Fault diagnosis method, device, equipment, storage medium and program product

Abstract

The present application relates to a fault diagnosis method, apparatus, device, storage medium, and program product. The method includes: collecting historical process data of a target system, constructing a fault detection model and a fault subspace library based on the historical process data, wherein the fault subspace library includes at least one candidate fault type; calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics; obtaining a sample to be tested, and calculating the current error statistics of the sample to be tested; if the current error statistics are not less than the fault threshold, reconstructing the sample to be tested based on the candidate fault type in sequence, and if the reconstructed error statistics of the reconstructed sample are less than the fault threshold, determining the candidate fault type currently used for reconstruction as the fault type of the sample to be tested. This method is applicable to target systems with dynamic and nonlinear characteristics, thus expanding its scope of application.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a fault diagnosis method, apparatus, device, storage medium, and program product. Background Art

[0002] With the development and progress of control theory and computer technology, military and civilian industrial systems have continued to grow in size and complexity. The development of sensor and computer technology has also made it possible to collect and store large amounts of process data. Consequently, data-driven fault diagnosis methods have been widely used in military and civilian industrial systems.

[0003] However, some military and civilian industrial systems are affected by time operation, and the process data collected from them have dynamic characteristics and nonlinearity. Traditional data-driven methods are difficult to apply to fault diagnosis of military and civilian industrial systems with the above characteristics. Summary of the Invention

[0004] Based on this, it is necessary to provide a fault diagnosis method, device, equipment, storage medium and program product that can be applied to dynamic data and nonlinear systems to address the above technical problems.

[0005] In a first aspect, the present application provides a fault diagnosis method, comprising:

[0006] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0007] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0008] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0009] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0010] In one embodiment, the constructing of a fault subspace library based on the historical process data includes:

[0011] Classifying the fault data in the historical process data to obtain multiple groups of fault data of different types, and determining the fault types obtained by classification as candidate fault types;

[0012] Based on the fault subspace extraction method, the fault features of each type of fault data are extracted, and the fault features are associated with the candidate fault types to obtain a corresponding fault subspace library.

[0013] In one embodiment, the constructing of a fault detection model based on the historical process data includes:

[0014] The fault detection model is constructed based on a canonical variable analysis method, wherein the input data of the fault detection model is the normal operating data, and the output of the fault detection model is a first canonical variable corresponding to the normal operating data.

[0015] In one embodiment, calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics, includes:

[0016] Calculating a squared prediction error statistic of the historical process data based on a residual between the historical process data and the first canonical variable, and determining the squared prediction error statistic as a historical error statistic;

[0017] A control limit of the historical error statistic is calculated based on a sampling distribution, and the control limit is determined as the fault threshold.

[0018] In one embodiment, obtaining a sample to be tested and calculating a current error statistic of the sample to be tested includes:

[0019] Preprocessing the sample to be tested, and calculating a second normalized variable of the preprocessed sample to be tested;

[0020] Based on the residual between the sample to be tested and the second standard variable, a square prediction error statistic of the sample to be tested is calculated, and the square prediction error statistic is determined as the current error statistic.

[0021] In one embodiment, the method further comprises:

[0022] When the current error statistic value is less than the fault threshold, determining the state of the sample to be tested as no fault;

[0023] After reconstructing the sample to be tested based on all candidate fault types in the fault subspace library, the fault manifestation of the sample to be tested is recorded when the statistical values ​​of the reconstruction errors of the reconstructed sample are not less than the fault threshold.

[0024] In a second aspect, the present application further provides a fault diagnosis device, comprising:

[0025] a construction module, configured to collect historical process data of a target system and construct a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0026] A first calculation module is configured to calculate historical error statistics of the target system based on the fault detection model, and calculate a fault threshold based on the historical error statistics;

[0027] A second calculation module is used to obtain a sample to be tested and calculate a current error statistic value of the sample to be tested;

[0028] A reconstruction module is used to reconstruct the samples to be tested in sequence based on the candidate fault types when the current error statistic is not less than the fault threshold, and to determine the candidate fault type currently used for reconstruction as the fault type of the sample to be tested when the reconstruction error statistic of the reconstructed sample is less than the fault threshold.

[0029] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0030] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0031] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0032] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0033] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0035] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0036] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0037] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0038] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0039] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0040] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0041] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0042] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0043] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0044] The aforementioned fault diagnosis method, apparatus, device, storage medium, and program product first acquire relevant process data. They then incorporate fault reconstruction technology into canonical variable analysis to extract the fault direction from the process data, thereby eliminating abnormal conditions and restoring normal conditions. This provides a complete fault diagnosis solution based on fault reconstruction and canonical variable analysis, supporting fault diagnosis in target systems with dynamic characteristics. This solution is applicable to various types of big data systems with dynamic and nonlinear characteristics, expanding its scope of application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 1 is a flow chart of a fault diagnosis method in one embodiment;

[0047] Figure 2 is a flow chart of a fault diagnosis method in another embodiment;

[0048] Figure 3 1 is a flow chart of a fault diagnosis method in another embodiment;

[0049] Figure 4 Schematic diagram of experimental results of a fault diagnosis method in one embodiment;

[0050] Figure 5 is a schematic diagram of experimental results of a fault diagnosis method in another embodiment;

[0051] Figure 6 Schematic diagram of experimental results of a fault diagnosis method in another embodiment;

[0052] Figure 7 is a structural block diagram of a fault diagnosis device in one embodiment;

[0053] Figure 8 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0055] With the development and advancement of control theory and computer technology, military and civilian industrial systems are growing in size and complexity. Failure to promptly detect and accurately diagnose a fault can cause unimaginable damage and loss. Fault diagnosis technology has evolved over half a century since 1967, resulting in numerous methods and classifications. Common fault diagnosis classifications fall into two categories: qualitative and quantitative.

[0056] Among them, qualitative methods mainly include: (1) graph theory methods; (2) expert systems; and (3) qualitative simulation. Quantitative methods mainly include: (1) analytical model-based methods; and (2) data-driven methods. In industrial systems such as military and civilian systems, system complexity is increasing proportionally with the number of systems. At the same time, the development of sensor technology and computer technology has made it possible to collect and store large amounts of process data.

[0057] In this context, fault diagnosis using analytical model-based methods is becoming increasingly difficult due to system modeling issues. Data-driven approaches primarily include: 1) machine learning methods such as neural networks and support vector machines; 2) multivariate statistical methods such as principal component analysis, canonical variate analysis (CVA), and independent component analysis; 3) signal processing methods such as wavelet transforms; and 4) information fusion. Multivariate statistical methods such as CVA are among the most commonly used data-driven fault diagnosis methods. Algorithms such as principal component analysis and independent component analysis typically perform fault diagnosis under the assumption that process data is steady and undisturbed. However, real-world military and civilian industrial systems are often subject to time-dependent fluctuations, and the process data collected from them exhibits dynamic characteristics. Therefore, the CVA fault diagnosis method has been developed specifically to address the dynamic nature of process data.

[0058] The most commonly used data-driven fault diagnosis methods are contribution diagrams and fault reconstruction. Although contribution diagrams have the advantages of being simple and not requiring prior knowledge, they also have three fatal drawbacks: (1) nonlinearity is a common characteristic of military and civilian industrial systems, but they cannot be applied to systems with nonlinear characteristics; (2) fault diagnosis methods based on contribution diagrams are based on the assumption that the statistics used have good fault detection effects; and (3) fuzzy effects between process variables may lead to confusion in fault diagnosis results.

[0059] Based on this, the present application provides a fault diagnosis method based on fault reconstruction, such as Figure 1 As shown, this embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0060] Step 101: Collect historical process data of the target system and build a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0061] Step 102, calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0062] Step 103: Obtain a sample to be tested and calculate the current error statistics of the sample to be tested;

[0063] Step 104: When the current error statistic is not less than the fault threshold, the sample to be tested is reconstructed in sequence based on the candidate fault type. When the reconstruction error statistic of the reconstructed sample is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0064] The target system refers to various types of systems with large amounts of process data and dynamic data, including military and civilian industrial systems, etc., and there is no specific restriction here. Figure 2 Historical process data refers to the large amount of process data collected from a dynamic, multivariable target system. This includes both normal operating data and fault data from various failure modes. After collecting the process data, data consolidation is performed to combine the fault data from various failure modes into a fault dataset. After this is completed, data preprocessing is performed, such as removing outliers.

[0065] Historical error statistics refer to error statistics derived from historical process data and the fault detection model. The test sample refers to the sample collected at the current moment during online fault diagnosis, and the current error statistics refer to the error statistics corresponding to the sample at the current moment. The reconstruction process is used to determine the fault type when a fault exists in the test sample.

[0066] For example, see Figure 2 This embodiment is divided into offline and online processes. The process of establishing the fault detection model, fault subspace library, and fault threshold calculation can be performed offline; when applied to fault diagnosis of test samples, it can be performed online. During fault diagnosis, the fault detection model, fault threshold, and fault subspace library previously acquired offline can be directly called.

[0067] The above-described fault diagnosis method first acquires relevant process data and then incorporates fault reconstruction technology into canonical variable analysis to extract the fault direction from the process data, thereby eliminating abnormal conditions and restoring normal conditions. This embodiment provides a complete fault diagnosis solution based on fault reconstruction and canonical variable analysis, supporting fault diagnosis for target systems with dynamic characteristics. It is applicable to military and civilian industrial systems with dynamic and nonlinear characteristics, expanding its scope of application.

[0068] In an exemplary embodiment, building a fault detection model based on the historical process data includes:

[0069] The fault detection model is constructed based on canonical variate analysis (CVA), wherein the input data of the fault detection model is the normal operation data, and the output of the fault detection model is a first canonical variate corresponding to the normal operation data.

[0070] Canonical variate analysis is a linear dimensionality reduction technique that reduces high-dimensional data by maximizing the correlation between two sets of variables. This technique yields a set of canonical variates that best explain the information contained in the set. Using canonical variate analysis, optimal predictions of future outputs can be made based on the system's past and present states, thereby achieving process identification. The modeling process for canonical variate analysis can be found in:

[0071] The historical process data is preprocessed to obtain the data matrix X. The variable state space model identified by CVA is:

[0072] Formula 1:

[0073] ;

[0074] Where, , A, and C are the measurement variable matrix, state matrix, and output matrix, respectively. and Respectively represent the historical process data System status and output at any moment. and is the modeling error which usually follows a non-Gaussian distribution.

[0075] In order to avoid the influence of different unit values, it is necessary to eliminate the dimensional effect of multiple variables. Standardize the preprocessed data matrix X:

[0076] Formula 2:

[0077] ;

[0078] Where E(X) represents the expectation of X.

[0079] If the output vector Depend on variables, using Extended output vector of past and future measurements , you can get the past and future output vectors and :

[0080] Formula 3:

[0081] ;

[0082] Get the past and future output Hankel matrices and , where the Hankel Matrix is ​​a matrix in which the elements on each sub-diagonal are equal.

[0083] Formula 4:

[0084] ;

[0085] Formula 5:

[0086] ;

[0087] Formula 6:

[0088] ;

[0089] Formula 7:

[0090] ;

[0091] in, and They are and The average value of and They are and estimated value. , is the number of samples.

[0092] and The covariance and cross-covariance matrices are:

[0093] Formula 8:

[0094] ;

[0095] The objective of CVA can be obtained by decomposing the Hankel matrix of the scaled singular value get:

[0096] Formula 9:

[0097] ;

[0098] Where, 、 and .

[0099] Among them, U is a An orthogonal matrix is ​​called a left singular vector; V is a The orthogonal matrix of is called the right singular vector. Σ is a The diagonal matrix is ​​a nonnegative real number matrix called the singular values, which are usually arranged in descending order. The Hankel matrix can help identify the dynamic characteristics of the system. Through singular value decomposition, the Hankel matrix can be used for data compression and dimensionality reduction.

[0100] At this point, the first standard variable can be obtained :

[0101] Formula 10:

[0102] ;

[0103] in, represents the transformation matrix and , which transforms past observation samples in historical process data into the canonical variable space.

[0104] In one embodiment, the calculating of the historical error statistics of the target system based on the fault detection model and the calculating of the fault threshold based on the historical error statistics include: calculating the square prediction error statistics of the historical process data based on the residuals of the historical process data and the first standard variable, and determining the square prediction error statistics as the historical error statistics; and calculating the control limits of the historical error statistics based on a sampling distribution, and determining the control limits as the fault threshold.

[0105] Specifically, the SPE statistic and its control limits are calculated, wherein the SPE statistic is used to measure the difference between the observed value and the value predicted by the CVA model.

[0106] The SPE statistic is calculated as follows:

[0107] Formula 11:

[0108] ;

[0109] in, Represents the residual (the actual observed value and the estimated value, that is, the residual of the first standard variable). Distributed Computing Control limits of statistics. It should be noted that the SPE statistical value here is the SPE statistical value of each data in the historical process data. Therefore, the data volume of the SPE statistical value is equal to the data volume of the historical process data.

[0110] In an exemplary embodiment, constructing a fault subspace library based on the historical process data includes: classifying the fault data in the historical process data to obtain multiple groups of fault data of different types, and determining the classified fault types as candidate fault types; extracting fault features of each type of fault data based on a fault subspace extraction method, associating the fault features with the candidate fault types, and obtaining a corresponding fault subspace library.

[0111] Based on the fault subspace extraction method:

[0112] if Represents the historical process data Fault samples, refer to Formula 12:

[0113] ;

[0114] in, represents the correction value of the canonical variate space, which satisfies the zero mean condition and is therefore usually ignored. represents the fault magnitude estimated in the canonical variable space. Formula 12 is rewritten as Formula 13:

[0115] ;

[0116] Where, represents the sample obtained by processing formula 12, represents the fault amplitude obtained by processing formula 12. The average fault data matrix is ​​calculated as follows:

[0117] Formula 14:

[0118] ;

[0119] Singular Value Decomposition :

[0120] Formula 15:

[0121] ;

[0122] Among them, the diagonal matrix With non-zero singular values ​​in descending order, choose The dimension of the fault subspace is the number of The minimum number of dimensions for which the statistic is within the control limits.

[0123] Assuming that a system has 10 types of faults, the fault subspace extraction method is used to extract the fault features of these 10 types of faults respectively and store them for use in online fault diagnosis.

[0124] In an exemplary embodiment, obtaining the sample to be tested and calculating the current error statistic of the sample to be tested include: preprocessing the sample to be tested and calculating the second normalized variable of the preprocessed sample to be tested; calculating the squared prediction error statistic of the sample to be tested based on the residual between the sample to be tested and the second normalized variable, and determining the squared prediction error statistic as the current error statistic.

[0125] See Figure 3 In an exemplary embodiment, the method further comprises:

[0126] Step 301: When the current error statistic value is less than the fault threshold, the state of the sample to be tested is determined to be fault-free.

[0127] Step 302: After reconstructing the sample to be tested based on all candidate fault types in the fault subspace library, if the statistical values ​​of the reconstruction errors of the reconstructed samples are not less than the fault threshold, the fault manifestation of the sample to be tested is recorded, and a fault of the type corresponding to the fault manifestation and the fault characteristics are newly created in the fault subspace library.

[0128] Samples collected at the current moment , first standardize the preprocessing:

[0129] Formula 16:

[0130] ;

[0131] Then calculate the second normalized variable of the current sample :

[0132] Formula 17:

[0133] ;

[0134] in, represents the output vector of the current sample extended by q past measurements, represent The average value of .

[0135] Formula 18:

[0136] ;

[0137] in, Represents the canonical variable of the current sample.

[0138] Online Sample The statistics are calculated as follows:

[0139] Formula 19:

[0140] ;

[0141] in, Represents the residual between the tested sample and the second canonical variable.

[0142] If the statistics of the current sample Lower than If the control limit is exceeded, no fault has occurred at the current moment, and the monitoring continues to see if a fault occurs at the next moment. Otherwise, a fault has occurred at the current moment, and the type of fault needs to be diagnosed.

[0143] Assume there are 10 fault modes. If the current sample collected is detected as a fault, the fault data for each of the 10 fault modes is used to reconstruct the current fault sample. If the reconstructed monitoring statistic for the second fault mode falls below the control limit, the fault for the second mode is considered to have occurred. If the reconstructed monitoring statistics for the current fault sample, using the fault data for all 10 fault modes, do not fall below the control limit, the fault is considered to be an unknown fault. Process engineers are required to identify the fault direction and refine the relevant fault subspace library. If the current fault is an unknown fault, a new fault of that type is created in the fault subspace library for subsequent fault detection.

[0144] In order to better explain the technology of this embodiment, the fault diagnosis effect of the canonical variable analysis based on fault reconstruction is clearly and intuitively demonstrated. Figure 4 、 Figure 5 、 Figure 6 TEP experimental results are shown.

[0145] exist Figure 4 、 Figure 5 、 Figure 6 In the figure, the solid line represents the statistic, the dotted line represents the control limit, and the set of points on the horizontal axis represents the reconstructed statistic. When the statistic is higher than the control limit, it is considered that a fault has occurred at the current moment and the fault diagnosis needs to be reconstructed. Figures 4 to 6 The faults in the are all reconstructed using the corresponding fault subspace. It is worth noting that the statistics above the control limit are almost all below the control limit after being reconstructed using the corresponding fault subspace. This shows that the canonical variable analysis fault diagnosis technology based on fault reconstruction has a good fault diagnosis effect.

[0146] This embodiment has the following beneficial effects:

[0147] Aiming at the fault diagnosis problem of dynamic multivariable target system process monitoring, a fault diagnosis technology based on canonical variable analysis and fault reconstruction is proposed.

[0148] Different from the canonical variable analysis fault diagnosis method based on contribution graph, it can be applied to target systems with dynamic and nonlinear characteristics, effectively reducing the use assumptions of CVA fault diagnosis technology and expanding the scope of application.

[0149] The proposed fault diagnosis technology uses TEP process data as experimental input, primarily based on eliminating the assumption that the adopted statistics have good fault detection performance and the fuzzy effects between process variables. Its effectiveness and accuracy are verified based on the digital results of TEP case studies. Compared to the canonical variable analysis fault diagnosis method based on contribution graphs, the proposed method has more comprehensive evaluation criteria and a more systematic process.

[0150] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0151] Based on the same inventive concept, embodiments of the present application also provide a fault diagnosis device for implementing the aforementioned fault diagnosis method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations in one or more embodiments of the fault diagnosis device provided below can be found in the above-mentioned limitations on the fault diagnosis method and will not be further elaborated here.

[0152] In an exemplary embodiment, Figure 7 As shown, a fault diagnosis device 600 is provided, comprising: a construction module 601, a first calculation module 602, a second calculation module 603 and a reconstruction module 604, wherein:

[0153] A construction module 601 is configured to collect historical process data of a target system and construct a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0154] A first calculation module 602 is configured to calculate historical error statistics of the target system based on the fault detection model, and calculate a fault threshold based on the historical error statistics;

[0155] The second calculation module 603 is used to obtain a sample to be tested and calculate the current error statistics of the sample to be tested;

[0156] The reconstruction module 604 is used to reconstruct the samples to be tested in sequence based on the candidate fault types when the current error statistic is not less than the fault threshold, and to determine the candidate fault type currently used for reconstruction as the fault type of the sample to be tested when the reconstruction error statistic of the reconstructed sample is less than the fault threshold.

[0157] The construction module 601 is further configured to classify the fault data in the historical process data to obtain multiple groups of fault data of different types, and determine the fault types obtained by classification as candidate fault types;

[0158] Based on the fault subspace extraction method, the fault features of each type of fault data are extracted, and the fault features are associated with the candidate fault types to obtain a corresponding fault subspace library.

[0159] The construction module 601 is further configured to construct the fault detection model based on a canonical variable analysis method, wherein the input data of the fault detection model is the normal operation data, and the output of the fault detection model is a first canonical variable corresponding to the normal operation data.

[0160] The construction module 601 is further configured to calculate a squared prediction error statistic of the historical process data based on a residual between the historical process data and the first canonical variable, and determine the squared prediction error statistic as a historical error statistic;

[0161] A control limit of the historical error statistic is calculated based on a sampling distribution, and the control limit is determined as the fault threshold.

[0162] The second calculation module 603 is further configured to preprocess the sample to be tested and calculate a second normalized variable of the preprocessed sample to be tested;

[0163] Based on the residual between the sample to be tested and the second standard variable, a square prediction error statistic of the sample to be tested is calculated, and the square prediction error statistic is determined as the current error statistic.

[0164] The reconstruction module 604 is further configured to:

[0165] When the current error statistic value is less than the fault threshold, determining the state of the sample to be tested as no fault;

[0166] After reconstructing the sample to be tested based on all candidate fault types in the fault subspace library, the fault manifestation of the sample to be tested is recorded when the statistical values ​​of the reconstruction errors of the reconstructed sample are not less than the fault threshold.

[0167] Each module in the above-mentioned fault diagnosis device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0168] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless means, and the wireless means can be implemented via Wi-Fi, a mobile cellular network, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements a fault diagnosis method. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0169] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0170] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0171] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0172] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0173] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0174] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0175] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0176] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0177] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0178] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0179] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0180] Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type;

[0181] Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics;

[0182] Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested;

[0183] When the current error statistic is not less than the fault threshold, the samples to be tested are reconstructed in sequence based on the candidate fault types. When the reconstruction error statistic of the reconstructed samples is less than the fault threshold, the candidate fault type currently used for reconstruction is determined as the fault type of the sample to be tested.

[0184] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0185] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0186] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A fault diagnosis method, characterized in that: The method comprises: Collecting historical process data of the target system, and building a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type; Calculating historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics; Obtaining a sample to be tested, and calculating a current error statistic of the sample to be tested; When the current error statistic is not less than the fault threshold, reconstructing the sample to be tested based on the candidate fault type in sequence, and when the reconstruction error statistic of the reconstructed sample is less than the fault threshold, determining the candidate fault type currently used for reconstruction as the fault type of the sample to be tested; The constructing of the fault detection model based on the historical process data includes: constructing the fault detection model based on a canonical variable analysis method, wherein the input data of the fault detection model is the normal operating data, and the output of the fault detection model is a first canonical variable corresponding to the normal operating data.

2. The method according to claim 1, characterized in that The constructing of a fault subspace library based on the historical process data includes: Classifying the fault data in the historical process data to obtain multiple groups of fault data of different types, and determining the fault types obtained by classification as candidate fault types; Based on the fault subspace extraction method, the fault features of each type of fault data are extracted, and the fault features are associated with the candidate fault types to obtain a corresponding fault subspace library.

3. The method according to claim 1, characterized in that The calculating of historical error statistics of the target system based on the fault detection model, and calculating a fault threshold based on the historical error statistics, includes: Calculating a squared prediction error statistic of the historical process data based on a residual between the historical process data and the first canonical variable, and determining the squared prediction error statistic as a historical error statistic; A control limit of the historical error statistic is calculated based on a sampling distribution, and the control limit is determined as the fault threshold.

4. The method according to claim 1, wherein The obtaining of the sample to be tested and calculating the current error statistic of the sample to be tested includes: Preprocessing the sample to be tested, and calculating a second normalized variable of the preprocessed sample to be tested; Based on the residual between the sample to be tested and the second standard variable, a square prediction error statistic of the sample to be tested is calculated, and the square prediction error statistic is determined as the current error statistic.

5. The method according to claim 1, wherein The method further comprises: When the current error statistic value is less than the fault threshold, determining the state of the sample to be tested as no fault; After reconstructing the sample to be tested based on all candidate fault types in the fault subspace library, the fault manifestation of the sample to be tested is recorded when the statistical values ​​of the reconstruction errors of the reconstructed sample are not less than the fault threshold.

6. A fault diagnosis device, characterized in that: The device comprises: a construction module, configured to collect historical process data of a target system and construct a fault detection model and a fault subspace library based on the historical process data, wherein the historical process data includes normal operation data and fault data, and the fault subspace library includes at least one candidate fault type; A first calculation module is configured to calculate historical error statistics of the target system based on the fault detection model, and calculate a fault threshold based on the historical error statistics; A second calculation module is used to obtain a sample to be tested and calculate a current error statistic value of the sample to be tested; a reconstruction module, configured to, if the current error statistic is not less than the fault threshold, sequentially reconstruct the sample to be tested based on the candidate fault types, and, if the reconstruction error statistic of the reconstructed sample is less than the fault threshold, determine the candidate fault type currently used for reconstruction as the fault type of the sample to be tested; The construction module is further used to construct the fault detection model based on the canonical variable analysis method, wherein the input data of the fault detection model is the normal operation data, and the output of the fault detection model is the first canonical variable corresponding to the normal operation data.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images