Deep osa-based key performance indicator fault detection methods, systems, and devices

Abstract

The application discloses a key performance indicator fault detection method, system and device based on deep OSA, complete and accurate KPI related subspace data sets and input related subspace data sets are extracted by using OSA; a correlation matrix is calculated based on the KPI related subspace data sets and the input related subspace data sets, so as to accurately describe the correlation between the input and the KPI; the KPI related subspace data sets are divided into multiple subspace data sets by deep singular value decomposition of the correlation matrix, the KPI related subspace data sets are deeply separated, and the micro characteristics of the data are mined, so that the accuracy of micro fault detection can be improved.

Description

Technical Field

[0001] This invention relates to the field of key performance indicators technology in industrial processes, specifically to a method, system, and equipment for detecting faults in key performance indicators based on deep orthogonal subspace analysis. Background Technology

[0002] With the formal implementation of Industry 4.0, modern industry has gradually improved its intelligence level. Furthermore, to achieve the "dual carbon" (carbon peaking and carbon neutrality) goals, provinces have strictly implemented the "dual carbon" strategy, explicitly shutting down some high-consumption, high-pollution, and low-capacity factories, and upgrading outdated industrial equipment, continuously promoting the intelligentization, large-scale operation, high efficiency, and greening of the industrial sector. At the same time, industrial processes involve numerous equipment units and complex operating mechanisms, leading to frequent failures. Key performance indicators (KPIs) are important indicators that are significantly related to the quality of the final product and industrial economic benefits during industrial production. The continuous nature of industrial production processes makes failures related to some KPIs prone to unplanned downtime and maintenance, resulting in huge economic losses, and even casualties. Therefore, it is urgent to study fault detection methods for complex industrial processes, monitor the operating status of industrial processes in real time, promptly detect faults, and assess their impact on KPIs.

[0003] In actual industrial processes, KPIs are subject to delays, making real-time data collection difficult. To address this, multiple regression methods utilize input variables to detect the state of KPIs. In the field of multiple regression, orthonormal subspace analysis (OSA) is a novel and effective algorithm. While this method can detect faults in industrial processes promptly and accurately, and correctly determine the relationship between faults and KPIs, KPI faults in actual industrial processes often develop slowly and gradually, resulting in frequent minor faults. The characteristics of minor faults are not obvious, and the corresponding variables do not show significant changes. Therefore, because OSA does not fully mine the deeper information of the data, the loading of different spaces may not fully reflect the effective direction of each variable, and early fault information may be difficult to extract, leading to potentially unsatisfactory detection results. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for detecting key performance indicators based on deep orthogonal subspace analysis.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] According to one aspect of the present invention, a method for detecting key performance indicators faults based on deep OSA is provided, comprising the following steps:

[0007] S1. Collect offline historical normal data of industrial operation process, including input dataset and KPI dataset (i.e. output dataset).

[0008] S2. Perform normalization and standardization processing on the input and output datasets respectively;

[0009] S3. Use OSA to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets, and divide the KPI dataset into input-related subspace datasets and input-unrelated subspace datasets.

[0010] S4. Calculate the correlation matrix (CM) between the KPI-related subspace dataset and the input-related subspace dataset.

[0011] S5. Perform deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions;

[0012] S6. Using principal component analysis, both the KPI unrelated subspace dataset and the input unrelated subspace dataset are divided into principal component space dataset and residual space dataset, respectively.

[0013] S7. Establish a deep OSA model;

[0014] S8. Utilize a deep OSA model to detect industrial operating status online.

[0015] Furthermore, S5 performs deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions, including:

[0016] S5.1 First-order decomposition: Perform singular value decomposition on the correlation matrix to divide it into two mutually orthogonal subspaces, and then obtain the mapping matrix between the two subspaces;

[0017] S5.2, Second-order decomposition: Singular value decomposition is performed on the correlation matrices in the two subspaces in S5.1, and then each correlation matrix is ​​divided into two mutually orthogonal subspaces, finally obtaining the mapping matrix of the four spaces;

[0018] S5.3, L-order decomposition: Following the methods in S5.1 and S5.2, singular value decomposition is performed on the correlation matrix in each space sequentially. The L-order decomposition divides the correlation matrix into... Subspace, and thus obtain Mapping matrix of subspaces;

[0019] S5.4 Map the KPI-related subspace dataset to the decomposed dataset. In the projection matrix of each subspace, the KPI-related subspace dataset is further divided into... Subspace datasets.

[0020] Furthermore, based on S5.4 and S6, the final deep OSA model is established: the input dataset is represented as... The output dataset is the sum of the KPI-related subspace dataset, the principal component space dataset of the KPI-irrelevant subspace, and the residual space dataset of the KPI-irrelevant subspace.

[0021] Furthermore, S8 utilizes a deep OSA model to monitor industrial operating status online, specifically including:

[0022] S8.1: Online acquisition of input and output data during industrial operation.

[0023] S8.2: Standardize and regulate the input data and KPI data acquired online;

[0024] S8.3: Calculate the input data acquired online. The Hotelling statistic in the KPI-related subspace, the Hotelling statistic in the principal component space of the KPI-unrelated subspace, and the SPE statistic in the residual space of the KPI-unrelated subspace are calculated. The Hotelling statistic in the principal component space of the input unrelated subspace and the SPE statistic in the residual space of the input unrelated subspace are calculated for the KPI data acquired online.

[0025] S8.4: Determine the industrial operating status: If the input data acquired online is in If at least one of the Hotelling statistics in each KPI-related subspace is higher than its corresponding confidence limit, then a KPI-related fault is considered to have occurred; otherwise, proceed to S8.5.

[0026] S8.5: If at least one of the following statistics—the Hotelling statistic in the principal space of the KPI-independent subspace and the SPE statistic in the residual space of the KPI-independent subspace—is higher than its corresponding confidence limit, then a KPI-independent fault is considered to have occurred; otherwise, proceed to S8.6.

[0027] S8.6: If at least one of the following statistics—the Hotelling statistic in the principal space of the input-irrelevant subspace and the SPE statistic in the residual space of the input-irrelevant subspace—is higher than its corresponding confidence limit, then an input-irrelevant fault is considered to have occurred. Otherwise, continue acquiring the input data and KPI data for the next time step online, and return to S8.1.

[0028] The preset confidence limit is calculated using kernel density estimation.

[0029] According to another aspect of the present invention, a fault detection system for key performance indicators based on deep OSA is provided, comprising:

[0030] The data acquisition module is used to collect offline historical normal data of industrial operation processes, including input datasets and KPI datasets;

[0031] The data processing module is used to perform normalization and standardization processing on the input dataset and the KPI dataset, respectively.

[0032] The first data partitioning module is used to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets using OSA, and to divide the output dataset into input-related subspace datasets and input-unrelated subspace datasets.

[0033] The data calculation module is used to calculate the correlation matrix between the KPI-related subspace dataset and the input-related subspace dataset;

[0034] The data decomposition module is used to perform deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions.

[0035] The second data partitioning module is used to divide the KPI unrelated subspace dataset and the input unrelated subspace dataset into the main component space dataset and the residual space dataset, respectively, using principal component analysis.

[0036] The model building module is used to build deep OSA models;

[0037] The detection module is used to detect the industrial operating status online using a deep OSA model.

[0038] Furthermore, the correlation matrix undergoes deep singular value decomposition to obtain KPI-related datasets in different directions, including:

[0039] First-order decomposition: Perform singular value decomposition on the correlation matrix to divide it into two mutually orthogonal subspaces, and then obtain the mapping matrix between the two subspaces;

[0040] Second-order decomposition: Singular value decomposition is performed on the correlation matrices in the two subspaces of the first-order decomposition, and then each correlation matrix is ​​divided into two mutually orthogonal subspaces, finally obtaining the mapping matrix of the four spaces.

[0041] l-order decomposition: Following the methods of first-order and second-order decomposition, singular value decomposition is performed on the correlation matrix in each space sequentially. Through l-order decomposition, the correlation matrix can be divided into... Subspace, and thus obtain Mapping matrix of subspaces;

[0042] Map the KPI-related subspace dataset to the decomposed dataset. In the projection matrix of each subspace, the KPI-related subspace dataset is further divided into... Subspace datasets.

[0043] According to another aspect of the present invention, an apparatus is provided, comprising: one or more processors; and a memory for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the method as described above.

[0044] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method described above.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] 1. This invention addresses the problem of minor fault detection by proposing a deep orthogonal subspace analysis algorithm. This algorithm utilizes OSA to extract complete and accurate KPI-related subspace datasets and input-related subspace datasets; it calculates a correlation matrix based on the KPI-related subspace datasets and input-related subspace datasets to accurately describe the correlation between inputs and KPIs; and it divides the KPI-related subspace datasets into multiple subspace datasets through deep singular value decomposition of the correlation matrix, performing deep separation of the KPI-related subspace datasets and mining the subtle features of the data.

[0047] 2. Compared with the traditional OSA method, the deep OSA proposed in this invention can improve the accuracy of minor fault detection, especially KPI-related minor faults. Experimental results demonstrate the effectiveness and feasibility of this invention. Attached Figure Description

[0048] Figure 1 This is a flowchart of a method for detecting faults in key performance indicators of industrial processes based on deep orthogonal subspace analysis, according to the present invention.

[0049] Figure 2 This is a graph showing the detection results of the deep OSA method proposed in this invention for the current anomaly of the #2A coal mill in an embodiment;

[0050] Figure 3 This is a graph showing the detection results of the OSA method for the abnormal current in the #2A coal mill in this embodiment.

[0051] Figure 4 The graph shows the detection results of the #2A coal mill current anomaly in the traditional partial least squares method embodiment. The solid line represents the Hotelling statistic in the KPI-related subspace of the online input data.

[0052] Figure 5 The graph shows the detection results of the current anomaly in the #2A coal mill of the improved partial least squares method in the embodiment. The solid line is the Hotling statistic graph in the KPI-related subspace of the online input data.

[0053] Figure 6 This is a schematic diagram of the device of the present invention. Detailed Implementation

[0054] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0055] Fault detection method for key performance indicators of industrial processes based on deep orthogonal subspace analysis, such as Figure 1 As shown, it includes the following steps:

[0056] S1: Select offline historical normal data of the thermal power generation process, including the input dataset. and the output dataset, i.e., the KPI dataset Where n represents the number of samples, and m and s represent the number of input variables and KPIs, respectively;

[0057] Input variables include #A coal mill current, #A coal mill rotary separator current, #A coal feeder instantaneous coal feed rate, #A coal feeder motor speed, #A coal feeder coal feeding command, #A coal mill cold primary air electric regulating damper position feedback, #A coal mill hot primary air electric regulating damper position feedback, #A coal mill rotary separator speed output, #A coal mill sealing air pressure, sealing fan current, and sealing fan outlet main pipe pressure;

[0058] KPIs include the differential pressure between the upper and lower grinding bowls of #A coal mill, the pressure of the air-coal mixture in #A coal mill, and the primary air volume output of #A coal mill (three-selection output).

[0059] The offline historical normal dataset for each input variable and KPI contains 1980 samples;

[0060] S2: Standardize both the input dataset and the KPI dataset so that the mean of the processed dataset is 0 and the variance is 1, thus obtaining the standardized input dataset. and KPI dataset ;

[0061] S3: Use OSA to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets, and divide the KPI dataset into input-related subspace datasets and input-unrelated subspace datasets;

[0062] (1)

[0063] In this context, the superscript T indicates to, and the superscript -1 indicates the inverse matrix. This is a KPI-related subspace dataset. For the KPI-independent subspace dataset, To input the relevant subspace dataset, For input unrelated subspace datasets, orthogonal to , and , orthogonal to and , For X to The transfer relationship dataset, For Y to The dataset of transfer relationships;

[0064] In this embodiment, OSA is used as a preprocessing method to initially extract the correlated subspace dataset between the input dataset and the KPI dataset, and to ensure that it is orthogonal to the uncorrelated subspace dataset between the input and KPI, which can improve the accuracy of the correlation matrix in S4.

[0065] S4: Calculation and The correlation matrix between them;

[0066] (2)

[0067] in, represent and The correlation matrix between them;

[0068] S5: Yes Perform deep singular value decomposition to obtain KPI-related datasets from different directions;

[0069] S5.1: First-order decomposition: for Perform singular value decomposition , It is divided into two mutually orthogonal subspaces;

[0070] (3)

[0071] in, It is by The former A matrix consisting of the eigenvectors corresponding to the largest eigenvalues. It is by After A matrix consisting of eigenvectors corresponding to each eigenvalue. , It is by A diagonal matrix composed of the eigenvalues ​​of . for The correlation matrix of one of the corresponding subspaces, for The correlation matrix of the corresponding other subspace, and These are the mapping matrices for the two subspaces mentioned above;

[0072] S5.2: Second-order decomposition: for and Perform singular value decomposition separately , , and They were each divided into two mutually orthogonal subspaces. It was divided into four subspaces;

[0073] (4)

[0074] in, It is by The former A matrix consisting of the eigenvectors corresponding to the largest eigenvalues. It is by After A matrix consisting of eigenvectors corresponding to each eigenvalue. , It is by A diagonal matrix composed of the eigenvalues ​​of . It is by The former A matrix consisting of the eigenvectors corresponding to the largest eigenvalues. It is by After A matrix consisting of eigenvectors corresponding to each eigenvalue. , It is by A diagonal matrix composed of the eigenvalues ​​of . , for The correlation matrix of one of the corresponding subspaces (also) (correlation matrix of the corresponding first subspace) for The correlation matrix of the corresponding other subspace (also) (correlation matrix of the corresponding second subspace) , for The correlation matrix of one of the corresponding subspaces (also) (correlation matrix of the corresponding third subspace) for The correlation matrix of the corresponding other subspace (also) (correlation matrix of the corresponding fourth subspace) , , and They are , , and The mapping matrix of the subspace it belongs to;

[0075] S5.3: l( L-order decomposition: Following the methods in S5.1 and S5.2, singular value decomposition is performed sequentially on the correlation matrix of each subspace. Through l-order decomposition, the following can be achieved: Classified to In each subspace:

[0076] (5)

[0077] in, for The correlation matrix of the corresponding k-th subspace, yes The mapping matrix of the subspace, , ;

[0078] S5.4: Will Mapped to the decomposed In the projection matrix of each subspace:

[0079] (6)

[0080] in, It is the score matrix of the k-th subspace. I represents the identity matrix, from which it can be seen that as... Depth decomposition, KPI-related information in the data is broken down into... Each subspace, implemented In-depth analysis is beneficial for separating fault information related to minor KPIs.

[0081] S6: Yes and Principal component analysis was performed separately to obtain the principal component space dataset and residual space dataset for both datasets.

[0082] (7)

[0083] in, and They are and The load matrix, and They represent and The number of principal components, and They are and The score matrix, and They are Principal space dataset and residual space dataset, and They are Principal space dataset and residual space dataset;

[0084] S7: Obtain the final deep OSA model based on S5.4 and S6;

[0085] (8)

[0086] in, ;

[0087] S8: Utilize the deep OSA model in S7 to detect abnormal current in #2A coal mill;

[0088] S8.1: Select detection data from the thermal power generation process. The sample at each sampling time includes the input data. and KPI data ;

[0089] S8.2: Standardize and regulate the detection data of the thermal power generation process to obtain standardized input data. and KPI data ;

[0090] S8.3: Calculate the input data in The Hotelling statistic in the KPI-related subspace, the Hotelling statistic in the principal component space of the KPI-unrelated subspace, and the SPE statistic in the residual space of the KPI-unrelated subspace are calculated. The Hotelling statistic in the principal component space of the input unrelated subspace and the SPE statistic in the residual space of the input unrelated subspace are calculated for the KPI data acquired online.

[0091] (9)

[0092] in, Is the input data in Hotelling statistic in a KPI-related subspace It is the Hotelling statistic of the input data in the principal dimension space of the KPI uncorrelated subspace. It is the SPE statistic of the input data in the residual space of the KPI uncorrelated subspace. It is the Hotelling statistic of KPI data in the principal dimension space of the input unrelated subspace. It is the SPE statistic of KPI data in the residual space of the uncorrelated subspace of the input. It is a KPI-related subspace dataset of the input data. It is a dataset of KPI-independent subspaces of the input data. It is the input-related subspace dataset of KPI data. It is a dataset of KPI data input-irrelevant subspaces. It is the score vector of the k-th KPI-related subspace dataset of the input data. It is the score vector of the principal component space dataset, which is a subspace of the input data's KPIs. It is the score vector of the principal component space dataset of the input unrelated subspace of the KPI data. It is the residual space dataset of the KPI-independent subspace of the input data. It is the residual space dataset of the input uncorrelated subspace of KPI data;

[0093] S8.4: Fault diagnosis in current thermal power generation process:

[0094] If the input data calculated by S8.3 is in Hotling statistic in a KPI-related subspace If at least one statistic is higher than its corresponding confidence limit, then a KPI-related fault is considered to have occurred; otherwise, proceed to S8.5.

[0095] S8.5: The Hotelling statistic of the input data calculated in S8.3 in the principal space of the KPI uncorrelated subspace. SPE statistic of input data in the residual space of the KPI uncorrelated subspace If at least one statistic is higher than its corresponding confidence limit, then a KPI-irrelevant fault is considered to have occurred; otherwise, proceed to S8.6.

[0096] S8.6: If the KPI data calculated in S8.3 contains the Hotelling statistic in the principal space of the uncorrelated input subspace... SPE statistic of KPI data in the residual space of the uncorrelated subspace of the input If at least one statistic is higher than its corresponding confidence limit, then an input irrelevant fault is considered to have occurred; otherwise, continue to check the input data and KPI data of the thermal power generation process at the next moment, and return to S8.1.

[0097] In this embodiment, the experimental data comes from a case of abnormal current in the #2A coal mill recorded at Zhejiang Energy Zhoushan Thermal Power Plant on July 3, 2019. The specific occurrence of this case is as follows: at 11:04:39 on July 3, 2019, the current in the #2A coal mill began to rise; at 11:04:42, the current began to decrease; at 11:14:45, the feeder inlet gate closed; and the coal mill stopped operating at 15:02:24. Therefore, this invention uses normal operating data (1980 samples) from 10:31:01 to 11:03:60 to construct the training model; and uses data (857 samples) from 10:58:45 to 11:13:01 for detection. First, a deep OSA model was established using 1980 normal samples from the thermal power generation process. Then, the deep OSA model was used to detect current anomalies in the #2A coal mill. The detection data included 857 samples (the first 354 were normal samples, and the last 503 were fault samples). To more effectively illustrate the performance of the deep OSA model, this implementation also compared the results with some existing regression algorithms on this problem. The methods compared included: OSA method, traditional partial least squares method, and improved partial least squares method. The detection results of the #2A coal mill current anomaly are as follows: Figures 2-5 As shown. Figure 2(a) is a Hotling statistic plot of the first KPI-related subspace of the online input data; (b) is a Hotling statistic plot of the second KPI-related subspace of the online input data; (c) is a Hotling statistic plot of the third KPI-related subspace of the online input data; (d) is a Hotling statistic plot of the fourth KPI-related subspace of the online input data; (e) is a Hotling statistic plot of the fifth KPI-related subspace of the online input data; (f) is a Hotling statistic plot of the sixth KPI-related subspace of the online input data; (g) is a Hotling statistic plot of the seventh KPI-related subspace of the online input data; and (h) is a Hotling statistic plot of the eighth KPI-related subspace of the online input data. Figure 3 (a) is a Hotling statistic plot in the KPI-related subspace of the online input data; (b) is a SPE statistic plot in the KPI-related subspace of the online input data.

[0098] Depend on Figure 2 It can be seen that the input data is in Hotling statistic in a KPI-related subspace An abnormal current was detected in the #2A coal mill within sampling points 355 to 857. Within the first 354 sampling points, the current did not exceed the confidence limit, indicating that this fault is output-related, consistent with the findings. Figures 3-5 It can be seen that the statistics of the input data in the output correlation subspace , , , No fault samples were detected in the early stages of the abnormal current in the #2A coal mill.

[0099] Table 1 shows the fault detection rate and false alarm rate for different methods of detecting current anomalies in the #2A coal mill. As can be seen from Table 1, for the #2A coal mill current anomaly, the input data... Hotling statistic in a KPI-related subspace The detection rates are all above 93%, especially The detection rate of all three methods reached 100%. The detection rates of the other three methods were all below 98.3%. From the above results, it can be concluded that the deep OSA model proposed in this invention can accurately detect current anomalies in the #2A coal mill, especially in the initial stage of the current anomaly.

[0100] Table 1. Fault detection rate and false alarm rate of abnormal current in #2A coal mill

[0101]

[0102] This embodiment provides a key performance indicator fault detection system based on deep OSA, including:

[0103] The data acquisition module is used to collect offline historical normal data of industrial operation processes, including input datasets and KPI datasets;

[0104] The data processing module is used to normalize and standardize the input and output datasets respectively, so that the mean of the processed dataset is 0 and the variance is 1.

[0105] The first data partitioning module is used to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets using OSA, and to divide the KPI dataset into input-related subspace datasets and input-unrelated subspace datasets.

[0106] The data calculation module is used to calculate the correlation matrix between the KPI-related subspace dataset and the input-related subspace dataset;

[0107] The data decomposition module is used to perform deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions, including:

[0108] First-order decomposition: Perform singular value decomposition on the correlation matrix to divide it into two mutually orthogonal subspaces, and then obtain the mapping matrix between the two subspaces;

[0109] Second-order decomposition: Singular value decomposition is performed on the correlation matrices in the two subspaces of the first-order decomposition, and then each correlation matrix is ​​divided into two mutually orthogonal subspaces, finally obtaining the mapping matrix of the four spaces.

[0110] l-order decomposition: Following the methods of first-order and second-order decomposition, singular value decomposition is performed on the correlation matrix in each space sequentially. Through l-order decomposition, the correlation matrix can be divided into... Subspace, and thus obtain Mapping matrix of subspaces;

[0111] Map the KPI-related subspace dataset to the decomposed dataset. In the projection matrix of each subspace, the KPI-related subspace dataset is further divided into... Subspace datasets.

[0112] The second data partitioning module is used to divide the KPI unrelated subspace dataset and the input unrelated subspace dataset into the main component space dataset and the residual space dataset, respectively, using principal component analysis.

[0113] The model building module is used to build deep OSA models.

[0114] The detection module is used to detect the industrial operating status online using a deep OSA model.

[0115] An embodiment of the device includes: one or more processors; and a memory for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the method described in any of the above-mentioned steps: extracting a complete and accurate KPI-related subspace dataset and an input-related subspace dataset using OSA; calculating a correlation matrix based on the KPI-related subspace dataset and the input-related subspace dataset to accurately describe the correlation between inputs and KPIs; dividing the KPI-related subspace dataset into multiple subspace datasets through deep singular value decomposition of the correlation matrix; performing deep separation on the KPI-related subspace dataset; and mining subtle features of the data.

[0116] This embodiment provides a computer-readable storage medium storing a computer program. When executed by a processor, the program implements the method described in any of the above-mentioned embodiments. It stores a key performance indicator (KPI) fault detection method based on deep OSA, calculates a correlation matrix based on a KPI-related subspace dataset and an input-related subspace dataset to accurately describe the correlation between the input and the KPI, divides the KPI-related subspace dataset into multiple subspace datasets through deep singular value decomposition of the correlation matrix, performs deep separation on the KPI-related subspace dataset, and mines subtle features of the data. Further details are as follows:

[0117] The computer system includes a central processing unit (CPU) 101, which performs various appropriate actions and processes based on programs stored in read-only memory (ROM) 102 or loaded from storage into random access memory (RAM) 103. RAM 103 also stores various programs and data required for system operation. The CPU 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.

[0118] The following components are connected to I / O interface 105: an input section 106 including a keyboard, mouse, etc.; an output section 107 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 108 including a hard disk, etc.; and a communication section 109 including a network interface card such as a LAN card, modem, etc. The communication section 109 performs communication processing via a network such as the Internet. A drive is also connected to I / O interface 105 as needed. A removable medium 111, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 110 as needed so that computer programs read from it can be installed into storage section 108 as needed.

[0119] In particular, according to embodiments of the present invention, the above-described reference process Figure 1 The described process can be implemented as a computer software program. For example, Embodiment 1 of the present invention includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit (CPU) 101, it performs the functions defined in the system of this application.

[0120] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0121] The box in the attached diagram Figure 6The diagram illustrates the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments 1 of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the figures. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0122] The units described in the embodiments of this invention can be implemented in software or hardware, and can also be located in a processor. The names of these units do not necessarily limit the specific unit itself. The described units or modules can also be located in a processor; for example, it can be described as: a key performance indicator fault detection system based on deep OSA, including: a data acquisition module; a data processing module; a first data partitioning module; a data calculation module; a data decomposition module; a second data partitioning module; a model building module; and a detection module. The names of these units do not necessarily limit the specific unit itself.

[0123] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0124] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0125] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, the above-described features have similar functions to (but are not limited to) those disclosed in this application.

[0126] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0127] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for detecting key performance indicators based on deep OSA, characterized in that, Includes the following steps: S1. Collect offline historical normal data of industrial operation process, including input dataset and KPI dataset; S2. Perform normalization and standardization processing on the input dataset and KPI dataset respectively; S3. Use OSA to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets, and divide the KPI dataset into input-related subspace datasets and input-unrelated subspace datasets. S4. Calculate the correlation matrix between the KPI-related subspace dataset and the input-related subspace dataset; S5. Perform deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions; S6. Using principal component analysis, both the KPI unrelated subspace dataset and the input unrelated subspace dataset are divided into principal component space dataset and residual space dataset, respectively. S7. Establish a deep OSA model; S8. Utilize deep OSA models to detect industrial operating status online; In S5, deep singular value decomposition is performed on the correlation matrix to obtain KPI-related datasets in different directions, including: S5.1 First-order decomposition: Perform singular value decomposition on the correlation matrix to divide it into two mutually orthogonal subspaces, and then obtain the mapping matrix between the two subspaces; S5.2, Second-order decomposition: Singular value decomposition is performed on the correlation matrices in the two subspaces in S5.1, and then each correlation matrix is ​​divided into two mutually orthogonal subspaces, finally obtaining the mapping matrix of the four spaces; S5.3, L-order decomposition: Following the methods in S5.1 and S5.2, singular value decomposition is performed on the correlation matrix in each space sequentially. The L-order decomposition divides the correlation matrix into... Subspace, and thus obtain Mapping matrix of subspaces; S5.4 Map the KPI-related subspace dataset to the decomposed dataset. In the projection matrix of each subspace, the KPI-related subspace dataset is further divided into... Subspace datasets; S8 utilizes a deep OSA model to monitor industrial operating status online, specifically including: S8.1: Online collection of input data and KPI data for industrial operations; S8.2: Standardize and regulate the input data and KPI data acquired online; S8.3: Calculate the input data acquired online. The Hotelling statistic in the KPI-related subspace, the Hotelling statistic in the principal component space of the KPI-unrelated subspace, and the SPE statistic in the residual space of the KPI-unrelated subspace are calculated. The Hotelling statistic in the principal component space of the input unrelated subspace and the SPE statistic in the residual space of the input unrelated subspace are calculated for the KPI data acquired online. S8.4: Determine the industrial operating status: If the input data acquired online is in If at least one of the Hotelling statistics in each KPI-related subspace is higher than its corresponding confidence limit, then an output-related fault is considered to have occurred; otherwise, proceed to S8.

5. S8.5: If at least one of the following statistics—the Hotelling statistic of the online-acquired input data in the principal space of the KPI uncorrelated subspace and the SPE statistic of the online-acquired input data in the residual space of the KPI uncorrelated subspace—is higher than its corresponding confidence limit, then an output uncorrelated fault is considered to have occurred; otherwise, proceed to S8.

6. S8.6: If at least one of the following statistics of the KPI data acquired online is higher than its corresponding confidence limit: the Hotelling statistic in the principal space of the input-unrelated subspace and the SPE statistic in the residual space of the input-unrelated subspace, then an input-unrelated fault is considered to have occurred. Otherwise, continue to acquire the input data and KPI data for the next time step online and return to S8.

1.

2. The method for detecting key performance indicators based on deep OSA according to claim 1, characterized in that, Based on S5.4 and S6, the final deep OSA model is established: the input dataset is represented as... The KPI dataset is represented as the sum of the KPI-related subspace dataset, the principal component space dataset of the KPI-irrelevant subspace, and the residual space dataset of the KPI-irrelevant subspace.

3. The method for detecting key performance indicators based on deep OSA according to claim 1, characterized in that, The mean of the processed dataset in S2 is 0, and the variance is 1.

4. A fault detection system for key performance indicators based on deep OSA, characterized in that, include: The data acquisition module is used to collect offline historical normal data of industrial operation processes, including input datasets and KPI datasets; The data processing module is used to perform normalization and standardization processing on the input dataset and the KPI dataset, respectively. The first data partitioning module is used to divide the input dataset into KPI-related subspace datasets and KPI-unrelated subspace datasets using OSA, and to divide the KPI dataset into input-related subspace datasets and input-unrelated subspace datasets. The data calculation module is used to calculate the correlation matrix between the KPI-related subspace dataset and the input-related subspace dataset; The data decomposition module is used to perform deep singular value decomposition on the correlation matrix to obtain KPI-related datasets in different directions. The second data partitioning module is used to divide the KPI unrelated subspace dataset and the input unrelated subspace dataset into the main component space dataset and the residual space dataset, respectively, using principal component analysis. The model building module is used to build deep OSA models; The detection module is used to detect the industrial operating status online using a deep OSA model; The correlation matrix is ​​subjected to deep singular value decomposition to obtain KPI-related datasets in different directions, including: First-order decomposition: Perform singular value decomposition on the correlation matrix to divide it into two mutually orthogonal subspaces, and then obtain the mapping matrix between the two subspaces; Second-order decomposition: Singular value decomposition is performed on the correlation matrices in the two subspaces of the first-order decomposition, and then each correlation matrix is ​​divided into two mutually orthogonal subspaces, finally obtaining the mapping matrix of the four spaces. l-order decomposition: Following the methods of first-order and second-order decomposition, singular value decomposition is performed on the correlation matrix in each space sequentially. Through l-order decomposition, the correlation matrix can be divided into... Subspace, and thus obtain Mapping matrix of subspaces; Map the KPI-related subspace dataset to the decomposed dataset. In the projection matrix of each subspace, the KPI-related subspace dataset is further divided into... Subspace datasets; Using deep OSA models, online monitoring of industrial operating status is performed, including: Online collection of input data and KPI data from industrial operations; Standardize and regulate the input data and KPI data acquired online; Calculate the input data acquired online. The Hotelling statistic in the KPI-related subspace, the Hotelling statistic in the principal component space of the KPI-unrelated subspace, and the SPE statistic in the residual space of the KPI-unrelated subspace are calculated. The Hotelling statistic in the principal component space of the input unrelated subspace and the SPE statistic in the residual space of the input unrelated subspace are calculated for the KPI data acquired online. Determine the industrial operating status: If the input data acquired online is in If at least one of the Hotelling statistics in each KPI-related subspace is higher than its corresponding confidence limit, then an output-related fault is considered to have occurred; otherwise, proceed to the next step. If at least one of the following statistics—the Hotelling statistic in the principal space of the KPI-independent subspace and the SPE statistic in the residual space of the KPI-independent subspace—is higher than its corresponding confidence limit, then an output inconsistency fault is considered to have occurred; otherwise, proceed to the next step. If at least one of the following statistics—the Hotelling statistic in the principal space of the input-irrelevant subspace and the SPE statistic in the residual space of the input-irrelevant subspace—is higher than its corresponding confidence limit, then an input-irrelevant fault is considered to have occurred. Otherwise, continue to acquire the input data and KPI data for the next time step online and return to step one.

5. An electronic device, comprising: One or more processors; A memory for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the method as described in any one of claims 1-3.

6. A computer-readable storage medium storing a computer program, characterized in that... When the computer program is executed by a processor, it implements the method as described in any one of claims 1-3.

Images