Shield tunneling machine abnormity diagnosis method and system based on maximum information nonparametric

Abstract

The invention provides a shield tunneling machine abnormity diagnosis method and system based on maximum information nonparameterization, and the method comprises the steps: obtaining a vibration signal of a cutterhead bearing in a shield tunneling machine in the tunneling process of the shield tunneling machine, and determining a space energy attenuation mode of a vibration waveform formed when the vibration signal is transmitted in an underground complex stratum; performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set; selecting target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information nonparameter; generating a collaborative diagnosis feature vector according to the space energy attenuation mode and the target abnormal frequency band energy distribution; and inputting the collaborative diagnosis feature vector into a preset random forest classifier, and outputting a diagnosis result matched with the tunneling abnormal state of the shield tunneling machine. The accuracy and the real-time performance of abnormal state diagnosis of the shield tunneling machine in the complex stratum are improved.

Description

Technical Field

[0001] The present application relates to the field of maximum information non-parameter technology, and in particular to a shield machine abnormality diagnosis method and system based on maximum information non-parameter. Background Art

[0002] When a shield machine is excavating in complex strata, abnormal conditions of the cutterhead bearing (such as tool wear, sudden changes in strata, etc.) will cause changes in the vibration signal characteristics. Due to the complex underground environment, the vibration signal is interfered by multiple factors, and traditional methods are difficult to accurately identify abnormalities.

[0003] At present, there are schemes that use a diagnostic method based on vibration signal spectrum analysis and support vector machine. This method first extracts the spectrum characteristics of the vibration signal through fast Fourier transform, then uses principal component analysis to reduce the dimension, and finally inputs the support vector machine model for classification. This method uses frequency domain energy distribution and machine learning algorithms to identify abnormal conditions to a certain extent.

[0004] This solution only relies on frequency domain features, ignoring the energy attenuation characteristics of vibration signals during spatial propagation, resulting in insufficient sensitivity to abnormal conditions such as formation mutations. In addition, principal component analysis is based on linear transformation, which makes it difficult to effectively mine the nonlinear association between vibration features and fault labels, affecting the accuracy of feature screening and thus reducing the reliability of diagnostic results. Summary of the invention

[0005] The present application provides a shield machine abnormality diagnosis method and system based on maximum information non-parameters, which are used to solve the problems of low accuracy and poor real-time performance in the diagnosis of abnormal states of shield machines in complex formations in the prior art.

[0006] In a first aspect, the present application provides a shield machine abnormality diagnosis method based on maximum information non-parameters, comprising:

[0007] During the tunneling process of the shield machine, a vibration signal of a cutterhead bearing in the shield machine is obtained, and a spatial energy attenuation mode of a vibration waveform formed when the vibration signal propagates in a complex underground stratum is determined;

[0008] Performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set;

[0009] Selecting a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter;

[0010] generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution;

[0011] The collaborative diagnosis feature vector is input into a preset random forest classifier, and a diagnosis result matching the abnormal excavation state of the shield machine is output.

[0012] Optionally, the selecting a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion includes:

[0013] Based on the multi-band energy distribution set, the energy distribution of the current frequency band is used as the first variable, and the preset abnormal state label set is used as the second variable, to construct a joint probability distribution table between the first variable and the second variable, wherein the abnormal state label set is pre-labeled based on energy distribution features corresponding to different abnormal types in historical data;

[0014] Based on the maximum information non-parameter in the maximum mutual information criterion, the occurrence probability of each energy value in the first variable and the occurrence probability of each abnormal label in the second variable are calculated respectively;

[0015] Based on the joint probability distribution table and each of the occurrence probabilities, a nonlinear correlation index value between the current frequency band energy distribution and the abnormal state is calculated;

[0016] The energy distribution corresponding to the frequency band whose nonlinear correlation index value is greater than the preset index threshold is marked as the target abnormal frequency band energy distribution.

[0017] Optionally, the calculating, based on the joint probability distribution table and each of the occurrence probabilities, a nonlinear correlation index value between the current frequency band energy distribution and the abnormal state includes:

[0018] Extracting the joint probability value of each energy value in the current frequency band energy distribution and the corresponding abnormal label from the joint probability distribution table one by one;

[0019] For each combination of energy value and abnormal label, multiply the probability of occurrence of the energy value in the first variable by the probability of occurrence of the abnormal label in the second variable to obtain a corresponding independent probability product result;

[0020] Performing a ratio operation on the joint probability value of each combination and the product of the independent probabilities, and taking the natural logarithm of the ratio operation result;

[0021] Multiplying the joint probability value of each combination by the corresponding natural logarithm result to obtain the contribution value of each combination;

[0022] The contribution values ​​of all combinations are accumulated to obtain the nonlinear correlation index value of the current frequency band.

[0023] Optionally, generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution includes:

[0024] Extracting energy decay rate parameters in the propagation direction, energy difference parameters between adjacent sensor nodes, and energy distribution symmetry parameters from the spatial energy decay pattern;

[0025] Extracting a frequency band span parameter, an energy concentration parameter and an energy fluctuation amplitude parameter from the target abnormal frequency band energy distribution;

[0026] The result of multiplying the energy attenuation rate parameter by the frequency band span parameter is used as a first weight coefficient, the result of multiplying the energy difference parameter by the energy concentration parameter is used as a second weight coefficient, and the result of multiplying the energy distribution symmetry parameter by the energy fluctuation amplitude parameter is used as a third weight coefficient;

[0027] The first weight coefficient, the second weight coefficient and the third weight coefficient are spliced ​​into a first multidimensional feature sequence in a preset order, and the overlapping area ratio parameter of the spatial energy attenuation pattern and the abnormal frequency band energy distribution is added to the first multidimensional feature sequence to generate a collaborative diagnosis feature vector.

[0028] Optionally, the adding of an overlapping area ratio parameter of the spatial energy attenuation pattern and the abnormal frequency band energy distribution to the first multidimensional feature sequence to generate a collaborative diagnosis feature vector includes:

[0029] Counting the number of overlapping intervals between a first energy distribution interval and a second energy distribution interval, the first energy distribution interval is an energy distribution interval in the spatial energy attenuation pattern, and the second energy distribution interval is an energy distribution area in the abnormal frequency band energy distribution;

[0030] Dividing the number of overlapping intervals by the total number of distribution intervals corresponding to the spatial energy attenuation pattern, to obtain a proportion parameter of the overlapping area;

[0031] Adding the proportion parameter of the overlapping area to the end of the last weight coefficient in the first multidimensional feature sequence to generate a second multidimensional feature sequence;

[0032] A collaborative diagnosis feature vector is generated based on the second multi-dimensional feature sequence.

[0033] Optionally, marking the energy distribution corresponding to the frequency band whose nonlinear correlation index is greater than a preset index threshold as target abnormal frequency band energy distribution includes:

[0034] Compare the nonlinear correlation index value of the current frequency band with a preset index threshold;

[0035] When the comparison result is that the nonlinear correlation index is greater than the energy distribution corresponding to the frequency band of the preset index threshold, marking the energy distribution corresponding to the current frequency band as a preliminary abnormal frequency band energy distribution;

[0036] The preliminary abnormal frequency band energy distribution is verified to generate a target abnormal frequency band energy distribution.

[0037] Optionally, the step of inputting the collaborative diagnosis feature vector into a preset random forest classifier and outputting a diagnosis result matching the abnormal tunneling state of the shield machine includes:

[0038] Inputting the collaborative diagnosis feature vector into a preset random forest classifier, wherein the preset random forest classifier decomposes the collaborative diagnosis feature vector into a plurality of independent feature items;

[0039] By presetting multiple decision paths in the random forest classifier, a hierarchical judgment is performed on each of the independent feature items;

[0040] According to the hierarchical judgment result, determine the abnormal type mark corresponding to the independent feature item;

[0041] Matching the abnormal type tags of all independent feature items with abnormal states in a preset abnormal state mapping table;

[0042] If more than half of the independent feature items match the same abnormal state, the abnormal state is output and used as the diagnosis result.

[0043] In a second aspect, the present application provides a shield machine abnormality diagnosis system based on maximum information non-parameters, comprising:

[0044] An acquisition module is used to acquire the vibration signal of the cutterhead bearing in the shield machine during the tunneling process of the shield machine, and determine the spatial energy attenuation mode of the vibration waveform formed when the vibration signal propagates in the complex underground strata;

[0045] A decomposition module, used for performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set;

[0046] A selection module, configured to select a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter;

[0047] A generating module, used for generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution;

[0048] An input module is used to input the collaborative diagnosis feature vector into a preset random forest classifier and output a diagnosis result that matches the abnormal excavation state of the shield machine.

[0049] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a shield machine abnormality diagnosis method based on maximum information non-parametric as described in any one of the first aspects.

[0050] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement a shield machine abnormality diagnosis method based on maximum information non-parameters as described in any one of the first aspects.

[0051] In the present application, a shield machine abnormality diagnosis method based on maximum information non-parameters is provided, and the method includes: during the tunneling process of the shield machine, obtaining the vibration signal of the cutter head bearing in the shield machine, and determining the spatial energy attenuation pattern of the vibration waveform formed when the vibration signal propagates in the complex underground strata; performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set; using the maximum mutual information criterion to select a target abnormal frequency band energy distribution from the multi-band energy distribution set, and the maximum mutual information criterion is determined by the maximum information non-parameters; generating a collaborative diagnosis feature vector based on the spatial energy attenuation pattern and the target abnormal frequency band energy distribution; inputting the collaborative diagnosis feature vector into a preset random forest classifier, and outputting a diagnosis result that matches the tunneling abnormal state of the shield machine.

[0052] The technical solution provided by this application has the following beneficial effects:

[0053] This application collects the vibration signal of the cutter head bearing in real time, analyzes its energy attenuation characteristics during the propagation process in complex underground strata, effectively captures the impact of stratum structure changes on the propagation of vibration waves, and provides a spatial dimension feature basis for abnormal diagnosis. Signal decomposition technology is used to convert the vibration signal into frequency domain energy distribution, completely retaining the fault feature information of different frequency bands, and avoiding feature omissions caused by single frequency band analysis. Based on the mutual information criterion constructed based on maximum information non-parameters, the characteristic frequency band with the strongest correlation with the abnormal state is accurately identified from multiple frequency bands, thereby improving the efficiency of screening abnormal features. The dual-dimensional features of the spatial energy attenuation pattern and the energy distribution of the abnormal frequency band are integrated to construct a composite feature vector with strong characterization capabilities, thereby enhancing the completeness of the classifier input features. Hierarchical decision-making is performed on multi-dimensional features through an integrated learning algorithm to achieve high-precision matching of abnormal types and ensure the reliability of the diagnosis results.

[0054] Furthermore, this application also constructs a joint probability distribution table of the vibration signal frequency band energy distribution and the preset abnormal label, calculates the independent probability of the energy value and the abnormal label respectively based on the maximum information non-parameter, and then quantifies the nonlinear correlation index between the frequency band energy and the abnormal state, and finally selects the frequency band with the index value exceeding the threshold as the target abnormal feature. This method breaks through the limitations of traditional linear analysis and realizes the accurate mining of the deep correlation between vibration signals and fault types.

[0055] In addition, this solution effectively extracts frequency band features in vibration signals that have strong nonlinear correlations with abnormal conditions by combining probability statistics with information theory, solving the problem of inaccurate fault feature extraction in complex geological environments and providing high-value feature input for subsequent diagnosis.

[0056] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Figure 1 A flowchart of a shield machine abnormality diagnosis method based on maximum information non-parameters provided in an embodiment of the present application;

[0059] Figure 2 A schematic diagram of the structure of a shield machine abnormality diagnosis system based on maximum information non-parameters provided in an embodiment of the present application;

[0060] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0062] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0063] Researchers have found that when a shield machine is excavating in complex strata, traditional vibration signal analysis methods are difficult to accurately capture the correlation characteristics between abnormal conditions of the cutter head bearing and stratum changes, resulting in insufficient fault diagnosis accuracy. Based on this, an embodiment of the present application provides a shield machine abnormality diagnosis method based on maximum information non-parameters. This method analyzes the spatial energy attenuation characteristics and multi-band energy distribution of vibration signals, uses the maximum mutual information criterion to screen key abnormal frequency bands, and fuses them to generate collaborative diagnosis feature vectors. Finally, a random forest classifier is used to achieve accurate matching of abnormal conditions. The technical solution of the present application can be applied to shield construction safety monitoring scenarios under complex geological conditions such as urban subway tunnels and underpasses of airport runways.

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0065] Figure 1 A flowchart of a shield machine abnormality diagnosis method based on maximum information non-parameters provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0066] Step 101: During the tunneling process of the shield machine, a vibration signal of a cutterhead bearing in the shield machine is obtained, and a spatial energy attenuation pattern of a vibration waveform formed when the vibration signal propagates in a complex underground stratum is determined.

[0067] In this step, the vibration signal of the cutterhead bearing represents the mechanical vibration waveform data generated by the cutterhead bearing during the tunneling of the shield machine, including time domain amplitude and frequency characteristics. The underground complex stratum refers to the non-homogeneous and variable rock and soil medium environment encountered during the tunneling of the shield machine. The spatial energy attenuation mode represents the quantitative characteristics of the energy attenuation of the vibration wave as the propagation distance increases during the propagation of the stratum, including parameters such as attenuation rate and directionality.

[0068] In an embodiment of the present application, a three-axis acceleration sensor installed on the cutter head bearing is used to collect vibration signals in real time, and anti-interference filtering technology is used to eliminate equipment noise; a sensor array arranged on the shield machine casing is used to receive stratum propagation signals, and a three-dimensional propagation model is established based on the wave equation theory; the signal energy envelope is extracted through wavelet time-frequency analysis, and the anisotropic attenuation coefficient is calculated in combination with the spatial position of the sensor, and finally a spatial energy attenuation model including parameters such as the main attenuation direction and energy gradient is constructed.

[0069] For example, in an airport runway underpass project, eight vibration sensors arranged in a ring were used for real-time monitoring. It was found that at the interface between the runway pavement concrete layer and the lower sand layer, the attenuation rate of the vibration signal in the vertical direction was higher than that in the horizontal direction. By comparing the signal strength differences of sensors at different positions, the area where the thickness of the runway structure layer changes was accurately identified.

[0070] Step 102: performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set.

[0071] In this step, multi-frequency domain decomposition represents the process of dividing the vibration signal into multiple sub-bands according to the frequency components. The multi-band energy distribution set represents the set of energy values ​​of each sub-band, characterizing the proportion of vibration energy of different frequency components, and is used to reflect the interference of complex underground strata structure. The dynamic changes of the multi-band energy distribution set are directly related to the vibration energy attenuation caused by the reflection and diffraction of the underground medium interface.

[0072] In an embodiment of the present application, the preprocessed vibration signal is decomposed by wavelet packet transform, and a 6-layer decomposition is realized by using wavelet basis function; the energy value of the reconstructed signal of each node is calculated, and the frequency band is converted into 24 sub-bands that conform to the human ear perception characteristics through scale division; energy normalization processing is used to eliminate amplitude differences, and finally a multi-band energy distribution set containing the relative energy proportion of each band is generated.

[0073] For example, in the above-mentioned project, decomposition found that the energy proportion of the vibration signal of the runway concrete layer in the high-frequency band was significantly higher than that of the sand layer, especially the energy ratio difference in the 315-400Hz frequency band reached 2.3 times (obtained by calculating the mean energy ratio of the frequency band under the same working conditions of the concrete layer and the sand layer). This feature was identified as a key indicator for stratum identification.

[0074] Step 103: Selecting a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter.

[0075] In this step, the maximum mutual information criterion represents an indicator for measuring the statistical dependence between two variables. The target abnormal frequency band energy distribution specifically refers to the vibration energy distribution characteristics within a specific frequency range that is strongly correlated with the abnormal state of the shield machine, which is screened out by the maximum mutual information criterion. The maximum information nonparametric representation is a probability distribution calculation method based on kernel density estimation.

[0076] In an embodiment of the present application, a frequency band-fault label database containing historical abnormal cases is constructed; the window kernel density estimation method is used to calculate the joint probability distribution; the maximum information coefficient algorithm improved based on the divergence theory is used to calculate the correlation between each frequency band and the abnormal type; the dynamic threshold screening method is used to determine the frequency band, and finally the target abnormal frequency band energy distribution that is strongly associated with the typical risk of runway underpass is output.

[0077] For example, in response to the unique pavement settlement risk of runway underpass, the system automatically screened out the 80-160Hz frequency band as the key monitoring object. This selection was based on the calculation of the correlation coefficient between energy mutation and settlement deformation in this frequency band in historical cases (coefficient > 0.85), and successfully warned of three potential settlement areas during actual excavation.

[0078] Step 104: Generate a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution.

[0079] In this step, the collaborative diagnosis feature vector represents a multi-dimensional feature array that fuses spatial propagation and frequency domain features.

[0080] In an embodiment of the present application, a canonical correlation analysis is performed on the three principal components of the spatial attenuation pattern and the six characteristic parameters of the abnormal frequency band; the linear inseparable features are converted to a high-dimensional space by kernel function mapping; the top seven most discriminative features are retained by ranking the feature importance; and a 13-dimensional collaborative diagnosis feature vector is constructed by adding a spatiotemporal correlation factor.

[0081] For example, when crossing the critical area of ​​the runway, an abnormal peak appeared in the cross-feature of "vertical attenuation rate × 160Hz frequency band energy" in the fusion feature vector generated by the system. After inversion analysis, it was found that there was a pavement joint defect at that location.

[0082] Step 105: Input the collaborative diagnosis feature vector into a preset random forest classifier, and output a diagnosis result that matches the abnormal excavation state of the shield machine.

[0083] In this step, the preset random forest classifier contains an ensemble learning model of 200 decision trees. The abnormal state of tunneling refers to the typical fault mode classification system exhibited by the shield machine when it works abnormally in complex strata. The diagnosis result is a structured decision report output by the system, which contains three-dimensional abnormality type determination, severity classification and location information. For example: Class B2 fault (confidence 89%), secondary warning, location: 368th ring + 1.2m, bearing abnormality at 0.8m to the right of the Y axis.

[0084] In the embodiment of the present application, a random forest classifier optimized by an integrated strategy is adopted, and 100 decision trees are set for parallel calculation; the feature weights are dynamically adjusted based on error estimation; a transfer learning mechanism is introduced to adapt to different geological conditions; and a triple diagnosis result including anomaly type, location coordinates, and risk level is output.

[0085] For example, during continuous excavation, the system warns in advance of possible differential settlement in the center area of ​​the runway based on the changing trend of the characteristic vector. The diagnostic result shows "settlement risk level II, located at the 11 o'clock direction of the cutterhead."

[0086] This solution achieves multi-dimensional and accurate perception of shield tunneling anomalies by establishing a collaborative analysis system of vibration signal spatial propagation and frequency domain characteristics. Engineering applications have shown that the system can stably identify typical anomalies such as tool wear and formation mutations, and the timeliness of early warning is improved compared with conventional methods. It provides a full-process diagnostic capability for abnormality discovery, positioning, and evaluation for complex formation construction, effectively reducing construction risks and optimizing maintenance decisions.

[0087] In order to solve the problem of insufficient accuracy in identifying abnormal frequency bands during tunneling of a shield machine in complex strata, in some embodiments, step 103: selecting a target abnormal frequency band energy distribution from the multi-band energy distribution set using the maximum mutual information criterion includes:

[0088] Step 201: Based on the multi-band energy distribution set, taking the energy distribution of the current frequency band as the first variable and the preset abnormal state label set as the second variable, construct a joint probability distribution table between the first variable and the second variable, wherein the abnormal state label set is pre-labeled based on the energy distribution characteristics corresponding to different abnormal types in the historical data.

[0089] In step 201, the joint probability distribution table refers to a two-dimensional table that records the co-occurrence frequency of each energy value and anomaly label, wherein the first variable is the interval division after the discretization of the frequency band energy value, and the second variable is a predefined anomaly type label set.

[0090] In an embodiment of the present application, the vibration signals in the historical data are first binned according to their energy values ​​and divided into several intervals of equal width. At the same time, a label system including typical anomalies such as tool wear and formation mutation is established. Then, the co-occurrence frequency of each energy interval and each abnormal label is counted, and the zero probability problem is avoided through Laplace smoothing, and finally a complete joint probability distribution table is constructed.

[0091] Step 202: Based on the maximum information non-parameter in the maximum mutual information criterion, the occurrence probability of each energy value in the first variable and the occurrence probability of each abnormal label in the second variable are calculated respectively.

[0092] In step 202, the occurrence probability includes the marginal probability of the energy value and the marginal probability of the abnormal label, which respectively represent the independent occurrence frequency of each energy interval and the abnormal label in the historical data.

[0093] In an embodiment of the present application, a kernel density estimation method is used to calculate the probability distribution of energy values, and discrete data is smoothed by a Gaussian kernel function; for abnormal label probabilities, maximum likelihood estimation is performed directly based on the frequency of label occurrence; both probability calculations are normalized to ensure that the sum of the probabilities is 1.

[0094] Step 203: Based on the joint probability distribution table and each of the occurrence probabilities, a nonlinear correlation index value between the current frequency band energy distribution and the abnormal state is calculated.

[0095] In step 203, the nonlinear correlation index value is a metric value that quantifies the strength of the association between the energy of the frequency band and the abnormal state.

[0096] In the embodiment of the present application, the ratio of the product of the joint probability and the marginal probability is first calculated, then the logarithm is taken and multiplied by the joint probability, and finally a weighted sum is performed on all possible energy-label combinations; the Monte Carlo integration method is used to handle the probability calculation problem of continuous variables to ensure the calculation accuracy of the index value.

[0097] Step 204: marking the energy distribution corresponding to the frequency band whose nonlinear correlation index value is greater than the preset index threshold as the target abnormal frequency band energy distribution.

[0098] In step 204, the preset indicator threshold is a judgment boundary determined by analyzing the difference in indicator value distribution between normal operating conditions and abnormal operating conditions in historical data.

[0099] In the embodiment of the present application, a curve analysis method is used to determine the optimal threshold so as to maximize the difference between the true positive rate and the false positive rate; a secondary verification is performed on the screened abnormal frequency band to ensure that its energy distribution pattern has a stable correlation characteristic with the target abnormality.

[0100] Here is a specific example:

[0101] In a certain airport runway underpass project, when the system screened the abnormal frequency bands for the collected cutterhead bearing vibration signals, it first established a label library containing 5 types of typical abnormalities such as pavement settlement and tool wear based on 6 months of historical construction data (the label setting is based on 32 abnormal event records that actually occurred). For the 24 frequency bands obtained by decomposing the vibration signal of the current excavation section (the frequency band division uses the engineering vibration analysis band), the 80-160Hz frequency band with energy fluctuations was selected as the analysis object, and the energy monitoring value of this frequency band for 10 consecutive minutes (sampling frequency 1000Hz, after sliding average processing) was divided into 20 equal-width intervals as the first variable, and a joint probability distribution table was constructed with the abnormal label library (a total of 100 groups of valid samples). The probability of occurrence of the energy value of this frequency band when the settlement is abnormal is 0.38 (based on the statistical analysis of 28 settlement events in historical data), and the marginal probability of the abnormal label is calculated to be 0.25 (the proportion of settlement in 32 abnormal events). The final calculated nonlinear correlation index value of this frequency band is 0.21 (calculated by the mutual information formula), which exceeds the preset threshold of 0.18 (determined by curve analysis of 100 sets of verification data). Therefore, this frequency band is marked as the target abnormal frequency band for pavement settlement monitoring in runway underpass projects.

[0102] In the embodiments of the present application, the scheme realizes the intelligent screening of abnormal frequency bands by establishing an accurate probability statistical model and an innovative nonlinear correlation measurement method, which not only improves the accuracy of diagnosis, but also reduces the false alarm rate, providing reliable technical support for underpass projects in sensitive areas such as airport runways.

[0103] In order to solve the problem of insufficient quantification accuracy of the correlation between frequency band energy and abnormal state in shield construction, in some embodiments, step 203: the nonlinear correlation index value between the current frequency band energy distribution and the abnormal state is calculated based on the joint probability distribution table and each occurrence probability, including:

[0104] Step 301: Extracting the joint probability value of each energy value in the current frequency band energy distribution and the corresponding abnormal label from the joint probability distribution table one by one.

[0105] In step 301, the joint probability value refers to the probability of a specific energy interval and a specific abnormal label appearing at the same time, reflecting the statistical law of the co-occurrence of the two.

[0106] In an embodiment of the present application, a sliding window method is used to extract data from a distribution table, and the window width is set to 10 consecutive sampling points, sliding 1 point each time to ensure data continuity; after frequency statistics of the energy-label combination in each window, Laplace smoothing technology is used to deal with the zero frequency problem, and finally a stable joint probability estimate is obtained.

[0107] Step 302: For each combination of energy value and abnormal label, multiply the probability of occurrence of the energy value in the first variable by the probability of occurrence of the abnormal label in the second variable to obtain a corresponding independent probability product result.

[0108] In step 302, the independent probability product result represents the theoretical co-occurrence probability when the energy value and the abnormal label are independent of each other.

[0109] In an embodiment of the present application, the theoretical probability is calculated by the marginal probability multiplication rule, wherein the probability of occurrence of the energy value is obtained by Gaussian kernel density estimation, and the probability of abnormal labels is estimated by historical frequency; the product result is normalized to ensure that the sum of the theoretical probabilities of all combinations is 1.

[0110] Step 303: Perform a ratio operation on the joint probability value of each combination and the product of the independent probabilities, and take the natural logarithm of the ratio operation result.

[0111] In step 303, the ratio operation reflects the degree of deviation between the actual observed probability and the theoretical independent probability, and the natural logarithm processing is used to amplify the difference.

[0112] In the embodiments of the present application, fixed-point arithmetic is used to ensure calculation accuracy, and an exception handling mechanism is provided to deal with division by zero errors; Taylor expansion approximation is used for logarithmic calculation to improve calculation efficiency while ensuring accuracy.

[0113] Step 304: multiply the joint probability value of each combination by the corresponding natural logarithm result to obtain the contribution value of each combination.

[0114] In step 304, the contribution value quantifies the contribution of a single energy-label combination to the overall correlation.

[0115] In the embodiment of the present application, a weight coefficient is introduced to adjust the contribution weight of each combination, and an appropriate gain is given to rare but important abnormal modes; parallel computing technology is used to accelerate the calculation of contribution values ​​of large-scale combinations.

[0116] Step 305: Accumulate the contribution values ​​of all combinations to obtain the nonlinear correlation index value of the current frequency band.

[0117] In the embodiment of the present application, a layered accumulation algorithm is used to avoid numerical overflow, and the final result is standardized so that its value range is limited to between 0 and 1, which is convenient for subsequent threshold comparison.

[0118] Here is a specific example:

[0119] In the specific implementation of a certain airport runway underpass project, when analyzing the vibration energy data in the 80-160Hz frequency band, the system first extracts data from the constructed joint probability distribution table (100 groups of samples containing 20 energy intervals and 5 types of abnormal labels). Take the combination of the energy interval [120,140] and the "pavement settlement" label as an example: through statistical historical data, it is found that this combination actually appears 28 times (total samples 100 times), and the calculated joint probability value is 0.28; the probability of occurrence of this energy interval is 0.35 (calculated based on the interval appearing 35 times in all samples), and the probability of occurrence of the "pavement settlement" label is 0.25 (the proportion of settlement in historical abnormal events), and the two are multiplied to obtain the independent probability product result of 0.0875. Dividing the joint probability value 0.28 by the independent probability 0.0875, the ratio is 3.2, which is about 1.16 after taking the natural logarithm, and then multiplied by the joint probability value 0.28, the contribution value of this combination is about 0.325. The system repeated the above calculation process for all 300 valid combinations of 20 energy intervals and 5 types of abnormal labels (some combinations were filtered out due to insufficient data), and finally accumulated a nonlinear correlation index value of 0.21 for this frequency band (this value was confirmed to be valid through statistical analysis of 100 groups of verification samples).

[0120] In the embodiments of the present application, the scheme realizes accurate characterization of the nonlinear relationship between vibration characteristics and construction anomalies by establishing a sophisticated probability statistical model and an innovative correlation quantification method, providing a reliable abnormal warning basis for shield construction in sensitive areas such as airport runways.

[0121] In order to solve the problem of insufficient fusion of multi-dimensional features in shield construction, in some embodiments, step 104: generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution includes:

[0122] Step 401: extracting energy attenuation rate parameters in the propagation direction, energy difference parameters between adjacent sensor nodes, and energy distribution symmetry parameters from the spatial energy attenuation pattern.

[0123] In step 401, the propagation direction specifically refers to the direction of the energy transfer path of the vibration waveform between the shield machine cutter bearing and the formation sensor. The energy attenuation rate parameter refers to the amount of energy loss per unit distance of the vibration wave in a specific propagation direction. Adjacent sensor nodes refer to vibration signal acquisition devices deployed in the formation outside the shield machine cutter bearing. The energy difference parameter is the standard deviation of the energy values ​​received by adjacent sensor nodes. The energy distribution symmetry parameter specifically refers to the distribution difference of the vibration wave energy at the symmetrical position of the cutter bearing center. It is used to quantify the asymmetric characteristics of the vibration energy in the spatial distribution and reflect the distortion effect of formation heterogeneity on the propagation of vibration waves. It is specifically calculated through three elements: the energy reception ratio of the symmetrical sensor node pair, the position offset of the energy extreme point, and the inclination angle of the main energy axis.

[0124] In the embodiment of the present application, the least squares method is used to fit the energy attenuation curve of each propagation direction to obtain the rate parameter; the energy difference between nodes is calculated based on the ring-arranged sensor array; the symmetry parameter is calculated by the mirror comparison method, and finally a three-dimensional spatial feature group is formed.

[0125] Step 402: extracting a frequency band span parameter, an energy concentration parameter and an energy fluctuation amplitude parameter from the target abnormal frequency band energy distribution.

[0126] In step 402, the frequency band span parameter is the frequency range width of the abnormal frequency band. The energy concentration parameter is the energy proportion of the first three main peaks in the frequency band. The energy fluctuation amplitude parameter is the extreme difference of the frequency band energy changing over time.

[0127] In the embodiment of the present application, the frequency band boundary calculation span is determined by spectrum envelope analysis; the peak search algorithm is used to extract the main energy components; the time domain statistical analysis is used to obtain the fluctuation characteristics, and finally a three-dimensional frequency domain feature group is constructed.

[0128] Step 403: The result of multiplying the energy attenuation rate parameter by the frequency band span parameter is used as the first weight coefficient, the result of multiplying the energy difference parameter by the energy concentration parameter is used as the second weight coefficient, and the result of multiplying the energy distribution symmetry parameter by the energy fluctuation amplitude parameter is used as the third weight coefficient.

[0129] In step 403, the weight coefficient is a reinforced feature obtained by cross-multiplication of features, and is used to highlight key diagnostic information.

[0130] In the embodiment of the present application, the spatial features and the frequency domain features are combined by Cartesian products, and the product coefficients are adjusted by a dynamic weighting algorithm to ensure that the dimensions of the features in each dimension are unified, and finally three sets of enhanced feature values ​​are generated.

[0131] Step 404: Concatenate the first weight coefficient, the second weight coefficient and the third weight coefficient into a first multidimensional feature sequence in a preset order, and add a parameter of the overlapping area ratio of the spatial energy attenuation pattern and the abnormal frequency band energy distribution to the first multidimensional feature sequence to generate a collaborative diagnosis feature vector.

[0132] In step 404, the overlapping area ratio parameter is the intersection ratio of the spatial energy distribution and the frequency band energy distribution in the time-frequency domain.

[0133] In the embodiment of the present application, the overlapping area is calculated by time-frequency matrix dot multiplication operation; the proportion parameter is obtained by area integration method; each feature is sorted according to diagnostic importance and then spliced ​​to finally generate a collaborative diagnosis feature vector.

[0134] Here is a specific example:

[0135] During the implementation of a certain airport runway underpass project, when the system performed feature fusion for the identified 80-160Hz target abnormal frequency band, it first extracted three key parameters from the spatial energy attenuation pattern: the vertical energy attenuation rate parameter was 0.15 per second (obtained by fitting the data of 5 vibration sensors arranged 1 to 5 meters behind the cutter head), the energy difference parameter of adjacent sensor nodes was 0.3 (calculated the standard deviation of the energy received by 8 circumferentially arranged sensors), and the energy distribution symmetry parameter was 0.85 (obtained by comparing the energy receiving ratio of sensors at symmetrical positions at 3 o'clock and 9 o'clock on the cutter head). At the same time, the frequency band span parameter 80Hz (the difference between the upper and lower limits of the frequency band determined by spectrum analysis), the energy concentration parameter 0.6 (the proportion of the energy of the first three main peaks in the frequency band to the total amount), and the energy fluctuation amplitude parameter 1.2 volts (the maximum and minimum amplitude difference within a 10-minute monitoring period) were extracted from the target abnormal frequency band. The attenuation rate is multiplied by the frequency band span to obtain the first weight coefficient of 12, the energy difference is multiplied by the concentration to obtain the second weight coefficient of 0.18, and the symmetry is multiplied by the fluctuation amplitude to obtain the third weight coefficient of 1.02. The overlapping area ratio parameter of 0.45 (the ratio of the area where the spatial energy distribution and the frequency band energy distribution appear at the same time) is calculated through the time-frequency analysis matrix, and finally a collaborative diagnosis vector containing these four features is generated.

[0136] In the embodiment of the present application, the scheme realizes the deep coordination of spatial propagation characteristics and frequency domain characteristics through an innovative multi-source feature fusion method, improves the reliability of abnormal diagnosis and early warning capabilities under complex working conditions, and provides effective technical support for shield construction in sensitive areas such as airport runways.

[0137] In order to solve the problem of insufficient representation of spatiotemporal correlation in the feature fusion process, in some embodiments, step 404: adding the overlapping area ratio parameter of the spatial energy attenuation pattern and the abnormal frequency band energy distribution to the multidimensional feature sequence to generate a collaborative diagnosis feature vector includes:

[0138] Step 501: Count the number of overlapping intervals between a first energy distribution interval and a second energy distribution interval, where the first energy distribution interval is an energy distribution interval in the spatial energy attenuation pattern, and the second energy distribution interval is an energy distribution area in the abnormal frequency band energy distribution.

[0139] In step 501, the number of overlapping intervals refers to the number of regions where the spatial energy distribution and the frequency band energy distribution appear simultaneously in the time-frequency domain, reflecting the temporal-spatial correlation strength of the two features.

[0140] In an embodiment of the present application, a time-frequency distribution matrix is ​​established, the heat map of spatial energy distribution and the spectrum map of frequency band energy distribution are compared point by point, and a sliding window matching algorithm is used to count the number of overlapping areas that meet the dual energy thresholds, and the window size is adaptively adjusted according to the sampling frequency.

[0141] Step 502: Divide the number of overlapping intervals by the total number of distribution intervals corresponding to the spatial energy attenuation pattern to obtain a proportion parameter of the overlapping area.

[0142] In the embodiment of the present application, the area integration method is used to calculate the total distribution interval, the dimension effect is eliminated through normalization processing, and a smoothing coefficient is introduced to process the boundary effect, and finally a standardized proportion parameter is obtained.

[0143] Step 503: adding the proportion parameter of the overlapping area to the end of the last weight coefficient in the first multidimensional feature sequence to generate a second multidimensional feature sequence.

[0144] In step 503, the second multi-dimensional feature sequence is a new sequence with enhanced spatiotemporal correlation features based on the original features.

[0145] In the embodiment of the present application, a feature buffer is designed to ensure data alignment, dynamic memory allocation technology is used to achieve feature expansion, and a verification mechanism is used to ensure data integrity.

[0146] Step 504: Generate a collaborative diagnosis feature vector based on the second multi-dimensional feature sequence.

[0147] In the embodiment of the present application, a feature compression algorithm is used to reduce the dimension, and the transmission efficiency is improved through quantization coding, and finally a standardized feature vector suitable for classifier input is generated.

[0148] Here is a specific example:

[0149] During the implementation of a certain airport runway underpass project, when the system generated the collaborative diagnosis feature vector, it first divided the spatial energy attenuation pattern into 20 detection intervals (from 1 meter to 5 meters behind the cutter head, with each interval being 0.2 meters), and at the same time divided the 80-160Hz abnormal frequency band into 15 frequency bands (each frequency band is 5.33Hz). Through time-frequency analysis, it was found that there were 9 overlapping intervals in the 3.2-4.0 meter depth interval and the 105-125Hz frequency band (obtained by comparing the spatial sensor data with the spectrum analysis results). The number of overlapping intervals 9 was divided by the total number of spatial intervals 20, and the overlapping area ratio parameter 0.45 was obtained. Then the three sets of weight coefficients [12, 0.18, 1.02] calculated previously were spliced ​​in sequence with the ratio parameter 0.45 to form a new four-dimensional feature sequence [12, 0.18, 1.02, 0.45].

[0150] In the embodiment of the present application, the scheme effectively improves the feature vector's ability to characterize complex working conditions through an innovative spatiotemporal correlation feature enhancement method, enabling the system to more accurately identify hidden engineering risks and providing reliable technical support for underground crossing projects in sensitive areas such as airport runways.

[0151] In order to solve the problem of high false alarm rate in the abnormal frequency band screening process, in some embodiments, step 204: marking the energy distribution corresponding to the frequency band whose nonlinear correlation index is greater than the preset index threshold as the target abnormal frequency band energy distribution includes:

[0152] Step 601: Compare the nonlinear correlation index value of the current frequency band with a preset index threshold.

[0153] In step 601, the preset index threshold is a criterion determined by analyzing the distribution difference of index values ​​between normal and abnormal working conditions in historical data.

[0154] In an embodiment of the present application, a kernel density estimation method is used to construct an indicator value probability distribution model, and the ratio of the true positive rate to the false positive rate is maximized by determining the optimal threshold point, and an appropriate safety margin is set according to the engineering safety level.

[0155] Step 602: When the comparison result is that the nonlinear correlation index is greater than the energy distribution corresponding to the frequency band of the preset index threshold, the energy distribution corresponding to the current frequency band is marked as a preliminary abnormal frequency band energy distribution.

[0156] In step 602, the preliminary abnormal frequency band energy distribution refers to candidate abnormal frequency band data obtained through preliminary threshold screening.

[0157] In the embodiment of the present application, a dynamic threshold comparison mechanism is established, a sliding window verification is performed on the index value of each frequency band, and a double buffering technology is used to store the candidate frequency band data to ensure the real-time and continuity of the screening process.

[0158] Step 603: verify the preliminary abnormal frequency band energy distribution to generate a target abnormal frequency band energy distribution.

[0159] In step 603, the verification process refers to verifying whether there are at least two consecutive frequency bands in the adjacent frequency bands whose index values ​​exceed the index threshold. If the consecutive frequency bands synchronously exceed the threshold, the mark of the preliminary abnormal frequency band energy distribution is retained; if not, the preliminary abnormal frequency band energy distribution is removed from the mark list.

[0160] In the embodiment of the present application, firstly, the continuous abnormality of the frequency band energy is tested by time series analysis, then the matching degree between the abnormal frequency band and the formation characteristics is verified in combination with geological exploration data, and finally a comprehensive evaluation is performed using a Bayesian network.

[0161] Here is a specific example:

[0162] During the implementation of a certain airport runway underpass project, when the system made an abnormal judgment on the 80-160Hz frequency band, it first compared the nonlinear correlation index value of 0.21 (calculated based on the mutual information of 100 groups of samples) calculated for the frequency band with the preset threshold value of 0.18 (determined by analyzing 50 groups of normal working conditions and 50 groups of abnormal working conditions). Since the index value exceeded the threshold, the system marked the frequency band as a preliminary abnormal frequency band. Subsequently, verification processing was carried out: by analyzing the monitoring data of three consecutive excavation rings (each ring advanced 1.2 meters), it was found that the energy value of the frequency band continued to maintain more than 1.5 times the baseline value (the baseline value was the average energy value of the first 30 rings during normal construction); at the same time, the geological radar data was compared to confirm that the energy abnormal area and the runway pavement joint position coincided 85% (calculated by the spatial position matching algorithm). Finally, the system officially determined the frequency band as the target abnormal frequency band.

[0163] In the embodiment of the present application, the scheme effectively reduces the false alarm rate of abnormal frequency band identification by establishing a strict two-stage verification mechanism, while ensuring the detection rate of important engineering risks, and provides a reliable frequency band screening method for underground crossing projects in sensitive areas such as airport runways.

[0164] In order to solve the accuracy problem of multi-feature collaborative decision-making in shield construction abnormality diagnosis, in some embodiments, step 105: inputting the collaborative diagnosis feature vector into a preset random forest classifier, and outputting a diagnosis result matching the abnormal state of the shield machine's tunneling, includes:

[0165] Step 701: input the collaborative diagnosis feature vector into a preset random forest classifier, and the preset random forest classifier decomposes the collaborative diagnosis feature vector into multiple independent feature items.

[0166] In step 701, the independent feature item refers to a single feature component with clear engineering significance in the collaborative diagnosis feature vector, including the feature parameters of the spatial energy attenuation mode and the feature parameters of the target abnormal frequency band energy distribution.

[0167] In an embodiment of the present application, a feature importance ranking algorithm is used to analyze the collaborative diagnosis feature vector, the independent contribution of each feature is determined based on the Gini coefficient, and the features with strong correlation are merged and processed through feature grouping technology, and finally decomposed into multiple independent feature items with clear physical meanings.

[0168] Step 702: Perform hierarchical judgment on each of the independent feature items by presetting multiple decision paths in the random forest classifier.

[0169] In step 702, hierarchical judgment refers to the step-by-step feature judgment process from the root node to the leaf node of the decision tree.

[0170] In the embodiment of the present application, each decision tree is set with 10 hierarchical judgment nodes, the information gain ratio is used as the splitting criterion, the tree depth is controlled by the pre-pruning technology, and multi-path parallel evaluation is performed on each feature item.

[0171] Step 703: Determine the abnormal type label corresponding to the independent feature item according to the hierarchical judgment result.

[0172] In step 703, the exception type tag is the exception category code corresponding to the tree leaf node of the decision tree.

[0173] In an embodiment of the present application, a labeling system including a 5-digit code is established, the first two digits represent the abnormal major category, and the last three digits represent the specific subcategory, and the final label is determined by a leaf node voting mechanism.

[0174] Step 704: Match the abnormal type tags of all independent feature items with the abnormal states in a preset abnormal state mapping table.

[0175] In step 704, the preset abnormal state mapping table is a correspondence table between feature marks and engineering abnormal states.

[0176] In the embodiment of the present application, a mapping relationship is constructed based on 300 groups of historical abnormal cases, a fuzzy matching algorithm is used to process atypical features, and a confidence threshold is set to filter low-reliability matches.

[0177] Step 705: If more than half of the independent feature items match the same abnormal state, the abnormal state is output and used as the diagnosis result.

[0178] In the embodiment of the present application, a weighted voting mechanism is adopted to assign higher weights to important features, and a double confirmation mechanism is set up to ensure the reliability of the results, and finally a structured diagnostic report containing the abnormality type, location and risk level is output.

[0179] Here is a specific example:

[0180] In a certain airport runway underpass project, the system inputs a collaborative diagnosis vector containing four features, including spatial attenuation parameter (12) and frequency band association parameter (0.21), into a preset random forest classifier. The classifier first decomposes the vector into four independent feature items (the parameters come from the previous feature fusion calculation). Through parallel judgment of 100 decision trees, the spatial attenuation parameter was judged as abnormal in 82 trees (the threshold was set to 1.5 times the historical normal value), and the frequency band association parameter was abnormal in 79 trees (based on the mutual information threshold of 0.18). According to the preset marking rules ("10" indicates pavement abnormality, "20" indicates tool abnormality), these two feature items are marked as "10012" (pavement joint abnormality) and "10015" (pavement thickness abnormality) respectively. In the abnormal state mapping table, these two marks correspond to the "runway pavement structure abnormality" category. Since more than 60 feature items (accounting for 60% of the total) matched this type of abnormality, the system finally output the diagnosis result of "runway pavement structure abnormality, risk level II".

[0181] In the embodiment of the present application, the scheme achieves high-precision diagnosis of shield construction anomalies through an innovative multi-level feature decision-making mechanism, improves the accuracy of identifying engineering risks under complex working conditions, and provides reliable technical support for underground crossing construction of major projects such as airport runways.

[0182] Figure 2 A schematic diagram of the structure of a shield machine abnormality diagnosis system based on maximum information non-parameters provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes:

[0183] The acquisition module 21 is used to acquire the vibration signal of the cutterhead bearing in the shield machine during the tunneling process of the shield machine, and determine the spatial energy attenuation mode of the vibration waveform formed when the vibration signal propagates in the complex underground strata.

[0184] The decomposition module 22 is used to perform multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set.

[0185] The selection module 23 is used to select a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter.

[0186] The generating module 24 is used to generate a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution.

[0187] The input module 25 is used to input the collaborative diagnosis feature vector into a preset random forest classifier, and output a diagnosis result that matches the abnormal excavation state of the shield machine.

[0188] Figure 2 The shield machine abnormality diagnosis system based on maximum information non-parameter can be executed Figure 1 The implementation principle and technical effect of the shield machine abnormality diagnosis method based on maximum information non-parameters described in the illustrated embodiment will not be repeated. The specific manner in which each module and unit performs operations in the shield machine abnormality diagnosis system based on maximum information non-parameters in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0189] In one possible design, Figure 2 The shield machine abnormality diagnosis system based on maximum information non-parameters of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0190] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0191] The processing component 32 is described above Figure 1 The embodiment provides a shield machine abnormality diagnosis method based on maximum information non-parameters.

[0192] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0193] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0194] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0195] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0196] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0197] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0198] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a shield machine abnormality diagnosis method based on maximum information non-parameters.

[0199] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0200] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0201] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0202] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A shield machine abnormality diagnosis method based on maximum information non-parameter, characterized in that: include: During the tunneling process of the shield machine, a vibration signal of a cutterhead bearing in the shield machine is obtained, and a spatial energy attenuation mode of a vibration waveform formed when the vibration signal propagates in a complex underground stratum is determined; Performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set; Selecting a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter; generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution; The collaborative diagnosis feature vector is input into a preset random forest classifier, and a diagnosis result matching the abnormal excavation state of the shield machine is output.

2. The method according to claim 1, characterized in that The selecting a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using the maximum mutual information criterion includes: Based on the multi-band energy distribution set, the energy distribution of the current frequency band is used as the first variable, and the preset abnormal state label set is used as the second variable, to construct a joint probability distribution table between the first variable and the second variable, wherein the abnormal state label set is pre-labeled based on energy distribution features corresponding to different abnormal types in historical data; Based on the maximum information non-parameter in the maximum mutual information criterion, the occurrence probability of each energy value in the first variable and the occurrence probability of each abnormal label in the second variable are calculated respectively; Based on the joint probability distribution table and each of the occurrence probabilities, a nonlinear correlation index value between the current frequency band energy distribution and the abnormal state is calculated; The energy distribution corresponding to the frequency band whose nonlinear correlation index value is greater than the preset index threshold is marked as the target abnormal frequency band energy distribution.

3. The method according to claim 2, characterized in that The calculating, based on the joint probability distribution table and each of the occurrence probabilities, a nonlinear correlation index value between the current frequency band energy distribution and the abnormal state comprises: Extracting the joint probability value of each energy value in the current frequency band energy distribution and the corresponding abnormal label from the joint probability distribution table one by one; For each combination of energy value and abnormal label, multiply the probability of occurrence of the energy value in the first variable by the probability of occurrence of the abnormal label in the second variable to obtain a corresponding independent probability product result; Performing a ratio operation on the joint probability value of each combination and the product of the independent probabilities, and taking the natural logarithm of the ratio operation result; Multiplying the joint probability value of each combination by the corresponding natural logarithm result to obtain the contribution value of each combination; The contribution values ​​of all combinations are accumulated to obtain the nonlinear correlation index value of the current frequency band.

4. The method according to claim 1, characterized in that The generating of the collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution includes: Extracting energy decay rate parameters in the propagation direction, energy difference parameters between adjacent sensor nodes, and energy distribution symmetry parameters from the spatial energy decay pattern; Extracting a frequency band span parameter, an energy concentration parameter and an energy fluctuation amplitude parameter from the target abnormal frequency band energy distribution; The result of multiplying the energy attenuation rate parameter by the frequency band span parameter is used as a first weight coefficient, the result of multiplying the energy difference parameter by the energy concentration parameter is used as a second weight coefficient, and the result of multiplying the energy distribution symmetry parameter by the energy fluctuation amplitude parameter is used as a third weight coefficient; The first weight coefficient, the second weight coefficient and the third weight coefficient are spliced ​​into a first multidimensional feature sequence in a preset order, and the overlapping area ratio parameter of the spatial energy attenuation pattern and the abnormal frequency band energy distribution is added to the first multidimensional feature sequence to generate a collaborative diagnosis feature vector.

5. The method according to claim 4, characterized in that The adding the overlapping area ratio parameter of the spatial energy attenuation pattern and the abnormal frequency band energy distribution to the first multidimensional feature sequence to generate a collaborative diagnosis feature vector includes: Counting the number of overlapping intervals between a first energy distribution interval and a second energy distribution interval, the first energy distribution interval being an energy distribution interval in the spatial energy attenuation pattern, and the second energy distribution interval being an energy distribution area in the abnormal frequency band energy distribution; Dividing the number of overlapping intervals by the total number of distribution intervals corresponding to the spatial energy attenuation pattern, to obtain a proportion parameter of the overlapping area; Adding the proportion parameter of the overlapping area to the end of the last weight coefficient in the first multidimensional feature sequence to generate a second multidimensional feature sequence; A collaborative diagnosis feature vector is generated based on the second multi-dimensional feature sequence.

6. The method according to claim 2, characterized in that The step of marking the energy distribution corresponding to the frequency band whose nonlinear correlation index is greater than the preset index threshold as the target abnormal frequency band energy distribution includes: Compare the nonlinear correlation index value of the current frequency band with a preset index threshold; When the comparison result is that the nonlinear correlation index is greater than the energy distribution corresponding to the frequency band of the preset index threshold, marking the energy distribution corresponding to the current frequency band as a preliminary abnormal frequency band energy distribution; The preliminary abnormal frequency band energy distribution is verified to generate a target abnormal frequency band energy distribution.

7. The method according to claim 1, characterized in that The step of inputting the collaborative diagnosis feature vector into a preset random forest classifier and outputting a diagnosis result matching the abnormal tunneling state of the shield machine comprises: Inputting the collaborative diagnosis feature vector into a preset random forest classifier, wherein the preset random forest classifier decomposes the collaborative diagnosis feature vector into a plurality of independent feature items; By presetting multiple decision paths in the random forest classifier, a hierarchical judgment is performed on each of the independent feature items; According to the hierarchical judgment result, determine the abnormal type mark corresponding to the independent feature item; Matching the abnormal type tags of all independent feature items with abnormal states in a preset abnormal state mapping table; If more than half of the independent feature items match the same abnormal state, the abnormal state is output and used as the diagnosis result.

8. A shield machine abnormality diagnosis system based on maximum information non-parameters, characterized in that: include: An acquisition module is used to acquire the vibration signal of the cutterhead bearing in the shield machine during the tunneling process of the shield machine, and determine the spatial energy attenuation mode of the vibration waveform formed when the vibration signal propagates in the complex underground strata; A decomposition module, used for performing multi-frequency domain decomposition on the vibration signal to obtain a multi-band energy distribution set; A selection module, configured to select a target abnormal frequency band energy distribution from the multi-frequency band energy distribution set by using a maximum mutual information criterion, wherein the maximum mutual information criterion is determined by a maximum information non-parameter; A generating module, used for generating a collaborative diagnosis feature vector according to the spatial energy attenuation pattern and the target abnormal frequency band energy distribution; An input module is used to input the collaborative diagnosis feature vector into a preset random forest classifier and output a diagnosis result that matches the abnormal excavation state of the shield machine.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a shield machine abnormality diagnosis method based on maximum information non-parameters as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a shield machine abnormality diagnosis method based on maximum information non-parameters as described in any one of claims 1 to 7 is implemented.

Images