An internet-based tunneling engineering blasting effect evaluation and analysis method

Abstract

The application discloses a tunnel mining engineering blasting effect evaluation analysis method based on the Internet, relates to the technical field of tunnel mining engineering, and comprises the following steps: synchronously fusing a multimodal feature set to generate a multidimensional blasting feature representation; performing multiscale time series decomposition and spectrum energy clustering on the multidimensional blasting feature representation to generate a blasting response mode, and performing statistics and spectrum analysis to generate a blasting effect evaluation index set; comparing the blasting effect evaluation index set with a standard template in a historical blasting engineering database to generate a deviation benchmark, and calculating a current blasting effect deviation degree by using a Bayesian dynamic inference method to generate a deviation vector; performing dynamic weight adjustment and feature contribution correction on the deviation vector to generate a comprehensive blasting effect score result; and analyzing and summarizing the comprehensive blasting effect score result to generate a blasting effect evaluation report. The application improves the blasting quality control and safety early warning reliability.

Description

Technical Field

[0001] This invention relates to the field of tunnel mining engineering technology, and in particular to an internet-based method for evaluating and analyzing the blasting effects in tunnel mining. Background Technology

[0002] Blasting operations in tunnel mining are a critical link affecting construction safety and efficiency, and their effectiveness directly impacts surrounding rock stability, over- and under-excavation control, and the quality of subsequent lining construction. With the development of sensor networks, geological exploration, and digital construction technologies, the monitoring and analysis of blasting effects has gradually shifted from manual experience-based assessment to data-driven quantitative analysis. Currently, common blasting effect evaluation techniques typically rely on single data sources such as on-site vibration monitoring, acoustic emission detection, video recognition, or ground-penetrating radar inversion, and statistically analyze blasting effects using indicators such as energy distribution, peak vibration velocity, and fragment size.

[0003] While existing blasting effect evaluation technologies can quantify some indicators, they still have limitations in multi-source data collaborative analysis and dynamic assessment. Existing methods often rely on fixed evaluation models or static weighting systems, making it difficult to adaptively adjust the contribution of each indicator based on changes in real-time monitoring data. Furthermore, due to the significant temporal and spectral characteristics of the blasting process, traditional methods struggle to extract implicit response patterns from the multi-scale time-frequency domain, resulting in an insufficient understanding of blasting energy transfer and surrounding rock response mechanisms. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an Internet-based method for evaluating and analyzing the blasting effect in tunnel mining projects, which solves the problems of insufficient multimodal data fusion and low real-time performance of blasting effect evaluation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an internet-based method for evaluating and analyzing the blasting effects in tunnel mining projects, comprising: Collect real-time monitoring data from multiple sources, perform preprocessing, and generate a multimodal feature set; Simultaneous fusion of multimodal feature sets generates a multidimensional explosive feature representation; The multi-dimensional blasting feature representation is decomposed into multi-scale time series and clustered with spectral energy to generate blasting response patterns. Statistical and spectral analysis is then performed to generate a set of blasting effect evaluation indicators. The blasting effect evaluation index set is compared with the standard template in the historical blasting project database to generate a deviation benchmark. The Bayesian dynamic inference method is then used to calculate the current blasting effect deviation and generate a deviation vector. Dynamic weight adjustment and feature contribution correction are applied to the deviation vector to generate a comprehensive blasting effect score. The comprehensive blasting effect score results are analyzed and summarized to generate a blasting effect evaluation report.

[0007] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method described in this invention, the multi-source real-time monitoring data includes vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data.

[0008] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating the multimodal feature set are as follows: Denoising, time synchronization, and event clipping are performed on multi-source real-time monitoring data to generate multi-source, multi-scale time series segments. Physical constraint feature extraction, time-frequency analysis, and visual texture feature extraction are performed on multi-source, multi-scale time-series segments to generate candidate multimodal feature vector sets; Sparse encoding, dimensionality reduction, and interpretability enhancement are performed on the candidate multimodal feature vector groups to generate a multimodal feature set.

[0009] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating multi-dimensional blasting feature representation are as follows: The multimodal feature set is initially sorted and aligned according to event ID and timestamp to generate an ordered multimodal feature list; The ordered multimodal feature list is normalized according to modality category and labeled with confidence weights to generate a multimodal feature set with confidence labels. Local weighted fusion is then performed according to time windows to generate a preliminary fused feature representation. Cross-modal attention adjustment is performed on the preliminary fusion feature representation to generate attention-weighted fusion features, and dimensionality reduction and structured encoding are performed to generate multidimensional burst feature representations.

[0010] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating the blasting response mode are as follows: The multidimensional blasting feature representation is divided into several fixed-length time windows according to the time series, generating a multi-scale time period set; Multi-scale time series decomposition is performed on the multi-scale time period set to generate a multi-scale time-frequency feature matrix; By using a multi-scale time-frequency feature matrix, the multi-scale time period set is labeled with categories to form clustered feature groups; Identify recurring high-energy patterns and anomalous energy patterns from clustered feature groups to generate blasting response patterns.

[0011] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating the blasting effect evaluation index set are as follows: Time-domain statistical analysis and multi-scale time-frequency decomposition of the blasting response mode are performed to generate multi-dimensional time-frequency statistical features. By utilizing multidimensional time-frequency statistical features, energy zoning damage indicators, fragmentation and block size distribution, and flyrock risk are calculated and correlated with the surrounding rock structural characteristics to generate a safety risk probability vector, which is then integrated to form a set of physical and risk features. A set of evaluation indicators for blasting effects is generated by performing Bayesian fusion and confidence weighting on the set of physical and risk characteristics.

[0012] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the standard template in the historical blasting project database is obtained by statistical analysis, energy distribution feature extraction and safety risk assessment normalization of the blasting effect evaluation indicators of previous tunnel mining projects.

[0013] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating the deviation vector are as follows: The blasting effect evaluation index set is compared with the standard template in the historical blasting project database, and the deviation of each index is calculated to generate a deviation benchmark. Based on the deviation benchmark, the deviations of each indicator are weighted by probability to form a weighted deviation distribution; By using the weighted deviation distribution, the comprehensive deviation of each indicator is calculated and a deviation vector is generated.

[0014] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating a comprehensive blasting effect score are as follows: The deviation vector is standardized and initial weights are calculated to generate a set of normalized deviation components. The normalized bias component set is corrected for feature contribution according to the causal relationship between indicators, and abnormal bias is suppressed by robust constraints to generate a weighted bias vector. The weighted bias vector is subjected to multi-scale weighted aggregation and time window smoothing calculation to generate a comprehensive bias score sequence. The comprehensive deviation score sequence is normalized and mapped and the confidence level is corrected to generate a comprehensive blasting effect score result.

[0015] As a preferred embodiment of the Internet-based tunnel mining blasting effect evaluation and analysis method of the present invention, the specific steps for generating the blasting effect evaluation report are as follows: The comprehensive blasting effect score results are decomposed into sub-items by time and batch, and trend analysis, anomaly identification and causal clue extraction are performed on the sub-items to generate a score sequence. The risk levels of each item are calculated using the scoring sequence, and then classified, summarized, and correlated to generate multi-audience handling suggestions and visual summaries, which are then integrated to form a set of decision response features. The system automatically writes, formats, and archives the report text for the decision response feature set, generating a blasting effect evaluation report.

[0016] The beneficial effects of this invention are as follows: by performing multi-scale temporal decomposition and spectral energy clustering on the fused multimodal feature data, hierarchical identification and energy partitioning analysis of the blasting energy release process at different time scales are realized, thereby extracting blasting response patterns with statistical stability, improving the accuracy of energy behavior identification and the interpretability of pattern representation, and enhancing the reliability of blasting quality control and safety early warning. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for evaluating and analyzing the blasting effects in internet-based tunnel mining projects.

[0019] Figure 2 A flowchart for generating multimodal feature sets.

[0020] Figure 3 A flowchart for generating multidimensional blasting feature representations.

[0021] Figure 4 A flowchart generated based on the comprehensive blasting effect scoring results. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides an internet-based method for evaluating and analyzing the blasting effects in tunnel mining, comprising the following steps: S1. Collect real-time monitoring data from multiple sources, perform preprocessing, and generate a multimodal feature set.

[0026] S1.1 Multi-source real-time monitoring data includes vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data.

[0027] Specifically, vibration sensors, acoustic acquisition sensors, image acquisition sensors, surrounding rock displacement sensors, and construction parameter acquisition sensors are deployed at the tunnel mining site. Before blasting operations, the time synchronization of each acquisition sensor is calibrated to ensure that the time reference of various monitoring signals is consistent. During blasting operations, vibration sensors collect vibration waveform data, acoustic acquisition sensors synchronously collect acoustic data, image acquisition sensors continuously collect image data of the blasting area, surrounding rock displacement sensors record surrounding rock displacement data in real time, and construction parameter acquisition sensors record construction parameter data such as drilling depth, charge amount, detonation sequence, and delay. The vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data are timestamped and formatted to generate multi-source real-time monitoring data.

[0028] S1.2. Denoise, synchronize time, and prune events from multi-source real-time monitoring data to generate multi-source, multi-scale time-series segments.

[0029] Specifically, high-pass filters, low-pass filters, or wavelet denoising methods are used to remove measurement noise and environmental interference from the vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data of the multi-source real-time monitoring data. Time synchronization of the vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data is performed using a unified time reference, for example, linear interpolation alignment is used for time errors within ±1 millisecond range. Based on the initiation time and duration of the blasting event, the synchronized vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data are segmented into several time periods corresponding to the blasting events, generating multi-source, multi-scale time-series segments. Each time period contains timestamp-aligned vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data for feature extraction.

[0030] S1.3 Perform physical constraint feature extraction, time-frequency analysis and visual texture feature extraction on multi-source and multi-scale time segments to generate candidate multimodal feature vector groups.

[0031] Specifically, acceleration, velocity, displacement, and energy release characteristics are calculated for multi-source, multi-scale time-series vibration waveform data and surrounding rock displacement data according to physical constraints. Instantaneous energy and momentum characteristics are obtained through integration and differentiation to form physical constraint characteristics. Short-time Fourier transform is performed on vibration waveform data, acoustic data, and surrounding rock displacement data for time-frequency analysis to extract frequency peaks, frequency band energy distribution, and spectral density characteristics within each time window, generating time-frequency characteristics. Local binary mode method is used to perform texture analysis on image data to extract texture directionality, roughness, and edge density characteristics, generating visual texture characteristics. Physical constraint characteristics, time-frequency characteristics, and visual texture characteristics are integrated according to timestamps and event IDs to generate candidate multimodal feature vector groups.

[0032] It should also be noted that physical constraint relationships refer to the relationships that limit and calculate the range of changes and correlation laws of data such as vibration waveforms, displacements, and energy based on the mechanical laws and rock mass motion characteristics during blasting. Among them, the rock mass motion characteristics specifically include the elastic deformation range, plastic deformation behavior, crack propagation speed, block displacement amplitude, local vibration frequency response, stress wave propagation speed, and damping characteristics of the rock mass under blasting action.

[0033] S1.4 Perform sparse coding, dimensionality reduction and interpretability enhancement on the candidate multimodal feature vector group to generate a multimodal feature set.

[0034] Specifically, a sparse representation method is used to retain the main feature components and remove redundant components from the candidate multimodal feature vector group to generate a sparse coding vector. The singular value decomposition method is used to reduce the dimensionality of the sparse coding vector to the example dimension to reduce data redundancy, resulting in a dimensionality-reduced feature vector. Through interpretability enhancement processing, the dimensionality-reduced feature vector is normalized and structured to make each dimensionality-reduced feature vector statistically meaningful, generating a multimodal feature set.

[0035] It should also be noted that the example dimension refers to the number of principal components retained after dimensionality reduction, which is used to retain the main feature information while reducing the data dimensionality; the statistical significance refers to the data distribution pattern or numerical characteristics reflected by the feature vector after dimensionality reduction, such as mean, variance, kurtosis or energy distribution, which can be used to quantify and compare the patterns of different observation data. Sparse representation is a linear algebra technique that selects only a small number of basis vectors from an overcomplete dictionary to represent the original vector, thereby preserving the main feature components and removing redundant information, achieving efficient feature compression and data representation. Singular value decomposition decomposes the original matrix into three matrices, including an eigenvector matrix, a diagonal singular value matrix, and another eigenvector matrix, extracting the main feature directions, thereby achieving data dimensionality reduction.

[0036] S2. Synchronously fuse the multimodal feature sets to generate a multidimensional explosive feature representation.

[0037] S2.1. Perform preliminary sorting and alignment of the multimodal feature set according to event ID and timestamp to generate an ordered multimodal feature list.

[0038] Specifically, the multimodal feature set is grouped according to event ID. The multimodal features within each event are sorted by timestamp from smallest to largest, and feature points with abnormal or missing timestamps are marked. Missing timestamps are filled using nearest neighbor interpolation, and abnormal timestamps are smoothed to ensure the continuity of the time series. After correcting the time series, the multimodal features within each event are arranged in chronological order to form an ordered sequence, and the event sequences are uniformly aligned to ensure that the multimodal features across events are comparable in the time dimension. The sequences are then merged to generate an ordered list of multimodal features.

[0039] S2.2 Normalize the ordered multimodal feature list according to modality category and label the confidence weight to generate a multimodal feature set with confidence label, and perform local weighted fusion according to time window to generate a preliminary fused feature representation.

[0040] Specifically, the ordered multimodal feature list is grouped according to modality category. For each modality, a normalization method is used to map the values ​​of the multimodal features to a uniform range, such as normalizing the amplitude to between 0 and 1. Confidence weights are calculated based on the feature signal-to-noise ratio and labeled for each feature, generating a multimodal feature set with confidence labels. This set is then divided according to time windows, for example, 0.5 seconds per window. The multimodal features within each time window are averaged according to their confidence weights to generate a locally fused feature vector. This process is repeated for all time windows, and the locally weighted fusion results are arranged in chronological order to form a preliminary fused feature representation.

[0041] S2.3. Perform cross-modal attention adjustment on the preliminary fusion feature representation to generate attention-weighted fusion features, and perform dimensionality reduction and structured encoding to generate multidimensional burst feature representation.

[0042] Specifically, a feature weight matrix is ​​constructed for each multimodal feature in the preliminary fusion feature representation according to modality category. Attention weights are assigned based on the correlation between features between modalities, and highly correlated modal features are given higher attention weights to generate attention-weighted fusion features. Principal component analysis is used to reduce the dimensionality of the attention-weighted fusion features, reducing the feature dimension to the example dimension to reduce redundancy. The dimensionality-reduced multimodal features are then structured and encoded, arranged according to time series and modality category to generate multidimensional burst feature representations.

[0043] S3. The multi-dimensional blasting feature representation is decomposed into multi-scale temporal sequence and clustered with spectral energy to generate blasting response patterns. Statistical and spectral analysis is then performed to generate a set of blasting effect evaluation indicators.

[0044] S3.1 Divide the multidimensional blasting feature representation into several fixed-length time windows according to the time series to generate a multi-scale time period set.

[0045] Specifically, the multidimensional blasting feature representation is sorted according to timestamp order and divided into fixed-length time windows, such as using data every second or every five seconds as a time window. For each time window, a feature subset corresponding to the time period is extracted from the multidimensional blasting feature representation to generate a preliminary time period set. The sliding window processing is repeated on the entire time series, and time windows of different lengths can be used to generate multi-scale time period sets to cover short-term impact response and long-term energy changes, thereby realizing hierarchical temporal slicing of the multidimensional blasting feature representation and forming a multi-scale time period set.

[0046] S3.2 Perform multi-scale time series decomposition on the multi-scale time period set to generate a multi-scale time-frequency feature matrix.

[0047] Specifically, for each time period in the multi-scale time period set, the wavelet transform method is used to calculate the time-frequency representation, obtaining the energy distribution matrix of each time period at different frequencies; the time-frequency representations of each time period are arranged in the order of time windows to form a preliminary time-frequency matrix; time-frequency calculations are performed on the time period sets with different time window lengths to generate the corresponding multi-scale time-frequency feature matrices, and each multi-scale time-frequency feature matrix is ​​normalized to maintain numerical consistency, thereby completing the multi-scale time series decomposition of the multi-scale time period set and obtaining the multi-scale time-frequency feature matrix.

[0048] S3.3. Using a multi-scale time-frequency feature matrix, the multi-scale time period set is labeled with categories to form cluster feature groups.

[0049] Specifically, the Fast Fourier Transform (FFT) method is used to calculate the spectral energy distribution characteristics of each time period vector in the multi-scale time-frequency feature matrix and construct a set of feature vectors. The hierarchical clustering method is used to divide the time periods into several categories according to energy similarity, including high-energy, medium-energy, and low-energy categories. Each cluster category is labeled, and time periods belonging to the same category are grouped into the same cluster feature group. The cluster feature groups are organized in chronological order, and the category identifier and corresponding time period information are recorded for each group to form a cluster feature group.

[0050] It should also be noted that hierarchical clustering is an unsupervised clustering method that organizes data into a tree-like hierarchical structure based on similarity by progressively merging or splitting samples.

[0051] S3.4 Identify recurring high-energy patterns and anomalous energy patterns from cluster feature groups to generate blasting response patterns.

[0052] Specifically, the frequency and position of each category in the cluster feature group on the time axis are statistically analyzed, and the recurring categories are selected as high-frequency patterns. Anomalous energy candidates are marked according to their distance from the cluster center. The time windows of adjacent or nearby high-frequency and anomalous candidates are combined to form pattern segments. The representative time-frequency characteristics and statistical descriptive quantities of each pattern segment are calculated, including spectral peak frequency, peak energy, duration and energy ratio. Finally, all pattern segments are output according to event ID and timestamp to generate a set of blasting response patterns.

[0053] It should also be noted that the frequency of occurrence of each category in the cluster feature group recorded on the time axis is statistically analyzed to generate a category frequency table; a frequency threshold is set, and categories with an occurrence frequency higher than the frequency threshold are marked as high-frequency candidates; high-frequency candidate categories are arranged in chronological order, and the occurrence of the same type in adjacent time windows is merged to eliminate duplicate calculations; the merged and filtered category set is output as the high-frequency pattern.

[0054] The specific steps for setting the frequency threshold are as follows: count the number of times each category appears in the cluster feature group, accumulate the frequency of each category by time period or sample batch, analyze the frequency distribution characteristics to determine the boundary of high-frequency mode, determine the range of frequency threshold value based on historical blasting data and actual observation data, for example, set it to 2 to 5 times to distinguish between repetitive mode and occasional mode, and mark the category that exceeds the frequency threshold as high-frequency mode.

[0055] S3.5 Perform time-domain statistical analysis and multi-scale time-frequency decomposition on the blasting response mode to generate multi-dimensional time-frequency statistical features.

[0056] Specifically, time-domain statistics are extracted from the time-series signal of each blasting response mode, including peak particle velocity, root mean square, pulse duration, energy accumulation curve, and pulse rise slope, and the time-domain statistical results are output. Short-time Fourier transform and discrete wavelet transform are performed on the time-series signal of each blasting response mode to obtain the short-time spectrum and multi-scale time-frequency representation, and the time-frequency matrix is ​​output. Spectral features are calculated on the time-frequency matrix, including peak frequency, peak amplitude, spectral energy centroid, band energy ratio, peak kurtosis, and spectral entropy. The spectral features and time-domain statistical results are then concatenated to form the original time-frequency feature vector. The original time-frequency feature vector is normalized and standardized, and aggregated by event ID and time window to generate multi-dimensional time-frequency statistical features.

[0057] S3.6. Utilize multidimensional time-frequency statistical features to calculate energy zoning damage indicators, fragmentation and block size distribution, and flyrock risk. Corresponding to the surrounding rock structural characteristics, generate a safety risk probability vector and integrate it to form a set of physical and risk features.

[0058] Specifically, the multidimensional time-frequency statistical characteristics are divided into several energy zones according to energy thresholds, and the peak energy, average energy, and energy accumulation in each zone are calculated to generate energy zone damage indicators. According to the fragmentation statistical method, the size of the fragmented blocks in the high-energy zone is measured and statistically analyzed to calculate the fragmentation and block diameter distribution. At the same time, the flyrock risk is estimated based on the energy distribution and flyrock ejection pattern, and the flyrock occurrence probability is correlated with the surrounding rock displacement and vulnerability data to generate a safety risk probability vector. The energy zone damage indicators, fragmentation and block diameter distribution, and safety risk probability vector are integrated according to event ID and time window to form a set of physical and risk characteristics.

[0059] It should also be noted that when setting energy thresholds, the energy distribution of multidimensional time-frequency statistical characteristics within each time window is statistically analyzed, and the overall range and fluctuation of peak energy, average energy, and energy accumulation are analyzed. The zoning boundaries are determined based on historical blasting data and safety standards. The value range can be based on the minimum to maximum energy of the observed data. For example, the energy can be divided into a low energy zone (0-0.3), a high energy zone (0.3-0.7), and an extremely high energy zone (0.7-1) to facilitate the subsequent calculation of peak energy, average energy, and energy accumulation within each zone. Fragmentation statistics is a method that quantifies the degree of rock fragmentation and the distribution characteristics of rock fragments by measuring their size, counting their number, and analyzing their distribution. Flying stone projection law refers to the law that the initial velocity, projection angle, and landing position of flying stones change with the amount of explosive, rock structure, and blasting method during blasting, and is used to predict the trajectory and landing point distribution of flying stones.

[0060] S3.7 Perform Bayesian fusion and confidence weighting on the set of physical and risk characteristics to generate a set of evaluation indicators for blasting effects.

[0061] Specifically, the physical and risk feature set is split into several sub-feature vectors according to the indicator category, and the prior probability distribution is calculated based on the confidence value of each sub-feature vector; Bayesian inference is used to fuse the probabilities of each sub-feature vector to calculate the posterior probability distribution; each indicator is assigned a weighted value according to the posterior probability distribution to generate a weighted feature vector; all weighted feature vectors are integrated according to the indicator category to generate a set of blasting effect evaluation indicators.

[0062] It should be noted that the energy change patterns at various time scales are obtained through multi-scale time series decomposition. Then, the energy density and spectral peak characteristics are adaptively clustered using the spectral energy clustering method to automatically extract blasting response modes with statistical stability. A multi-dimensional blasting effect evaluation index set is generated through statistical and spectral analysis. This achieves time-frequency decoupling and energy partitioning identification of blasting signals, enabling multi-source data such as vibration, acoustics, and surrounding rock response to be transformed into physically interpretable blasting behavior characteristics, thereby improving the accuracy and dynamic adaptability of blasting effect evaluation.

[0063] It should be noted that after generating the set of evaluation indicators for blasting effects, dynamic deviation analysis is performed based on the confidence level of each indicator. The current embodiment uses a weighted and normalized approach to perform the analysis, thereby effectively connecting the Bayesian fusion results to the calculation of sub-risk levels, the generation of disposal suggestions, and the processing of visual summaries.

[0064] S4. Compare the blasting effect evaluation index set with the standard template in the historical blasting project database to generate a deviation benchmark, and use the Bayesian dynamic inference method to calculate the current blasting effect deviation and generate a deviation vector.

[0065] S4.1 The standard template in the historical blasting project database is obtained by statistical analysis, energy distribution feature extraction and safety risk assessment normalization of the blasting effect evaluation indicators of previous tunnel mining projects.

[0066] Specifically, the evaluation indicators of blasting effects from previous tunnel mining projects are collected and cleaned. The indicators are standardized according to unified units and names, outliers are removed, and the normalized indicators are subjected to spectral energy feature extraction and statistical calculation, including mean, variance and reference limits. The indicators are then grouped and statistically analyzed according to project type or geological category. The normalized mean, variance and reference limits of each blasting effect evaluation indicator are organized into a standard template in the historical blasting project database.

[0067] S4.2 Compare the blasting effect evaluation index set with the standard templates in the historical blasting project database item by item according to index category, calculate the deviation of each index, and generate the deviation benchmark.

[0068] Specifically, the units of each indicator in the standard template of the blasting effect evaluation index set and the historical blasting project database are standardized and normalized; corresponding indicators are matched one by one according to the indicator category, such as matching vibration energy index, acoustic energy index, and surrounding rock displacement index; for each indicator, the deviation benchmark is calculated, and the expression is: ; in, Indicates the first Deviation of the indicator This indicates the first of the evaluation indicators for blasting effects. The value of each indicator This indicates the standard template in the historical blasting project database that corresponds to the first... The value of each indicator An index representing the category of the indicator; All indicator deviations are summarized by indicator category to generate a deviation benchmark.

[0069] S4.3. Based on the deviation benchmark, the deviations of each indicator are weighted by probability to form a weighted deviation distribution.

[0070] Specifically, based on the deviation benchmark, the deviation of each indicator is adjusted according to the corresponding confidence weight to obtain a weighted deviation value; the weighted deviations of all indicators are normalized so that the sum reflects the overall deviation distribution; the normalized weighted deviation values ​​are arranged according to the indicator category to form a weighted deviation distribution.

[0071] S4.4. Using the weighted deviation distribution, summarize and calculate the comprehensive deviation of each indicator to generate a deviation vector.

[0072] Specifically, based on the weighted deviation distribution, the weighted deviation values ​​are summed according to the indicator category to obtain the comprehensive deviation degree of each indicator. The summed comprehensive deviation degree is then normalized, and the normalized value of each indicator is used as a component of the deviation vector. During the processing, normalization intervals are set for each indicator, for example, the normalization intervals for vibration energy, acoustic energy, surrounding rock displacement, and flyrock risk are 0-1. The comprehensive deviation degrees of each indicator are arranged and combined according to the indicator category to generate the deviation vector.

[0073] S5. Perform dynamic weight adjustment and feature contribution correction on the deviation vector to generate a comprehensive blasting effect score.

[0074] S5.1 Standardize the deviation vector and calculate the initial weights to generate a normalized set of deviation components.

[0075] Specifically, based on the deviation vector, the mean normalization and variance standardization processes are performed on each indicator component to ensure that each indicator component is consistent with the mean and standard deviation, thereby adjusting the standardization scale and obtaining the standardized components. The standardized components are then calculated proportionally according to the indicator category, and the initial weight of each indicator is determined by the proportion of each component in the sum, so that the sum of all initial weights is 1. During the processing, standardized means and standard deviations can be set separately for example indicators. For example, the mean of the vibration energy indicator is 0.5 and the standard deviation is 0.1, the mean of the acoustic energy indicator is 0.3 and the standard deviation is 0.05, the mean of the surrounding rock displacement indicator is 0.2 and the standard deviation is 0.05, and the mean of the flyrock risk indicator is 0.1 and the standard deviation is 0.02. The standardized components and their corresponding initial weights are combined to generate a set of normalized deviation components.

[0076] S5.2 Correct the characteristic contribution of the normalized deviation component set according to the causal relationship between the indicators, and suppress abnormal deviations through robustness constraints to generate a weighted deviation vector.

[0077] Specifically, a causal matrix is ​​established by normalizing the set of deviation components according to the causal relationship between the indicators. The values ​​of each indicator component are proportionally adjusted through the matrix to match the contribution with the relevant indicators. Robust constraints are applied to the adjusted components, such as limiting the deviation amplitude to the example range, compressing or smoothing indicators that exceed the example range, and combining the indicator components in the original order to generate a weighted deviation vector.

[0078] It should also be noted that the causal relationship between indicators refers to the statistical relationship in the analysis of blasting effects in tunnel mining engineering, where the change of one indicator has a temporal sequence and dependence on the change of another indicator. That is, the change of the former can predict or explain the trend of the latter to a certain extent. For example, abnormal fluctuations in the blasting vibration velocity index may cause changes in the surrounding rock fissure ratio index, indicating that the blasting vibration velocity has a causal influence on the surrounding rock fissure ratio. The causal relationship can be determined through time series analysis of historical monitoring data. Granger causality tests or partial correlation analysis are often used to quantify the intensity of this influence in order to reflect the direction and degree of interaction between the indicators. Example ranges are usually determined based on statistical analysis of historical blasting projects. For example, the energy zoning damage index range can be 0.0–1.0, the fragmentation and block size distribution index range can be 0–100, the flyrock risk index range can be 0–1, and the surrounding rock vulnerability index range can be 0–1. The specific values ​​are determined based on different project types and statistical samples.

[0079] S5.3 Perform multi-scale weighted aggregation and time window smoothing calculation on the weighted bias vector to generate a comprehensive bias score sequence.

[0080] Specifically, within short-scale (e.g., 1 second), medium-scale (e.g., 5 seconds), and long-scale (e.g., 30 seconds) time windows, the components are weighted and summed according to the initial weights of the indicators to obtain the scale aggregation sequences for each scale. The aggregation sequences for each scale are normalized to eliminate the dimensional differences between scales, and the normalized sequences are linearly combined according to the preset scale weights to generate multi-scale aggregation sequences. The multi-scale aggregation sequences are smoothed using time windows, such as using a sliding window average or an exponentially weighted moving average (example window length is 3 to 5 time windows) on the multi-scale aggregation sequences, and can be supplemented with median filtering or Gaussian filtering to suppress peak anomalies, generating a comprehensive deviation score sequence.

[0081] It should also be explained that the specific steps for setting the scale weights are as follows: determine the initial weight ratio based on the relative importance of each scale to the overall deviation representation, for example, the short scale can be set to 0.3, the medium scale to 0.5, and the long scale to 0.2; normalize the weights of each scale to ensure that the sum of the weights of each scale is 1; when generating the multi-scale aggregation sequence, linearly combine the normalized weights with the normalized aggregation sequence of the corresponding scale to form the final weighted sequence.

[0082] S5.4 Normalize and map the comprehensive deviation score sequence and correct the confidence level to generate the comprehensive blasting effect score result.

[0083] Specifically, the comprehensive deviation score sequence is normalized and mapped according to a predetermined score range. The minimum-maximum normalization method is used to convert the comprehensive deviation score sequence into a normalized score sequence, for example, between 0 and 100 points. Based on the confidence level of each item and the historical fluctuation variance corresponding to each time point in the normalized score sequence, a confidence factor sequence is calculated. For example, when the variance is small and the confidence level of the item is high, the confidence factor value is close to 1. The normalized score sequence and the confidence factor sequence are weighted and fused so that the time points with higher confidence levels have a greater weight in the final score, generating a comprehensive blasting effect score result.

[0084] S6. Analyze and summarize the comprehensive blasting effect scoring results, and generate a blasting effect evaluation report.

[0085] S6.1 Decompose the comprehensive blasting effect score results into sub-item tables according to time and batch, and perform trend analysis, anomaly identification and causal clue extraction on the sub-item tables to generate a score sequence.

[0086] Specifically, the comprehensive blasting effect score is decomposed into sub-item tables by time and batch. Sub-item scores are extracted for each batch based on blasting batch and timestamp index to form the sub-item table output. Trend analysis is performed on the sub-item tables. For each sub-item sequence, a sliding window moving average and linear regression are applied to calculate the trend slope, and a sub-item sequence with the trend slope is output. Anomaly identification is performed on the sub-item sequences with trend slopes. Z-score detection and cumulative sum test are used to identify abrupt change points and label abnormal time windows, outputting sub-item sequences with anomaly labels. For sub-item sequences with anomaly labels, time lag correlation coefficients and Granger causality tests are used to test the precedence-lag relationship between sub-item sequences. The causal relationships are summarized to form causal clues, and a score sequence containing the trend slope, anomaly labels, and causal clues is output.

[0087] S6.2 Calculate the risk level of each item using the scoring sequence, and perform classification summary and correlation analysis to generate multi-audience handling suggestions and visual summaries, and integrate them to form a set of decision response features.

[0088] Specifically, the risk probability distribution density is calculated based on the time series value of each item in the scoring sequence, and the scoring values ​​are mapped to risk level labels according to the preset risk threshold, generating a risk label table containing the risk levels of each item. Using the risk label table as input, and combining the risk concern elements and handling authority mapping table for different audience categories, multi-audience handling suggestion texts are generated according to the risk level classification rules. For example, when the risk level is high, the suggestion "immediately stop blasting and verify the charge parameters" is generated, and when the risk level is medium, the suggestion "adjust the borehole spacing and re-measure the vibration parameters" is generated. Based on the risk label table and multi-audience handling suggestions, a visual summary is constructed, and a visual summary chart is generated through color grading, time line, and blasting batch correlation diagram. The risk label table, multi-audience handling suggestions, and visual summary chart are integrated by time index and blasting batch index to form a decision response feature set.

[0089] It should also be explained that the specific steps for setting the risk boundary threshold are as follows: Based on the distribution of the comprehensive blasting effect score results in the historical blasting data sample, the minimum, maximum and percentile values ​​of the score are statistically analyzed, and the boundary point of the risk level is determined by analyzing the distribution density curve; the score value range is divided into several level ranges, for example, the risk level is divided into three ranges: low risk, medium risk and high risk; according to the actual engineering safety standards and historical deviation tolerance, the critical point of each range is set as the risk boundary threshold; in the example, when the score value range is 0 to 1, the risk boundary threshold can be set to 0.35 and 0.7, where samples with a score value less than 0.35 are classified as low risk, samples with a score value between 0.35 and 0.7 are classified as medium risk, and samples with a score value greater than 0.7 are classified as high risk. The distribution density curve is a continuous curve used to describe the probability distribution of a set of data within a range of values. The horizontal axis represents the data value, and the vertical axis represents the probability density of the corresponding value. The height and shape of the curve can reflect the concentration range and dispersion of the data. The risk level classification rule is based on the determination of the risk level according to the interval position of the comprehensive blasting effect score result relative to the preset risk boundary threshold.

[0090] S6.3 Automated writing, formatting, and archiving of the report text for the decision response feature set to generate a blasting effect evaluation report.

[0091] Specifically, the decision response feature set is organized according to the report chapter structure, including an overview of blasting parameters, comprehensive blasting effect scoring results, sub-risk levels, and multi-audience handling suggestions; corresponding text paragraphs are generated based on the content of each chapter, and titles, tables, figures, and annotations are set according to predefined document formats; automated typesetting and paragraph numbering are performed on the text paragraphs, charts and data are associated and labeled, and the text, tables, and figures are integrated to form the complete report text; the complete report text is checked for format consistency and metadata is archived, and saved as a standardized document file, such as PDF or sample Word format, to generate a blasting effect evaluation report.

[0092] It should also be noted that the document format includes determining the report title style, chapter hierarchy and numbering rules, body text font and size, paragraph spacing and line spacing settings, table border and cell alignment, figure title and numbering specifications, as well as the format of notes and references; according to the content type of each chapter, assign corresponding style templates and save them as reusable document template files, such as sample Word templates or PDF templates, so that they can be directly applied when generating the report text, achieving a unified document layout and formatting standard.

[0093] It should be noted that, in order to support remote monitoring and data sharing, the blasting effect evaluation and analysis method for tunnel mining engineering based on the Internet in this embodiment can upload the blasting response data collected by on-site sensors to a remote server via the Internet, so as to realize remote analysis, visualization and report generation. At the same time, the data collection, preprocessing and scoring calculation can be completed locally on-site, or can be transmitted in real time or in batches via the Internet, thus taking into account both on-site processing efficiency and remote management needs.

[0094] In summary, this invention achieves hierarchical identification and energy partitioning of the blasting energy release process at different time scales by performing multi-scale temporal decomposition and spectral energy clustering on the fused multimodal feature data. This extracts blasting response patterns with statistical stability, improves the accuracy of energy behavior identification and the interpretability of pattern representation, and enhances the reliability of blasting quality control and safety early warning.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet, characterized in that: include, Collect real-time monitoring data from multiple sources, perform preprocessing, and generate a multimodal feature set; Simultaneous fusion of multimodal feature sets generates a multidimensional explosive feature representation; The multidimensional blasting feature representation is used to generate blasting response patterns through multi-scale temporal decomposition and spectral energy clustering. The specific steps are as follows. The multidimensional blasting feature representation is divided into several fixed-length time windows according to the time series, generating a multi-scale time period set; Multi-scale time series decomposition is performed on the multi-scale time period set to generate a multi-scale time-frequency feature matrix; By using a multi-scale time-frequency feature matrix, the multi-scale time period set is labeled with categories to form clustered feature groups; The process involves identifying recurring high-energy and anomalous energy patterns from clustered feature groups, generating blasting response patterns, and performing statistical and spectral analysis to generate a set of blasting effect evaluation indicators. The specific steps are as follows: Time-domain statistical analysis and multi-scale time-frequency decomposition of the blasting response mode are performed to generate multi-dimensional time-frequency statistical features. By utilizing multidimensional time-frequency statistical features, energy zoning damage indicators, fragmentation and block size distribution, and flyrock risk are calculated and correlated with the surrounding rock structural characteristics to generate a safety risk probability vector, which is then integrated to form a set of physical and risk features. Bayesian fusion and confidence weighting are performed on the set of physical and risk characteristics to generate a set of evaluation indicators for blasting effects; The blasting effect evaluation index set is compared with the standard template in the historical blasting project database to generate a deviation benchmark. The Bayesian dynamic inference method is then used to calculate the current blasting effect deviation and generate a deviation vector. Dynamic weight adjustment and feature contribution correction are applied to the deviation vector to generate a comprehensive blasting effect score. The comprehensive blasting effect score results are analyzed and summarized to generate a blasting effect evaluation report.

2. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The multi-source real-time monitoring data includes vibration waveform data, acoustic data, image data, surrounding rock displacement data, and construction parameter data.

3. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The specific steps for generating the multimodal feature set are as follows: Denoising, time synchronization, and event clipping are performed on multi-source real-time monitoring data to generate multi-source, multi-scale time series segments. Physical constraint feature extraction, time-frequency analysis, and visual texture feature extraction are performed on multi-source, multi-scale time-series segments to generate candidate multimodal feature vector sets; Sparse encoding, dimensionality reduction, and interpretability enhancement are performed on the candidate multimodal feature vector groups to generate a multimodal feature set.

4. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The specific steps for generating the multidimensional explosive feature representation are as follows: The multimodal feature set is initially sorted and aligned according to event ID and timestamp to generate an ordered multimodal feature list; The ordered multimodal feature list is normalized according to modality category and labeled with confidence weights to generate a multimodal feature set with confidence labels. Local weighted fusion is then performed according to time windows to generate a preliminary fused feature representation. Cross-modal attention adjustment is performed on the preliminary fusion feature representation to generate attention-weighted fusion features, and dimensionality reduction and structured encoding are performed to generate multidimensional burst feature representations.

5. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The standard templates in the historical blasting project database are obtained by statistical analysis, energy distribution feature extraction, and safety risk assessment normalization of blasting effect evaluation indicators from previous tunnel mining projects.

6. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The specific steps for generating the deviation vector are as follows: The blasting effect evaluation index set is compared with the standard template in the historical blasting project database, and the deviation of each index is calculated to generate a deviation benchmark. Based on the deviation benchmark, the deviations of each indicator are weighted by probability to form a weighted deviation distribution; By using the weighted deviation distribution, the comprehensive deviation of each indicator is calculated and a deviation vector is generated.

7. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The specific steps for generating the comprehensive blasting effect score are as follows: The deviation vector is standardized and initial weights are calculated to generate a normalized set of deviation components. The normalized bias component set is corrected for feature contribution according to the causal relationship between indicators, and abnormal bias is suppressed by robust constraints to generate a weighted bias vector. The weighted bias vector is subjected to multi-scale weighted aggregation and time window smoothing calculation to generate a comprehensive bias score sequence. The comprehensive deviation score sequence is normalized and mapped and the confidence level is corrected to generate a comprehensive blasting effect score result.

8. The method for evaluating and analyzing the blasting effect in tunnel mining engineering based on the Internet as described in claim 1, characterized in that: The specific steps for generating the blasting effect evaluation report are as follows: The comprehensive blasting effect score results are decomposed into sub-items by time and batch, and trend analysis, anomaly identification and causal clue extraction are performed on the sub-items to generate a score sequence. The risk levels of each item are calculated using the scoring sequence, and then classified, summarized, and correlated to generate multi-audience handling suggestions and visual summaries, which are then integrated to form a set of decision response features. The system automatically writes, formats, and archives the report text for the decision response feature set, generating a blasting effect evaluation report.

Images