Shield tunneling machine abnormal data detection method and system based on weighted adaptive variable coefficient

Abstract

The invention discloses a shield tunneling machine abnormal data detection method and system based on a weighted adaptive variable coefficient. The method comprises the following steps: acquiring data of a plurality of different sensors on a shield tunneling machine; preprocessing based on the sensor data to obtain preprocessed data; obtaining a corresponding volatility index and a variation coefficient in a window based on the preprocessed data; obtaining an adaptive weight corresponding to the preprocessed data based on the volatility index; obtaining weighted self-adaptive variable coefficients of all the preprocessed data based on the variable coefficients in the window and the self-adaptive weights; obtaining an anomaly detection threshold value based on historical normal data and prior experience of the shield tunneling machine; obtaining abnormal data based on comparison between the weighted adaptive variable coefficient and an abnormal detection threshold value; and cleaning and extracting fault features based on the abnormal data. Abnormality detection and efficient cleaning of the multi-parameter data of the sensor of the shield tunneling machine are achieved, and the accuracy and reliability of the data of the shield tunneling machine are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal data detection for shield machines, and more particularly to a method and system for detecting abnormal data for shield machines based on weighted adaptive coefficient of variation. Background Art

[0002] The shield machine is a key equipment in tunnel construction, and its operating status directly affects the safety and efficiency of the project. However, due to the complex operating environment, the sensor data of the shield machine is often interfered by noise, abnormal values ​​and fault signals. The existing technology mainly processes the data through the following methods:

[0003] 1. Traditional data cleaning methods: Simple statistical methods such as mean filtering and median filtering are usually used for data cleaning. These methods are effective for stable and low-noise data, but it is difficult to maintain the integrity and effectiveness of the data when processing nonlinear, non-stationary and multi-dimensional data, and it is easy to cause the loss of fault characteristics.

[0004] 2. Threshold method: A fixed threshold is set based on experience to detect abnormal data that exceeds the threshold range. However, the data of the shield machine fluctuates greatly under different construction environments, and the fixed threshold is difficult to adapt to the dynamically changing complex environment, which can easily lead to misjudgment and missed judgment.

[0005] 3. Anomaly detection based on machine learning: Some existing technologies have introduced machine learning models for fault detection. Although this can improve the accuracy of detection, the complexity of the model and its strong dependence on the training process make it difficult to apply it to online monitoring and early warning in real time.

[0006] The disadvantages of the above-mentioned existing technologies mainly include: it is difficult to adapt to the dynamically changing construction environment; there are limitations when processing multi-dimensional data and it is impossible to fully integrate the information of various parameters; it is difficult to achieve a balance between real-time and accuracy.

[0007] Therefore, how to effectively integrate the multi-parameter data of the shield machine and then achieve efficient detection and cleaning of abnormal data of the shield machine is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0008] In view of this, the present invention provides a shield machine abnormal data detection method and system based on weighted adaptive coefficient of variation, which realizes abnormal detection and efficient cleaning of multi-parameter data of shield machine sensors, improves the accuracy and reliability of shield machine data, and provides strong support for real-time fault warning and safe operation of shield machines.

[0009] In order to achieve the above object, the present invention adopts the following technical solution:

[0010] The shield machine abnormal data detection method based on weighted adaptive coefficient of variation includes:

[0011] Obtain data from multiple different sensors on the shield machine;

[0012] Performing preprocessing based on the sensor data to obtain preprocessed data;

[0013] Obtaining corresponding volatility index and coefficient of variation within the window based on the preprocessed data;

[0014] Obtaining an adaptive weight corresponding to the preprocessed data based on the volatility index;

[0015] Obtaining a weighted adaptive coefficient of variation of all the preprocessed data based on the coefficient of variation within the window and the adaptive weight;

[0016] The anomaly detection threshold is obtained based on the historical normal data of the shield machine and prior experience;

[0017] Obtaining abnormal data based on comparing the weighted adaptive coefficient of variation with the abnormality detection threshold;

[0018] Cleaning is performed based on the abnormal data and fault features are extracted.

[0019] Preferably, the sensor data includes: shield machine pressure data, shield machine temperature data, shield machine vibration data and shield machine current data.

[0020] Preferably, the volatility index acquisition method is:

[0021] Based on the preprocessed data, the mean value μ in the window of the i-th preprocessed data is obtained. i (t);

[0022] Based on the mean μ within the window i (t) Get the volatility index δ of the i-th preprocessed data at time t i (t):

[0023]

[0024] Among them, N represents the total window data length in the sliding window, k represents the current window step length, and X i' (tk) represents the i-th preprocessed data at time tk.

[0025] Preferably, the method for obtaining the coefficient of variation within the window is:

[0026] Based on the volatility indicator δ i (t) and the mean μ in the window i (t) Obtain the coefficient of variation CV within the window corresponding to the i-th preprocessed data i (t):

[0027]

[0028] in, Represents a small positive number, preventing the denominator from being zero.

[0029] Preferably, the adaptive weight is specifically:

[0030] w i (t) = α·w i (t-1)+(1-α)·f(δ i (t));

[0031] Among them, w i (t) represents the adaptive weight of the i-th preprocessed data at time t, α represents the smoothing coefficient, and w i (t-1) represents the adaptive weight of the i-th preprocessed data at time t-1, and f represents the mapping function.

[0032] Preferably, the method for obtaining the abnormality detection threshold is:

[0033] The normal weighted adaptive coefficient of variation is calculated based on the historical normal data of the shield machine and prior experience;

[0034] Based on the normal weighted adaptive variation coefficient, a normal weighted adaptive variation coefficient mean μ is obtained. WACV' ;

[0035] Based on the normal weighted adaptive coefficient of variation mean μ WACV' Get the weighted adaptive coefficient of variation standard deviation σ WACV' ;

[0036] Based on μ WACV' and σ WACV' Get the upper threshold θ for anomaly detection upper and the anomaly detection threshold θ lower :

[0037] θ upper =μ WACV' +q·σ WACV'

[0038] θ lower =μ WACV' -q·σ WACV'

[0039] Where q represents the threshold adjustment coefficient;

[0040] The abnormality detection upper threshold θ upper and the anomaly detection lower threshold θ lower Together they constitute the anomaly detection threshold.

[0041] Preferably, the specific method for obtaining abnormal data is:

[0042] Based on the comparison between the weighted adaptive coefficient of variation and the anomaly detection threshold:

[0043] If the weighted adaptive variation coefficient is greater than the abnormality detection upper threshold or less than the abnormality detection lower threshold, it is determined to be the abnormal data; otherwise, it is determined to be normal data.

[0044] Preferably, the method further comprises: dividing the warning level based on the proportion of the weighted adaptive coefficient of variation exceeding the abnormality detection threshold:

[0045] Level 1 warning: the weighted adaptive coefficient of variation is greater than the upper threshold of abnormal detection or less than the lower threshold of abnormal detection by more than 10-20%;

[0046] Level 2 warning: the weighted adaptive coefficient of variation is greater than the upper threshold of abnormal detection or less than the lower threshold of abnormal detection by more than 20-30%;

[0047] Level 3 warning: the weighted adaptive coefficient of variation is greater than the anomaly detection upper threshold or less than the anomaly detection lower threshold by more than 30%.

[0048] Preferably, cleaning and extracting fault features based on the abnormal data specifically includes:

[0049] Performing interpolation processing based on the abnormal data to obtain cleaned data;

[0050] extracting the fault feature based on the cleaning data;

[0051] The fault characteristics include: time characteristics, parameter characteristics and statistical characteristics.

[0052] The shield machine abnormal data detection system based on weighted adaptive coefficient of variation includes: a data acquisition and processing module, a coefficient of variation acquisition module, a threshold acquisition module, a result output module and a result processing module;

[0053] The data acquisition and processing module is used to obtain a plurality of different sensor data on the shield machine; perform preprocessing based on the sensor data to obtain preprocessed data;

[0054] The variation coefficient acquisition module is used to obtain the corresponding volatility index and the variation coefficient within the window based on the preprocessed data; obtain the adaptive weight corresponding to the preprocessed data based on the volatility index; and obtain the weighted adaptive variation coefficient of all the preprocessed data based on the variation coefficient within the window and the adaptive weight;

[0055] The threshold acquisition module is used to obtain an anomaly detection threshold based on the weighted adaptive coefficient of variation;

[0056] The result output module is used to obtain abnormal data based on the comparison between the weighted adaptive coefficient of variation and the abnormal detection threshold;

[0057] The result processing module is used to clean and extract fault features based on the abnormal data.

[0058] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a shield machine abnormal data detection method and system based on weighted adaptive coefficient of variation, which has the following beneficial effects:

[0059] 1. Improve the accuracy of anomaly detection: Through the weighted adaptive coefficient of variation method, the multi-parameter data of the shield machine is comprehensively analyzed to make full use of the fault sensitivity of each parameter; through the adaptive weight adjustment mechanism, the algorithm can respond to changes in the operating status and environment in real time, improve the accuracy of anomaly detection, and reduce the probability of misjudgment and missed judgment.

[0060] 2. Enhanced real-time and robustness: The present invention uses a sliding window to dynamically monitor and analyze the data, and combines it with a real-time adjusted threshold to quickly detect abnormal points in the data; WACV has low computational complexity and is suitable for online real-time monitoring. It has strong robustness and can effectively cope with complex and changeable environmental interference during the operation of the shield machine.

[0061] 3. Strong multi-parameter data fusion capability: The present invention adopts the weighted coefficient of variation method to fuse the multi-parameter data of the shield machine, which can effectively process multi-dimensional, nonlinear and non-stationary data, and improve the overall effect of data cleaning and fault detection; through adaptive weight adjustment, a reasonable evaluation of the volatility of each parameter data is achieved, thereby improving the ability of multi-parameter data fusion.

[0062] 4. The early warning mechanism is flexible and effective: Through the real-time fault early warning mechanism, the system can divide the early warning levels according to the degree of change of the WACV value, trigger the early warning in time, and notify the operator to take corresponding measures to avoid the expansion of the fault; the graded early warning provides flexible risk warnings, which helps to ensure the safe operation of the shield machine.

[0063] 5. Data cleaning and fault feature extraction: The present invention can not only detect and clean abnormal data, but also extract key fault features, providing a high-quality data basis for subsequent fault diagnosis and analysis, thereby improving the accuracy and efficiency of fault diagnosis.

[0064] 6. Easy for engineering application: The method of the present invention has low computational complexity, is easy to implement, and is suitable for deployment and promotion in actual projects. Its real-time online monitoring characteristics enable it to be directly applied to the monitoring system of the shield machine, providing a practical solution for status monitoring and fault warning of engineering equipment.

[0065] 7. The present invention has significant advantages in improving the accuracy, real-time and reliability of shield machine operation data, can effectively solve the deficiencies of existing technologies in data cleaning and anomaly detection, and provide strong support for the safe operation and maintenance management of shield machines. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0067] Figure 1 A flow chart of the shield machine abnormal data detection method based on weighted adaptive coefficient of variation provided by the present invention.

[0068] Figure 2 This is a flow chart of the method for obtaining anomaly detection threshold provided by the present invention.

[0069] Figure 3 A schematic diagram of the structure of a shield machine abnormal data detection system based on weighted adaptive coefficient of variation provided in Example 3 of the present invention.

[0070] Figure 4 A schematic diagram of the structure of a shield machine abnormal data detection system based on weighted adaptive coefficient of variation provided in Example 4 of the present invention.

[0071] Figure 5 This is a structural block diagram of the computer device provided by the present invention. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] Example 1

[0074] like Figure 1As shown, the embodiment of the present invention discloses a shield machine abnormal data detection method based on weighted adaptive coefficient of variation, comprising:

[0075] Obtain data from multiple different sensors on the shield machine;

[0076] Perform preprocessing based on the sensor data to obtain preprocessed data;

[0077] Based on the preprocessed data, the corresponding volatility index and coefficient of variation within the window are obtained;

[0078] Obtaining adaptive weights corresponding to preprocessed data based on volatility indicators;

[0079] Obtaining weighted adaptive coefficient of variation of all preprocessed data based on the coefficient of variation within the window and the adaptive weight;

[0080] The anomaly detection threshold is obtained based on the historical normal data of the shield machine and prior experience;

[0081] Based on the comparison between the weighted adaptive coefficient of variation and the anomaly detection threshold, the anomaly data is obtained;

[0082] Clean and extract fault features based on abnormal data.

[0083] Example 2

[0084] The embodiment of the present invention discloses a shield machine abnormal data detection method based on weighted adaptive coefficient of variation, comprising:

[0085] Get data from multiple sensors on the shield machine:

[0086] Preferably, m different sensors are set on the shield machine to be detected, and the data collected by each sensor at time t is X i (t), where i = 1, 2, ..., m represents different sensors and t represents time, which reflects the physical quantity of the shield machine during operation.

[0087] Preferably, in this embodiment, a pressure sensor, a temperature sensor, a vibration sensor and a current sensor are provided to correspondingly collect and obtain shield machine pressure data, shield machine temperature data, shield machine vibration data and shield machine current data for subsequent analysis and processing.

[0088] Preprocess the sensor data to obtain preprocessed data:

[0089] Preferably, the preprocessing adopts data normalization to eliminate the influence of different sensor data magnitudes.

[0090] Preferably, this embodiment adopts the minimum-maximum normalization method:

[0091]

[0092] Among them, X i' (t) represents the normalized sensor data, i.e., preprocessed data, X min,i and X max,i Respectively represent X i The minimum and maximum values ​​of (t).

[0093] Based on the preprocessed data, the corresponding volatility index and coefficient of variation within the window are obtained:

[0094] Preferably, the volatility index acquisition method is:

[0095] Based on the preprocessed data, get the mean μ of the i-th preprocessed data in the window i (t);

[0096] Based on the mean μ within the window i (t) Get the volatility index δ of the i-th preprocessed data at time t i (t):

[0097]

[0098] Among them, N represents the total window data length in the sliding window, k represents the current window step length, and X i' (tk) represents the i-th preprocessed data at time tk.

[0099] Preferably, the mean μ in the window of the i-th preprocessed data is i (t) is:

[0100]

[0101] Preferably, the method for obtaining the coefficient of variation within the window is:

[0102] Based on the volatility indicator δ i (t) and the mean μ in the window i (t) Get the coefficient of variation CV in the window corresponding to the i-th preprocessed data i (t):

[0103]

[0104] in, Represents a small positive number, preventing the denominator from being zero.

[0105] Preferably, the mean μ in the window i (t) and volatility indicator δ i (t) corresponds to the central tendency and dispersion of the data.

[0106] Based on the volatility index, the adaptive weight corresponding to the preprocessed data is obtained:

[0107] Preferably, the initial weight w of each preprocessed data is set i (0):

[0108]

[0109] Initial weight w i (0) is the basic value for setting weights, reflecting the i-th preprocessed data X i The initial sensitivity for fault detection is assigned based on experience or historical data. Over time, the weights are dynamically adjusted using a smoothing adjustment formula.

[0110] Preferably, the adaptive weight is specifically:

[0111] w i (t) = α·w i (t-1)+(1-α)·f(δ i (t));

[0112] Among them, w i (t) represents the adaptive weight of the i-th preprocessed data at time t, α represents the smoothing coefficient, which is used to control the speed of adaptive adjustment, 0<α<1, w i (t-1) represents the adaptive weight of the i-th preprocessed data at time t-1, and f represents the mapping function.

[0113] Preferably, the present invention uses the smoothing coefficient α and the volatility index δ i (t), realizes the adaptive adjustment of weights, so that the algorithm can respond to changes in operating status and environment in real time, improves the accuracy of anomaly detection, and reduces the probability of misjudgment and missed judgment; through the adaptive adjustment of weights, a reasonable assessment of the volatility of each parameter data is achieved, and the ability of multi-parameter data fusion is improved.

[0114] The weighted adaptive coefficient of variation of all preprocessed data is obtained based on the coefficient of variation within the window and the adaptive weight:

[0115] Preferably, the coefficient of variation CV in the window corresponding to all preprocessed data is i (t) and adaptive weight w i (t) is fused to obtain the overall weighted adaptive coefficient of variation WACV(t) of all preprocessed data:

[0116]

[0117] Preferably, by fusing the coefficient of variation CV in different windows corresponding to the data of multiple sensors i(t) It realizes the fusion of multi-parameter data of shield machines, can effectively process multi-dimensional, nonlinear and non-stationary data, and improves the overall effect of data cleaning and fault detection.

[0118] Preferably, the weighted adaptive coefficient of variation WACV(t) comprehensively reflects the overall volatility of multi-parameter data of the shield machine and is used for anomaly detection.

[0119] Based on the historical normal data of the shield machine, the anomaly detection threshold is obtained:

[0120] Preferably, Figure 2 As shown in the figure, the method for obtaining the anomaly detection threshold is:

[0121] The normal weighted adaptive coefficient of variation WACV'(t) is calculated based on the historical normal data of the shield machine and prior experience;

[0122] Based on the normal weighted adaptive coefficient of variation WACV'(t), the normal weighted adaptive coefficient of variation mean μ is obtained WACV' ;

[0123] Based on the normal weighted adaptive coefficient of variation mean μ WACV' Get the standard deviation of the normal weighted adaptive coefficient of variation σ WACV' ;

[0124] Based on μ WACV' and σ WACV' Get the upper threshold θ for anomaly detection upper and the anomaly detection threshold θ lower :

[0125] θ upper =μ WACV' +q·σ WACV'

[0126] θ lower =μ WACV' -q·σ WACV'

[0127] Where q represents the threshold adjustment coefficient;

[0128] Anomaly detection upper threshold θ upper and the anomaly detection threshold θ lower Together they form the anomaly detection threshold.

[0129] Preferably, the normal weighted adaptive coefficient of variation mean μ WACV' for:

[0130]

[0131] Where T represents the total window length, that is, the total step length.

[0132] Preferably, the standard deviation of the normal weighted adaptive coefficient of variation σ WACV' for:

[0133]

[0134] Preferably, update μ in real time WACV' and σ WACV' , to achieve dynamic adjustment of anomaly detection thresholds.

[0135] The present invention utilizes a sliding window to dynamically monitor and analyze data, and combines a threshold value adjusted in real time to quickly detect abnormal points in the data.

[0136] Based on the comparison between the weighted adaptive coefficient of variation and the anomaly detection threshold, the anomaly data is obtained:

[0137] Preferably, the specific method for obtaining abnormal data is:

[0138] Comparison between weighted adaptive coefficient of variation and anomaly detection threshold:

[0139] If the weighted adaptive coefficient of variation WACV(t) is greater than the anomaly detection upper threshold θ upper Or less than the anomaly detection lower threshold θ lower , it is judged as abnormal data; otherwise it is judged as normal data.

[0140] Preferably, the method further comprises: dividing the warning level based on the proportion of the weighted adaptive coefficient of variation exceeding the abnormality detection threshold:

[0141] Level 1 warning: The weighted adaptive coefficient of variation WACV(t) is greater than the anomaly detection upper threshold θ upper Or less than the anomaly detection lower threshold θ lower Exceed 10-20%;

[0142] Secondary warning: The weighted adaptive coefficient of variation WACV(t) is greater than the upper threshold of anomaly detection θ upper Or less than the anomaly detection lower threshold θ lower Exceed 20-30%;

[0143] Level 3 warning: The weighted adaptive coefficient of variation WACV(t) is greater than the upper threshold of anomaly detection θ upper Or less than the anomaly detection lower threshold θ lower More than 30%.

[0144] Preferably, early warning processing: trigger the early warning mechanism, record abnormal information, and take corresponding measures.

[0145] Preferably, through the real-time fault warning mechanism, the system can divide different warning levels according to the degree of change of the WACV value, trigger the warning in time, and notify the operator to take corresponding measures to avoid the expansion of the fault. The graded warning provides flexible risk warnings and helps to ensure the safe operation of the shield machine.

[0146] Clean and extract fault features based on abnormal data.

[0147] Preferably, interpolation processing is performed based on abnormal data to obtain cleaned data

[0148]

[0149] Among them, X i” (t-1) represents the abnormal data at time t-1, X i” (t+1) represents the abnormal data at time t+1.

[0150] Preferably, extracting fault features based on the cleaned data;

[0151] Fault characteristics include: time characteristics, parameter characteristics and statistical characteristics.

[0152] Preferably, the time characteristics include: the frequency and duration of the abnormality; the parameter characteristics include: the rate of change of each parameter during the abnormality; and the statistical characteristics include: the WACV mean and the WACV standard deviation.

[0153] For the detected abnormal data, the system uses interpolation and other methods to clean it to ensure the continuity of the data. The cleaned data is used to extract fault features, including time features and parameter features, to provide high-quality data for subsequent fault diagnosis.

[0154] Example 3

[0155] like Figure 3 As shown, the shield machine abnormal data detection system based on weighted adaptive coefficient of variation includes: a data acquisition and processing module, a coefficient of variation acquisition module, a threshold acquisition module, a result output module and a result processing module;

[0156] The data acquisition and processing module is used to obtain data from multiple sensors on the shield machine; preprocess the sensor data to obtain preprocessed data;

[0157] A coefficient of variation acquisition module is used to obtain the corresponding volatility index and the coefficient of variation within the window based on the preprocessed data; obtain the adaptive weight corresponding to the preprocessed data based on the volatility index; obtain the weighted adaptive coefficient of variation of all preprocessed data based on the coefficient of variation within the window and the adaptive weight;

[0158] A threshold acquisition module, used to obtain an anomaly detection threshold based on a weighted adaptive coefficient of variation;

[0159] A result output module is used to obtain abnormal data based on the comparison between the weighted adaptive coefficient of variation and the abnormal detection threshold;

[0160] The result processing module is used to clean and extract fault features based on abnormal data.

[0161] Preferably, each module in the system of the present invention can implement the above-mentioned method accordingly.

[0162] Example 4

[0163] like Figure 4 As shown, the shield machine abnormal data detection system based on weighted adaptive coefficient of variation includes: a data acquisition and processing module, a coefficient of variation acquisition module, a threshold acquisition module, a result output module and a result processing module;

[0164] The data acquisition and processing module is used to obtain data from multiple sensors on the shield machine; preprocess the sensor data to obtain preprocessed data;

[0165] A coefficient of variation acquisition module is used to obtain the corresponding volatility index and the coefficient of variation within the window based on the preprocessed data; obtain the adaptive weight corresponding to the preprocessed data based on the volatility index; obtain the weighted adaptive coefficient of variation of all preprocessed data based on the coefficient of variation within the window and the adaptive weight;

[0166] A threshold acquisition module, used to obtain an anomaly detection threshold based on a weighted adaptive coefficient of variation;

[0167] A result output module is used to obtain abnormal data based on the comparison between the weighted adaptive coefficient of variation and the abnormal detection threshold;

[0168] The result processing module is used to clean and extract fault features based on abnormal data.

[0169] Preferably, each module in the system of the present invention can implement the above-mentioned method accordingly.

[0170] Preferably, it also includes: a graded warning module;

[0171] The graded warning module is used to classify warning levels based on the proportion of weighted adaptive coefficient of variation exceeding the anomaly detection threshold:

[0172] Level 1 warning: The weighted adaptive coefficient of variation WACV(t) is greater than the anomaly detection upper threshold θ upper Or less than the anomaly detection lower threshold θ lower Exceed 10-20%;

[0173] Secondary warning: The weighted adaptive coefficient of variation WACV(t) is greater than the upper threshold of anomaly detection θ upper Or less than the anomaly detection lower threshold θ lower Exceed 20-30%;

[0174] Level 3 warning: The weighted adaptive coefficient of variation WACV(t) is greater than the upper threshold of anomaly detection θ upper Or less than the anomaly detection lower threshold θ lower More than 30%.

[0175] Early warning processing: trigger the early warning mechanism, record abnormal information, and take corresponding measures.

[0176] Example 5

[0177] Based on the same inventive concept, the present invention also provides a computer device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0178] Memory, used to store computer programs;

[0179] The processor, when used to execute the program stored in the memory, can implement the shield machine abnormal data detection method based on weighted adaptive coefficient of variation as in Example 1 or 2.

[0180] like Figure 5 As shown, the electronic device may include: a processor 41, a communication interface 42, a memory 43 and a communication bus 44, wherein the processor 41, the communication interface 42 and the memory 43 communicate with each other through the communication bus 44. The processor 41 may call the logic instructions in the memory 43 to execute the shield machine abnormal data detection method based on the weighted adaptive coefficient of variation in embodiment 1 or 2.

[0181] In addition, the logic instructions in the above-mentioned memory 43 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0182] It can be seen from the above technical solutions that the present invention discloses a shield machine abnormal data detection method and system based on weighted adaptive coefficient of variation, which has the following beneficial effects:

[0183] 1. Improve the accuracy of anomaly detection: Through the weighted adaptive coefficient of variation method, the multi-parameter data of the shield machine is comprehensively analyzed to make full use of the fault sensitivity of each parameter; through the adaptive weight adjustment mechanism, the algorithm can respond to changes in the operating status and environment in real time, improve the accuracy of anomaly detection, and reduce the probability of misjudgment and missed judgment.

[0184] 2. Enhanced real-time and robustness: The present invention uses a sliding window to dynamically monitor and analyze the data, and combines it with a real-time adjusted threshold to quickly detect abnormal points in the data; WACV has low computational complexity and is suitable for online real-time monitoring. It has strong robustness and can effectively cope with complex and changeable environmental interference during the operation of the shield machine.

[0185] 3. Strong multi-parameter data fusion capability: The present invention adopts the weighted coefficient of variation method to fuse the multi-parameter data of the shield machine, which can effectively process multi-dimensional, nonlinear and non-stationary data, and improve the overall effect of data cleaning and fault detection; through adaptive weight adjustment, a reasonable evaluation of the volatility of each parameter data is achieved, thereby improving the ability of multi-parameter data fusion.

[0186] 4. The early warning mechanism is flexible and effective: Through the real-time fault early warning mechanism, the system can divide the early warning levels according to the degree of change of the WACV value, trigger the early warning in time, and notify the operator to take corresponding measures to avoid the expansion of the fault; the graded early warning provides flexible risk warnings, which helps to ensure the safe operation of the shield machine.

[0187] 5. Data cleaning and fault feature extraction: The present invention can not only detect and clean abnormal data, but also extract key fault features, providing a high-quality data basis for subsequent fault diagnosis and analysis, thereby improving the accuracy and efficiency of fault diagnosis.

[0188] 6. Easy for engineering application: The method of the present invention has low computational complexity, is easy to implement, and is suitable for deployment and promotion in actual projects. Its real-time online monitoring characteristics enable it to be directly applied to the monitoring system of the shield machine, providing a practical solution for status monitoring and fault warning of engineering equipment.

[0189] 7. The present invention has significant advantages in improving the accuracy, real-time and reliability of shield machine operation data, can effectively solve the deficiencies of existing technologies in data cleaning and anomaly detection, and provide strong support for the safe operation and maintenance management of shield machines.

[0190] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0191] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A shield machine abnormal data detection method based on weighted adaptive coefficient of variation, characterized in that: include: Obtain data from multiple different sensors on the shield machine; Performing preprocessing based on the sensor data to obtain preprocessed data; Obtaining corresponding volatility index and coefficient of variation within the window based on the preprocessed data; Obtaining an adaptive weight corresponding to the preprocessed data based on the volatility index; Obtaining a weighted adaptive coefficient of variation of all the preprocessed data based on the coefficient of variation within the window and the adaptive weight; The anomaly detection threshold is obtained based on the historical normal data of the shield machine and prior experience; Obtaining abnormal data based on comparing the weighted adaptive coefficient of variation with the abnormality detection threshold; Cleaning is performed based on the abnormal data and fault features are extracted.

2. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 1 is characterized in that: The sensor data includes: shield machine pressure data, shield machine temperature data, shield machine vibration data and shield machine current data.

3. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 1 is characterized in that: The volatility index acquisition method is: Based on the preprocessed data, the mean value μ in the window of the i-th preprocessed data is obtained. i (t); Based on the mean μ within the window i (t) Get the volatility index δ of the i-th preprocessed data at time t i (t): Among them, N represents the total window data length in the sliding window, k represents the current window step length, and X i' (tk) represents the i-th preprocessed data at time tk.

4. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 3 is characterized in that: The method for obtaining the coefficient of variation within the window is: Based on the volatility indicator δ i (t) and the mean μ in the window i (t) Obtain the coefficient of variation CV within the window corresponding to the i-th preprocessed data i (t): in, Represents a small positive number, preventing the denominator from being zero.

5. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 4 is characterized in that: The adaptive weight is specifically: w i (t)=α·w i (t-1)+(1-α)·f(δ i (t)); Among them, w i (t) represents the adaptive weight of the i-th preprocessed data at time t, α represents the smoothing coefficient, and w i (t-1) represents the adaptive weight of the i-th preprocessed data at time t-1, and f represents the mapping function.

6. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 5 is characterized in that: The method for obtaining the anomaly detection threshold is: The normal weighted adaptive coefficient of variation is calculated based on the historical normal data of the shield machine and prior experience; Based on the normal weighted adaptive variation coefficient, a normal weighted adaptive variation coefficient mean μ is obtained. WACV' ; Based on the normal weighted adaptive coefficient of variation mean μ WACV' Get the weighted adaptive coefficient of variation standard deviation σ WACV' ; Based on μ WACV' and σ WACV' Get the upper threshold θ for anomaly detection upper and the anomaly detection threshold θ lower : i upper =μ WACV' +q·s WACV' i lower =μ WACV' -q·s WACV' Where q represents the threshold adjustment coefficient; The abnormality detection upper threshold θ upper and the anomaly detection lower threshold θ lower Together they constitute the anomaly detection threshold.

7. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 6 is characterized in that: The specific method to obtain abnormal data is: Based on the comparison between the weighted adaptive coefficient of variation and the anomaly detection threshold: If the weighted adaptive variation coefficient is greater than the abnormality detection upper threshold or less than the abnormality detection lower threshold, it is determined to be the abnormal data; otherwise, it is determined to be normal data.

8. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 7 is characterized in that: Also includes: Based on the proportion of the weighted adaptive coefficient of variation exceeding the anomaly detection threshold, the warning level is divided: Level 1 warning: the weighted adaptive coefficient of variation is greater than the upper threshold of abnormal detection or less than the lower threshold of abnormal detection by more than 10-20%; Level 2 warning: the weighted adaptive coefficient of variation is greater than the upper threshold of abnormal detection or less than the lower threshold of abnormal detection by more than 20-30%; Level 3 warning: the weighted adaptive coefficient of variation is greater than the anomaly detection upper threshold or less than the anomaly detection lower threshold by more than 30%.

9. The shield machine abnormal data detection method based on weighted adaptive coefficient of variation according to claim 1 is characterized in that: Cleaning and extracting fault features based on the abnormal data specifically includes: Performing interpolation processing based on the abnormal data to obtain cleaned data; extracting the fault feature based on the cleaning data; The fault characteristics include: time characteristics, parameter characteristics and statistical characteristics.

10. A shield machine abnormal data detection system based on weighted adaptive coefficient of variation, applied to a shield machine abnormal data detection method based on weighted adaptive coefficient of variation as claimed in any one of claims 1 to 9, characterized in that: include: Data acquisition and processing module, coefficient of variation acquisition module, threshold acquisition module, result output module and result processing module; The data acquisition and processing module is used to obtain multiple different sensor data on the shield machine; Performing preprocessing based on the sensor data to obtain preprocessed data; The coefficient of variation acquisition module is used to obtain the corresponding volatility index and the coefficient of variation within the window based on the preprocessed data; Obtaining an adaptive weight corresponding to the preprocessed data based on the volatility index; Obtaining a weighted adaptive coefficient of variation of all the preprocessed data based on the coefficient of variation within the window and the adaptive weight; The threshold acquisition module is used to obtain an anomaly detection threshold based on the weighted adaptive coefficient of variation; The result output module is used to obtain abnormal data based on the comparison between the weighted adaptive coefficient of variation and the abnormal detection threshold; The result processing module is used to clean and extract fault features based on the abnormal data.

Images