.17%, which is less than the set 10% threshold; the absolute value of the grouting pressure deviation was 0.02 MPa, which is less than the set 0.1 MPa threshold; and the absolute value of the ambient temperature deviation was 0.3℃, which is less than the set 2℃ threshold. Therefore, the system determined that the abnormal compensation was completed, and the curing process returned to normal operation.

[0118] By constructing an intelligent compensation and fault-tolerance mechanism for sand column solidification anomalies, rapid response and adaptive repair of sudden process anomalies during solidification are achieved. Specifically, by inputting the initial parameter status and real-time monitoring status into the anomaly response mapping model, intelligent decision-making from anomaly identification to the optimal handling plan is realized, avoiding the lag and subjectivity of traditional manual judgment and significantly improving the timeliness and accuracy of anomaly handling. Secondly, a dynamic feedback control loop is established. By setting an adaptively adjustable execution feedback cycle, it is ensured that anomaly compensation measures can be fully verified and adjusted, forming a closed-loop control process of "detection-decision-execution-verification-readjustment", which greatly enhances the stability and reliability of the system. In addition, a comprehensive fault-tolerance and redundancy design is prepared. When the parameter deviation cannot be controlled within the allowable range after the compensation measures are executed, a hardware-level switching mechanism can be automatically triggered, such as the backup switching of the bacterial solution tank and grouting pipeline. This effectively avoids the interruption of the entire solidification process due to the failure of a single equipment, significantly improves the continuous operation capability and engineering quality stability of sand solidification projects, provides strong technical support for engineering implementation under complex working conditions, and greatly reduces engineering risks and economic losses caused by process anomalies.

[0119] Example 2

[0120] Please see Figure 7This embodiment discloses a MIP intensity prediction system based on machine learning and multi-feature fusion. The system can implement the method of the above embodiment, including a historical sand column solidification process data acquisition module, a sand column solidification quality index mapping model construction module, a sand column solidification abnormal measure parameter mapping model construction module, a current sand column solidification parameter mapping module, a current solidification parameter adjustment and easily abnormal solidification parameter determination module, and a current abnormal measure parameter setting and measure execution feedback module.

[0121] The historical sand column solidification process data acquisition module is set with quality indicators such as average unconfined compressive strength and calcium carbonate distribution difference coefficient, as well as several influencing factor parameters, in order to collect historical sand column solidification data.

[0122] The sand column solidification quality index mapping model construction module constructs the final sand column solidification quality index mapping model based on the data collected by the historical sand column solidification process data acquisition module.

[0123] The sand column solidification anomaly response parameter mapping model construction module identifies several parameters prone to anomalies, including bacterial activity, grouting pressure, and ambient temperature. It then sets corresponding countermeasures and their parameters, and collects historical data based on this to construct the final sand column solidification anomaly response parameter mapping model.

[0124] The current sand column solidification parameter mapping module collects the current sand column solidification parameter data in real time and inputs it into the mapping model in the sand column solidification quality index mapping model construction module for mapping.

[0125] The current curing parameter adjustment and easily abnormal curing parameter judgment module sets dual thresholds for compressive strength and distribution difference coefficient. If the mapping result in the current sand column curing parameter mapping module meets the dual thresholds, the curing parameters are repeatedly adjusted and mapped again until the result meets the requirements. Then, the deviation of easily abnormal curing parameters is collected and calculated in real time. If the deviation meets the corresponding threshold, no measures need to be taken. Otherwise, the current abnormal measure parameter setting and measure execution feedback module is executed.

[0126] The current abnormal measure parameter setting and measure execution feedback module inputs the real-time and preset parameters of easily abnormal curing collected in the current curing parameter adjustment and easily abnormal curing parameter judgment module into the mapping model in the sand column curing abnormal measure parameter mapping model construction module for mapping. After performing compensation operation based on the mapping result, it provides feedback to verify the recovery status of easily abnormal parameters. If the expected results are not achieved, hardware backup switching is initiated; otherwise, compensation is completed.

[0127] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0128] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present invention.

Claims

1. A method for predicting the intensity of MICO based on machine learning and multi-feature fusion, characterized in that, Includes the following steps: S1. Set the quality indicators of average unconfined compressive strength and calcium carbonate distribution difference coefficient, and set several influencing factor parameters to collect historical sand column solidification data. S2. Construct a final sand column solidification quality index mapping model based on the data collected in S1; S3. Identify several parameters that are prone to anomalies, including bacterial activity, grouting pressure, and ambient temperature, then set corresponding countermeasures and their parameters, and collect historical data based on this to construct the final sand column solidification anomaly response parameter mapping model. Specifically, it includes: S31. Based on the operational steps of several historical sand column solidification processes, the types of parameters prone to abnormalities during the sand column solidification process are defined, resulting in a set of easily abnormal solidification parameter types. This set includes bacterial activity, grouting pipeline pressure, and ambient temperature. Then, based on this set, corresponding countermeasures and parameter types are defined for each easily abnormal solidification parameter, resulting in a set of solidification parameter abnormality countermeasures and a set of solidification parameter abnormality countermeasures. The set includes replenishing bacterial solution, executing a backflushing procedure, and activating a cooling device. The set includes the bacterial solution flow rate, replenishment time, total bacterial solution replenishment volume, backflushing direction, flushing medium type, flushing pressure, single flushing duration, flushing cycle count, flushing flow rate, cooling power, target cooling temperature, and cooling duration. S32. Based on the set of easily abnormal curing parameter types, the set of measures to deal with abnormal curing parameters, and the set of measures to deal with abnormal curing parameters, collect the values ​​of various easily abnormal curing parameters before and after the abnormality and the parameter data of the corresponding measures taken in the several historical sand column curing operation steps in S31, and obtain the historical pre-abnormal curing parameter dataset, the historical post-abnormal curing parameter dataset, and the historical abnormal measure parameter dataset. S33. Based on the historical pre-abnormal solidification parameter dataset, the historical post-abnormal solidification parameter dataset, and the historical abnormal measure parameter dataset, construct a mapping model with the input of the solidification parameter data before and after the abnormality and the output of the abnormal measure parameter data, and obtain the final sand column solidification abnormal measure parameter mapping model. S4. Real-time acquisition of current sand column solidification parameter data and input into the mapping model in S2 for mapping; S5. Set dual thresholds for compressive strength and distribution difference coefficient. If the mapping result in S4 meets the dual thresholds, repeatedly adjust the curing parameters and remap until the result meets the requirements. Then collect and calculate the deviation of easily abnormal curing parameters in real time. If the deviation meets the corresponding threshold, no measures need to be taken. Otherwise, proceed to S6. S6. Input the real-time parameters that are prone to anomalies and the preset parameters collected in S5 into the mapping model in S3 for mapping. Perform compensation operation based on the mapping result and then provide feedback to verify the recovery status of the parameters prone to anomalies. If the expected results are not achieved, start hardware backup switching; otherwise, the compensation is completed.

2. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 1, characterized in that, S1 includes the following steps: S11. Define several index types for measuring the quality of sand columns after curing using the MICP method, and obtain a set of sand column curing quality index types; the set of sand column curing quality index types includes the average unconfined compressive strength of the cured sand column and the calcium carbonate distribution difference coefficient. Then, several parameter types that affect the quality after curing are defined during the solidification process of sand columns using the MICP method, resulting in a set of sand column solidification parameter types. The set of sand column solidification parameter types includes bacterial solution characteristics, cementing solution characteristics, environmental parameters, grouting method, grouting interval time, number of grouting cycles, soil properties, and geometric parameters of the sand column before solidification. S12. Based on the sand column solidification quality index type set and sand column solidification parameter type set, collect the sand column solidification quality index data and sand column solidification parameter data corresponding to multiple historical MICP sand column solidification operations to obtain historical sand column solidification quality index dataset and historical sand column solidification parameter dataset.

3. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 2, characterized in that, S2 includes the following steps: S21. Based on the historical sand column solidification quality index dataset and the historical sand column solidification parameter dataset, construct a mapping model with various types of sand column solidification parameter data as input and various types of sand column solidification quality index data as output, and obtain the final sand column solidification quality index mapping model.

4. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 3, characterized in that, S4 includes the following steps: S41. Based on the set of sand column curing parameter types, collect the corresponding sand column curing parameter data in real time during the current sand column curing process to obtain the current sand column curing parameter dataset. S42. Input the current sand column solidification parameter dataset into the final sand column solidification quality index mapping model for mapping to obtain the current sand column solidification quality index dataset.

5. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 4, characterized in that, S5 includes the following steps: S51. Set the corresponding unconfined compressive strength threshold and calcium carbonate distribution difference coefficient threshold according to the current requirements for sand column solidification. S52. If the unconfined compressive strength data in the current sand column solidification quality index dataset is less than the unconfined compressive strength threshold, or the calcium carbonate distribution difference coefficient data in the current sand column solidification quality index dataset is greater than or equal to the calcium carbonate distribution difference coefficient threshold, the current sand column solidification parameter dataset is repeatedly adjusted. In each repetition, the adjusted current sand column solidification parameter dataset is input again into the final sand column solidification quality index mapping model for mapping, until the unconfined compressive strength data in the mapping result is greater than or equal to the unconfined compressive strength threshold and the calcium carbonate distribution difference coefficient data in the mapping result is less than the calcium carbonate distribution difference coefficient threshold, thus obtaining the current final sand column solidification parameter dataset; otherwise, no adjustment is required, and the current sand column solidification parameter dataset is used as the current final sand column solidification parameter dataset. S53. Based on the current requirements for sand column solidification, set the deviation threshold between each type of easily abnormal solidification parameter and the corresponding preset value to obtain the current set of deviation thresholds for easily abnormal solidification parameters.

6. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 5, characterized in that, The S5 also includes: S54. Set the parameters for the current sand column curing based on the current final sand column curing parameter dataset; after setting, start the current sand column curing operation; during the operation, collect the corresponding abnormal curing parameter data in real time according to the abnormal curing parameter type set to obtain the current real-time abnormal curing parameter set; then obtain the preset values ​​corresponding to the current various abnormal curing parameters based on the current final sand column curing parameter dataset to obtain the current initial abnormal curing parameter set. Based on the current set of deviation thresholds for easily abnormal solidified parameters, if the absolute value of the difference between the corresponding parameters in the current real-time set of easily abnormal solidified parameters and the current initial set of easily abnormal solidified parameters is greater than or equal to the corresponding deviation threshold, execute S6; otherwise, no action is required.

7. The MIP intensity prediction method based on machine learning and multi-feature fusion according to claim 6, characterized in that, S6 includes the following steps: S61. Input the current initial set of easily abnormal solidification parameters and the current real-time set of easily abnormal solidification parameters into the final sand column solidification abnormality measure parameter mapping model for mapping, and obtain the current abnormality measure parameter dataset. S62. Set the execution feedback cycle; set and execute the corresponding parameters during the replenishment of bacterial solution, execution of backwashing procedure and start-up of cooling device according to the current abnormal measure parameter dataset; after the execution feedback cycle, collect the corresponding easily abnormal solidification parameter data in real time again to obtain the current compensated easily abnormal solidification parameter set; If the absolute value of the difference between the corresponding parameters in the current compensated set of parameters prone to anomalies and the current initial set of parameters prone to anomalies is less than the corresponding deviation threshold, the anomaly compensation is completed; otherwise, execute S63. S63. Switch from the currently used bacterial solution tank to the standby bacterial solution tank and from the currently used grouting pipeline to the standby pipeline.

8. A system for implementing the MIP intensity prediction method based on machine learning and multi-feature fusion as described in any one of claims 1-7, characterized in that: It includes a historical sand column solidification process data acquisition module, a sand column solidification quality index mapping model construction module, a sand column solidification abnormal measure parameter mapping model construction module, a current sand column solidification parameter mapping module, a current solidification parameter adjustment and easily abnormal solidification parameter judgment module, and a current abnormal measure parameter setting and measure execution feedback module. The historical sand column solidification process data acquisition module is set with average unconfined compressive strength and calcium carbonate distribution difference coefficient, as well as several influencing factor parameters, in order to collect historical sand column solidification data. The sand column solidification quality index mapping model construction module constructs the final sand column solidification quality index mapping model based on the data collected by the historical sand column solidification process data acquisition module. The sand column solidification anomaly response parameter mapping model construction module identifies several parameters prone to anomalies, including bacterial activity, grouting pressure, and ambient temperature. It then sets corresponding countermeasures and their parameters, and collects historical data based on this to construct the final sand column solidification anomaly response parameter mapping model. The current sand column solidification parameter mapping module collects the current sand column solidification parameter data in real time and inputs it into the mapping model in the sand column solidification quality index mapping model construction module for mapping. The current curing parameter adjustment and easily abnormal curing parameter judgment module sets dual thresholds for compressive strength and distribution difference coefficient. If the mapping result in the current sand column curing parameter mapping module meets the dual thresholds, the curing parameters are repeatedly adjusted and mapped again until the result meets the requirements. Then, the deviation of easily abnormal curing parameters is collected and calculated in real time. If the deviation meets the corresponding threshold, no measures need to be taken. Otherwise, the current abnormal measure parameter setting and measure execution feedback module is executed. The current abnormal measure parameter setting and measure execution feedback module inputs the real-time parameters of easily abnormal curing collected in the current curing parameter adjustment and easily abnormal curing parameter judgment module, as well as the preset parameters, into the mapping model in the sand column curing abnormal measure parameter mapping model construction module for mapping. After performing compensation operation based on the mapping result, it provides feedback to verify the recovery status of easily abnormal parameters. If the expected results are not achieved, hardware backup switching is initiated; otherwise, compensation is completed.

Images