Displacement observation data-based roadway support opportunity and support parameter determination method

Abstract

The invention discloses a roadway support opportunity and support parameter determination method based on displacement observation data, and the method comprises the steps: obtaining roadway surface displacement-time sequence data through displacement monitoring equipment, and extracting key morphological characteristics, such as curve slope change and inflection point position, through a displacement-time curve characteristic recognition model; and constructing a roadway displacement state evaluation index system to obtain a comprehensive evaluation index value. Meanwhile, a support parameter prediction model based on a BP neural network is established, the feature data serve as input, and support parameters such as the length and diameter of the anchor rods and the distance between the anchor cables are predicted after training. The comprehensive displacement state evaluation index value is compared with a preset threshold value, and when the comprehensive displacement state evaluation index value reaches or exceeds the threshold value, the predicted supporting parameters are adopted for roadway supporting. According to the method, the roadway support time and parameters can be accurately determined, and the roadway support decision scientificity and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of roadway support, and in particular to a method for determining the timing and parameters of roadway support based on displacement observation data. Background Technology

[0002] As a crucial infrastructure in mining and underground engineering construction, roadways' stability directly impacts project safety and construction efficiency. During roadway construction and use, various factors such as geological conditions and mining activities cause deformation and displacement of the surrounding rock. Without timely and appropriate support, this can easily lead to collapses and other safety accidents. Therefore, accurately determining the timing and parameters for roadway support is critical to ensuring roadway stability. Traditional roadway support decisions rely heavily on engineering experience, lacking scientific quantitative basis and failing to adapt to complex and changing engineering environments. While technological advancements have led to the development of analytical methods based on monitoring data, significant room for improvement remains in the depth and accuracy of data processing, as well as the scientific rigor of support parameter predictions.

[0003] The primary problem with existing technologies is their simplistic and crude data processing methods. Most methods merely analyze the numerical magnitude of displacement data, failing to fully explore the deeper characteristics inherent in the displacement-time curve. This results in an inability to comprehensively reflect the deformation patterns of the surrounding rock in the tunnel, making it difficult to accurately determine the timing of support. For example, they cannot effectively identify key morphological features such as abrupt changes in the slope of the displacement curve and the appearance of inflection points, leading to a lag in the judgment of tunnel deformation trends. Secondly, in terms of support parameter prediction, existing models lack a comprehensive consideration of the coupled effects of multiple factors. Traditional models often focus only on the influence of a single factor or a few factors on support parameters, neglecting the inherent correlation between support parameters such as anchor bolt length, diameter, anchor cable spacing, and shotcrete thickness, as well as the corrective effect of factors such as surrounding rock grade on support parameters. This leads to deviations between the predicted support parameters and actual engineering needs, resulting in either insufficient support threatening tunnel safety or excessive support causing resource waste. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a method for determining the timing and parameters of roadway support based on displacement observation data.

[0005] The technical solution adopted in this invention is a method for determining the timing and parameters of roadway support based on displacement observation data, comprising the following steps:

[0006] Step S1: Real-time monitoring of the surface displacement of the roadway is carried out using displacement monitoring equipment to obtain continuous displacement-time series data and form an initial displacement-time dataset;

[0007] Step S2: Using the displacement-time curve feature recognition model, process the initial displacement-time dataset, extract the preset calibration features of the displacement-time curve from the data, including curve slope change, inflection point position, curvature change amplitude feature parameters, and define the dataset including the preset calibration features as the feature dataset.

[0008] Step S3: Based on the feature parameters in the feature dataset, construct a roadway displacement state evaluation index system, and calculate the comprehensive displacement state evaluation index value by allocating weights to each feature parameter.

[0009] Step S4: Establish a support parameter prediction model based on a BP neural network. Use the feature dataset as input data and the roadway support parameters, including anchor bolt length, anchor bolt diameter, anchor cable spacing, and shotcrete thickness, as output data to train the BP neural network and obtain the trained support parameter prediction model.

[0010] Step S5: Input the feature data formed by processing the real-time displacement-time data through the displacement-time curve feature recognition model into the trained support parameter prediction model to obtain the predicted roadway support parameters.

[0011] Step S6: Compare the calculated comprehensive displacement state evaluation index value with the preset support trigger threshold. When the comprehensive displacement state evaluation index value reaches or exceeds the support trigger threshold, the predicted roadway support parameters are used to support the roadway.

[0012] Furthermore, in the displacement-time curve feature recognition model, the slope change characteristics of the displacement-time curve are quantified using the following formula: Where K is the quantized value of the slope change, d i Let t represent the displacement monitoring value at the i-th time point. i Let represent the i-th time point, and n be the total number of displacement-time data points; this quantization value is used to characterize the drastic change in the slope of the displacement-time curve, and the weight of the slope change quantization value in the feature dataset is ω. K ω K These are the weight coefficients determined using the analytic hierarchy process (AHP).

[0013] Furthermore, in the displacement-time curve feature recognition model, the inflection point position of the displacement-time curve is determined using the following formula:

[0014]

[0015] Where I is the inflection point index value, and sign is the sign function; the inflection point index value corresponds to the location point on the displacement-time curve where the slope change trend changes. The displacement value and time value at this location point are used as calibration features in the feature dataset, and the weight of this feature is ω when constructing the roadway displacement state evaluation index system. I ω I Determined by principal component analysis.

[0016] Furthermore, in the support parameter prediction model based on the BP neural network, a correction coefficient α for the surrounding rock grade of the roadway is introduced, and the calculation formula for the output layer of the BP neural network is adjusted as follows: Among them, P j For the j-th predicted roadway support parameter, w kj v represents the connection weight between the k-th neuron in the output layer and the j-th support parameter output node. ik x represents the connection weight between the i-th neuron in the hidden layer and the k-th neuron in the output layer. i is the i-th feature data of the input layer, f is the activation function, m is the number of neurons in the hidden layer, and l is the number of feature data in the input layer; the roadway surrounding rock grade correction coefficient α is determined by the lithology and integrity parameters of the roadway surrounding rock through a pre-established surrounding rock grade classification table, and is used to correct the support parameters predicted by the BP neural network.

[0017] Furthermore, when constructing the roadway displacement state evaluation index system, the comprehensive displacement state evaluation index value is calculated using the following formula: Where E is the comprehensive displacement state evaluation index value, ω j Let y be the weight of the j-th feature parameter. j Let y be the standardized value of the j-th feature parameter, and p be the total number of feature parameters; the standardized value y j It is obtained by mapping the values ​​of each feature parameter to the interval [0, 1], and the weight ω of each feature parameter is... j The comprehensive displacement state evaluation index value, determined by the entropy weight method, is used to characterize the comprehensive state of roadway displacement and serves as the basis for determining whether roadway support is necessary.

[0018] Furthermore, in the BP neural network-based support parameter prediction model, the correlation between roadway support parameters is modeled, and the correlation correction coefficient β is calculated using the following formula. ij : Where, β ij P is the correlation correction coefficient between the i-th support parameter and the j-th support parameter. is Let i be the value of the i-th support parameter in the s-th training data set. Let be the average value of the i-th support parameter in the training data, and q be the number of training data sets; the correlation correction coefficient is used to correct the correlation of the support parameters predicted by the BP neural network, so that the predicted support parameters conform to the mutual relationship in actual engineering, where i, j = 1, 2, ..., n, and n is the total number of roadway support parameters.

[0019] Furthermore, step S3 includes the following sub-steps:

[0020] Step S31: Based on the slope change, inflection point location, and curvature change amplitude characteristic parameters of the displacement-time curve, define corresponding sub-evaluation indicators. Each sub-evaluation indicator is used to reflect the change characteristics of roadway displacement from different perspectives.

[0021] Step S32: The weight of each sub-evaluation index in the roadway displacement state evaluation index system is determined by using the analytic hierarchy process. This weight determination process is achieved by constructing a judgment matrix and calculating the eigenvectors and eigenvalues ​​of the matrix, which is used to quantify the importance of each sub-evaluation index to the comprehensive displacement state evaluation.

[0022] Step S33: Perform dimensionless processing on each characteristic parameter, converting characteristic parameters with different dimensions into comparable values, so that each characteristic parameter has the same status when calculating the comprehensive displacement state evaluation index value;

[0023] Step S34: Based on each sub-evaluation index and its weight, the dimensionless characteristic parameters are weighted and summed to obtain the comprehensive displacement state evaluation index value, which comprehensively reflects the overall displacement state of the roadway.

[0024] Furthermore, step S4 includes the following sub-steps:

[0025] Step S41: Determine the network structure of the BP neural network, including the number of input layer nodes, the number of hidden layers and the number of hidden layer nodes, and the number of output layer nodes. The number of input layer nodes corresponds to the number of feature parameters in the feature dataset, and the number of output layer nodes corresponds to the number of roadway support parameters. The number of hidden layers and nodes are determined by trial and error. Construct a neural network architecture for support parameter prediction.

[0026] Step S42: Initialize the connection weights and thresholds of the BP neural network. Random numbers are used to assign values ​​to the connection weights and thresholds between neurons in the network to provide initial parameters for the training of the neural network.

[0027] Step S43: Divide the feature dataset into a training set and a test set. Use the training set to train the BP neural network through forward and backward propagation. By adjusting the connection weights and thresholds, the output error of the neural network is gradually reduced until the preset training stopping condition is met.

[0028] Step S44: Use a test set to verify the performance of the trained BP neural network, calculate the error between the output of the neural network on the test set and the actual roadway support parameters, and determine whether the prediction ability of the neural network meets the requirements.

[0029] Furthermore, step S5 includes the following sub-steps:

[0030] Step S51: Input the real-time acquired displacement-time data into the displacement-time curve feature recognition model, and extract the preset calibration features of the displacement-time curve according to the processing flow of the model to form real-time feature data;

[0031] Step S52: Convert the real-time feature data to meet the input requirements of the trained support parameter prediction model, ensuring that the data can be correctly input into the neural network;

[0032] Step S53: Input the real-time feature data after format conversion into the trained support parameter prediction model based on BP neural network. After the neural network calculates, the predicted roadway support parameters are obtained.

[0033] Step S54: Check the rationality of the predicted roadway support parameters. Based on the actual situation and experience of the roadway project, determine whether the predicted support parameters are within a reasonable range. If they are not within a reasonable range, make corresponding corrections.

[0034] Beneficial Effects: This invention proposes a method for determining the timing and parameters of roadway support based on displacement observation data. This method utilizes a displacement-time curve feature recognition model to deeply mine the key morphological features contained in the displacement-time curve. By accurately extracting features such as curve slope changes, inflection point locations, and curvature change amplitudes, it comprehensively and meticulously reflects the deformation patterns of the roadway surrounding rock. Compared to traditional methods that only focus on the magnitude of data values, this approach can keenly capture the trend changes and abrupt changes in roadway deformation, thus providing a more scientific and reliable basis for accurately determining the timing of support and avoiding support delays caused by lagging judgments of deformation trends. Addressing the lack of multi-factor consideration in support parameter prediction, this method constructs a support parameter prediction model based on a BP neural network and implements several innovative optimizations. A roadway surrounding rock grade correction coefficient is introduced to fully consider the influence of factors such as surrounding rock lithology and integrity on support parameters; simultaneously, the correlation between support parameters is modeled, making the predictions of parameters such as anchor bolt length, diameter, anchor cable spacing, and shotcrete thickness more consistent with the interrelationships in actual engineering projects. This allows the predicted support parameters to closely align with actual engineering needs, ensuring both the safety and stability of the roadway and avoiding resource waste caused by over-support, significantly improving the scientific rigor and practicality of support parameter prediction. Through advanced data processing methods and optimized prediction models, this method achieves precise determination of roadway support timing and parameters, providing strong support for the safe and efficient construction of roadway projects, effectively improving the scientific nature and accuracy of roadway support decisions, and reducing engineering risks and costs. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method steps of the present invention;

[0036] Figure 2 This is a diagram showing the unit composition for implementing the method of the present invention. Detailed Implementation

[0037] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] like Figure 1 As shown, the method for determining the timing and parameters of roadway support based on displacement observation data is presented.

[0039] Step S1: Real-time monitoring of the surface displacement of the roadway is carried out using displacement monitoring equipment to obtain continuous displacement-time series data and form an initial displacement-time dataset;

[0040] Specifically, this step, as the initial stage of the entire method, plays a crucial role in acquiring raw data. The selection of displacement monitoring equipment is critical to data quality. High-precision total stations typically offer millimeter-level accuracy, meeting the accuracy requirements of roadway displacement monitoring. Laser rangefinders can be set to a measurement frequency of once per minute or even higher, ensuring the capture of subtle changes in roadway displacement. The displacement-time series data acquired by these devices covers key information such as roadway roof subsidence and sidewall convergence, accumulating continuously from the start of monitoring, providing fundamental data support for subsequent analysis. The completeness and continuity of the initial displacement-time dataset directly impact subsequent judgments on roadway deformation trends and are the cornerstone of the entire support decision-making process.

[0041] During implementation, the placement of displacement monitoring equipment must adhere to scientific principles. For the tunnel roof, a monitoring point is typically placed every 10-20 meters to monitor roof subsidence; for the sidewalls, monitoring points are symmetrically placed at the waistline to monitor the convergence displacement of the sidewalls. During equipment installation, stability and reliability must be ensured to avoid data errors caused by external interference. Simultaneously, a data acquisition system should be established to achieve real-time connection between the displacement monitoring equipment and the data storage terminal, ensuring that the acquired displacement-time data can be recorded and saved promptly and accurately, forming an initial displacement-time dataset, providing a reliable data source for subsequent analysis.

[0042] Step S2: Using the displacement-time curve feature recognition model, process the initial displacement-time dataset, extract the preset calibration features of the displacement-time curve from the data, including curve slope change, inflection point position, curvature change amplitude feature parameters, and define the dataset including the preset calibration features as the feature dataset.

[0043] Specifically, this step aims to extract deeper information from the initial data. The displacement-time curve feature recognition model analyzes the data using a specific algorithm. Changes in the curve slope reflect the trend of displacement velocity in the roadway; an increasing slope indicates acceleration, while a decreasing slope indicates deceleration. Inflection points signify a change in the roadway deformation trend, potentially indicating changes in surrounding rock stability. The magnitude of curvature changes reflects the severity of displacement changes. Extracting these key morphological features transforms the raw data into more valuable analytical information, providing a basis for judging the roadway deformation state. The feature dataset focuses on key morphological features, eliminating redundant information, making subsequent analysis more targeted and efficient.

[0044] During implementation, the initial displacement-time dataset must first be imported into the displacement-time curve feature recognition model. The model analyzes the data point-by-point according to the established algorithms and rules. For extracting curve slope changes, the slope differences across different time periods are analyzed based on the calculation logic of displacement changes at adjacent time points. Inflection point locations are determined by identifying points where the slope trend changes. The calculation of curvature change amplitude is based on an algorithm analyzing the curve's curvature. After model processing, the extracted feature parameters, such as curve slope changes, inflection point locations, and curvature change amplitude, are integrated to form a structured feature dataset. This dataset can be directly used to subsequently construct a roadway displacement state evaluation index system and train a support parameter prediction model.

[0045] Step S3: Based on the feature parameters in the feature dataset, construct a roadway displacement state evaluation index system, and calculate the comprehensive displacement state evaluation index value by allocating weights to each feature parameter.

[0046] Specifically, constructing a roadway displacement state evaluation index system involves the systematic processing of characteristic datasets, which is of great significance. Based on characteristic parameters such as curve slope changes, inflection point locations, and curvature change amplitudes in the characteristic dataset, corresponding sub-evaluation indices are defined, each reflecting roadway displacement characteristics from different dimensions. Through a scientific weight allocation method, the importance of each sub-evaluation index in the system is determined, making the evaluation results more consistent with reality. Dimensionless processing of the characteristic parameters eliminates the influence of dimensional differences, ensuring that each parameter has equal weight in the calculation. The final comprehensive displacement state evaluation index value integrates information from multiple key characteristic parameters, comprehensively and objectively reflecting the overall state of roadway displacement and providing a quantitative basis for support decisions.

[0047] In practical implementation, the first step is to establish a sub-evaluation index system based on characteristic parameters, such as constructing a velocity change index based on curve slope changes and a trend change index based on inflection point positions. Next, the analytic hierarchy process (AHP) is used to determine weights and construct a judgment matrix. The relative importance of each sub-evaluation index is judged through expert scoring or engineering experience, and the eigenvectors and eigenvalues ​​of the matrix are calculated to obtain the weight coefficients of each sub-evaluation index. Then, dimensionless methods such as normalization are used to transform characteristic parameters of different dimensions into comparable values. Finally, based on each sub-evaluation index and its weight, the dimensionless characteristic parameters are weighted and summed to calculate the comprehensive displacement state evaluation index value, thus completing the construction of the roadway displacement state evaluation index system and the calculation of index values.

[0048] Step S4: Establish a support parameter prediction model based on a BP neural network. Use the feature dataset as input data and the roadway support parameters, including anchor bolt length, anchor bolt diameter, anchor cable spacing, and shotcrete thickness, as output data to train the BP neural network and obtain the trained support parameter prediction model.

[0049] Specifically, the support parameter prediction model established in this step, based on a backpropagation (BP) neural network, is the core of achieving accurate support parameter prediction. The BP neural network learns from the feature dataset to establish a mapping relationship between input features and roadway support parameters. Using support parameters such as anchor bolt length, anchor bolt diameter, anchor cable spacing, and shotcrete thickness as outputs, the model can predict reasonable support parameters based on roadway deformation characteristics. Through training the neural network, adjusting the connection weights and thresholds within the network optimizes model performance and improves prediction accuracy and reliability. The trained model can quickly and accurately predict support parameters based on real-time roadway deformation feature data, providing a scientific basis for roadway support.

[0050] In the implementation process, the structure of the BP neural network is first determined. The number of input layer nodes is set based on the number of feature parameters in the feature dataset, and the number of output layer nodes corresponds to the number of roadway support parameters. The number of hidden layers and nodes is adjusted through trial and error, combined with actual training results and prediction accuracy. Next, the connection weights and thresholds of the neural network are initialized, typically using random numbers. Then, the feature dataset is divided into training and testing sets. The neural network is trained using the training set through forward and backward propagation. During training, the connection weights and thresholds are continuously adjusted to gradually reduce the output error until a preset training stopping condition is met, such as reaching a certain number of training iterations or the error being less than a set value. Finally, the performance of the trained model is verified using the testing set. By calculating the error between the model's output and the actual support parameters, the model's predictive ability is judged. If the requirements are met, a usable, trained support parameter prediction model is obtained.

[0051] Step S5: Input the feature data formed by processing the real-time displacement-time data through the displacement-time curve feature recognition model into the trained support parameter prediction model to obtain the predicted roadway support parameters.

[0052] Specifically, this step achieves a crucial transformation from real-time monitoring data to support parameter prediction. Real-time displacement-time data reflects the current deformation state of the roadway. After processing by the displacement-time curve feature recognition model, key morphological features are extracted to form feature data, which contains core information about the current roadway deformation. This feature data is then input into the trained support parameter prediction model, and using the established input-output mapping relationship, the predicted roadway support parameters are calculated. This prediction method based on real-time data can respond promptly to roadway deformation, providing immediate parameter references for roadway support decisions and ensuring the timeliness and effectiveness of support measures.

[0053] In practice, after real-time displacement-time data is acquired, it is immediately input into the displacement-time curve feature recognition model. Following the model's processing flow, key morphological features such as curve slope changes, inflection point locations, and curvature change amplitudes are extracted to form real-time feature data. Since the trained support parameter prediction model has specific requirements for the input data format, the real-time feature data needs to be format-converted to conform to the model's input specifications. After format conversion, the real-time feature data is input into the support parameter prediction model. The neurons within the model perform layer-by-layer calculations and processing on the input data based on pre-trained connection weights and thresholds. After a forward propagation process, the model finally outputs predicted roadway support parameters, such as anchor bolt length and diameter.

[0054] Step S6: Compare the calculated comprehensive displacement state evaluation index value with the preset support trigger threshold. When the comprehensive displacement state evaluation index value reaches or exceeds the support trigger threshold, the predicted roadway support parameters are used to support the roadway.

[0055] Specifically, this step is the final decision-making stage of the entire method and is of decisive significance. The comprehensive displacement state evaluation index comprehensively reflects the overall displacement state of the roadway, and the preset support trigger threshold is a critical value determined based on factors such as roadway geological conditions and engineering design requirements. Comparing the two allows us to determine whether the roadway needs support. When the comprehensive displacement state evaluation index reaches or exceeds the support trigger threshold, it indicates that the roadway displacement has reached a dangerous level, and the stability of the surrounding rock is threatened. At this time, it is necessary to promptly apply the predicted roadway support parameters to ensure roadway safety. This decision-making method based on quantitative index comparison avoids the uncertainty of subjective judgment, making support decisions more scientific and reasonable.

[0056] During implementation, the first step is to determine a reasonable support trigger threshold based on the actual conditions of the tunnel, either by organizing an expert team or by drawing on engineering experience, combined with tunnel geological survey data and engineering design standards. After obtaining the calculated comprehensive displacement state evaluation index value, it is compared with the preset support trigger threshold. If the comprehensive displacement state evaluation index value reaches or exceeds the support trigger threshold, a detailed support construction plan is immediately developed according to the predicted tunnel support parameters, including specific operational steps and technical requirements for anchor bolt installation, anchor cable arrangement, and shotcrete construction. Subsequently, construction personnel are organized to carry out tunnel support construction according to the plan, strictly controlling quality during construction to ensure that the support measures effectively enhance tunnel stability and guarantee the safe operation of the tunnel project.

[0057] Preferably, in the displacement-time curve feature recognition model, the slope change feature of the displacement-time curve is quantified using the following formula: Where K is the quantized value of the slope change, d i Let t represent the displacement monitoring value at the i-th time point. i Let represent the i-th time point, and n be the total number of displacement-time data points. This quantized value characterizes the drastic change in the slope of the displacement-time curve, and is used as one of the calibration parameters in the feature dataset for subsequent calculations. The weight of the quantized slope change value in the feature dataset is ω. K ω K These are the weight coefficients determined using the analytic hierarchy process (AHP).

[0058] Specifically, in tunnel support, the slope change of the displacement-time curve directly reflects the dynamic trend of tunnel deformation rate, which is crucial for determining the timing of support. Through a specific quantification method, the abstract concept of slope change is transformed into a concrete numerical value. This value can serve as a key parameter in the feature dataset, participating in the subsequent construction of the tunnel displacement state evaluation index system and the prediction of support parameters. In implementation, based on the time series characteristics of displacement-time data, the displacement and time changes at adjacent time points are analyzed and calculated to obtain the quantified value of the slope change. Simultaneously, the analytic hierarchy process (AHP) is used to determine the weight of this quantified value in the feature dataset, clarifying its importance in the entire evaluation system. This makes subsequent analysis more targeted and scientific, ensuring that the analysis of tunnel deformation rate changes accurately guides support decisions.

[0059] Preferably, in the displacement-time curve feature recognition model, the inflection point position of the displacement-time curve is determined using the following formula:

[0060]

[0061] Where I is the inflection point index value, and sign is the sign function; the inflection point index value corresponds to the location point on the displacement-time curve where the slope change trend changes. The displacement value and time value at this location point are used as calibration features in the feature dataset, and the weight of this feature is ω when constructing the roadway displacement state evaluation index system. I ω I Determined by principal component analysis.

[0062] Specifically, during roadway deformation, the inflection point of the displacement-time curve corresponds to a change in the roadway deformation trend. Accurately identifying the inflection point location helps to predict changes in the stability of the surrounding rock in advance, buying time for support decisions. The inflection point index value, obtained through a specific calculation method, precisely locates the point on the curve where the slope change trend changes, and its corresponding displacement and time values ​​become key features in the feature dataset. During implementation, based on the changing patterns of the displacement-time data, a specific algorithm is used to analyze the data point by point to identify the inflection point location. Furthermore, principal component analysis is used to determine the weight of this feature in constructing the roadway displacement state evaluation index system, highlighting the importance of the inflection point location in reflecting the roadway deformation state, making subsequent judgments on the roadway deformation trend more accurate, and providing strong support for scientific and rational support decisions.

[0063] Preferably, in the support parameter prediction model based on BP neural network, a correction coefficient α for the surrounding rock grade of the roadway is introduced, and the calculation formula of the output layer of the BP neural network is adjusted as follows: Among them, P j For the j-th predicted roadway support parameter, w kj v represents the connection weight between the k-th neuron in the output layer and the j-th support parameter output node. ik x represents the connection weight between the i-th neuron in the hidden layer and the k-th neuron in the output layer. i is the i-th feature data of the input layer, f is the activation function, m is the number of neurons in the hidden layer, and l is the number of feature data in the input layer; the roadway surrounding rock grade correction coefficient α is determined by the lithology and integrity parameters of the roadway surrounding rock through a pre-established surrounding rock grade classification table, and is used to correct the support parameters predicted by the BP neural network to improve the accuracy of the prediction.

[0064] Specifically, the support parameter prediction model based on a BP neural network is optimized by introducing a correction coefficient for the surrounding rock grade. The surrounding rock grade is influenced by various factors such as lithology and integrity, and different grades of surrounding rock have different requirements for support parameters. This correction coefficient, determined based on the actual parameters of the surrounding rock through a pre-established surrounding rock grade classification table, effectively corrects the support parameters predicted by the BP neural network, making the prediction results more closely match actual engineering needs. During implementation, after constructing the BP neural network, relevant parameters of the surrounding rock grade are used as additional inputs or correction criteria. The correction coefficient is introduced during the calculation process of the neural network output layer to adjust the predicted support parameters such as anchor bolt length and diameter. This approach fully considers the influence of surrounding rock characteristics on support parameters, compensates for the shortcomings of traditional neural network models in considering various factors, improves the accuracy of support parameter prediction, and ensures that the support scheme effectively guarantees roadway stability.

[0065] Preferably, when constructing the roadway displacement state evaluation index system, the comprehensive displacement state evaluation index value is calculated using the following formula: Where E is the comprehensive displacement state evaluation index value, ω j Let y be the weight of the j-th feature parameter. j Let y be the standardized value of the j-th feature parameter, and p be the total number of feature parameters; the standardized value y j This is obtained by mapping each feature parameter value to the interval [0, 1], and the weight ω of each feature parameter is... j The comprehensive displacement state evaluation index value, determined by the entropy weight method, is used to characterize the comprehensive state of roadway displacement and serves as the basis for determining whether roadway support is necessary.

[0066] Specifically, the calculation method for the comprehensive displacement state evaluation index value is described in the construction of the roadway displacement state evaluation index system. This index value is a key basis for determining whether the roadway needs support, and its calculation integrates information from multiple feature parameters in the feature dataset. By standardizing each feature parameter and mapping it to the [0,1] interval, the influence of dimensional differences is eliminated, ensuring that each parameter has equal status in the calculation. At the same time, the entropy weight method is used to determine the weight of each feature parameter, objectively allocating weights based on the information entropy of the data itself, avoiding interference from subjective factors. In the implementation process, feature parameters such as curve slope changes and inflection point positions in the feature dataset are first standardized. Then, based on the weights calculated by the entropy weight method, the standardized feature parameters are weighted and summed to obtain the comprehensive displacement state evaluation index value. This value comprehensively reflects the overall state of roadway displacement, providing a quantitative and scientific basis for subsequent comparison with the preset support trigger threshold to determine the support timing.

[0067] Preferably, in the BP neural network-based support parameter prediction model, the correlation between roadway support parameters is modeled, and the correlation correction coefficient β is calculated using the following formula. ij : Where, β ij P is the correlation correction coefficient between the i-th support parameter and the j-th support parameter. is Let i be the value of the i-th support parameter in the s-th training data set. Let be the average value of the i-th support parameter in the training data, and q be the number of training data sets; the correlation correction coefficient is used to correct the correlation of the support parameters predicted by the BP neural network, so that the predicted support parameters are more consistent with the actual relationship in engineering, where i, j = 1, 2, ..., n, and n is the total number of roadway support parameters.

[0068] Specifically, the modeling of the correlation between support parameters in the support parameter prediction model based on BP neural networks is crucial. In actual tunnel support engineering, support parameters such as anchor bolt length, anchor bolt diameter, anchor cable spacing, and shotcrete thickness are not independent but intrinsically correlated. Reasonably considering these correlations can make the predicted support parameters more consistent with engineering realities. By calculating correlation correction coefficients, the support parameters predicted by the BP neural network are corrected for correlation, making the combination of parameters more scientific and reasonable. During implementation, based on a large amount of training data, the value patterns of each support parameter under different working conditions are analyzed, and correlation correction coefficients between parameters are calculated. After the neural network predicts the initial support parameters, the parameters are adjusted according to the correlation correction coefficients to optimize the combination of support parameters, avoiding insufficient or excessive support due to parameter mismatches, and improving the overall effectiveness and reliability of the support scheme.

[0069] Preferably, step S3 includes the following sub-steps:

[0070] Step S31: Based on the slope change, inflection point location, and curvature change amplitude characteristic parameters of the displacement-time curve, define corresponding sub-evaluation indicators. Each sub-evaluation indicator is used to reflect the change characteristics of roadway displacement from different perspectives.

[0071] Step S32: The weight of each sub-evaluation index in the roadway displacement state evaluation index system is determined by using the analytic hierarchy process. This weight determination process is achieved by constructing a judgment matrix and calculating the eigenvectors and eigenvalues ​​of the matrix, which is used to quantify the importance of each sub-evaluation index to the comprehensive displacement state evaluation.

[0072] Step S33: Perform dimensionless processing on each characteristic parameter, converting characteristic parameters with different dimensions into comparable values, so that each characteristic parameter has the same status when calculating the comprehensive displacement state evaluation index value;

[0073] Step S34: Based on each sub-evaluation index and its weight, the dimensionless characteristic parameters are weighted and summed to obtain the comprehensive displacement state evaluation index value, which comprehensively reflects the overall displacement state of the roadway.

[0074] Preferably, step S4 includes the following sub-steps:

[0075] Step S41: Determine the network structure of the BP neural network, including the number of input layer nodes, the number of hidden layers and the number of hidden layer nodes, and the number of output layer nodes. The number of input layer nodes corresponds to the number of feature parameters in the feature dataset, and the number of output layer nodes corresponds to the number of roadway support parameters. The number of hidden layers and the number of nodes are determined by trial and error to construct a suitable neural network architecture for support parameter prediction.

[0076] Step S42: Initialize the connection weights and thresholds of the BP neural network. Random numbers are used to assign values ​​to the connection weights and thresholds between neurons in the network to provide initial parameters for the training of the neural network.

[0077] Step S43: Divide the feature dataset into a training set and a test set. Use the training set to train the BP neural network through forward and backward propagation. By adjusting the connection weights and thresholds, the output error of the neural network is gradually reduced until the preset training stopping condition is met.

[0078] Step S44: Use a test set to verify the performance of the trained BP neural network, calculate the error between the output of the neural network on the test set and the actual roadway support parameters, and determine whether the prediction ability of the neural network meets the requirements.

[0079] Preferably, step S5 includes the following sub-steps:

[0080] Step S51: Input the real-time acquired displacement-time data into the displacement-time curve feature recognition model, and extract the preset calibration features of the displacement-time curve according to the processing flow of the model to form real-time feature data;

[0081] Step S52: Convert the real-time feature data to meet the input requirements of the trained support parameter prediction model, ensuring that the data can be correctly input into the neural network;

[0082] Step S53: Input the real-time feature data after format conversion into the trained support parameter prediction model based on BP neural network. After the neural network calculates, the predicted roadway support parameters are obtained.

[0083] Step S54: Check the rationality of the predicted roadway support parameters. Based on the actual situation and experience of the roadway project, determine whether the predicted support parameters are within a reasonable range. If they are not within a reasonable range, make corresponding corrections.

[0084] like Figure 2 As shown, a method for determining the timing and parameters of roadway support based on displacement observation data is implemented through different units, including:

[0085] The data acquisition unit is used to acquire real-time displacement-time series data of the roadway surface displacement through displacement monitoring equipment;

[0086] The feature recognition unit, connected to the data acquisition unit, is used to process the acquired displacement-time series data using a displacement-time curve feature recognition model, and extract preset calibration features to form a feature dataset.

[0087] The status evaluation unit, connected to the feature recognition unit, is used to construct a roadway displacement status evaluation index system based on the feature dataset and calculate the comprehensive displacement status evaluation index value.

[0088] The model training unit, connected to the feature recognition unit, is used to establish a support parameter prediction model based on a BP neural network and to train it using a feature dataset.

[0089] The parameter prediction unit is connected to the feature recognition unit and the model training unit respectively, and is used to input real-time feature data into the trained support parameter prediction model to obtain the predicted roadway support parameters.

[0090] The support decision unit is connected to the state evaluation unit and the parameter prediction unit respectively. It is used to compare the comprehensive displacement state evaluation index value with the preset support trigger threshold. When the threshold is reached or exceeded, the predicted roadway support parameters are used for roadway support.

[0091] The method for determining the timing and parameters of roadway support based on displacement observation data, which deeply integrates displacement-time curve feature recognition and BP neural network technology, has shown significant advantages in the field of roadway support and effectively overcomes many drawbacks of the background technology.

[0092] To address the limitations of simplistic and crude data processing methods in the background technology, this method employs a displacement-time curve feature recognition model to comprehensively and deeply mine the potential information in displacement-time data. Unlike traditional methods that only focus on the magnitude of displacement values, this model emphasizes extracting key morphological features such as curve slope changes, inflection point locations, and curvature variation amplitudes. These features accurately depict the trends and abrupt changes in the deformation of the surrounding rock in the tunnel, leading to a more accurate understanding of the tunnel deformation patterns. For example, by identifying inflection point locations, it can keenly capture key nodes where the tunnel deformation trend changes, providing a scientific basis for judging the timing of support, avoiding support delays caused by untimely assessment of deformation trends, and significantly improving the timeliness and accuracy of support timing judgment.

[0093] In addressing the shortcomings of traditional support parameter prediction models that lack multi-factor consideration, a support parameter prediction model based on a backpropagation neural network plays a crucial role. This model introduces a correction coefficient for the surrounding rock grade, fully considering the influence of factors such as rock lithology and integrity on support parameters. Simultaneously, it models and analyzes the correlations between support parameters such as anchor bolt length and diameter, anchor cable spacing, and shotcrete thickness. This approach ensures that the predicted support parameters are no longer isolated but comprehensively consider various practical engineering factors and the interrelationships between parameters. Compared to the limitations of traditional models that focus on only a few factors, this method predicts support parameters that better align with actual engineering needs, ensuring roadway safety while avoiding resource waste caused by over-support, significantly improving the scientific rigor and practicality of support parameter prediction.

[0094] In summary, the method for determining the timing and parameters of roadway support based on displacement observation data, through data processing techniques and advanced prediction models, effectively overcomes the shortcomings of the prior art, achieves accurate determination of the timing and parameters of roadway support, significantly improves the scientificity and accuracy of roadway support decision-making, reduces engineering risks and costs, and provides strong technical support for the safe and efficient construction of roadway projects.

[0095] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0096] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for determining the timing and parameters of roadway support based on displacement observation data, characterized in that, Includes the following steps: Step S1: Real-time monitoring of the surface displacement of the roadway is carried out using displacement monitoring equipment to obtain continuous displacement-time series data and form an initial displacement-time dataset; Step S2: Using the displacement-time curve feature recognition model, process the initial displacement-time dataset, extract the preset calibration features of the displacement-time curve from the data, including curve slope change, inflection point position, curvature change amplitude feature parameters, and define the dataset including the preset calibration features as the feature dataset. Step S3: Based on the feature parameters in the feature dataset, construct a roadway displacement state evaluation index system, and calculate the comprehensive displacement state evaluation index value by allocating weights to each feature parameter. Step S4: Establish a support parameter prediction model based on a BP neural network. Use the feature dataset as input data and the roadway support parameters, including anchor bolt length, anchor bolt diameter, anchor cable spacing, and shotcrete thickness, as output data to train the BP neural network and obtain the trained support parameter prediction model. Step S5: Input the feature data formed by processing the real-time displacement-time data through the displacement-time curve feature recognition model into the trained support parameter prediction model to obtain the predicted roadway support parameters. Step S6: Compare the calculated comprehensive displacement state evaluation index value with the preset support trigger threshold. When the comprehensive displacement state evaluation index value reaches or exceeds the support trigger threshold, the predicted roadway support parameters are used to support the roadway.

2. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, In the displacement-time curve feature recognition model, the slope change feature of the displacement-time curve is quantified using the following formula: Where K is the quantized value of the slope change, di represents the displacement monitoring value at the i-th time point, and t i Let represent the i-th time point, and n be the total number of displacement-time data points; this quantization value is used to characterize the drastic change in the slope of the displacement-time curve, and the weight of the slope change quantization value in the feature dataset is ω. K ω K These are the weight coefficients determined using the analytic hierarchy process (AHP).

3. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, In the displacement-time curve feature recognition model, the inflection point position of the displacement-time curve is determined using the following formula: Where I is the inflection point index value, and sign is the sign function; the inflection point index value corresponds to the location point on the displacement-time curve where the slope change trend changes. The displacement value and time value at this location point are used as calibration features in the feature dataset, and the weight of this feature is ω when constructing the roadway displacement state evaluation index system. I ω I Determined by principal component analysis.

4. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, In the support parameter prediction model based on the BP neural network, a correction coefficient α for the surrounding rock grade of the roadway is introduced, and the calculation formula for the output layer of the BP neural network is adjusted as follows: Among them, P j For the j-th predicted roadway support parameters, w kj v represents the connection weight between the k-th neuron in the output layer and the j-th support parameter output node. ik x represents the connection weight between the i-th neuron in the hidden layer and the k-th neuron in the output layer. i is the i-th feature data of the input layer, f is the activation function, m is the number of neurons in the hidden layer, and l is the number of feature data in the input layer; the roadway surrounding rock grade correction coefficient α is determined by the lithology and integrity parameters of the roadway surrounding rock through a pre-established surrounding rock grade classification table, and is used to correct the support parameters predicted by the BP neural network.

5. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, When constructing the roadway displacement state evaluation index system, the comprehensive displacement state evaluation index value is calculated using the following formula: Where E is the comprehensive displacement state evaluation index value, ω j Let y be the weight of the j-th feature parameter. j Let y be the standardized value of the j-th feature parameter, and p be the total number of feature parameters; the standardized value y j It is obtained by mapping the values ​​of each feature parameter to the interval [0, 1], and the weight ω of each feature parameter is... j The comprehensive displacement state evaluation index value, determined by the entropy weight method, is used to characterize the comprehensive state of roadway displacement and serves as the basis for determining whether roadway support is necessary.

6. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, In the BP neural network-based support parameter prediction model, the correlation between roadway support parameters is modeled, and the correlation correction coefficient β is calculated using the following formula. ij : Where, β ij P is the correlation correction coefficient between the i-th support parameter and the j-th support parameter. is Let i be the value of the i-th support parameter in the s-th training data set. Let be the average value of the i-th support parameter in the training data, and q be the number of training data sets; the correlation correction coefficient is used to correct the correlation of the support parameters predicted by the BP neural network, so that the predicted support parameters conform to the mutual relationship in actual engineering, where i, j = 1, 2, ..., n, and n is the total number of roadway support parameters.

7. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, Step S3 includes the following sub-steps: Step S31: Based on the slope change, inflection point location, and curvature change amplitude characteristic parameters of the displacement-time curve, define corresponding sub-evaluation indicators. Each sub-evaluation indicator is used to reflect the change characteristics of roadway displacement from different perspectives. Step S32: The weight of each sub-evaluation index in the roadway displacement state evaluation index system is determined by using the analytic hierarchy process. This weight determination process is achieved by constructing a judgment matrix and calculating the eigenvectors and eigenvalues ​​of the matrix, which is used to quantify the importance of each sub-evaluation index to the comprehensive displacement state evaluation. Step S33: Perform dimensionless processing on each characteristic parameter, converting characteristic parameters with different dimensions into comparable values, so that each characteristic parameter has the same status when calculating the comprehensive displacement state evaluation index value; Step S34: Based on each sub-evaluation index and its weight, the dimensionless characteristic parameters are weighted and summed to obtain the comprehensive displacement state evaluation index value, which comprehensively reflects the overall displacement state of the roadway.

8. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, Step S4 includes the following sub-steps: Step S41: Determine the network structure of the BP neural network, including the number of input layer nodes, the number of hidden layers and the number of hidden layer nodes, and the number of output layer nodes. The number of input layer nodes corresponds to the number of feature parameters in the feature dataset, and the number of output layer nodes corresponds to the number of roadway support parameters. The number of hidden layers and nodes are determined by trial and error. Construct a neural network architecture for support parameter prediction. Step S42: Initialize the connection weights and thresholds of the BP neural network. Random numbers are used to assign values ​​to the connection weights and thresholds between neurons in the network to provide initial parameters for the training of the neural network. Step S43: Divide the feature dataset into a training set and a test set. Use the training set to train the BP neural network through forward and backward propagation. By adjusting the connection weights and thresholds, the output error of the neural network is gradually reduced until the preset training stopping condition is met. Step S44: Use a test set to verify the performance of the trained BP neural network, calculate the error between the output of the neural network on the test set and the actual roadway support parameters, and determine whether the prediction ability of the neural network meets the requirements.

9. The method for determining the timing and parameters of roadway support based on displacement observation data according to claim 1, characterized in that, Step S5 includes the following sub-steps: Step S51: Input the real-time acquired displacement-time data into the displacement-time curve feature recognition model, and extract the preset calibration features of the displacement-time curve according to the processing flow of the model to form real-time feature data; Step S52: Convert the real-time feature data to meet the input requirements of the trained support parameter prediction model, ensuring that the data can be correctly input into the neural network; Step S53: Input the real-time feature data after format conversion into the trained support parameter prediction model based on BP neural network. After the neural network calculates, the predicted roadway support parameters are obtained. Step S54: Check the rationality of the predicted roadway support parameters. Based on the actual situation and experience of the roadway project, determine whether the predicted support parameters are within a reasonable range. If they are not within a reasonable range, make corresponding corrections.

Images