A visual management method and system for a single-lane tunneling working face auxiliary transportation line

Abstract

This application relates to a visualization management method and system for auxiliary transportation lines in a single-lane tunneling face. The method includes: acquiring time-series data of equipment parameters and performing spatiotemporal alignment on the time-series data to obtain an environmental equipment data matrix; inputting the environmental equipment data matrix into a reliability assessment model to calculate the exploration accuracy confidence level and obtain a confidence score list; adjusting the operating parameters of the exploration equipment based on the confidence score list and the environmental equipment data matrix to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results; and rendering a basic digital twin tunnel model based on the real-time exploration results and the confidence score list to obtain a real-time integrated visualization interface. This method can perceive and quantify the real-time interference caused by extreme environmental factors on the accuracy of the detection equipment, thereby improving the reliability of visualization.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to a visualization management method and system for auxiliary transportation lines in a single-lane tunneling face. Background Technology

[0002] With the deepening development of the Industrial Internet of Things (IIoT) and digital twin technologies, intelligent mine operation support technologies characterized by intelligent sensing and adaptive control have emerged. These technologies emphasize real-time interaction and dynamic optimization of the production environment and equipment status, leading to the current methods of visual monitoring and proactive early warning management for critical underground transportation links. In traditional technical models, auxiliary transportation safety in narrow tunnels generally relies on fixed-location sensing devices and video surveillance systems for environmental and obstacle detection. The processing logic is relatively static, primarily presenting the collected distance signals or video images directly and issuing simple threshold alarms. Managers then rely on experience to assess risks and dispatch vehicles. However, this traditional monitoring method, dependent on fixed parameters and human experience, has significant limitations. It cannot perceive and quantify the real-time interference caused by extreme environmental factors such as dust and humidity on the accuracy of detection equipment. Therefore, when the environment deteriorates, the reliability of the detection results may quietly decrease or even fail, creating monitoring blind spots. Often, tracing can only be done after an accident occurs, making it difficult to implement pre-accident warnings and process interventions, leaving the transportation link constantly exposed to hidden risks caused by the environment. Summary of the Invention

[0003] Therefore, it is necessary to provide a visual management method and system for auxiliary transportation lines in single-lane tunneling faces that can sense and quantify the real-time interference caused by extreme environmental factors on the accuracy of detection equipment, in order to address the above-mentioned technical problems.

[0004] Firstly, this application provides a visual management method for auxiliary transportation routes in a single-lane tunneling face, including:

[0005] Acquire the timing data of device parameters and perform spatiotemporal alignment on the timing data of device parameters to obtain the environmental device data matrix;

[0006] The environmental equipment data matrix is ​​input into the reliability assessment model to calculate the detection accuracy confidence level and obtain a list of confidence scores.

[0007] Based on the confidence score list and the environmental equipment data matrix, the operating parameters of the exploration equipment are adjusted to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results.

[0008] Based on real-time exploration results and a list of confidence scores, the basic digital twin tunnel model is rendered to obtain a real-time integrated visualization interface.

[0009] Furthermore, the reliability assessment model was obtained through the following methods, including:

[0010] The historical data streams of the equipment are aggregated to obtain historical multi-source data; the historical data streams include environmental parameter time series data, raw output data of the exploration equipment, event logs, and baseline anchor point coordinates;

[0011] Based on historical multi-source data, a training set is constructed to obtain a historical training sample set, which is then divided into a training set, a test set, and a validation set.

[0012] Based on the training set and the test set, the reliability assessment model framework is trained using weighted multi-class log loss as the loss function to obtain a preliminary reliability level classification model.

[0013] Based on the validation set, the detection accuracy confidence level output by the preliminary reliability level classification model is probabilistically calibrated to obtain the reliability assessment model.

[0014] Furthermore, based on historical multi-source data, a training set is constructed to obtain a historical training sample set, including:

[0015] Feature extraction is performed on historical multi-source data to obtain a feature matrix, and consistency checks are performed on the feature matrices corresponding to multiple devices to obtain consistency labels;

[0016] The exploration data of the exploration equipment to the physical benchmark anchor point is extracted from historical multi-source data to obtain benchmark exploration data. Based on the benchmark exploration data and the coordinates of the benchmark anchor point, error analysis is performed to obtain high confidence labels.

[0017] Based on the event logs, the feature matrix of the corresponding event is marked as unreliable to obtain negative sample labels. Based on the feature matrix, feature analysis is performed to obtain health labels.

[0018] Based on consistency labels, high confidence labels, negative sample labels, and health labels, confidence fusion is performed to obtain the reliability level label vector and sample weight vector of the corresponding feature matrix.

[0019] By integrating the feature matrix, reliability level label vector, and sample weight vector, a historical training sample set is obtained.

[0020] Furthermore, based on the confidence score list and the environmental equipment data matrix, the operating parameters of the exploration equipment are adjusted to obtain an adaptive parameter adjustment instruction set, including:

[0021] By comparing the detection accuracy confidence level and the confidence threshold in the confidence score list, the target operating mode of the corresponding detection equipment is obtained; the target operating modes include normal monitoring mode, active optimization mode, and security enhancement mode.

[0022] If the target's operating mode is normal monitoring mode, then set the real-time operating parameters of the probe device to reserved parameters;

[0023] If the target working mode is active optimization mode, then query the adjustment baseline of the corresponding environmental equipment data matrix from the parameter optimization knowledge base, and superimpose the adjustment baseline and the dynamic adjustment amount to obtain the optimized adjustment parameters;

[0024] If the target operating mode is security-enhanced mode, then obtain the fixed strong configuration parameters;

[0025] By integrating retained parameters, optimized adjustment parameters, and fixed strong configuration parameters, an adaptive parameter adjustment instruction set is obtained.

[0026] Furthermore, the dynamic adjustment amount is calculated as the difference between the exploration accuracy confidence score and the target confidence score. The formula for the dynamic adjustment amount is:

[0027]

[0028] in, To dynamically adjust the amount, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, The sampling period is The control error at the current moment, Here, t represents the error integral term, and t is the current discrete time step. The integral anti-saturation weight, where i is the error index. This is the filtered error signal. This is the error signal from the previous moment.

[0029] Furthermore, the control error is calculated using the following formula:

[0030]

[0031] in, The control error at the current moment, The target confidence score, This represents the confidence level of the actual exploration accuracy at the current moment.

[0032] Furthermore, based on the real-time exploration results and confidence score list, the basic digital twin tunnel model is rendered to obtain a real-time integrated visualization interface, which also includes:

[0033] Based on the confidence score list and warning threshold, the warning level of the detection area corresponding to each detection device is determined, and an independent warning level list is obtained;

[0034] Based on the independent early warning level list and real-time transportation dynamic information, the transportation equipment and the investigation area are associated to obtain a risk association pair list;

[0035] Based on the risk association pair list, the corresponding early warning decisions are matched from the decision rule base to obtain the early warning decision list;

[0036] Based on the early warning decision list, early warning information is generated, which is used to instruct the corresponding transportation equipment to avoid risks.

[0037] Secondly, this application also provides a visual management system for auxiliary transportation lines in a single-lane tunneling face, including:

[0038] The alignment module is used to acquire device parameter timing data and perform spatiotemporal alignment on the device parameter timing data to obtain an environmental device data matrix.

[0039] The accuracy module is used to input the environmental equipment data matrix into the reliability assessment model, calculate the detection accuracy confidence level, and obtain a list of confidence scores.

[0040] The instruction module is used to adjust the operating parameters of the exploration equipment based on the confidence score list and the environmental equipment data matrix, and obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results;

[0041] The rendering module is used to render the basic digital twin tunnel model based on real-time exploration results and a list of confidence scores, resulting in a real-time integrated visualization interface.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the method provided in the first aspect of this application.

[0043] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of the method provided in the first aspect of this application.

[0044] The aforementioned visualization management method and system for auxiliary transportation lines in a single-lane tunneling face acquires time-series data of equipment parameters and performs spatiotemporal alignment on this data to obtain an environmental equipment data matrix. This matrix is ​​then input into a reliability assessment model to calculate the exploration accuracy confidence level, resulting in a confidence score list. Based on the confidence score list and the environmental equipment data matrix, the operating parameters of the exploration equipment are adjusted, resulting in an adaptive parameter adjustment instruction set. This set instructs the corresponding exploration equipment to provide real-time exploration results. Based on the real-time exploration results and the confidence score list, a basic digital twin tunnel model is rendered to obtain a real-time integrated visualization interface. By introducing a data-driven dynamic equipment reliability assessment model, real-time environmental data is transformed into a quantitative score for the detection accuracy of each exploration device, thereby enabling continuous diagnosis of the exploration equipment's status. Through an adaptive control algorithm, the core operating states of the exploration equipment, such as laser power and image processing parameters, are adjusted in real-time, allowing the equipment to proactively optimize performance in harsh environments to maintain effective detection capabilities, effectively improving the reliability of visualization. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the process of a visual management method for auxiliary transportation lines in a single-lane tunneling face, provided by an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the structure of a visual management system for an auxiliary transportation line in a single-lane tunneling face, provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] In one embodiment, such as Figure 1As shown, a visual management method for auxiliary transportation routes in a single-lane tunneling face is provided. This embodiment illustrates the method using a terminal as an example. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0050] Step 101: Obtain the timing data of the device parameters and perform spatiotemporal alignment on the timing data of the device parameters to obtain the environmental device data matrix.

[0051] Among them, equipment parameter time-series data refers to parameter data continuously collected from equipment, including environmental sensors, detection devices, or monitoring devices, arranged in chronological order. This data may include environmental parameters, equipment status, and detection outputs, used to describe the real-time operation of the equipment and changes in the surrounding environment. Spatiotemporal alignment is a data processing operation designed to synchronize and calibrate data from different devices, different time points, or different spatial locations, ensuring data consistency in timestamps and spatial coordinates, and eliminating data inconsistencies caused by device heterogeneity or acquisition delays. The environmental equipment data matrix is ​​a dataset organized in matrix form, where rows represent time points, columns represent different device or parameter dimensions, and matrix elements contain aligned parameter values, serving as a unified input for subsequent analysis and integrating multi-source data information.

[0052] The terminal simultaneously reads the latest uploaded device parameter time-series data from all devices in the network. At this time, the data is raw, unprocessed streaming data. Since there may be slight deviations in the clocks of each device or different sampling periods, a unified time axis is required. A reference time is selected, and through interpolation or resampling techniques, the data sequence of each device is mapped onto this unified time axis. This ensures that when analyzing the state at any instant, all device data used are from the same moment, avoiding misjudgments caused by time asynchrony. Each detection device has its specific installation position in the tunnel. Based on the spatial coordinates of the device, its measurement data is transformed into a common coordinate system. Optionally, a distance measurement data needs to be combined with the device's own position and converted into the absolute position of the target in the common coordinate system. The observation results of all devices are unified on the same map, so that the detection data of different devices on the same object can be mutually verified and integrated. After completing the spatiotemporal alignment, all aligned data are organized according to time points and data dimensions and filled into a two-dimensional matrix, finally generating an environmental device data matrix.

[0053] Step 102: Input the environmental equipment data matrix into the reliability assessment model, calculate the exploration accuracy confidence level, and obtain a list of confidence scores.

[0054] Specifically, the reliability assessment model is a pre-trained machine learning model used to evaluate the reliability of the output data from the exploration equipment. Trained on historical data, the model quantifies the exploration accuracy of the equipment in the current environment, providing an objective credibility metric. The exploration accuracy confidence score is a numerical score output by the model, representing the reliability of the measurement results of a specific exploration equipment at a given time. A high score indicates high reliability, while a low score suggests potential errors or anomalies. The confidence score list is a list or vector containing the confidence score for each exploration equipment, used for subsequent decision-making, such as parameter adjustment or early warning.

[0055] The terminal takes the environmental equipment data matrix as input and inputs it into the reliability assessment model. The model calculates the confidence score of each device through training capabilities, ensuring that the score reflects the true reliability and obtaining a structured list of confidence scores.

[0056] Step 103: Based on the confidence score list and the environmental equipment data matrix, adjust the operating parameters of the exploration equipment to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results.

[0057] Specifically, the operating parameters of a detection device refer to the dynamically adjustable settings of the device, such as sampling frequency, sensitivity threshold, or operating mode. These parameters affect the detection behavior and data quality of the device. An adaptive parameter adjustment instruction set is a collection of instructions or commands used to instruct a specific detection device on how to adjust its operating parameters. This instruction set is dynamically generated based on confidence levels and environmental data, aiming to optimize the device's detection capabilities.

[0058] The terminal reads the detection accuracy confidence score of each detection device in the confidence score list, compares each score with the preset confidence threshold, determines the target working mode for each detection device based on the comparison results, and adopts different strategies to generate new operating parameters based on the target working mode determined for each device. The parameters generated for all detection devices are summarized, and the parameter values ​​are encapsulated into instructions that the devices can recognize and execute, forming a unified adaptive parameter adjustment instruction set.

[0059] Step 104: Based on the real-time exploration results and confidence score list, render the basic digital twin tunnel model to obtain a real-time integrated visualization interface.

[0060] The real-time exploration results refer to the latest data fed back by the exploration equipment after receiving adaptive commands. This data reflects the true state of the current environment or equipment and is dynamically updated to ensure the timeliness of the visualized content. The basic digital twin tunnel model is a virtual 3D tunnel model that simulates the geometry and equipment layout of a physical single-tunnel excavation face. As the core of the digital twin, it provides the basic framework for rendering. The real-time integrated visualization interface is a graphical interface that dynamically displays the real-time status of the tunnel, equipment confidence levels, and early warning information. It is used for human-computer interaction to help intuitively grasp the overall situation.

[0061] The terminal associates the data from the real-time detection results with the corresponding virtual objects in the digital twin model. Based on the confidence score list, it sets visualization attributes for the elements in the model. For example, a rule can be defined: device icons with a confidence score higher than 0.8 are displayed in green, those between 0.5 and 0.8 are displayed in yellow, and those below 0.5 are displayed in red and flashing. The data values ​​themselves can also be displayed in the form of numerical labels or dashboards. According to the above binding and mapping rules, the entire scene is dynamically drawn, the position and state of objects in the model are adjusted according to real-time data, the color is changed according to the confidence score, low reliability areas are highlighted, data labels, trend curves, and early warning information are drawn on the model, and a complete and interactive real-time integrated visualization interface is output and displayed on the monitoring screen or the terminal's display module.

[0062] This embodiment provides a visualization management method for auxiliary transportation lines in a single-lane tunneling face. It acquires time-series data of equipment parameters and performs spatiotemporal alignment on this data to obtain an environmental equipment data matrix. This matrix is ​​then input into a reliability assessment model to calculate the exploration accuracy confidence level, resulting in a confidence score list. Based on the confidence score list and the environmental equipment data matrix, the operating parameters of the exploration equipment are adjusted to obtain an adaptive parameter adjustment instruction set. This instruction set instructs the corresponding exploration equipment to provide real-time exploration results. Based on the real-time exploration results and the confidence score list, a basic digital twin tunnel model is rendered to obtain a real-time integrated visualization interface. By introducing a data-driven dynamic equipment reliability assessment model, real-time environmental data is transformed into a quantitative score for the detection accuracy of each exploration device, thereby achieving continuous diagnosis of the exploration equipment's status. Through an adaptive control algorithm, the core operating states of the exploration equipment, such as laser power and image processing parameters, are adjusted in real-time, enabling the exploration equipment to proactively optimize performance in harsh environments to maintain effective detection capabilities, effectively improving the reliability of visualization.

[0063] In one embodiment, the reliability assessment model is obtained through the following methods:

[0064] Step 201: Summarize the historical data stream of the device to obtain historical multi-source data; the historical data stream includes environmental parameter time series data, original output data of the exploration device, event logs and reference anchor point coordinates.

[0065] Among them, historical data streams refer to the raw data sets continuously recorded from various devices over a period of time and arranged in chronological order, serving as the fundamental raw materials for building models. Historical multi-source data refers to a comprehensive dataset formed by aggregating historical data streams from different sources and types, reflecting the multi-source nature of the data and aiming to provide a comprehensive informational context for model training.

[0066] The terminal collects four types of historical data streams from databases or logs. These include environmental parameter time-series data, such as temperature and humidity data recorded by environmental sensors over time; raw output data from detection equipment, such as radar and camera data, which are unprocessed detection signals or raw readings; event logs, which include specific events such as equipment alarms, maintenance records, and the start / end of transportation tasks; and reference anchor point coordinates, which include coordinate data of precisely located physical reference points set in the tunnel. The different types and sources of data are correlated and packaged according to timelines and other criteria to form a structured historical multi-source data set.

[0067] Step 202: Based on historical multi-source data, construct a training set to obtain a historical training sample set, and divide the historical training sample set into a training set, a test set, and a validation set.

[0068] Specifically, the historical training sample set refers to a collection of samples constructed using historical multi-source data, directly used for training machine learning models. Each sample contains input features and corresponding labels. The training set is a subset of samples from the historical training sample set, specifically used for directly training the model and tuning its parameters. The validation set is another subset of samples from the historical training sample set, used to evaluate model performance during training to adjust hyperparameters or select the optimal model and prevent overfitting. The test set is a completely independent set of samples from the historical training sample set, used to evaluate the model's generalization ability after training and tuning.

[0069] Based on historical multi-source data, the terminal constructs labeled training samples through feature extraction, label allocation and other operations. All samples are merged into a historical training sample set. According to a predetermined ratio, optionally, 70% training set, 15% validation set and 15% test set are used. Using random sampling and other methods, the historical training sample set is divided into three non-overlapping parts: training set, validation set and test set.

[0070] Step 203: Based on the training set and the test set, the reliability assessment model framework is trained using weighted multi-class log loss as the loss function to obtain a preliminary reliability level classification model.

[0071] Specifically, weighted multi-class log loss is a loss function used to train classification models. It measures the difference between the probability distribution predicted by the model and the true labels. Weighting means that different importance can be assigned to errors in different classes, making it suitable for imbalanced datasets. A reliability assessment model framework refers to a machine learning model structure with undetermined parameters that has not yet been trained. A preliminary reliability rating classification model refers to a model with preliminary classification capabilities obtained after training the reliability assessment model framework on a training set using a specified loss function.

[0072] The terminal loads the reliability assessment model framework into the training environment and initializes its parameters. Training set data is input into the model framework in batches. The model performs forward propagation for each sample, outputs the probability distribution of the predicted reliability level, and calculates the weighted multi-class log loss value between the predicted result and the true label. The larger the loss value, the greater the prediction error. Backpropagation is performed using an optimization algorithm, and the model parameters are adjusted according to the loss value to minimize the loss function. This process is repeated until the model performance converges. During training, the model performance is periodically evaluated using a test set to check for overfitting. If overfitting occurs, the training strategy needs to be adjusted. When the training reaches the predetermined conditions, the final model parameters are saved, resulting in a preliminary reliability level classification model. This model has learned to predict the reliability level from the input data, but the output confidence level may not be accurate enough.

[0073] Step 204: Based on the validation set, perform probability calibration on the detection accuracy confidence level output by the preliminary reliability level classification model to obtain the reliability assessment model.

[0074] The validation set refers to a subset of data from the historical training sample set, independent of the training and test sets. It is not used for model training or testing but is specifically used for model tuning and evaluation, such as probability calibration. Exploration accuracy confidence refers to the probability value output by the preliminary reliability classification model for each prediction result, representing the model's confidence in the reliability of its predictions. However, without calibration, the confidence level may not reflect the true probability; for example, a model outputting a 90% confidence level might only have an actual accuracy of 70%. Probability calibration is a post-processing technique used to adjust the confidence level output by the model to better match the true probability distribution. The reliability assessment model refers to the preliminary reliability classification model after probability calibration. Its output exploration accuracy confidence level has higher credibility and interpretability and can be directly used for subsequent decision-making.

[0075] The terminal inputs validation set data into the preliminary reliability classification model, allowing the model to predict each sample and record the original detection accuracy confidence level output by the model. The confidence level predicted by the model is compared with the true label of the sample on the validation set to analyze the deviation between the confidence level and the true accuracy. Optionally, a reliability curve is plotted to visualize the relationship between confidence level and accuracy. Based on the deviation analysis, a calibration method is selected to fit a mapping function. Optionally, Platt scaling is used to map the original confidence level to the calibrated confidence level using logistic regression. The calibration function is trained using validation set data to adjust the original confidence level to a value closer to the true probability. The calibration function is integrated into the preliminary model to form a reliability assessment model. When new data is input, the model first outputs the original confidence level and then corrects it using the calibration function. The validation set or other data are used to check whether the calibrated confidence level is more accurate.

[0076] This embodiment effectively improves the prediction accuracy of the detection accuracy of the exploration equipment by introducing a deep learning model, thereby improving the reliability and relevance of the visualization.

[0077] In one embodiment, a training set is constructed based on historical multi-source data to obtain a historical training sample set, including:

[0078] Step 301: Extract features from historical multi-source data to obtain a feature matrix, and perform consistency checks on the feature matrices corresponding to multiple devices to obtain consistency labels.

[0079] Historical multi-source data refers to a comprehensive dataset obtained by aggregating historical data streams from devices. This includes time-series environmental parameter data, raw output data from the probe devices, event logs, and baseline anchor coordinates. It serves as the raw input for feature extraction, providing rich contextual information. Feature extraction is a data processing operation that extracts meaningful and quantifiable features from the raw historical multi-source data. These features better reflect the device's operating status and environmental impact. The feature matrix is ​​a structured dataset, typically organized in matrix form, with rows representing each sample and columns representing the extracted feature dimensions. This simplifies the raw data and facilitates model learning. Consistency testing is an evaluation operation used to compare the consistency of output data from multiple devices under the same or similar conditions. Optionally, it can determine whether the devices are reliable in their collaboration by calculating the correlation coefficient or variance between the data. The consistency label is a value representing the degree of consistency among the data from multiple devices. A high score indicates high consistency between devices, while a low score suggests potential deviations or anomalies.

[0080] The terminal extracts features from historical multi-source data, calculates statistical features such as mean and variance from environmental parameter time-series data, and extracts features such as distance and intensity from the original output data of the exploration equipment. The features are organized into a feature matrix, with each sample corresponding to one row. Consistency checks are performed on the feature matrices corresponding to multiple devices, comparing the feature values ​​extracted by different devices at the same time point. Through cluster analysis, the differences or similarities between features are calculated. If the device data is highly consistent, a high consistency label is assigned; otherwise, a low consistency label is assigned.

[0081] Step 302: Extract the exploration data of the exploration equipment on the physical reference anchor point from the historical multi-source data to obtain the reference exploration data. Based on the reference exploration data and the coordinates of the reference anchor point, perform error analysis to obtain high confidence labels.

[0082] Specifically, historical multi-source data refers to a comprehensive dataset obtained by aggregating historical data streams from equipment, including time-series environmental parameter data, raw output data from exploration equipment, event logs, and reference anchor point coordinates, serving as the raw input for feature extraction. Exploration data from exploration equipment to physical reference anchor points refers to the raw data obtained by exploration equipment, such as lidar or cameras, measuring known and fixed physical reference points in the tunnel, including distance or angle readings. Reference exploration data is a subset of the extracted data, specifically targeting the measurement results of reference anchor points, and is used as input for error analysis because the coordinates of the reference anchor points are known, providing a true value reference. Reference anchor point coordinates are the precise, known coordinates of the physical reference point in the tunnel space, serving as the ground truth value for error analysis. Error analysis is a comparative operation that calculates the difference between the reference exploration data and the reference anchor point coordinates to quantify the measurement deviation of the exploration equipment. A high confidence label is a label value used to identify the confidence level of a sample. The characteristic is that the label does not directly represent the magnitude of the error value, but rather indicates that the sample has high reference value. That is, since the coordinates of the reference anchor point are fixed, any error clearly reflects the inherent accuracy deviation of the exploration equipment in the current environment. Therefore, such samples can accurately characterize the performance of the equipment and have a high confidence level.

[0083] The terminal filters all measurement records of each physical benchmark anchor point from historical multi-source data to obtain benchmark exploration data. Using database query methods, the data is filtered by equipment number and anchor point number to ensure that only exploration data related to the benchmark anchor point is retained. Error analysis is performed by comparing each benchmark exploration data with the corresponding benchmark anchor point coordinates to calculate the error value. Statistical methods are used to analyze the deviation pattern, but the focus is not on the error value itself, but on the clarity of the error. Based on the error analysis results, a high-confidence label is assigned to each sample. As long as the sample successfully performs an error comparison, regardless of the error size, a high-confidence label is assigned because the sample provides direct evidence of equipment accuracy deviation. A list of high-confidence labels is generated, corresponding to the samples in the feature matrix.

[0084] Step 303: Based on the event log, mark the feature matrix of the corresponding event as unreliable to obtain negative sample labels, and perform feature analysis based on the feature matrix to obtain health labels.

[0085] Specifically, event logs refer to historical event data, such as equipment failures, alarms, or maintenance records, providing key time-point information to identify abnormal situations. Negative sample labels are labels generated based on event logs, indicating that a specific sample is unreliable. Optionally, if the event log records a failure of a probed device at a certain point in time, the corresponding feature matrix is ​​marked as a negative sample. Feature analysis is a data mining operation that analyzes equipment status trends from the feature matrix, such as using machine learning methods or statistical indicators to assess equipment health. A health label is a value representing the operational health status of the equipment; a high score indicates normal operation, while a low score suggests potential degradation or abnormality.

[0086] The terminal scans event records in the logs, finds the feature matrix at the corresponding time point, marks these matrices as unreliable, generates negative sample labels, ensures that the training set contains clear counterexamples, performs feature analysis based on the feature matrix, performs clustering or regression analysis on the features, calculates the device health index, and assigns health labels based on the degree to which the features deviate from the normal range.

[0087] Step 304: Based on the consistency label, high confidence label, negative sample label and health label, perform confidence fusion to obtain the reliability level label vector and sample weight vector of the corresponding feature matrix.

[0088] Here, consistency labels represent labels indicating data consistency across multiple devices. High-confidence labels represent labels indicating samples with high reference value. Negative sample labels are unreliable sample labels based on event logs. Health labels are device health status labels obtained based on feature analysis. Confidence fusion is an integration operation that combines multiple label sources to generate a unified reliability assessment, aiming to resolve label conflicts and balance the contributions of each source. The reliability level label vector is a vector representing the overall reliability level of each sample, with vector elements corresponding to the probability distribution of different reliability categories. The sample weight vector is a list of weight values ​​representing the importance of each sample during training; samples with higher weights have a greater impact on the model.

[0089] The terminal collects all label sources, including consistency labels, high-confidence labels, negative sample labels, and health labels, ensuring a one-to-one correspondence between labels and samples in the feature matrix. Confidence fusion is performed using either a weighted average or model fusion method. For example, the weighted average method assigns weights to each label type, with higher weights for high-confidence labels and lower weights for negative sample labels. A comprehensive reliability score is calculated for each sample. Based on the fusion result, a reliability level label vector is generated for each sample, with the probability value obtained by normalizing the fusion score. Weights are assigned based on label consistency or uncertainty, and the variance of the label source is calculated. If multiple labels conflict, the weight of that sample is reduced, or information entropy is used; the higher the entropy, the lower the weight. Entropy is calculated based on the distribution of the reliability level vector, resulting in a reliability level label vector and a sample weight vector, both of which match the sample dimension of the feature matrix.

[0090] Step 305: Integrate the feature matrix, reliability level label vector, and sample weight vector to obtain the historical training sample set.

[0091] The feature matrix is ​​the data matrix after feature extraction, where rows represent samples and columns represent feature dimensions. The reliability level label vector is the reliability label vector for each sample. The sample weight vector is the weight value for each sample. The historical training sample set is a complete training dataset, where each sample includes a feature vector, a reliability level label vector, and sample weights, and is directly used for model training.

[0092] The terminal ensures that the feature matrix, reliability level label vector, and sample weight vector are aligned in the sample dimension. It uses an index matching method to ensure that the features, labels, and weights of the i-th sample correspond to the same data point. It employs structured data integration, combining each row of the feature matrix (i.e., the feature vector of a sample) with the corresponding reliability level label vector and sample weight value to form a complete sample unit. It uses data encapsulation techniques, such as creating a list or dictionary structure, where each element contains a feature vector, label vector, and weight value, to organize all sample units into a standard format for easy reading by machine learning libraries. It checks whether the integrated dataset is complete, verifies the consistency of the sample number and the absence of missing values ​​through statistical methods, and generates a historical training sample set that can be directly input into the model training process.

[0093] This embodiment generates a batch of well-labeled training samples, providing the reliability assessment model with the most direct and reliable learning material on equipment accuracy deviations, greatly enhancing the accuracy and interpretability of model evaluation, and thus improving the reliability of visualization.

[0094] In one embodiment, based on the confidence score list and the environmental device data matrix, the operating parameters of the detection device are adjusted to obtain an adaptive parameter adjustment instruction set, including:

[0095] Step 401: Compare the detection accuracy confidence level and confidence threshold in the confidence score list to obtain the target working mode of the corresponding detection device; the target working mode includes normal monitoring mode, active optimization mode and security enhancement mode.

[0096] The detection accuracy confidence score is an output of the reliability assessment model; it's a numerical value representing the reliability assessment of the current measurement accuracy of a detection device. A higher value indicates higher reliability. The confidence threshold is one or two preset critical values ​​used to divide consecutive confidence scores into different decision intervals, typically with a lower threshold and an upper threshold. The target operating mode is an operational strategy assigned to the detection device based on the confidence score comparison results; it's a decision label, not a specific parameter. The normal monitoring mode is suitable for stable conditions with high confidence, aiming to maintain the status quo and operate energy-efficiently. The proactive optimization mode is suitable for situations with moderate confidence but room for improvement, aiming to fine-tune parameters to improve performance or efficiency. The safety enhancement mode is suitable for abnormal or high-risk situations with low confidence, aiming to prioritize safety and adopt conservative strategies.

[0097] The terminal reads the detection accuracy confidence score of each detection device from the confidence score list and compares it with a preset confidence score threshold. This is a rule-based judgment process. Optionally, if the detection accuracy confidence score is greater than the upper threshold, the device is determined to be reliable and its target working mode is set to normal monitoring mode. If the detection accuracy confidence score is between the upper and lower thresholds, the device is determined to have room for optimization and its target working mode is set to active optimization mode. If the detection accuracy confidence score is lower than the lower threshold, the device is determined to be unreliable and poses a risk, and its target working mode is set to security enhancement mode.

[0098] Step 402: If the target working mode is normal monitoring mode, then set the real-time operating parameters of the probe device to reserved parameters.

[0099] Specifically, real-time operating parameters refer to the set of configuration parameters that the device is currently using, such as sampling frequency, power, and sensitivity. Reserved parameters refer to the decision made in normal monitoring mode not to make any changes to the device parameters, but to continue using its current real-time operating parameters.

[0100] When the terminal determines that the target operating mode of a device is normal monitoring mode, it generates an instruction that explicitly sets the device's reserved parameters to its current real-time operating parameters, meaning that the device will continue to operate using the existing parameters in the next control cycle.

[0101] Step 403, or if the target working mode is active optimization mode, then query the adjustment baseline of the corresponding environmental equipment data matrix from the parameter optimization knowledge base, and superimpose the adjustment baseline and the dynamic adjustment amount to obtain the optimization adjustment parameters.

[0102] Specifically, the parameter optimization knowledge base is a database storing various historical optimal parameters, recording the best parameter benchmarks that the probe device can adopt under different environmental conditions. The adjustment baseline is a set of basic optimized parameter values ​​retrieved from the parameter optimization knowledge base; it represents the recommended configuration for the current environmental conditions. The dynamic adjustment amount is a relatively small parameter correction value, usually calculated by a more refined algorithm based on recent errors or trends, used for fine-tuning the adjustment baseline. The optimized adjustment parameters are the set of parameters obtained by superimposing the adjustment baseline and the dynamic adjustment amount, aiming to further improve the device's performance on its current good foundation.

[0103] The terminal uses the current environmental equipment data matrix, i.e. the current environmental and equipment status snapshot, as the query condition to retrieve the matching or most similar scenario from the parameter optimization knowledge base and obtain the corresponding adjustment baseline parameters. The queried adjustment baseline is superimposed with the dynamically calculated adjustment amount in real time and a simple arithmetic sum is performed to obtain a new set of optimization adjustment parameters.

[0104] Step 404, or if the target operating mode is security-enhanced mode, then obtain the fixed strong configuration parameters.

[0105] Among them, the fixed strong configuration parameters are a set of preset, very conservative but extremely reliable parameter configurations. Their design goal is to maximize the robustness and security of the device, even at the cost of sacrificing some performance or energy consumption.

[0106] When the device is determined to be in enhanced security mode, the terminal directly calls the fixed enhanced configuration parameters prepared for this type of device from the preset parameter configuration table. Without complicated calculations, it can quickly switch to the safest operating mode when it is deemed unreliable or high-risk. This can minimize the possibility of false alarms and missed alarms, prevent security incidents caused by device perception errors, and prioritize security.

[0107] Step 405: Integrate the retained parameters, optimized adjustment parameters, and fixed strong configuration parameters to obtain the adaptive parameter adjustment instruction set.

[0108] The adaptive parameter adjustment instruction set is a collection of multiple instructions, each for a specific exploration device, explicitly indicating the operating parameters it should use in the next cycle.

[0109] The terminal iterates through all the probe devices, summarizes the parameters determined by each device according to its target working mode, encapsulates these parameters into an instruction format that the device can recognize and execute, and combines the instructions of all devices into a complete adaptive parameter adjustment instruction set.

[0110] This embodiment improves detection accuracy or efficiency by restoring equipment parameters to an optimal state and making adaptive fine-tuning, enabling reliability assessment and intelligent decision-making to actually apply to physical equipment, forming a complete perception-decision closed loop, and realizing adaptive optimization of transportation route safety monitoring visualization.

[0111] In one embodiment, the dynamic adjustment amount is calculated as the difference between the detection accuracy confidence score and the target confidence score, and the formula for the dynamic adjustment amount is:

[0112]

[0113] in, To dynamically adjust the amount, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, The sampling period is The control error at the current moment, Here, t represents the error integral term, and t is the current discrete time step. The integral anti-saturation weight, where i is the error index. This is the filtered error signal. This is the error signal from the previous moment.

[0114] Specifically, the dynamic adjustment amount is a parameter correction value that needs to be superimposed on the adjustment baseline in active optimization mode. It is a value calculated by the control algorithm and is used to finely and adaptively adjust the equipment parameters, so that the operating parameters of the probe equipment can not only adapt to the environment, but also be dynamically optimized according to real-time performance errors, thereby more accurately approaching the target performance.

[0115] Control error is the difference between the target confidence score and the actual exploration accuracy confidence score at the current moment. It is the core input of the entire control algorithm and reflects the gap between the current performance and the expected performance. A positive value indicates that the actual confidence score is lower than the target, and the equipment performance needs to be improved; a negative value indicates that the actual confidence score is higher than the target, and the equipment load can be appropriately reduced.

[0116] The proportional gain coefficient is an adjustable constant used to amplify the control effect of the proportional term. It determines the strength of the response to the current error. The larger the value, the faster the correction speed of the current error. However, too large a value may lead to over-response or even instability.

[0117] The integral gain coefficient is an adjustable constant used to amplify the control action of the integral term. It determines the strength of the correction for the accumulation of historical errors and is used to eliminate steady-state errors. As long as the error exists, the integral term will continue to accumulate, thereby generating a sufficiently large control quantity to eventually eliminate the error.

[0118] The differential gain coefficient is an adjustable constant used to amplify the control effect of the differential term, determine the prediction of the error change trend and the damping strength, reduce the overshoot of the system by suppressing the rate of error change, improve stability, and make the response smoother.

[0119] The sampling period is the time interval between two consecutive calculations. It is a fixed time value that determines the sampling frequency of process variables and the speed of control updates.

[0120] The integral term of the error is the sum of all control errors from the initial time to the current time, representing the overall cumulative effect of historical errors and forming the basis for the integral term calculation. The integral anti-saturation weight is a weighting factor that varies between 0 and 1, used to assign different importance to different historical errors when summing the integral term. The older the error, the lower the weight, preventing integral saturation. When the error cannot be eliminated due to some limitation, the integral term will continue to accumulate to a very large value. Once the error reverses, it takes a long time to exit the saturation state. Adding weights can weaken the influence of older errors, avoid excessive expansion of the integral term, and improve response speed. The filtered error signal is the signal after the original control error has been processed by a digital filter. High-frequency, rapid fluctuations or noise in the original error signal are filtered out. The derivative term is very sensitive to noise; directly using the difference of the original error will amplify the noise, causing severe jitter in the control quantity. Using the filtered error for differentiation allows the derivative term to truly reflect the trend of error change, rather than noise.

[0121] At each sampling moment, the terminal reads the current actual detection accuracy confidence level, calculates the control error at the current moment, inputs the current error into a digital filter to obtain a smoothed error signal, and simultaneously saves the filtering error at the previous moment. It calculates a proportional term to provide an immediate and proportional correction force for the current error, calculates an integral term to calculate the weighted sum of historical errors (the weights can be dynamically adjusted according to the error magnitude, time distance, or whether saturation is present to eliminate continuous steady-state deviations), calculates a differential term to calculate the rate of change of the filtered error to predict the future trend of the error, playing a damping role and suppressing overshoot. The three outputs are added together to obtain the dynamic adjustment amount, which is then output and usually limited to prevent excessive single adjustment.

[0122] This embodiment ensures a rapid response to current performance deviations through a proportional term, while the integral term gradually accumulates and eliminates small, persistent errors that cannot be completely corrected by the proportional term alone, ultimately stabilizing the actual confidence level precisely at the target value. The derivative term senses the trend of error changes and applies a reverse pull when overshoot and oscillation may occur due to excessively fast response, making the entire parameter adjustment process smoother and convergent faster. Filtering the error before differentiation effectively suppresses the interference of measurement noise on control and avoids unnecessary and severe parameter fluctuations. Introducing an integral anti-saturation weight avoids the delayed response problem caused by the ineffective accumulation of the integral term when the equipment capacity has reached its limit, thus improving control quality.

[0123] In one embodiment, the control error is calculated using the following formula:

[0124]

[0125] in, The control error at the current moment, The target confidence score, This represents the confidence level of the actual exploration accuracy at the current moment.

[0126] Specifically, when e(t) > 0, it indicates that the actual confidence level is lower than the target, and the equipment parameters need to be adjusted positively to improve its performance. When e(t) < 0, it indicates that the actual confidence level is higher than the target, and the parameters can be adjusted negatively to optimize resources. The error value e(t) will be used as the input of the PID controller to calculate the proportional, integral, and derivative terms, ultimately generating the dynamic adjustment quantity u(t), achieving fine control of the equipment parameters.

[0127] In one embodiment, after rendering the basic digital twin tunnel model based on real-time exploration results and a confidence score list to obtain a real-time integrated visualization interface, the method further includes:

[0128] Step 701: Based on the confidence score list and the warning threshold, determine the warning level of the exploration area corresponding to each exploration device to obtain an independent warning level list.

[0129] The confidence score list is a list containing the detection accuracy confidence score calculated by the reliability assessment model for each detection device, reflecting the current reliability of the data from each device. The warning threshold is one or more preset confidence cutoff values ​​used to map continuous confidence scores to discrete warning levels. The detection area refers to the physical spatial range of the transportation route monitored by a single detection device. The warning level represents the risk level label for a specific detection area. The independent warning level list is a list where each entry specifies the current warning level for a particular detection device and its corresponding detection area; in this case, the risk assessment is conducted independently for each area.

[0130] The terminal reads the confidence score list to obtain the current confidence level of each detection device. It compares the confidence level of each device with the preset warning threshold, which is a rule-based judgment process. For example: if the confidence level is greater than the attention threshold, the warning level is normal; if the danger threshold is not less than the confidence level but less than the attention threshold, the warning level is attention; if the confidence level is less than the danger threshold, the warning level is danger. A warning level is assigned to each detection device-detection area pair, forming an independent warning level list.

[0131] Step 702: Based on the independent early warning level list and real-time transportation dynamic information, the transportation equipment and the investigation area are associated to obtain a risk association pair list.

[0132] Specifically, real-time transportation dynamic information refers to the real-time status data of mobile units in a transportation system, mainly including the identity, real-time location, operating speed, and planned route of the transportation equipment. Transportation equipment is a mobile unit that performs material transportation tasks on a transportation route. The risk association pair list is a list where each entry is an association pair in the form of (transportation equipment, exploration area, warning level), clearly indicating which equipment is about to enter or is currently in a risky area, and what the risk level is.

[0133] The terminal acquires real-time information such as the location and trajectory of all transportation equipment, analyzes the real-time location and preset route of each transportation equipment, predicts which exploration areas it will enter in the near future, and binds the warning level of the exploration area that the equipment is about to enter to the equipment. Only when the warning level is not normal will an association pair be generated. All abnormal association pairs are summarized to form a risk association pair list.

[0134] Step 703: Based on the risk association pair list, match the corresponding early warning decisions from the decision rule base to obtain the early warning decision list.

[0135] Specifically, the decision rule base is a database that stores logical rules, defining standardized response strategies that the system should adopt under different risk scenarios. Early warning decisions are specific operational instructions generated from the rule base for specific risk pairs. The early warning decision list is a list where each entry specifies which risk pair needs to be addressed, i.e., which device should execute what specific early warning decision.

[0136] The terminal iterates through each association pair in the risk association pair list. For each association pair, it performs a query and matching in the decision rule base, generates one or more specific early warning decisions for each successfully matched risk association pair, and summarizes all generated decisions into an early warning decision list.

[0137] Step 704: Based on the early warning decision list, generate early warning information. The early warning information is used to instruct the corresponding transportation equipment to avoid risks.

[0138] Warning information refers to packaged data packets or instructions that can be understood and executed by the transportation equipment or its control system. Risk avoidance refers to the purpose of warning information, namely, to avoid potential accidents by changing the behavior of the transportation equipment.

[0139] Based on the warning decision list, the terminal converts each decision into a specific control command or prompt message, including: command type, such as speed control, lighting control, or communication notification; target device, specifying the number of one or more transport devices receiving the command; command parameters, such as target speed value, stopping time, etc.; and validity period, indicating the validity period of the command. The encapsulated warning information is sent to the corresponding target transport device in real time and accurately through the vehicle-to-ground communication system. After receiving the warning information, the transport device's onboard control system automatically executes the command to complete the risk avoidance action.

[0140] This embodiment ensures that the results of all previous analyses, judgments, and decisions can be applied to the transportation equipment, directly guiding it to take evasive actions, thereby proactively and effectively preventing safety accidents and greatly improving the safety of the auxiliary transportation line of the entire single-lane tunneling face.

[0141] In one embodiment, if the data stream from the detection device indicates the presence of an obstacle in the tunnel, the obstacle is directly visualized in the digital twin model, and an early warning message is issued.

[0142] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0143] Based on the same inventive concept, this application also provides a visualization management system for auxiliary transportation lines in single-lane tunneling faces, used to implement the visualization management method for auxiliary transportation lines in single-lane tunneling faces described above. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the visualization management system for auxiliary transportation lines in single-lane tunneling faces provided below can be found in the limitations of the visualization management method for auxiliary transportation lines in single-lane tunneling faces described above, and will not be repeated here.

[0144] In one exemplary embodiment, such as Figure 2 As shown, a visual management system 800 for auxiliary transportation routes in a single-lane tunneling face is provided, including:

[0145] Alignment module 801 is used to acquire device parameter timing data and perform spatiotemporal alignment on the device parameter timing data to obtain an environmental device data matrix;

[0146] The accuracy module 802 is used to input the environmental equipment data matrix into the reliability assessment model, calculate the detection accuracy confidence level, and obtain a list of confidence scores.

[0147] The instruction module 803 is used to adjust the operating parameters of the exploration equipment based on the confidence score list and the environmental equipment data matrix to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results.

[0148] The rendering module 804 is used to render the basic digital twin tunnel model based on real-time exploration results and a list of confidence scores, so as to obtain a real-time integrated visualization interface.

[0149] Furthermore, the system also includes a training module for:

[0150] The historical data streams of the equipment are aggregated to obtain historical multi-source data; the historical data streams include environmental parameter time series data, raw output data of the exploration equipment, event logs, and baseline anchor point coordinates;

[0151] Based on historical multi-source data, a training set is constructed to obtain a historical training sample set, which is then divided into a training set, a test set, and a validation set.

[0152] Based on the training set and the test set, the reliability assessment model framework is trained using weighted multi-class log loss as the loss function to obtain a preliminary reliability level classification model.

[0153] Based on the validation set, the detection accuracy confidence level output by the preliminary reliability level classification model is probabilistically calibrated to obtain the reliability assessment model.

[0154] Furthermore, the training module is also used for:

[0155] Feature extraction is performed on historical multi-source data to obtain a feature matrix, and consistency checks are performed on the feature matrices corresponding to multiple devices to obtain consistency labels;

[0156] The exploration data of the exploration equipment to the physical benchmark anchor point is extracted from historical multi-source data to obtain benchmark exploration data. Based on the benchmark exploration data and the coordinates of the benchmark anchor point, error analysis is performed to obtain high confidence labels.

[0157] Based on the event logs, the feature matrix of the corresponding event is marked as unreliable to obtain negative sample labels. Based on the feature matrix, feature analysis is performed to obtain health labels.

[0158] Based on consistency labels, high confidence labels, negative sample labels, and health labels, confidence fusion is performed to obtain the reliability level label vector and sample weight vector of the corresponding feature matrix.

[0159] By integrating the feature matrix, reliability level label vector, and sample weight vector, a historical training sample set is obtained.

[0160] Furthermore, instruction module 803 is also used for:

[0161] By comparing the detection accuracy confidence level and the confidence threshold in the confidence score list, the target operating mode of the corresponding detection equipment is obtained; the target operating modes include normal monitoring mode, active optimization mode, and security enhancement mode.

[0162] If the target's operating mode is normal monitoring mode, then set the real-time operating parameters of the probe device to reserved parameters;

[0163] If the target working mode is active optimization mode, then query the adjustment baseline of the corresponding environmental equipment data matrix from the parameter optimization knowledge base, and superimpose the adjustment baseline and the dynamic adjustment amount to obtain the optimized adjustment parameters;

[0164] If the target operating mode is security-enhanced mode, then obtain the fixed strong configuration parameters;

[0165] By integrating retained parameters, optimized adjustment parameters, and fixed strong configuration parameters, an adaptive parameter adjustment instruction set is obtained.

[0166] Furthermore, the dynamic adjustment amount is calculated as the difference between the exploration accuracy confidence score and the target confidence score. The formula for the dynamic adjustment amount is:

[0167]

[0168] in, To dynamically adjust the amount, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, The sampling period is The control error at the current moment, Here, t represents the error integral term, and t is the current discrete time step. The integral anti-saturation weight, where i is the error index. This is the filtered error signal. This is the error signal from the previous moment.

[0169] Furthermore, the control error is calculated using the following formula:

[0170]

[0171] in, The control error at the current moment, The target confidence score, This represents the confidence level of the actual exploration accuracy at the current moment.

[0172] Furthermore, the system also includes an early warning module for:

[0173] Based on the confidence score list and warning threshold, the warning level of the detection area corresponding to each detection device is determined, and an independent warning level list is obtained;

[0174] Based on the independent early warning level list and real-time transportation dynamic information, the transportation equipment and the investigation area are associated to obtain a risk association pair list;

[0175] Based on the risk association pair list, the corresponding early warning decisions are matched from the decision rule base to obtain the early warning decision list;

[0176] Based on the early warning decision list, early warning information is generated, which is used to instruct the corresponding transportation equipment to avoid risks.

[0177] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a visual management method for auxiliary transportation lines in a single-lane tunneling face as described above.

[0178] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0179] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0180] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A visual management method for auxiliary transportation lines in a single-lane tunneling face, characterized in that, The method includes: Acquire device parameter time-series data and perform spatiotemporal alignment on the device parameter time-series data to obtain an environmental device data matrix; The environmental equipment data matrix is ​​input into the reliability assessment model to calculate the detection accuracy confidence level and obtain a list of confidence scores. Based on the confidence score list and the environmental equipment data matrix, the operating parameters of the exploration equipment are adjusted to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results. Based on the real-time exploration results and the confidence score list, the basic digital twin tunnel model is rendered to obtain a real-time integrated visualization interface.

2. The method according to claim 1, characterized in that, The reliability assessment model was obtained through the following methods, including: The historical data streams of the device are aggregated to obtain historical multi-source data; the historical data streams include environmental parameter time-series data, original output data of the exploration device, event logs, and reference anchor point coordinates; Based on the historical multi-source data, a training set is constructed to obtain a historical training sample set, and the historical training sample set is divided into a training set, a test set, and a validation set. Based on the training set and the test set, the reliability assessment model framework is trained using weighted multi-class log loss as the loss function to obtain a preliminary reliability level classification model. Based on the validation set, the detection accuracy confidence level output by the preliminary reliability level classification model is probabilistically calibrated to obtain the reliability assessment model.

3. The method according to claim 2, characterized in that, The process of constructing a training set based on the historical multi-source data to obtain a historical training sample set includes: Feature extraction is performed on the historical multi-source data to obtain a feature matrix, and consistency checks are performed on the feature matrices corresponding to multiple devices to obtain consistency labels; The exploration data of the exploration device on the physical reference anchor point is extracted from the historical multi-source data to obtain the reference exploration data. Based on the reference exploration data and the coordinates of the reference anchor point, error analysis is performed to obtain high confidence labels. Based on the event log, the feature matrix of the corresponding event is marked as unreliable to obtain negative sample labels, and feature analysis is performed based on the feature matrix to obtain health labels; Based on the consistency label, the high confidence label, the negative sample label, and the health label, confidence fusion is performed to obtain the reliability level label vector and sample weight vector corresponding to the feature matrix; The historical training sample set is obtained by integrating the feature matrix, the reliability level label vector, and the sample weight vector.

4. The method according to claim 1, characterized in that, The step of adjusting the operating parameters of the exploration equipment based on the confidence score list and the environmental equipment data matrix to obtain an adaptive parameter adjustment instruction set includes: The detection accuracy confidence level and confidence threshold in the confidence score list are compared to obtain the target working mode corresponding to the detection device; the target working mode includes normal monitoring mode, active optimization mode and security enhancement mode; If the target working mode is the normal monitoring mode, then the real-time operating parameters of the detection device are set as reserved parameters; Alternatively, if the target working mode is the active optimization mode, then the adjustment baseline corresponding to the environmental equipment data matrix is ​​queried from the parameter optimization knowledge base, and the adjustment baseline and the dynamic adjustment amount are superimposed to obtain the optimized adjustment parameters; Alternatively, if the target operating mode is the security enhancement mode, then obtain the fixed strong configuration parameters; By integrating the reserved parameters, the optimized adjustment parameters, and the fixed strong configuration parameters, the adaptive parameter adjustment instruction set is obtained.

5. The method according to claim 4, characterized in that, The dynamic adjustment amount is calculated based on the difference between the detection accuracy confidence score and the target confidence score, and the formula for the dynamic adjustment amount is: in, To dynamically adjust the amount, This is the proportional gain coefficient. This is the integral gain coefficient. The differential gain coefficient, The sampling period is The control error at the current moment, Here, t represents the error integral term, and t is the current discrete time step. The integral anti-saturation weight, where i is the error index. This is the filtered error signal. This is the error signal from the previous moment.

6. The method according to claim 5, characterized in that, The control error is calculated using the following formula: in, The control error at the current moment, The target confidence score, This represents the confidence level of the actual exploration accuracy at the current moment.

7. The method according to claim 1, characterized in that, After rendering the basic digital twin tunnel model based on the real-time exploration results and the confidence score list to obtain a real-time integrated visualization interface, the process further includes: Based on the confidence score list and the warning threshold, the warning level of the exploration area corresponding to each exploration device is determined to obtain an independent warning level list; Based on the independent early warning level list and real-time transportation dynamic information, the transportation equipment and the exploration area are associated to obtain a risk association pair list; Based on the risk association pair list, the corresponding early warning decisions are matched from the decision rule base to obtain the early warning decision list; Based on the aforementioned warning decision list, warning information is generated, which is used to instruct the corresponding transportation equipment to avoid risks.

8. A visual management system for auxiliary transportation lines in a single-lane tunneling face, characterized in that, The system includes: The alignment module is used to acquire device parameter timing data and perform spatiotemporal alignment on the device parameter timing data to obtain an environmental device data matrix. The accuracy module is used to input the environmental equipment data matrix into the reliability assessment model, calculate the exploration accuracy confidence level, and obtain a list of confidence scores. The instruction module is used to adjust the operating parameters of the exploration equipment based on the confidence score list and the environmental equipment data matrix to obtain an adaptive parameter adjustment instruction set; the adaptive parameter adjustment instruction set is used to instruct the corresponding exploration equipment to provide real-time exploration results. The rendering module is used to render the basic digital twin tunnel model based on the real-time exploration results and the confidence score list to obtain a real-time integrated visualization interface.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images