Modeling and fault diagnosis method and device of hydraulic drive cutterhead system of shield machine
Through the combination of adaptive AR model and Hotelling T2 statistics, fault prediction and diagnosis of the hydraulic drive cutter plate system of the shield machine is realized, and the fault problem in the existing technology is solved, which is difficult to describe under complex geological conditions, and the stability of the system is improved. and fault detection accuracy.
Patent Information
- Application Number
- CN202111276020.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-29
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-10-29
AI Technical Summary
The prior art is difficult to effectively predict and avoid the failure of the hydraulic drive cutter plate system of the shield machine under complex geological conditions, resulting in the system being unable to work normally.
Using the hydraulically driven shield machine cutter plate system modeling method based on adaptive AR model, the AR model is constructed by real-time data acquisition, and fault diagnosis is performed using online identification and Hotelling T2 statistics to achieve an approximate description of the system's nonlinear dynamic characteristics and Failure prediction.
It improves the adaptability and accuracy of the model, can predict faults in advance under various operating conditions, ensures the normal operation of the cutting board system, and enhances the stability and reliability of the system.
Smart Images

Figure BDA0003329422550000021 
Figure BDA0003329422550000022 
Figure BDA0003329422550000027
Abstract
Description
Technical Field
[0001] The invention relates to the field of fault diagnosis of a cutter head system of a shield machine, and in particular to a modeling and fault diagnosis method of a cutter head system of a hydraulically driven shield machine based on an adaptive AR model. Background Art
[0002] As one of the key components of shield machine equipment, the cutterhead drive system is the main working device for tunneling operations. The cutterhead speed of the shield machine is not high, but due to the large diameter of the cutterhead and the complex geological structure of the construction, the cutterhead drive system is required to have functions such as high power, large output torque, wide output speed variation range, impact resistance, two-way rotation of the cutterhead and escape. At the same time, under the condition of meeting the use requirements, it has the working characteristics of reducing installed power, saving energy and reducing consumption. The cutterhead drive system must also have high reliability and good stability.
[0003] During the tunneling process, the shield machine moves forward axially along the designated route. Due to the complexity of the geological conditions of the soil layer and the unpredictability of many factors during the construction process, such as encountering hard rock geology during tunneling, the cutterhead cannot cut, and it is more likely to cause the blade to break. Whether the cutterhead system can work normally directly affects the tunneling process of the entire shield machine. Therefore, the tunneling system of the shield machine must be able to provide early warning for various sudden situations. The cutterhead is an important component of the shield machine. Fault analysis of the cutterhead system is the primary and basic task to ensure that the shield machine cutterhead system can work normally in a complex environment. The cutterhead system of the shield machine is a complex system with strong nonlinearity, uncertainty and coupling. The main variables reflecting the working state of the shield machine cutterhead are the cutterhead speed, the cutterhead hydraulic system pressure and the cutterhead torque, and the geological conditions of different tunneling sections will affect the working state of the cutterhead. Therefore, it is difficult to establish a mathematical model based on the mechanism analysis method, and the performance of the mechanism model depends on the accuracy of physical parameters, some of which are sensitive to the actual working environment. When the external environment changes significantly, the mechanism model may not be able to describe the operating status of the cutter head system. Summary of the invention
[0004] The technical problem to be solved by the present invention is to provide a modeling and fault diagnosis method and device for a hydraulically driven cutter head system of a shield machine in view of the deficiencies in the prior art, so as to predict in advance the possible faults of the hydraulically driven cutter head system of the shield machine under various working conditions, thereby avoiding these faults in advance and ensuring the normal operation of the cutter head system.
[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a modeling method of a hydraulically driven cutterhead system of a shield machine, comprising:
[0006] S1. Real-time collection of data reflecting the operating status of the cutterhead system in the shield machine as identification data, and extraction of identification data of several continuous working sections within a time window as modeling data of the cutterhead system of the hydraulically driven shield machine;
[0007] The identification data includes the cutter head rotation speed, the cutter head hydraulic system pressure and the cutter head torque;
[0008] S2. Construct the AR model structure of the hydraulically driven shield machine cutter system:
[0009] Y(τ)=Y zero +α1Y(τ-1)+α2Y(τ-2)+…+α p Y(τ-p)+δ(τ);
[0010] Where: τ is the sampling time, T zero is the model offset, α1~α p is the regression coefficient matrix; p is the AR model order; is the identification data; δ(τ) is the modeling error, w=(Y zero ,α1,α2,...,α p ) is the parameter set to be identified, i.e., the coefficients of the AR model;
[0011] S3. Estimate the cutterhead system AR model parameter w, and take the model order corresponding to the minimum AIC value as the optimal order p of the cutterhead system AR model * :
[0012]
[0013]
[0014] in, Y k K(Y k (p+1),Y k (p+2),...,Y k (N k )) T , Representation Matrix The sum of the squares of the elements in Y k (1)~Y k (N k ) represents the identification data of the shield machine cutter system obtained by sampling in a certain time window; k = 1, 2, ..., K, k is the number of the continuous advancement operation data segment in the window, N k is the length of the kth data segment, and K is the number of consecutive sampled data segments in the time window;
[0015] The fault diagnosis analysis of the cutterhead system involved in the present invention is a predictive analysis, which requires the cutterhead characteristic model to meet the requirements of rapid calculation, easy prediction, and easy parameter estimation, and the AR model can meet these requirements well. In order to overcome the problem that it is difficult to describe the dynamic characteristics of the system when the geological conditions change by using the cutterhead system mechanism model, the present invention proposes to use an online identified adaptive AR model to describe the dynamic characteristics of the cutterhead system to ensure that the model has good local feature description capabilities and adaptability to changes in geological conditions. Since the hydraulically driven shield machine cutterhead system is a complex system with nonlinear characteristics, the use of a fixed parameter AR model alone can only describe the local characteristics of the cutterhead system during the advancement of the shield machine, and the nonlinear characteristics of the system will be lost to a certain extent, and the accuracy of the model will be reduced. In view of the problem that the nonlinear description ability of the AR model is insufficient, the present invention proposes a shield machine cutterhead system modeling method based on an adaptive AR model, that is, according to the different characteristics of the system dynamic characteristics in different operating time periods, the entire working range of the cutterhead system is divided into several small intervals (time windows), and then, for each small interval, a linearized AR model of the system is established. Since the prediction model can update the model parameters in real time according to the changes in the operating conditions in different time periods using the latest process input and output data, it is possible to achieve adaptive modeling of the shield machine cutter head system, thereby achieving an approximate description of the nonlinear dynamic characteristics of the entire working range of the system. Compared with the modeling method that directly adopts the nonlinear model, the cutter head system modeling method based on the adaptive AR model of the present invention converts the nonlinear optimization problem into a linear optimization problem, and it is more reasonable to divide the system working range according to the adaptability of the model. The present invention does not consider external control input, and uses all variables related to the cutter head system as the output signal of the AR model to obtain the prediction residual information of each signal. The residual information removes the autocorrelation characteristics of the signal to a certain extent, is suitable for statistical control quantity analysis, improves the modeling accuracy, and predicts in advance the possible faults of the hydraulically driven cutter head system of the shield machine under various working conditions, thereby avoiding these faults in advance and ensuring the normal operation of the cutter head system.
[0016] In the present invention, the optimal order p * The corresponding AR model is the diagnosis model M.
[0017] The present invention also provides a shield machine hydraulic drive cutter head system modeling method, comprising the following steps:
[0018] 1) Perform fault analysis on the modeling error of the fault diagnosis model M and establish an online error monitoring model. The modeling error of the training sample data is recorded as E train , E train is an n×3 matrix, That is E train There are n samples and 3 observed variables;
[0019] 2) Modeling error E train After decentralization and standardization, the error matrix Es is calculated: YesE train The mean of YesE train The standard deviation of
[0020] 3) Get the covariance matrix Ecor of the standardized error matrix Es:
[0021] 4) Perform eigendecomposition on the covariance matrix Ecor and select the number of principal components r;
[0022] 5) Calculate Hotelling's T using the following formula: 2 Control Limits: Among them, F α (r,nr) is a noncentral F distribution with parameter α and degrees of freedom r and nr;
[0023] 6) Online collection of the real-time data Y(m) of the shield machine cutterhead system at the current mth moment, and calculation of the prediction error based on the AR model of the shield machine cutterhead system Will Standardize, get
[0024] 7) Calculation T 2 Statistics And make fault judgment: If Then the data at the mth moment has a fault; if Then the data at the mth moment is normal;
[0025] Among them, the diagnostic model M is established using the above modeling method.
[0026] The present invention aims at the hydraulic drive cutterhead system of the shield machine, adopts a modeling method based on an adaptive AR model to establish a dynamic model of the cutterhead system of the shield machine, and performs online fault diagnosis on the cutterhead system. The method for modeling and fault diagnosis of the dynamic characteristics of the cutterhead system of the shield machine based on the adaptive AR model proposed by the present invention adopts a moving window mode to update the model parameters online, performs statistical analysis on the modeling error of the model, and adopts Hotelling T 2 The residual control chart is established by online calculation of the prediction error of the adaptive AR model for new sample data. 2 Statistics are used to monitor out-of-control points, through the out-of-control point T 2The contribution rate determines the fault variable, realizes the online monitoring and fault diagnosis of the multivariable statistical process of the cutter head system, and improves the accuracy of fault detection.
[0027] In the above step 4), the calculation formula for the number of principal components r is: When f(r)≥98%, the corresponding r is the number of principal components; f(r) is the cumulative variance contribution rate of the principal component; λ i is the i-th eigenvalue of the covariance matrix Ecor. By extracting the principal element, redundant variables can be removed and the accuracy of fault diagnosis can be improved.
[0028] In the above step 7), P is the first r columns of R, containing all the principal element information of R; R is a 3×3 eigenvector matrix of S, and S is the covariance matrix E COR The eigenvalue matrix of .
[0029] In step 7), when it is determined that a fault has occurred, the following steps are further performed:
[0030] definition calculate Each variable e m,j Score z for the i-th pivot m,i Contributions: abs represents the absolute value function, p(j,i) is the element in the jth row and ith column of P; j = 1, 2, 3;
[0031] calculate The jth variable pair in T 2 The sum of the contributions of the r principal components of the statistic: Here m is the sampling time of the detection data, and j is the sequence number of the variable in the detection data;
[0032] Compare The total contribution rate of the three variables, namely CONTABC m,1 ,CONTABC m,2 ,CONTABC m,3 The larger the value, the greater the contribution to the fault. By comparing the sizes of these three values, the variable corresponding to the larger value is determined to be the main variable causing the fault.
[0033] In step 7), when the data at the mth moment is normal, the following operation is also performed: obtaining a data group within a new time window, and returning to execute step 1).
[0034] As an inventive concept, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the steps of the method of the present invention.
[0035] As an inventive concept, the present invention also provides a computer-readable storage medium having a computer program / instruction stored thereon; the computer program / instruction, when executed by a processor, implements the steps of the method of the present invention.
[0036] As an inventive concept, the present invention also provides a computer program product, including a computer program / instruction; when the computer program / instruction is executed by a processor, the steps of the method of the present invention are implemented.
[0037] Compared with the prior art, the invention has the following beneficial effects: the invention not only has higher model accuracy, but also can make the model adapt to the changes of the external environment, and can well describe the nonlinear dynamic characteristics of the shield machine. The fault diagnosis method of the cutter head system of the hydraulic driven shield machine based on the adaptive AR model can be online constructed and updated based on Hotelling T 2 The residual control chart of the statistical quantity can realize real-time monitoring of out-of-control points and locate fault variables, and has high utilization value. DETAILED DESCRIPTION
[0038] Embodiment 1 of the present invention comprises the following steps:
[0039] 1) In order to establish the AR model of the hydraulically driven shield machine cutterhead system, the data reflecting the operating status of the cutterhead system in the shield machine are collected in real time as identification data, including the cutterhead speed, the cutterhead hydraulic system pressure and the cutterhead torque, and the data sampling period is set to 1 second. Considering the intermittent working characteristics of the shield machine, that is, the continuous working time of the shield machine in one advancement operation is not long enough, resulting in a small number of sampling data in a continuous working section, the data of several continuous working sections in a time window can be extracted as the modeling data of the hydraulically driven shield machine cutterhead system.
[0040] 2) Determine the AR model structure of the hydraulically driven shield machine cutterhead system. The AR model of the shield machine cutterhead system uses the actual value of the cutterhead speed (rpm), the cutterhead hydraulic system pressure (bar), and the cutterhead torque (KNm) as variables. The AR model structure of the hydraulically driven shield machine cutterhead system is constructed as follows:
[0041] Y(τ)=Y zero +α1Y(τ-1)+α2Y(τ-2)+…+α p Y(τ-p)+δ(τ) (1)
[0042] Where: τ is the sampling time, Y zero is the model offset, α1~α pis the regression coefficient matrix; p is the model order; it includes the three signals of the cutter head speed measurement, the cutter head hydraulic system pressure and the cutter head torque; δ(τ) is the modeling error, denoted by w = (Y zero ,α1,α2,...,α p ) is the parameter set to be identified, that is, the coefficients of the AR model.
[0043] 3) The least squares method is used to estimate the cutterhead system AR model parameter w, and the model order corresponding to the minimum AIC value is taken as the optimal order of the cutterhead system AR model. Suppose the data of the shield machine cutterhead system sampled in a certain time window is Y k (1)~Y k (N k ), k = 1, 2, ..., K, k is the number of the continuous advancement operation data segment in the window, N k is the length of the kth data segment, and K is the number of consecutive sampled data segments in the window. Based on the cutterhead system AR model (1) and the sampled data, the one-step forward prediction of the cutterhead system variable Y is calculated.
[0044]
[0045] in,
[0046]
[0047]
[0048] W=w T (5)
[0049] Predicting outputs through a minimal cutterhead system model The actual output Y k The variance of the cutterhead AR model is obtained by:
[0050]
[0051] in:
[0052] Y k =(Y k (p+1),Y k (p+2),...,Y k (N k )) T (7)
[0053] is the matrix composed of the measured data of the k-th section cutter head system variables; Representation Matrix The square sum of the elements in . The parameter estimates of the cutterhead system AR model are obtained from equations (1)–(7) as follows:
[0054]
[0055] in:
[0056]
[0057]
[0058] The order p of the cutterhead system AR model is obtained by searching for the order corresponding to the minimum value of the following AIC:
[0059]
[0060]
[0061] In the formula, p * is the optimal order of the AR model of the cutterhead system.
[0062] The fault diagnosis process of the hydraulically driven cutterhead system of the shield machine of the second embodiment of the present invention includes:
[0063] 1) For the K groups of data obtained in the current window, the cutterhead system AR optimization model established through the above steps 1)-3) is denoted as M, which is called the “current model”.
[0064] 2) Perform fault analysis on the modeling error of M and establish an online error monitoring model. The modeling error of the training sample data is recorded as E train , which is an n×3 matrix, That is E train There are n samples and 3 observed variables in
[0065]
[0066] 3) Modeling error E train Decentralization and standardization
[0067]
[0068] in: YesE train The mean of YesE train The standard deviation of .
[0069] 4) Obtain the covariance matrix Ecor (3×3 matrix) of the standardized error matrix Es
[0070]
[0071] 5) Perform eigendecomposition on the covariance matrix Ecor and select the number of principal components
[0072]
[0073] Where S is a 3×3 diagonal matrix, also E COR The eigenvalue matrix of the matrix, the elements on its diagonal are its eigenvalues λ1, λ2, λ3, and satisfy λ1≥λ2≥λ3, R is a 3×3 eigenvector matrix of S, P is the first r (≤3) columns of R, containing all the principal element information of R, is the remaining 3-column non-pivot information of R, r is determined by the cumulative contribution rate method, and the cumulative variance contribution rate of the first r principal components is recorded as:
[0074]
[0075] When f(r)≥98%, the corresponding r is the number of principal components.
[0076] 6) Calculate Hotelling's T 2 Control limits
[0077] For the process variable with n samples and r principal components, Hotelling's T 2 The statistic follows an F distribution with r and nr degrees of freedom, then the T with a confidence level of α = 99% 2 The upper limit of the statistic is
[0078]
[0079] 7) Online process variable monitoring and fault diagnosis of cutter head system
[0080] a) Online collection of the real-time data Y(m) of the shield machine cutterhead system at the current mth moment, and calculation of the prediction error according to the above optimized shield machine cutterhead system AR model (1)
[0081]
[0082] is a 1×3 vector and normalizes it
[0083]
[0084] b) Calculation T 2 Statistics and fault diagnosis
[0085]
[0086] like If the data at the mth moment has a fault, go to 8) to perform fault signal analysis;
[0087] like Then the data at the mth moment is normal, go to 9).
[0088] in, That is, λ1,…,λ r (r≤3) is a diagonal matrix consisting of diagonal elements.
[0089] 8) In case of variable failure The principal component score vector of , as well as the contribution value of each variable to each principal component score, is used to determine the variable that has failed. To this end, define It is the score vector corresponding to the data at the mth moment, and the fault variable is determined by the following steps:
[0090] a) Calculation Each variable e m,j (j=1,2,3) Score z for the i-th pivot m,i Contributions:
[0091]
[0092] In the above formula, abs represents the absolute value function, and p(j,i) is the element in the j-th row and i-th column of P.
[0093] b) Calculation The jth (j=1,2,3) variable pair in T 2 The sum of the contributions of the r principal components of the statistic
[0094]
[0095] according to The total contribution rate of the three variables, namely CONTABC m,1 ,CONTABC m,2 ,CONTABC m,3 , determine the variable that caused the failure.
[0096] 9) The AR model parameters of the hydraulically driven shield machine cutter head system are updated online using a moving window update mode. When a data set within a new time window is obtained, the "current model" is updated again from step 1) in claim 2, an adaptive AR model of the shield machine cutter head system is established, and the fault diagnosis process of the hydraulically driven shield machine cutter head system is performed.
[0097] The present invention takes a hydraulically driven cutterhead system of a shield machine as an example to further illustrate its specific implementation method.
[0098] 1. First, conduct online adaptive AR modeling of the hydraulic drive cutterhead system of the shield machine using the following steps:
[0099] 1) To establish the AR model of the hydraulic drive cutterhead system of the shield machine, the input and output data of the shield machine cutterhead system subsystem are collected as identification data, including the actual value of the cutterhead speed (rpm), the cutterhead hydraulic system pressure (bar), and the cutterhead torque (KNm), and the data sampling period is set to 1 second. Considering the intermittent working characteristics of the shield machine, that is, the continuous working time of the shield machine in one advancement operation is not long enough, resulting in a small number of sampling data in a continuous working section, 10 continuous working data segments of the shield machine in a certain working period are extracted, and the first 9 data segments are used as a training set window to establish the AR model. The total number of data contained in the training window is 2926, and the last data segment is used as a test window. The total number of data contained in the test window is 50.
[0100] 2) Determine the AR model structure of the hydraulically driven shield machine cutterhead system. The AR model of the shield machine cutterhead system uses the actual value of the cutterhead speed (rpm), the cutterhead hydraulic system pressure (bar), and the cutterhead torque (KNm) as variables. The AR model structure of the hydraulically driven shield machine cutterhead system is constructed as follows:
[0101] Y(τ)=Y zero +α1Y(τ-1)+α2Y(τ-2)+…+α p Y(τ-p)+δ(τ) (1)
[0102] Where: τ is the sampling time, Y zero is the model offset, α1~α p is the regression coefficient matrix; p is the model order; it includes the three signals of the cutter head speed measurement, the cutter head hydraulic system pressure and the cutter head torque; δ(τ) is the modeling error, denoted by w = (Y zero ,α1,α2,...,α p ) is the parameter set to be identified, that is, the coefficients of the AR model.
[0103] 3) The least squares method is used to estimate the cutterhead system AR model parameter w, and the model order corresponding to the minimum AIC value is taken as the optimal order of the cutterhead system AR model. Suppose the data of the shield machine cutterhead system sampled in a certain time window is Y k (1)~Y k (N k ), k = 1, 2, ..., K, k is the number of the continuous advancement operation data segment in the window, N K is the length of the kth data segment, and K is the number of consecutive sampled data segments in the window. Based on the cutterhead system AR model (1) and the sampled data, the one-step forward prediction of the cutterhead system variable Y is calculated:
[0104]
[0105] in,
[0106]
[0107]
[0108] W=w T (5)
[0109] Predicting outputs through a minimal cutterhead system model The actual output Y k The variance of the cutterhead AR model is obtained by:
[0110]
[0111] in:
[0112] Y k =(Y k (p+1),Y k (p+2),...,Y k (N k )) T (7)
[0113] is the matrix composed of the measured data of the k-th section cutter head system variables; Representation Matrix The square sum of the elements in . The parameter estimates of the cutterhead system AR model are obtained from equations (1)–(7) as follows:
[0114]
[0115] in:
[0116]
[0117]
[0118] The order p of the cutterhead system AR model is obtained by searching for the order corresponding to the minimum value of the following AIC:
[0119]
[0120]
[0121] In the formula, p * is the optimal order of the AR model of the cutterhead system.
[0122] The fault diagnosis example is as follows:
[0123] 1) For the K groups of data obtained in the current window, the cutterhead system AR optimization model established through steps 1)-3) in 1 above is denoted as M, which is called the "current model".
[0124] 2) Perform fault analysis on the modeling error of M and establish an online error monitoring model. The modeling error of the training sample data is recorded as E train , which is an n×3 matrix, That is E train There are n samples and 3 observed variables in
[0125]
[0126] 3) Modeling error E train Decentralization and standardization
[0127]
[0128] in: YesE train The mean of YesE train The standard deviation of .
[0129] 4) Obtain the covariance matrix Ecor (3×3 matrix) of the standardized error matrix Es
[0130]
[0131] 5) Perform eigendecomposition on the covariance matrix Ecor and select the number of principal components
[0132]
[0133] Where S is a 3×3 diagonal matrix, also E COR The eigenvalue matrix of the matrix, the elements on its diagonal are its eigenvalues λ1, λ2, λ3, and satisfy λ1≥λ2≥λ3, R is a 3×3 eigenvector matrix of S, P is the first r)≤3) columns of R, containing all the principal element information of R, is the remaining 3-column non-pivot information of R, r is determined by the cumulative contribution rate method, and the cumulative variance contribution rate of the first r principal components is recorded as:
[0134]
[0135] When f(r)≥98%, the corresponding r is the number of principal components.
[0136] 6) Calculate Hotelling's T 2 Control limits
[0137] For the process variable with n samples and r principal components, Hotelling's T 2The statistic follows an F distribution with r and nr degrees of freedom, then the T with a confidence level of α = 99% 2 The upper limit of the statistic is
[0138]
[0139] 7) Online process variable monitoring and fault diagnosis of cutter head system
[0140] a) Online collection of the real-time data Y(m) of the shield machine cutterhead system at the current mth moment, and calculation of the prediction error according to the above optimized shield machine cutterhead system AR model (1)
[0141]
[0142] is a 1×3 vector and normalizes it
[0143]
[0144] b) Calculation T 2 Statistics and fault diagnosis
[0145]
[0146] like If the data at the mth moment has a fault, go to 8) to perform fault signal analysis;
[0147] like Then the data at the mth moment is normal, go to 9).
[0148] in, That is, λ1,…,λ r (r≤3) is a diagonal matrix consisting of diagonal elements.
[0149] 8) In case of variable failure The principal component score vector of , as well as the contribution value of each variable to each principal component score, is used to determine the variable that has failed. To this end, define It is the score vector corresponding to the data at the mth moment, and the fault variable is determined by the following steps:
[0150] a) Calculation Each variable e m,j (j=1,2,3) Score z for the i-th pivot m,i Contributions:
[0151]
[0152] In the above formula, abs represents the absolute value function, and p(j,i) is the element in the j-th row and i-th column of P.
[0153] b) Calculation The jth (j=1,2,3) variable pair in T 2 The sum of the contributions of the r principal components of the statistic
[0154]
[0155] according to The total contribution rate of the three variables, namely CONTABC m,1 ,CONTABC m,2 ,CONTABC m,3 , determine the variable that caused the failure.
[0156] 9) The AR model parameters of the hydraulically driven shield machine cutter head system are updated online using a moving window update mode. When a data set within a new time window is obtained, the "current model" is updated again from step 1) in claim 2, an adaptive AR model of the shield machine cutter head system is established, and the fault diagnosis process of the hydraulically driven shield machine cutter head system is performed.
[0157] According to the above modeling method and parameter optimization method, the order and model parameters of the cutterhead system AR model in a certain time window and the minimum AIC of the AR model are shown in Tables 1 and 2. The model parameter matrix is a 16×3 matrix.
[0158] Table 1 Optimized model order
[0159] Minimum AIC Optimized AR model order <![CDATA[-4.3677×10 3 ]]> p=5
[0160] Table 2 AR model parameter matrix W
[0161] 0.144497 11.924168 -159.604171 0.285527 5.274801 124.241982 -0.000215 0.418427 11.075902 -0.000000 0.006110 0.174290 0.058281 -0.386585 -8.789628 0.000543 0.051582 0.906708 -0.000002 0.003106 0.103720 0.322682 2.457694 57.953336 0.000118 -0.030970 -0.505649 0.000001 0.004150 0.093698 -0.055004 -1.961273 -62.372033 0.000465 -0.008260 -0.3231890 -0.000016 0.004577 0.131535 0.266828 1.517280 52.876673 0.000210 0.044626 0.140772 -0.000003 -0.001346 -0.001878
[0162] According to the above step 3), the mean and variance of the standardized error of AR modeling are obtained, as shown in Table 3:
[0163] Table 3 Mean and variance of standardized error of AR model modeling
[0164] Cutter speed measurement Cutter hydraulic system pressure Cutter torque Mean 1.258412e-13 -6.050201e-13 -6.053151e-13 variance 0.320764 1.091060 1.463026
[0165] According to the above step 4), the covariance matrix E of the AR model modeling error is calculated cor , as shown in Table 4:
[0166] Table 4 Covariance matrix of AR model modeling error
[0167] 1.000000 0.001688 -0.002055 0.001688 1.000000 0.981032 -0.002055 0.981032 1.000000
[0168] According to the above step 5), the covariance matrix is decomposed to obtain the eigenvalue matrix S and the eigenvector matrix R, as shown in Table 7 and Table 8:
[0169] Table 7 Eigenvalue matrix S
[0170] 1.981032 0 0 0 1.000007 0 0 0 0.018960
[0171] Table 8 Eigenvector matrix R
[0172] 0.000264 -0.999996 0.002699 -0.707107 -0.002095 -0.7071034 -0.707107 0.001721 0.707104
[0173] According to the above step 5), it can be known that the number of main components r=2.
[0174] According to the above step 6) and selecting the confidence level α = 0.9999, we can calculate
[0175] Table 9 shows 5 sets of test data with a sampling period of 1 second.
[0176] Table 9 Test data X test
[0177] 1.39999997615814 76 1500 1.37999999523163 76 1500 1.37999999523163 80 1610 1.37999999523163 82 1670 1.47999999523163 86 1770
[0178] According to the above step 7), calculate the T of the 5 sets of prediction data 2 The statistics are shown in Table 10.
[0179] Table 10 T of the fifth group of test data 2 Statistics
[0180] 0.716599 2.549266 1.057479 1.058521 24.517601
[0181] From Table 10, we can see that the T 2 The statistic value is greater than the upper limit Therefore, this group of test data has a fault. According to the above step 8), the total contribution of each variable is calculated, as shown in Table 11.
[0182] Table 11 Total contribution rate of each variable to failure in the fifth group of test data
[0183] 99.984346 0.007692 0.007962
[0184] It can be concluded from the above table that the cutter disc speed in the test data plays a dominant role in the occurrence of the fault, that is, the cutter disc speed fails at this moment.
Claims
1. A fault diagnosis method for a hydraulically driven cutterhead system of a shield machine, characterized in that: The following steps are involved: 1) Perform fault analysis on the modeling error of the fault diagnosis model M and establish an online error monitoring model. The modeling error of the training sample data is recorded as E train , E train is an n×3 matrix, That is E train There are n samples and 3 observed variables; 2) Modeling error E train After decentralization and standardization, the error matrix Es is calculated: YesE train The mean of YesE train The standard deviation of 3) Get the covariance matrix Ecor of the standardized error matrix Es: 4) Perform eigendecomposition on the covariance matrix Ecor and select the number of principal components r; 5) Calculate Hotelling using the following formula ′ sT 2 Control Limits: where F α (r,nr) is a noncentral F distribution with parameter α and degrees of freedom r and nr; 6) Online collection of the real-time data Y(m) of the shield machine cutterhead system at the current mth moment, and calculation of the prediction error based on the AR model of the shield machine cutterhead system Will Standardize, get 7) Calculation T 2 Statistics And make fault judgment: If Then the data at the mth moment has a fault; if Then the data at the mth moment is normal; Among them, the establishment process of the fault diagnosis model M includes: S1. Real-time data reflecting the operation status of the cutterhead system in the shield machine is collected as identification data, and identification data of several continuous working sections in a time window are extracted as modeling data of the cutterhead system of the hydraulically driven shield machine; the identification data includes the cutterhead rotation speed, the cutterhead hydraulic system pressure and the cutterhead torque; S2. Construct the AR model structure of the hydraulically driven shield machine cutter system: Y(τ)=Y zero +α1Y(τ-1)+α2Y(τ-2)+…+α p γ(τ-p)+δ(τ); Where: τ is the sampling time, Y zero is the model offset, α1~α p is the regression coefficient matrix; p is the AR model order; Y is the identification data; δ(τ) is the modeling error, w=(Y zero ,α1,α2,...,α p ) is the parameter set to be identified, i.e., the coefficients of the AR model; S3. Estimate the cutterhead system AR model parameter w, and take the model order corresponding to the minimum AIC value as the optimal order p of the cutterhead system AR model * : in, AND k (And k (p+1),Y k (p+2),...,And k (N k )) T , Representation Matrix The sum of the squares of the elements in Y k (1)~Y k (N k ) represents the identification data of the shield machine cutter system obtained by sampling in a certain time window; k = 1, 2, ..., K, k is the number of the continuous advancement operation data segment in the window, N k is the length of the kth data segment, and K is the number of consecutive sampled data segments in the time window; With the optimal order p * The corresponding AR model is the fault diagnosis model M.
2. The fault diagnosis method of the hydraulic drive cutterhead system of a shield machine according to claim 1 is characterized in that: In step 4), the calculation formula for the number of principal components r is: When f(r)≥98%, the corresponding r is the number of principal components; f(r) is the cumulative variance contribution rate of the principal component; λ i is the i-th eigenvalue of the covariance matrix Ecor.
3. The fault diagnosis method of the hydraulic drive cutterhead system of a shield machine according to claim 2 is characterized in that: In step 7), P is the first r columns of R, containing all the principal element information of R; R is a 3×3 eigenvector matrix of S, and S is the covariance matrix E COR The eigenvalue matrix of .
4. The shield machine hydraulic drive cutterhead system modeling method according to any one of claims 1 to 3, characterized in that: In step 7), when it is determined that a fault has occurred, the following steps are further performed: definition calculate Each variable e m,j Score z for the i-th pivot m,i Contributions: abs represents the absolute value function. p(j,i) is the element in the jth row and ith column of P; j = 1, 2, 3; calculate The jth variable pair in T 2 The sum of the contributions of the r principal components of the statistic: Here m is the sampling time of the detection data, and j is the sequence number of the variable in the detection data; Compare The total contribution rate of the three variables, namely CONTABC m,1 ,CONTABC m,2 ,CONTABC m,3 The greater the total contribution rate, the greater the contribution to the failure; by comparing CONTABC m,1 ,CONTABC m,2 ,CONTABC m,3 The variable with the larger value is determined as the main variable causing the fault.
5. The fault diagnosis method of the hydraulic drive cutterhead system of a shield machine according to claim 1, characterized in that: In step 7), when the data at the mth moment is normal, the following operation is also performed: obtaining a data group within a new time window, and returning to execute step 1).
6. A computer device comprising a memory, a processor and a computer program stored in the memory; characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.
7. A computer-readable storage medium having a computer program / instruction stored thereon; characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.
8. A computer program product comprising a computer program / instructions; characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.
