A method, device and equipment for fault monitoring of a magnetic suspension system

Abstract

This invention discloses a fault monitoring method, apparatus, and device for a magnetic suspension system. In this solution, fault monitoring of the magnetic suspension system can be achieved by calculating the fault assessment value of each pair of suspension electromagnets under each fault condition and comparing it with the fault threshold. Furthermore, the target parameter vector used in calculating the fault assessment value is a parameter vector that satisfies the constraint condition of the lowest fault missed detection rate, thereby enabling this solution to more accurately detect faulty suspension electromagnets and ensure the safety of the magnetic suspension system.

Description

Technical Field

[0001] This invention relates to the field of fault monitoring technology, and more specifically, to a fault monitoring method, apparatus, and equipment for a magnetic suspension system. Background Technology

[0002] A magnetic suspension balance is a device that uses magnetic force to suspend an aircraft model and provides necessary data for calculating some aerodynamic parameters of the aircraft during wind tunnel experiments. With the continuous development and maturation of wind tunnel magnetic suspension balance technology and its successful application in wind tunnel testing, the safe and reliable operation of the magnetic suspension balance suspension system has received increasing attention. During wind tunnel testing with a magnetic suspension balance, a malfunction in the magnetic suspension system will inevitably lead to the uncontrolled crash of the test model, endangering the lives of the test personnel. If a malfunction in the magnetic suspension system can be detected and an alarm can be triggered, personnel casualties can be largely avoided. Therefore, how to accurately monitor malfunctions in magnetic suspension systems is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0003] The purpose of this invention is to provide a fault monitoring method, device, and equipment for magnetic suspension systems, so as to accurately monitor the faults of magnetic suspension systems.

[0004] To achieve the above objectives, the present invention provides a fault monitoring method for a magnetic suspension system, comprising:

[0005] Acquire the input and output data of each pair of suspended electromagnets at each time interval;

[0006] Determine a data-driven stable kernel representation and a target parameter vector for each type of fault; wherein the target parameter vector is a parameter vector that satisfies the minimum fault miss rate constraint.

[0007] Using the input data, the output data, the stable kernel representation, and the target parameter vector, the fault residual value corresponding to each fault is determined;

[0008] The target fault assessment value corresponding to each fault is determined by the fault residual value and the fault-free residual value corresponding to each fault, and the fault monitoring result is determined by comparing the target fault assessment value corresponding to each fault with the fault threshold.

[0009] The input data includes current data, and the output data includes attitude data and acceleration data.

[0010] The process of determining the data-driven stable kernel representation includes:

[0011] Construct a data-driven input-output data model;

[0012] Based on the data-driven input-output data model and historical data, a stable kernel representation based on data-driven principles is obtained.

[0013] The construction of the data-driven input-output data model includes:

[0014] Based on the state data, input data, output data, process noise data, and output noise data of the magnetic suspension system when it is fault-free, construct the corresponding first state data matrix, first input data matrix, first output data matrix, first process noise data matrix, and first output noise data matrix.

[0015] Construct a second input data matrix corresponding to the input data, construct a second output data matrix corresponding to the output data, construct a second process noise data matrix corresponding to the process noise data, and construct a second output noise data matrix corresponding to the output noise data. The second input data matrix, the second output data matrix, the second process noise data matrix, and the second output noise data matrix are all Hankel matrices.

[0016] Construct a third input data matrix corresponding to the input data, and construct a third process noise data matrix corresponding to the process noise data. Both the third input data matrix and the third process noise data matrix are Toeplitz matrices.

[0017] The data-driven input-output data model is constructed using the first state data matrix, the second input data matrix, the second output data matrix, the second process noise data matrix, the second output noise data matrix, the third input data matrix, the third process noise data matrix, and the system extended observability matrix.

[0018] The process of obtaining a data-driven stable kernel representation based on the data-driven input-output data model and historical data includes:

[0019] A fourth input data matrix and a third output data matrix are constructed based on historical data; both the fourth input data matrix and the third output data matrix are Hankel matrices.

[0020] A first target matrix is ​​constructed using the fourth input data matrix, the third output data matrix, the second input data matrix, and the second output data matrix, and QR decomposition is performed on the first target matrix to generate the decomposition result;

[0021] Using a data-driven input-output data model and the first decomposition result, a second target matrix is ​​determined, and the stable kernel representation is obtained by performing SVD decomposition on the second target matrix.

[0022] The method for constructing the target parameter vector for each type of fault includes:

[0023] A fault-free confidence set is constructed based on the fault-free data of each pair of suspended electromagnets, and a fault confidence set is constructed based on the fault data of each pair of suspended electromagnets.

[0024] Using the fault-free confidence set and the fault confidence set, determine the target parameter vector for each fault that satisfies the minimum fault missed detection rate constraint.

[0025] The step of determining the target parameter vector for each fault that satisfies the minimum fault miss rate constraint using the fault-free confidence set and the fault confidence set includes:

[0026] Set a first upper limit for the false alarm rate and a second upper limit for the false alarm rate.

[0027] The minimum fault miss rate constraint is constructed based on the mean of the first state detection variable under fault-free conditions in the fault-free confidence set, as well as the first upper limit value and the second upper limit value.

[0028] Using the mean and variance of the fault-free state detection variables in the fault-free confidence set and the mean and variance of the fault-free second state detection variables in the fault confidence set, calculate the target parameter vector for each type of fault that satisfies the minimum fault missed detection rate constraint.

[0029] The process of determining the fault monitoring result based on the comparison between the target fault assessment value and the fault threshold includes:

[0030] Determine whether the target fault assessment value for each type of fault is greater than the corresponding fault threshold.

[0031] If the target fault assessment value corresponding to a target fault is greater than the fault threshold, it is determined that a target fault has occurred in the suspension electromagnet, and an alarm message is generated.

[0032] To achieve the above objectives, the present invention further provides a fault monitoring device for a magnetic suspension system, comprising:

[0033] The acquisition module is used to acquire the input and output data of each pair of suspended electromagnets at each time interval;

[0034] The parameter vector determination module is used to determine the data-driven stable kernel representation and the target parameter vector for each type of fault; wherein, the target parameter vector is a parameter vector that satisfies the minimum fault false detection rate constraint.

[0035] The fault residual value determination module is used to determine the fault residual value corresponding to each fault by using the input data, the output data, the stability kernel representation and the target parameter vector;

[0036] The target fault assessment value determination module is used to determine the target fault assessment value corresponding to each fault by using the fault residual value and the fault-free residual value corresponding to each fault.

[0037] The comparison module is used to determine the fault monitoring result based on the comparison between the target fault assessment value and the fault threshold corresponding to each fault.

[0038] To achieve the above objectives, the present invention further provides an electronic device, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to implement the steps of the above-described fault monitoring method when executing the computer program.

[0041] As can be seen from the above solutions, the present invention provides a fault monitoring method, apparatus, and device for a magnetic suspension system. In this application, it is necessary to acquire the input and output data of each pair of suspension electromagnets at each time interval, and determine a data-driven stable kernel representation and a target parameter vector for each type of fault. The target parameter vector is a parameter vector that satisfies the minimum fault miss rate constraint. Then, using the input data, output data, stable kernel representation, and target parameter vector, the fault residual value corresponding to each fault is determined. The target fault evaluation value corresponding to each fault is determined by the fault residual value and the fault-free residual value. The fault monitoring result is determined based on the comparison between the target fault evaluation value and the fault threshold. Therefore, by calculating the fault evaluation value of the suspension electromagnet under each fault and comparing it with the fault threshold, this application can achieve fault monitoring of the magnetic suspension system. Furthermore, the target parameter vector used in calculating the fault evaluation value satisfies the minimum fault miss rate constraint, thereby enabling the application to more accurately detect faulty suspension electromagnets and ensure the safety of the magnetic suspension system. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1This is a schematic flowchart of a fault monitoring method for a magnetic suspension system disclosed in an embodiment of the present invention;

[0044] Figure 2 This is a schematic flowchart of a specific fault monitoring method for a magnetic suspension system disclosed in an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of the structure of a fault monitoring device for a magnetic suspension system disclosed in an embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram of an electronic device structure disclosed in an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] This invention discloses a fault monitoring method, device, and equipment for a magnetic suspension system, so as to accurately monitor the faults of the magnetic suspension system.

[0049] See Figure 1 The present invention provides a schematic flowchart of a fault monitoring method for a magnetic suspension system, which includes:

[0050] S11. Obtain the input and output data of each pair of suspended electromagnets at each time interval;

[0051] The magnetic suspension system typically includes five pairs of independent suspension electromagnets, which control the X-axis, Y-axis forward, Y-axis backward, Z-axis forward, and Z-axis backward directions, respectively. However, each pair of suspension electromagnets experiences the same fault in practice. Therefore, in this application, the same monitoring scheme can be used to monitor the faults of each pair of suspension electromagnets and obtain the final fault monitoring results. The time interval range in this application can be customized according to actual conditions and is not specifically limited here. Furthermore, the input data in this embodiment includes current data, and the output data includes attitude data and acceleration data.

[0052] It should be noted that most current fault detection schemes for magnetic suspension systems are based on single-variable fault detection, failing to fully utilize the system's operational data. For example, most magnetic control systems rely on gap data for fault detection. Single data points cannot contain all fault information; therefore, fault detection results derived solely from gap data processing undoubtedly increase the false negative rate. This solution utilizes the current, attitude, and acceleration data of each pair of suspension electromagnets to achieve multivariate fault detection. The current data includes current data from all five pairs of electromagnets; the attitude data includes five degrees of freedom (translation along the x, y, and z axes, pitch, and yaw); and the acceleration data includes filtered acceleration data for all five degrees of freedom. Through comprehensive analysis of this multivariate data, this solution reduces the false negative rate and improves fault detection effectiveness.

[0053] S12. Determine the data-driven stable kernel representation and the target parameter vector for each type of fault; wherein, the target parameter vector is a parameter vector that satisfies the minimum fault false detection rate constraint.

[0054] In this scheme, when monitoring the magnetic suspension system for faults, a minimum missed detection rate is required. Therefore, this scheme pre-sets constraints, specifically minimum missed detection rate constraints. When determining the target parameter vector corresponding to each fault, the target parameter vector must satisfy the minimum missed detection rate constraint, so that the determined target parameter vector is the parameter vector that minimizes the missed detection rate. Furthermore, the fault types in this embodiment are diverse, such as sensor faults (current sensor malfunction, attitude sensor fault), actuator faults (electromagnet short circuit or open circuit), etc., which are not specifically limited here.

[0055] S13. Using the input data, output data, stable kernel representation, and target parameter vector, determine the fault residual value corresponding to each fault.

[0056] Specifically, after obtaining the input and output data, this application constructs an input-output data matrix based on the input and output data. The input-output data matrix, the stable kernel representation, and the target parameter vector corresponding to each fault are then input into the residual generator to obtain the fault residual value corresponding to each fault, so as to calculate the fault evaluation value through the fault residual value of each fault.

[0057] It should be noted that the residual generator in the traditional scheme is based on model design. Since the magnetic suspension balance in this scheme is a nonlinear system and a system with variable model parameters, this scheme uses a data-driven stabilization kernel to build the residual generator in order to improve feasibility. This stabilization kernel can be obtained directly from the data without relying on the model, so it has high feasibility.

[0058] S14. Determine the target fault assessment value corresponding to each fault by using the fault residual value and the fault-free residual value corresponding to each fault, and determine the fault monitoring result based on the comparison result between the target fault assessment value corresponding to each fault and the fault threshold.

[0059] It is understood that, when calculating the fault-free residual value, this application needs to calculate the state detection variable under fault-free conditions based on the input data, output data, and stable kernel representation under fault-free conditions, and then calculate the mean of the state detection variable. The fault-free residual value corresponding to each fault is then calculated using the mean and the target parameter vector for each fault. In order to calculate the target fault evaluation value corresponding to each fault using the fault residual value and the fault-free residual value corresponding to each fault, the target fault evaluation value for each fault is used to determine whether the suspension electromagnet has a certain type of fault.

[0060] In summary, this application can monitor the faults of the magnetic suspension system by calculating the fault assessment value of each pair of suspension electromagnets under each fault condition and comparing it with the fault threshold. Furthermore, the target parameter vector used in calculating the fault assessment value is a parameter vector that satisfies the constraint condition of the lowest fault missed detection rate, thereby enabling this application to more accurately detect faulty suspension electromagnets and ensure the safety of the magnetic suspension system.

[0061] See Figure 2 The present invention provides a schematic flowchart of a specific fault monitoring method for a magnetic suspension system, which includes:

[0062] S21. Obtain the input and output data of each pair of suspended electromagnets at each time interval;

[0063] In this scheme, the input data is represented as u(k), and the output data is represented as y(k); where k is the sampling time, u(k) represents the input data at the k-th sampling time, and y(k) represents the output data at the k-th sampling time; the time interval in this scheme is [ks,k], where s represents the data length. The input data of the magnetic levitation electromagnet in this time interval is defined as:

[0064] u s (k)=[u T (ks) u T (k-s+1) … u T (k)] T Similarly, the output data for this time interval is defined as: y s (k)=[y T (ks) y T (k-s+1) … y T (k)]T It should be noted that this scheme can set a sliding time window, let k = k + 1. In this way, the input and output data of each time interval can be obtained to perform real-time fault monitoring of the suspended electromagnet.

[0065] S22. Construct a data-driven input-output data model, and obtain a data-driven stable kernel representation based on the data-driven input-output data model and historical data.

[0066] In this embodiment, the process of constructing a data-driven input-output data model specifically includes:

[0067] S2211. Based on the state data, input data, output data, process noise data and output noise data of the magnetic suspension system when there is no fault, construct the corresponding first state data matrix, first input data matrix, first output data matrix, first process noise data matrix and first output noise data matrix.

[0068] Specifically, this scheme first sets the state-space model of the magnetic suspension balance suspension system as follows:

[0069]

[0070] Where x(k)∈R n Let n be the system's state data, and n be the dimension of the state data. k is the input data for the system. u Let y(k) ∈ R be the dimension of the input data. m The output data of the system is given by m, where m is the dimension of the output data; A, B, C, and D in the formula are system parameter matrices of a certain dimension; w(k)∈R n The system's process noise data; v(k)∈R m This is the system's output noise data.

[0071] Furthermore, this solution also needs to collect process data at fault-free times. This process data includes the aforementioned state data x(k), input data u(k), output data y(k), process noise data w(k), and output noise data v(k). Then, based on these data, a data-driven first data matrix is ​​constructed, such as using the state data x(k) to construct the first state data matrix X. k Construct the first input data matrix U using the input data u(k). k Construct the first output data matrix Y using the output data y(k). k Construct the first process noise data matrix W using process noise data w(k). k The first output noise data matrix V is constructed using the output noise data v(k). kHere, we only construct the first input data matrix U using the input data u(k). k Taking this as an example, the first input data matrix is ​​specifically as follows:

[0072] in, It means k u The first data matrix is ​​a real matrix or vector of dimension ×N, where N is the length of the data. N should be large enough to contain more information about the process. The construction method of the other first data matrix is ​​the same as that of the first input data matrix, so it will not be described in detail here.

[0073] S2212. Construct a second input data matrix corresponding to the input data, construct a second output data matrix corresponding to the output data, construct a second process noise data matrix corresponding to the process noise data, construct a second output noise data matrix corresponding to the output noise data, and the second input data matrix, the second output data matrix, the second process noise data matrix, and the second output noise data matrix are all Hankel matrices.

[0074] This scheme constructs the first input data matrix U k First output data matrix Y k The first process noise data matrix w(k) and the first output noise data matrix V k Next, a corresponding second data matrix needs to be constructed, which includes: the Hankel matrix corresponding to the input data u(k); and the second input data matrix U. k,s The Hankel matrix corresponding to the output data matrix y(k): the second output data matrix Y k,s The Hankel matrix corresponding to the process noise data w(k): the second process noise data matrix W k,s The Hankel matrix corresponding to the output noise data v(k): the second output noise data matrix V k,s Here, we will only focus on constructing the second output data matrix Y. k,s Let's take an example to illustrate:

[0075]

[0076] Where s is the data length, and s≥n; other second data matrices can be constructed in this way, which will not be elaborated here.

[0077] S2213. Construct a third input data matrix corresponding to the input data, and construct a third process noise data matrix corresponding to the process noise data. Both the third input data matrix and the third process noise data matrix are Toeplitz matrices.

[0078] This scheme also requires constructing the Toeplitz matrix corresponding to the data u(k): the third input data matrix H u,s Construct the Toeplitz matrix corresponding to the process noise data w(k): the third process noise data matrix H w,s The third input data matrix is ​​specifically as follows: The noise data matrix for the third process is as follows: In the formula, A, B, C, and D are all system parameter matrices.

[0079] S2214. Construct a data-driven input-output data model using the first state data matrix, the second input data matrix, the second output data matrix, the second process noise data matrix, the second output noise data matrix, the third input data matrix, the third process noise data matrix, and the system extended observability matrix.

[0080] In this scheme, the system extended observability matrix is ​​represented by Γ. s It means, and

[0081] The data-driven input / output data model constructed in this scheme is as follows:

[0082] Y k,s =Γ s X k +H u,s U k,s +H w,s W k,s +V k,s .

[0083] Furthermore, this plan makes Where I is the identity matrix, the above data-driven input-output data model can be reconstructed as: The data-driven stable kernel representation κ in this application d Specifically:

[0084] Among them, Ψ s ⊥ This represents the specific implementation of the stable kernel representation, and Ψ s ⊥ For matrix Ψ s The orthogonal complement.

[0085] In this embodiment, the process of obtaining a data-driven stable kernel representation based on a data-driven input-output data model and historical data includes:

[0086] S2221. Construct a fourth input data matrix and a third output data matrix based on historical data; both the fourth input data matrix and the third output data matrix are Hankel matrices.

[0087] In this embodiment, it is necessary to construct Hankel matrices for past input and output data based on historical data. Here, the fourth input data matrix U is used respectively. k-p-1,p and the third output data matrix Y k-p-1,p Let p represent the length of the historical data and satisfy p ≥ n.

[0088] S2222: Construct a first target matrix using the fourth input data matrix, the third output data matrix, the second input data matrix, and the second output data matrix, and perform QR decomposition on the first target matrix to generate the decomposition result;

[0089] In this embodiment, an intermediate matrix can be constructed based on the fourth input data matrix and the third output data matrix. The intermediate matrix is ​​as follows: The first target matrix constructed based on the intermediate matrix, the second input data matrix, and the second output data matrix is ​​as follows: For the first target matrix The decomposition results obtained from the QR decomposition are as follows: Among them, R in the decomposition result 11 R 21 R 22 R 31 R 32 R 33 Q1, Q2 and Q3 are all submatrices generated after QR decomposition.

[0090] S2223. Using the data-driven input-output data model and the first decomposition result, determine the second target matrix, and obtain a stable kernel representation by performing SVD decomposition on the second target matrix.

[0091] In this scheme, R in the decomposition results 33 Q3 is considered as H in the reconstructed input-output data model. w,s W k,s +V k,s ,but The second objective matrix is ​​then obtained as follows: After performing singular value decomposition (SVD) on the second objective matrix, we obtain: Where U1, U2, D1, D2, V1, and V2 are all submatrices generated after SVD decomposition, and V1 is a submatrix. T It is the transpose of V1. This is the transpose of V2. The stable kernel is then represented by Ψ. s ⊥Specifically:

[0092]

[0093] in, For Γ s The orthogonal complement, and satisfying

[0094] S23. Determine the target parameter vector for each type of fault; where the target parameter vector is a parameter vector that satisfies the minimum fault missed detection rate constraint.

[0095] In this embodiment, the method for constructing the target parameter vector for each type of fault specifically includes:

[0096] S231. Construct a fault-free confidence set based on the fault-free data of each pair of suspended electromagnets, and construct a fault confidence set based on the fault data of each pair of suspended electromagnets.

[0097] In this embodiment, before constructing the confidence set, the state detection variables of the wind tunnel magnetic suspension balance suspension system are defined as follows: Where z(k) is the state detection variable based on input and output data, and the target input data matrix u is defined. s (k), define the target output data matrix y s (k), where the target input data matrix is ​​specifically u s (k)=[u T (ks) u T (k-s+1) … u T (k)] T The target output data matrix is ​​specifically: y s (k)=[y T (ks) y T (k-s+1) …y T (k)] T .

[0098] Furthermore, when constructing the confidence set based on the mean and covariance of the collected input and output data, this application specifically divides it into two cases: no-fault and faulty cases, which are explained in detail here:

[0099] Assume that, under fault-free conditions, the state detection variable z(k) satisfies: Among them, CS h For a fault-free confidence set, ζ z Let z(k) be the set of distributions. and Σ h These are the mean and variance of the state detection variables when there are no faults. and Σ hIt is calculated from health data, which can usually be selected from the fault-free data when the system is first put into operation, or from the data of a fault-free operation.

[0100] Understandably, in the initial stage of operation of the magnetic levitation balance system, the lack of fault data makes it difficult to analyze the data distribution characteristics of faults. Therefore, this solution will pre-design a fault detection system based on the data characteristics during normal operation to detect most potential faults. Assume that the magnetic levitation system will experience M typical faults during operation, and let f be the fault signal when the i-th fault occurs. i (k), i = 1, 2, ..., M. Without loss of generality, we assume that each fault has the same probability of occurrence, and model the faults using the mean and covariance matrices. Let z(k) under fault conditions satisfy: in, For fault confidence sets of fault data, Let be the state detection variable at the i-th fault; for The mean, Σ fi for The variance of the fault data can be used directly to build a fault confidence set when available fault data is available. If there is no fault data or little fault data, the mean and covariance of M fault cases can be defined to build a confidence set. When new fault data becomes available, the data is recorded and the confidence set is updated offline.

[0101] S232. Using the fault-free confidence set and the fault confidence set, determine the target parameter vector for each fault that satisfies the minimum fault missed detection rate constraint.

[0102] The specific calculation method of the target parameter vector in this scheme is as follows: set a first upper limit value for the false alarm rate and a second upper limit value for the false alarm rate; construct the minimum false alarm rate constraint condition based on the mean of the first state detection variable under fault-free conditions in the fault-free confidence set, as well as the first and second upper limit values; and calculate the target parameter vector of each fault that satisfies the minimum false alarm rate constraint condition by using the mean and variance of the state detection variable under fault-free conditions in the fault-free confidence set and the mean and variance of the second state detection variable under fault conditions in the fault confidence set.

[0103] Specifically, this solution must meet a minimum false alarm rate (FAR) when performing fault monitoring. The minimum FAR of the magnetic suspension system can be expressed as the false alarm rate P. F Minimize the failure rate P when fixed M That is, consider when P F When P is fixed and acceptable, how can we make P... M Minimization. The design of the residual generator based on the lowest false negative rate in this scheme ultimately transforms into obtaining the parameter vector g.i (g i (≠0) ensures that the parameter vector satisfies the minimum fault missed detection rate constraint, which is specifically:

[0104] minα i

[0105]

[0106] In the formula, z is a simplified description of z(k), and β i ∈(0,1) represents the false alarm rate P. F The upper limit of α i ∈(0,1) represents the upper limit of the fault missed detection rate. To distinguish between these two upper limits, β is... i Let α represent the first upper limit value of the i-th type of fault. i Let be the second upper limit value for the i-th type of fault, and the objective parameter vector in this scheme is determined as follows:

[0107] S2321, to Performing SVD decomposition yields the following results: Wherein, U in the formula i S i V i All are submatrices generated after SVD decomposition.

[0108] S2322 Solve the objective equation system The result is v i .

[0109] Where, λ m,i The largest eigenvalue, For S i The inverse matrix, where Ξ is the first intermediate parameter. The second intermediate parameter is calculated as follows: The third intermediate parameter is calculated as follows:

[0110] S2323, According to the formula Calculate the fourth intermediate parameter g i .

[0111] S2324, According to the formula Calculate the target parameter vector The target parameter vector is the target parameter vector that minimizes the false negative rate.

[0112] S24. Using the input data, output data, stable kernel representation, and target parameter vector, determine the fault residual value corresponding to each fault.

[0113] Specifically, after obtaining input and output data for a certain time interval, this application constructs the corresponding target input matrix u based on the input data for that time interval. s (k), construct the target output data matrix y based on the output data. s (k) then the input and output data matrices are obtained based on the target input matrix and the target output matrix. The residual generator in this scheme is specifically as follows: Let i be the target parameter vector corresponding to the i-th type of fault. for transpose, Ψ s ⊥ For the stable kernel representation, the fault residual value r corresponding to the i-th fault can be obtained through this residual generator. i (k).

[0114] S25. Determine the target fault assessment value corresponding to each fault by using the fault residual value and the fault-free residual value corresponding to each fault, and determine the fault monitoring result based on the comparison result between the target fault assessment value corresponding to each fault and the fault threshold.

[0115] In this embodiment, the fault-free residual value is: r i,h ,and The mean of the state detection variables under fault-free conditions recorded in the fault-free confidence set is used; therefore, the fault-free residual value can be pre-calculated and stored for direct use during fault monitoring. The specific method for calculating the target fault assessment value for each type of fault in this scheme is as follows: Among them, J(r) i Let be the target fault assessment value corresponding to the i-th type of fault. Furthermore, when determining the fault monitoring result based on the comparison between the target fault assessment value and the fault threshold, this scheme specifically determines whether the target fault assessment value corresponding to each fault is greater than the corresponding fault threshold. If there is a target fault whose target fault assessment value is greater than the fault threshold, then it is determined that a target fault has occurred in the suspension electromagnet, and an alarm message is generated. Specifically, this scheme can set the fault detection and assessment unit of the magnetic suspension system as follows:

[0116]

[0117] In other words: if the i-th target fault evaluation value of the suspended electromagnet is J(r) i () greater than the fault threshold J th If the i-th type of fault has occurred in the suspension electromagnet, an alarm message will be generated immediately for timely warning. If the i-th target fault assessment value of the suspension electromagnet is less than or equal to the fault threshold J, the alarm message will be generated immediately. thThis indicates that the suspension electromagnet is not faulty; in this embodiment, a fault threshold J can be set. th The value is 1.

[0118] The fault monitoring device, equipment, and storage medium provided in the embodiments of the present invention are described below. The fault monitoring device, equipment, and storage medium described below can be referred to in conjunction with the fault monitoring method described above.

[0119] See Figure 3 The present invention provides a schematic diagram of a fault monitoring device for a magnetic suspension system, comprising:

[0120] The acquisition module 11 is used to acquire the input and output data of each pair of suspended electromagnets at each time interval;

[0121] The parameter vector determination module 12 is used to determine the data-driven stable kernel representation and the target parameter vector for each type of fault; wherein, the target parameter vector is a parameter vector that satisfies the minimum fault false detection rate constraint.

[0122] The fault residual value determination module 13 is used to determine the fault residual value corresponding to each fault by using the input data, the output data, the stability kernel representation and the target parameter vector;

[0123] The target fault assessment value determination module 14 is used to determine the target fault assessment value corresponding to each fault by using the fault residual value and the fault-free residual value corresponding to each fault.

[0124] The comparison module 15 is used to determine the fault monitoring result based on the comparison result between the target fault assessment value and the fault threshold corresponding to each fault.

[0125] The parameter vector determination module includes:

[0126] Model construction unit, used to construct data-driven input-output data models;

[0127] The stable kernel representation determination unit is used to obtain a data-driven stable kernel representation based on the data-driven input-output data model and historical data.

[0128] The model construction unit includes:

[0129] The first construction subunit is used to construct the corresponding first state data matrix, first input data matrix, first output data matrix, first process noise data matrix and first output noise data matrix based on the state data, input data, output data, process noise data and output noise data of the magnetic suspension system when it is fault-free.

[0130] The second construction subunit is used to construct a second input data matrix corresponding to the input data, a second output data matrix corresponding to the output data, a second process noise data matrix corresponding to the process noise data, and a second output noise data matrix corresponding to the output noise data. The second input data matrix, the second output data matrix, the second process noise data matrix, and the second output noise data matrix are all Hankel matrices.

[0131] The third construction subunit is used to construct a third input data matrix corresponding to the input data and a third process noise data matrix corresponding to the process noise data. Both the third input data matrix and the third process noise data matrix are Toeplitz matrices.

[0132] The fourth construction subunit is used to construct the data-driven input-output data model using the first state data matrix, the second input data matrix, the second output data matrix, the second process noise data matrix, the second output noise data matrix, the third input data matrix, the third process noise data matrix, and the system extended observability matrix.

[0133] The stable kernel representation determination unit includes:

[0134] The fifth construction subunit is used to construct a fourth input data matrix and a third output data matrix based on historical data; both the fourth input data matrix and the third output data matrix are Hankel matrices.

[0135] The sixth construction subunit is used to construct a first target matrix using the fourth input data matrix, the third output data matrix, the second input data matrix, and the second output data matrix;

[0136] The first decomposition subunit is used to perform QR decomposition on the first target matrix and generate decomposition results;

[0137] The second decomposition subunit is used to determine the second target matrix using a data-driven input-output data model and the first decomposition result, and to obtain the stable kernel representation by performing SVD decomposition on the second target matrix.

[0138] The parameter vector determination module includes:

[0139] The confidence set construction unit is used to construct a fault-free confidence set based on the fault-free data of each pair of suspended electromagnets, and to construct a fault confidence set based on the fault data of each pair of suspended electromagnets.

[0140] The parameter vector determination unit is used to determine the target parameter vector of each fault that satisfies the minimum fault missed detection rate constraint by using the fault-free confidence set and the fault confidence set.

[0141] The parameter vector determination unit includes:

[0142] The sub-unit is configured to set the first upper limit for the false alarm rate and the second upper limit for the missed fault rate.

[0143] The seventh construction subunit is used to construct the minimum fault missed detection rate constraint condition based on the mean of the first state detection variable under fault-free conditions in the fault-free confidence set, as well as the first upper limit value and the second upper limit value.

[0144] The calculation subunit is used to calculate the target parameter vector of each fault that satisfies the minimum fault missed detection rate constraint by using the mean and variance of the fault-free state detection variables in the fault-free confidence set and the mean and variance of the fault-free second state detection variables in the fault confidence set.

[0145] The comparison module includes:

[0146] The judgment unit is used to determine whether the target fault assessment value corresponding to each fault is greater than the corresponding fault threshold; if there is a target fault assessment value corresponding to a target fault that is greater than the fault threshold, it is determined that the suspension electromagnet has a target fault and an alarm message is generated.

[0147] See Figure 4 The present invention provides a schematic diagram of an electronic device structure, comprising:

[0148] Memory 21 is used to store computer programs;

[0149] The processor 22 is configured to implement the steps of the fault monitoring method described in any of the above method embodiments when executing the computer program.

[0150] In this embodiment, the device can be a PC (Personal Computer), or a terminal device such as a tablet computer, PDA, or portable computer.

[0151] The device may include a memory 21, a processor 22, and a bus 23.

[0152] The memory 21 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 21 can be an internal storage unit of the device, such as the hard disk of the device. In other embodiments, the memory 21 can also be an external storage device of the device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 21 can include both internal and external storage units of the device. The memory 21 can be used not only to store application software and various types of data installed on the device, such as program code executing fault monitoring methods, but also to temporarily store data that has been output or will be output.

[0153] In some embodiments, processor 22 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 21 or process data, such as program code for executing a fault monitoring method.

[0154] This bus 23 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0155] Furthermore, the device may also include a network interface 24, which may optionally include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), typically used to establish communication connections between the device and other electronic devices.

[0156] Optionally, the device may further include a user interface 25, which may include a display, an input unit such as a keyboard, and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the device and to display a visual user interface.

[0157] Figure 4 Only devices with components 21-25 are shown; those skilled in the art will understand that... Figure 4 The structure shown does not constitute a limitation on the device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0158] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the fault monitoring method described in any of the above method embodiments.

[0159] The storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0160] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0161] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those 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 invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault monitoring method for a magnetic suspension system, characterized in that, include: Acquire the input and output data of each pair of suspended electromagnets at each time interval; Determine a data-driven stable kernel representation and a target parameter vector for each type of fault; wherein the target parameter vector is a parameter vector that satisfies the minimum fault miss rate constraint. Using the input data, the output data, the stable kernel representation, and the target parameter vector, the fault residual value corresponding to each fault is determined; The target fault assessment value corresponding to each fault is determined by the fault residual value and the fault-free residual value corresponding to each fault, and the fault monitoring result is determined by comparing the target fault assessment value corresponding to each fault with the fault threshold. The process of determining a stable kernel representation based on data-driven principles includes: Based on the state data, input data, output data, process noise data, and output noise data of the magnetic suspension system when it is fault-free, construct the corresponding first state data matrix, first input data matrix, first output data matrix, first process noise data matrix, and first output noise data matrix. Construct a second input data matrix corresponding to the input data, construct a second output data matrix corresponding to the output data, construct a second process noise data matrix corresponding to the process noise data, and construct a second output noise data matrix corresponding to the output noise data. The second input data matrix, the second output data matrix, the second process noise data matrix, and the second output noise data matrix are all Hankel matrices. Construct a third input data matrix corresponding to the input data, and construct a third process noise data matrix corresponding to the process noise data. Both the third input data matrix and the third process noise data matrix are Toeplitz matrices. The data-driven input-output data model is constructed using the first state data matrix, the second input data matrix, the second output data matrix, the second process noise data matrix, the second output noise data matrix, the third input data matrix, the third process noise data matrix, and the system extended observability matrix. A fourth input data matrix and a third output data matrix are constructed based on historical data; both the fourth input data matrix and the third output data matrix are Hankel matrices. A first target matrix is ​​constructed using the fourth input data matrix, the third output data matrix, the second input data matrix, and the second output data matrix, and QR decomposition is performed on the first target matrix to generate the decomposition result; Using a data-driven input-output data model and the first decomposition result, a second target matrix is ​​determined, and the stable kernel representation is obtained by performing SVD decomposition on the second target matrix.

2. The fault monitoring method according to claim 1, characterized in that, The input data includes current data, and the output data includes attitude data and acceleration data.

3. The fault monitoring method according to claim 1, characterized in that, The method for constructing the target parameter vector for each type of fault includes: A fault-free confidence set is constructed based on the fault-free data of each pair of suspended electromagnets, and a fault confidence set is constructed based on the fault data of each pair of suspended electromagnets. Using the fault-free confidence set and the fault confidence set, determine the target parameter vector for each fault that satisfies the minimum fault missed detection rate constraint.

4. The fault monitoring method according to claim 3, characterized in that, The step of determining the target parameter vector for each fault that satisfies the minimum fault miss rate constraint using the fault-free confidence set and the fault confidence set includes: Set a first upper limit for the false alarm rate and a second upper limit for the false alarm rate. The minimum fault miss rate constraint is constructed based on the mean of the first state detection variable under fault-free conditions in the fault-free confidence set, as well as the first upper limit value and the second upper limit value. Using the mean and variance of the fault-free state detection variables in the fault-free confidence set and the mean and variance of the fault-free second state detection variables in the fault confidence set, calculate the target parameter vector for each type of fault that satisfies the minimum fault missed detection rate constraint.

5. The fault monitoring method according to any one of claims 1 to 4, characterized in that, The fault monitoring results are determined based on the comparison between the target fault assessment value and the fault threshold, including: Determine whether the target fault assessment value for each type of fault is greater than the corresponding fault threshold. If the target fault assessment value corresponding to a target fault is greater than the fault threshold, it is determined that a target fault has occurred in the suspension electromagnet, and an alarm message is generated.

6. A fault monitoring device for a magnetic suspension system, characterized in that, include: The acquisition module is used to acquire the input and output data of each pair of suspended electromagnets at each time interval; The parameter vector determination module is used to determine the data-driven stable kernel representation and the target parameter vector for each type of fault; wherein, the target parameter vector is a parameter vector that satisfies the minimum fault false detection rate constraint. The fault residual value determination module is used to determine the fault residual value corresponding to each fault by using the input data, the output data, the stability kernel representation and the target parameter vector; The target fault assessment value determination module is used to determine the target fault assessment value corresponding to each fault by using the fault residual value and the fault-free residual value corresponding to each fault. The comparison module is used to determine the fault monitoring result based on the comparison between the target fault assessment value and the fault threshold corresponding to each fault. The parameter vector determination module includes: The first construction subunit is used to construct the corresponding first state data matrix, first input data matrix, first output data matrix, first process noise data matrix and first output noise data matrix based on the state data, input data, output data, process noise data and output noise data of the magnetic suspension system when it is fault-free. The second construction subunit is used to construct a second input data matrix corresponding to the input data, a second output data matrix corresponding to the output data, a second process noise data matrix corresponding to the process noise data, and a second output noise data matrix corresponding to the output noise data. The second input data matrix, the second output data matrix, the second process noise data matrix, and the second output noise data matrix are all Hankel matrices. The third construction subunit is used to construct a third input data matrix corresponding to the input data and a third process noise data matrix corresponding to the process noise data. Both the third input data matrix and the third process noise data matrix are Toeplitz matrices. The fourth construction subunit is used to construct the data-driven input-output data model using the first state data matrix, the second input data matrix, the second output data matrix, the second process noise data matrix, the second output noise data matrix, the third input data matrix, the third process noise data matrix, and the system extended observability matrix. The fifth construction subunit is used to construct a fourth input data matrix and a third output data matrix based on historical data; both the fourth input data matrix and the third output data matrix are Hankel matrices. The sixth construction subunit is used to construct a first target matrix using the fourth input data matrix, the third output data matrix, the second input data matrix, and the second output data matrix; The first decomposition subunit is used to perform QR decomposition on the first target matrix and generate decomposition results; The second decomposition subunit is used to determine the second target matrix using a data-driven input-output data model and the first decomposition result, and to obtain the stable kernel representation by performing SVD decomposition on the second target matrix.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the fault monitoring method as described in any one of claims 1 to 5 when executing the computer program.

Images