Industrial process operation data filling and repairing method and system

Abstract

The invention provides an industrial process operation data filling and repairing method and system, and relates to the technical field of data filling. The global long-term trend of data is captured through cyclic matrix nuclear norm minimization, local short-term fluctuation characteristics of the data are described in combination with time-adaptive Laplacian regularization, and correlation among variables is represented by graph structure regularization of adaptive weight. Balance parameters are dynamically adjusted through a self-adaptive weight mechanism, different time sequence characteristics are adapted, and the interpolation effect is optimized. And finally, solving the optimization model by adopting an alternating direction multiplier method, and generating filled and repaired data. Meanwhile, the global trend, the local fluctuation and the relation between variables of the data are considered, and the accuracy and the reliability of data filling are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of data filling, and in particular, to a method and system for filling and repairing operation data of an industrial process. Background Art

[0002] In the modern industrial system, the intelligent and refined management of industrial production highly depends on the real-time and accurate operation data during the production process. From a microscopic perspective, the equipment operation parameters and process index data collected by various sensors on the production line provide first-hand information for operators to timely understand the equipment status and adjust the production process; from a macroscopic perspective, a large amount of industrial data, after being integrated and analyzed, can provide strong support for the strategic decision-making and resource allocation of enterprises, helping enterprises gain advantages in the fierce market competition.

[0003] However, the extreme complexity of the industrial production environment poses many challenges to the data collection process. In a high-temperature environment, the electronic components of sensors are easily affected by thermal stress, resulting in a decrease in measurement accuracy or even failure, and thus data loss; a strong magnetic environment will interfere with the signal transmission of sensors, causing errors or loss of the collected data; strong vibration may cause physical damage to the sensors and prevent normal data collection. In addition, the instability of the industrial network, such as signal interference and network congestion, often leads to data transmission interruption, further exacerbating the data loss problem.

[0004] The negative impact of data loss on industrial production is multi-faceted. At present, data filling and repair technology is an important means to solve the problem of industrial data loss. There are various traditional data filling and repair methods, but each has obvious limitations.

[0005] Therefore, how to improve the accuracy and reliability of industrial data filling has become an urgent technical problem to be solved. Summary of the Invention

[0006] In order to improve the accuracy and reliability of industrial data filling, the present application provides a method and system for filling and repairing operation data of an industrial process.

[0007] In a first aspect, a method for filling and repairing operation data of an industrial process provided by the present application adopts the following technical solution: A method for filling and repairing operation data of an industrial process includes: Collecting partial observation data in the industrial process, where the partial observation data contains missing data; Based on the minimization of the nuclear norm of the circulant matrix, constructing a global data trend model, and performing interpolation on the partial observation data through the operation of minimizing the nuclear norm of the circulant matrix; Introduce a time - adaptive Laplacian regularization module to dynamically adjust the local smoothing constraint weights according to the time intervals of the data, and capture the short - term fluctuation characteristics of partial observed data; Construct a graph - structure regularization module with adaptive weights to learn the correlation weights between variables through the adjacency matrix and graph structure, making the imputation results of related variables tend to be consistent; Design an adaptive weight mechanism to dynamically adjust the balance parameter between the global trend and the local trend according to the local stationarity of the data; Integrate the global trend model, the time - adaptive Laplacian regularization module, the graph - structure regularization module, and the adaptive weight mechanism to construct a unified optimization problem; Use the alternating direction multiplier method to iteratively solve the optimization problem until the convergence condition is met, and output the complete imputed data.

[0008] Optionally, the minimization of the circulant matrix nuclear norm includes: Convert the partial observed data into the form of a circulant matrix; Use the fast Fourier transform to calculate the nuclear norm of the circulant matrix, and achieve global trend modeling and missing data imputation by minimizing the nuclear norm.

[0009] Optionally, the step of introducing a time - adaptive Laplacian regularization module to dynamically adjust the local smoothing constraint weights according to the time intervals of the data and capture the short - term fluctuation characteristics of partial observed data includes: Construct a Laplacian matrix, and determine the adjacent relationship between data according to the Laplacian matrix; Introduce a time - weighted factor to weight - correct the Laplacian regularization term according to the time intervals between data points; Through the time - weighted factor, adjust the smoothing constraint of the data according to the time interval, and capture the short - term fluctuation characteristics of partial observed data in the Laplacian matrix.

[0010] Optionally, the variable similarity weight in the graph - structure regularization module with adaptive weights is defined as: where is the scale factor for controlling the similarity, and represent two different variables, is the normalization constant to ensure that the sum of all weights is 1.

[0011] Optionally, the adaptive weight mechanism is used to dynamically adjust the balance parameter γ Specifically, it includes: Calculate the variance within a local time window as a measure of local stationarity for a given time series , within a time window centered at with a length of , the local variance can be calculated as: where is the mean of the data within this time window; Generate an adaptive weight by mapping through the Sigmoid function according to the local stationarity measure: where is the adaptive weight at time point , and are adjustable parameters used to control the shape and range of the mapping function, and the value range of the Sigmoid function is between (0, 1); Based on the adaptive weight, dynamically adjust the balance parameter γ to achieve the dynamic balance of global and local trend weights. Assuming the original fixed γ value is , then the dynamically adjusted γ value is calculated as: where represents the value of when paying more attention to the global trend.

[0012] Optionally, the steps of integrating the global trend model, time - adaptive Laplacian regularization module, graph - structure regularization module, and adaptive weight mechanism to construct a unified optimization problem include: Integrate the global trend model, time - adaptive Laplacian regularization module, graph - structure regularization module, and adaptive weight mechanism to construct a unified optimization model, and the optimization model is expressed as: where the global trend data is , the local trend data is , the variable relationship data is , and n represents the number of variables x.

[0013] Optionally, the steps of using the alternating direction multiplier method to iteratively solve the optimization problem until the convergence condition is met and outputting the complete interpolated data include: Introduce an auxiliary variable ; , , , and represent the optimization problem as: Construct the augmented Lagrangian function as follows: where is the Lagrange multiplier, is the penalty parameter; Update X: Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update the Lagrange multiplier: Iteratively update the above sub-problems until the convergence condition is met, and output the complete imputed data.

[0014] In a second aspect, the present application provides an operating data filling and repairing system for an industrial process, including: A data acquisition module for acquiring partial observed data in the industrial process, where the partial observed data contains missing data; A global trend module for constructing a data global trend model based on minimizing the nuclear norm of a circulant matrix, and imputing the partial observed data through the nuclear norm minimization operation of the circulant matrix; A short-term feature module for introducing a time-adaptive Laplacian regularization module to dynamically adjust the local smoothing constraint weight according to the time interval of the data and capture the short-term fluctuation characteristics of the partial observed data; A regularization module, used to construct a graph structure regularization module with adaptive weights, learn the correlation weights between variables through the adjacency matrix and the graph structure, and make the imputation results of related variables tend to be consistent; An adaptive weight module, used to design an adaptive weight mechanism, and dynamically adjust the balance parameter between the global trend and the local trend according to the local stationarity of the data; An optimization problem module, used to integrate the global trend model, the time - adaptive Laplacian regularization module, the graph structure regularization module and the adaptive weight mechanism to construct a unified optimization problem; An output module, used to iteratively solve the optimization problem by using the alternating direction method of multipliers until the convergence condition is met, and output the complete imputed data.

[0015] In a third aspect, the present application provides a computer device, the device includes: a memory, a processor, and when the processor runs the computer instructions stored in the memory, it executes the method as described above.

[0016] In a fourth aspect, the present application provides a computer - readable storage medium, including instructions, when the instructions run on a computer, the computer is made to execute the method as described above.

[0017] In summary, the present application includes the following beneficial technical effects: The present application captures the global long - term trend of the data through minimizing the circulant matrix nuclear norm, combines the time - adaptive Laplacian regularization to describe the local short - term fluctuation characteristics of the data, and uses the graph structure regularization with adaptive weights to characterize the correlation between variables. Dynamically adjust the balance parameter through the adaptive weight mechanism to adapt to different time - series characteristics and optimize the imputation effect. Finally, use the alternating direction method of multipliers to solve the optimization model and generate the filled and repaired data. Considering the global trend, local fluctuations and relationships between variables of the data significantly improves the accuracy and reliability of data filling. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the computer device structure of the hardware operating environment involved in the solution of the embodiment of the present application; Figure 2 It is a schematic flowchart of the first embodiment of the method for filling and repairing the operation data of the industrial process of the present application; Figure 3 It is a Laplacian matrix diagram of the first embodiment of the method for filling and repairing the operation data of the industrial process of the present application; Figure 4 It is a structural block diagram of the first embodiment of the system for filling and repairing the operation data of the industrial process of the present application. Detailed Embodiments

[0019] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be 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 the present application and are not used to limit the present application.

[0020] Referring to Figure 1 , Figure 1 is a schematic structural diagram of a computer device for the hardware operating environment involved in the solution of the embodiment of the present application.

[0021] As Figure 1 shown, the computer device may include: a processor 1001, such as a Central Processing Unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0022] Those skilled in the art can understand that Figure 1 the structure shown in

[0023] does not constitute a limitation on the computer device, and may include more or fewer components than shown in the figure, or combine certain components, or different component arrangements. Figure 1 As

[0024] shown, in the Figure 1 computer device shown, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the present application may be provided in the computer device. The computer device calls the industrial process operation data filling and repairing program stored in the memory 1005 through the processor 1001 and executes the industrial process operation data filling and repairing method provided by the embodiment of the present application.

[0025] An embodiment of the present application provides a method for filling and repairing operation data of an industrial process. Referring to Figure 2 , Figure 2 which is a schematic flowchart of the first embodiment of the method for filling and repairing operation data of the industrial process of the present application.

[0026] In this embodiment, the method for filling and repairing operation data of the industrial process includes the following steps: Step S10: Collect partial observed data in the industrial process, where the partial observed data contains missing data.

[0027] It should be noted that in the industrial process, industrial process data modeling is crucial for the optimization and regulation of the industrial production process. The data filling and repairing problem can be summarized as follows: For any partial observed data , whose observation index set is Ω, the goal is to fill the missing data from . In this context, represents the orthogonal projection supported on Ω, while represents the orthogonal projection supported on the complement of Ω. The operator is described as follows: where .

[0028] Step S20: Based on the minimization of the nuclear norm of the circulant matrix, construct a global trend model of the data, and interpolate the partial observed data through the minimization operation of the nuclear norm of the circulant matrix.

[0029] It should be noted that the minimization of the nuclear norm of the circulant matrix includes: converting the partial observed data into the form of a circulant matrix; calculating the nuclear norm of the circulant matrix using the fast Fourier transform, and realizing global trend modeling and missing data interpolation by minimizing the nuclear norm.

[0030] In specific implementation, industrial data has the low-rank property and can be completed through a low-rank model. The traditional low-rank model is as follows: where is the repaired data, is the observed data, is the observation set. The following example can be used to more intuitively understand this model: There is the following partial observed data: Then its observation set Ω is Using the model The following repaired matrix can be obtained: However, it can be proved by using the definition of convex function that the optimization function of the model is a non-convex function, and this optimization problem is an NP-hard problem. Therefore, the nuclear norm of the matrix is used to convexify the optimization function to transform it into a convex function. The definition of the nuclear norm of the matrix is as follows: where represents the k-th largest singular value of the matrix. Therefore, the optimization problem becomes: It can be proved by using the definition of convex function and Courant-Fisher theorem that this optimization function is a convex function. However, in order to pursue the global trend of the data, in this embodiment, the nuclear norm of the data matrix is transformed into the nuclear norm of the circulant matrix of the data. The definition of the circulant matrix is as follows: Therefore, the optimization problem becomes: When performing the nuclear norm minimization operation on the circulant matrix of the data, since each row and each column of the circulant matrix contains the complete information of the data, the nuclear norm minimization of the circulant matrix can be used to complete the data imputation and pursue the global trend of the data. In addition, the nuclear norm of the circulant matrix can be calculated by using the discrete Fourier transform as follows: The use of the discrete Fourier transform can greatly improve the calculation efficiency.

[0031] Step S30: Introduce a time-adaptive Laplacian regularization module to dynamically adjust the local smoothing constraint weight according to the time interval of the data and capture the short-term fluctuation characteristics of some observed data.

[0032] It can be understood that the step of introducing a time-adaptive Laplacian regularization module to dynamically adjust the local smoothing constraint weight according to the time interval of the data and capture the short-term fluctuation characteristics of some observed data includes: constructing a Laplacian matrix, determining the adjacent relationship between data according to the Laplacian matrix; introducing a time weighting factor to weight and correct the Laplacian regularization term according to the time interval between data points; and adjusting the smoothing constraint of the data according to the time interval through the time weighting factor to capture the short-term fluctuation characteristics of some observed data in the Laplacian matrix.

[0033] In specific implementation, the traditional Laplacian time regularization idea is as follows: First, the definition of the Laplacian matrix is given: Among them is the degree matrix of the graph, is the adjacency matrix of the graph. Consider the Laplacian matrix as shown in Figure 3 : Its Laplacian matrix is: Each column in the matrix represents the adjacency situation of the corresponding data. 2 represents adjacent to two data, and -1 represents adjacent to this data. Its first column is: It is a special form of the Laplacian kernel. Here, the definition of the Laplacian kernel is given: Among them, τ represents the degree of the graph, reflecting how many data each data is connected to. Using the definition of the Laplacian kernel and the relationship between circular convolution and circulant matrix, the definition of Laplacian regularization is given: Taking Figure 3 as an example, its Lx matrix is: Its second norm is equivalent to the distance between each data and its adjacent data. Therefore, through this regularization term, the local trend of the data can be pursued, and the smoothness of the interpolated data can be pursued. However, in the actual process, the situation of different sampling time intervals will be encountered, and this regularization term cannot handle this situation.

[0034] To solve the problem that traditional Laplacian regularization is only applicable to uniform time intervals (all variable sampling frequencies are the same), the present invention introduces a time weighting factor to improve Laplacian regularization. The basic idea is as follows: Let the time interval between adjacent data points in the time series be , and we design a time weighting factor to weight and correct the Laplacian regularization term as follows: Among them is the time weighting factor. When the sampling time interval is small, the weight of this term is large, forcing the adjacent points to be smoother; when the sampling time interval is large, the weight of this term is small, allowing a larger time change. Using the time weighting factor, the data with a short time interval is more strongly constrained, and the data with a long time interval is allowed to have a larger fluctuation, which can improve the smoothness and accuracy of interpolation.

[0035] Step S40: Construct a graph structure regularization module with adaptive weights, and learn the correlation weights between variables through the adjacency matrix and the graph structure, so that the interpolation results of related variables tend to be consistent.

[0036] It should be noted that the variable similarity weight in the graph structure regularization module with adaptive weights is defined as: where is a scale factor for controlling the similarity degree, and represent two different variables, is a normalization constant used to ensure that the sum of all weights is 1.

[0037] It can be understood that for the problem that the traditional method ignores the mutual relationship between variables during imputation and each sequence is imputed independently, the method proposed in this embodiment adopts an adaptive-weighted graph structure-based variable relationship representation to ensure that related variables have similar imputation results. The specific idea is as follows: The regularization term uses the variable to represent the similarity of variables, and its definition is as follows: where is a scale factor for controlling the similarity degree, is a normalization constant used to ensure that the sum of all weights is 1 (i.e., ).

[0038] Therefore, the weight-adaptive graph regularization term is defined as: If has a large value, it indicates that the similarity between variable i and variable j is relatively high, otherwise it indicates that their similarity is relatively low. By adaptively updating the value of the weight , the model can better capture the relationship between variables. When calculating the imputation result, for variable pairs with being large, the model will tend to give similar imputation values because they are closely related in the graph structure and are more likely to have similar change trends or characteristics in practical significance. n represents the number of variables X.

[0039] Step S50: Design an adaptive weight mechanism to dynamically adjust the balance parameter between the global trend and the local trend according to the local stationarity of the data.

[0040] It can be understood that the adaptive weight mechanism is used to dynamically adjust the balance parameter γ , and specifically includes: calculating the variance within a local time window as a measure of local stationarity. For a given time series , within a time window centered on with a length of , the local variance It can be calculated as: where, is the mean value of the data within the time window; According to the local stationarity metric, an adaptive weight is generated through mapping by the Sigmoid function: where, is the adaptive weight at time point , and are adjustable parameters used to control the shape and range of the mapping function, and the value range of the Sigmoid function is between (0, 1); based on the adaptive weight, the balance parameter γ is dynamically adjusted to achieve the dynamic balance of the global and local trend weights. Assuming the original fixed γ value is , then the dynamically adjusted γ value is calculated as: where, represents the value of when paying more attention to the global trend.

[0041] Step S60: Integrate the global trend model, the time-adaptive Laplacian regularization module, the graph structure regularization module, and the adaptive weight mechanism to construct a unified optimization problem.

[0042] It should be noted that by integrating the global trend model, the time-adaptive Laplacian regularization module, the graph structure regularization module, and the adaptive weight mechanism, a unified optimization model is constructed, and the optimization model is expressed as: where the global trend data is , the local trend data is , the variable relationship data is , and n represents the number of variables X.

[0043] Step S70: Use the alternating direction multiplier method to iteratively solve the optimization problem until the convergence condition is satisfied, and output the complete interpolated data.

[0044] The step of using the alternating direction multiplier method to iteratively solve the optimization problem until the convergence condition is satisfied and outputting the complete interpolated data includes: Introduce an auxiliary variable ; , , , and represent the optimization problem as: Construct the augmented Lagrangian function as follows: where is the Lagrange multiplier, is the penalty parameter; Update X: Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update : Fix the remaining variables and solve the following optimization problem: Update the Lagrange multiplier: Iteratively update the above sub-problems until the convergence condition is met, and output the complete imputed data.

[0045] In a specific implementation, the method further includes: dividing the risk levels of the data according to the filled and repaired data, and outputting the classification results; the risk levels include high risk, medium risk, and low risk, which respectively correspond to different data fluctuation ranges.

[0046] In a specific implementation, this embodiment proposes an industrial process operation data filling and repairing method considering global / local characteristics, which simultaneously pursues the global trend and local trend of data based on cyclic matrix nuclear norm minimization and time-adaptive Laplacian regularization, and additionally uses a variable relationship representation of a graph structure with adaptive weights (graph structure regularization) and an adaptive weight mechanism to consider the relationship between variables and dynamically adjust the balance parameter: Cyclic matrix nuclear norm minimization is used for data filling and capturing the global trend of data.

[0047] Local trend of data: The adjacent relationship between data is obtained by using the graph-based Laplacian matrix, so as to construct a Laplacian regularization term, but it can only process data with a fixed sampling frequency. Therefore, a time-adaptive Laplacian regularization module is introduced to handle the situation where the sampling frequencies of different variables are inconsistent, and the Laplacian regularization is corrected by introducing a time weighting factor.

[0048] Variable relationship characterization module based on graph structure: The similarity weights between variables are described by using the graph structure and the adjacency matrix, so that the imputation results of related variables tend to be similar.

[0049] Balanced parameter adaptive weight mechanism module: Dynamically adjust the balanced parameter according to the local stationarity of the data γ to achieve the dynamic balance of the global and local trend weights.

[0050] Finally, a unified optimization problem is constructed and solved by using the alternating direction method of multipliers (ADMM).

[0051] In this embodiment, the global long-term trend of the data is captured by minimizing the nuclear norm of the circulant matrix, the local short-term fluctuation characteristics of the data are described by combining the time-adaptive Laplacian regularization, and the correlation between variables is characterized by using the graph structure regularization with adaptive weights. The balanced parameter is dynamically adjusted through the adaptive weight mechanism to adapt to different time series characteristics and optimize the imputation effect. Finally, the alternating direction method of multipliers is used to solve the optimization model to generate the data after filling and repair. Considering the global trend, local fluctuation and relationship between variables of the data, the accuracy and reliability of data filling are significantly improved.

[0052] In addition, an embodiment of the present application also proposes a computer-readable storage medium, on which a program for filling and repairing the operation data of an industrial process is stored. When the program for filling and repairing the operation data of the industrial process is executed by a processor, the steps of the method for filling and repairing the operation data of the industrial process as described above are implemented.

[0053] Refer to Figure 4 , Figure 4 which is the structural block diagram of the first embodiment of the operation data filling and repairing system of the industrial process of the present application.

[0054] As Figure 4 shown, the operation data filling and repairing system proposed by the embodiment of the present application includes: A data acquisition module 10, configured to acquire partial observed data in an industrial process, where the partial observed data includes missing data; A global trend module 20, configured to construct a data global trend model based on minimizing the nuclear norm of the circulant matrix, and perform imputation on the partial observed data through the operation of minimizing the nuclear norm of the circulant matrix; The short-term feature module 30 is used to introduce a time-adaptive Laplacian regularization module, dynamically adjust the local smoothing constraint weights according to the time interval of the data, and capture the short-term fluctuation features of partial observed data; The regularization module 40 is used to construct a graph structure regularization module with adaptive weights, learn the correlation weights between variables through the adjacency matrix and the graph structure, and make the imputation results of related variables tend to be consistent; The adaptive weight module 50 is used to design an adaptive weight mechanism, and dynamically adjust the balance parameter between the global trend and the local trend according to the local stationarity of the data; The optimization problem module 60 is used to integrate the global trend model, the time-adaptive Laplacian regularization module, the graph structure regularization module and the adaptive weight mechanism to construct a unified optimization problem; The output module 70 is used to iteratively solve the optimization problem by using the alternating direction multiplier method until the convergence condition is satisfied, and output the complete imputed data.

[0055] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present application. In specific applications, those skilled in the art can set according to needs, and the present application does not make any restrictions on this.

[0056] In this embodiment, the global long-term trend of the data is captured by minimizing the nuclear norm of the circulant matrix, the local short-term fluctuation features of the data are described by combining the time-adaptive Laplacian regularization, and the correlation between variables is characterized by the graph structure regularization with adaptive weights. The balance parameter is dynamically adjusted through the adaptive weight mechanism to adapt to different time series characteristics and optimize the imputation effect. Finally, the alternating direction multiplier method is used to solve the optimization model to generate the data after filling and repairing. Considering the global trend, local fluctuations and relationships between variables of the data, the accuracy and reliability of data filling are significantly improved.

[0057] It should be noted that the above-described work process is only illustrative and does not limit the protection scope of the present application. In practical applications, those skilled in the art can select some or all of them according to actual needs to achieve the purpose of the solution of this embodiment, and no restrictions are made here.

[0058] In addition, for the technical details not described in detail in this embodiment, reference can be made to the method for filling and repairing the operation data of the industrial process provided in any embodiment of the present application, and details are not described here again.

[0059] In addition, it should be noted that in this text, the terms "including", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or system comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or system including such element.

[0060] The serial numbers of the embodiments of the present application above are only for description and do not represent the superiority or inferiority of the embodiments.

[0061] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as a read-only memory (ROM) / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present application. The above is only the preferred embodiment of the present application and does not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A method for filling and repairing operation data of an industrial process, characterized in that: include: Collecting partial observation data of an industrial process, wherein the partial observation data includes missing data; Based on the minimization of the nuclear norm of the circulant matrix, a global trend model of the data is constructed, and some observation data are interpolated through the minimization operation of the nuclear norm of the circulant matrix; A time-adaptive Laplace regularization module is introduced to dynamically adjust the local smoothing constraint weights according to the time interval of the data to capture the short-term fluctuation characteristics of some observed data; Construct an adaptive weighted graph structure regularization module to learn the correlation weights between variables through the adjacency matrix and graph structure, so that the interpolation results of related variables tend to be consistent; Design an adaptive weight mechanism to dynamically adjust the balance parameters between global and local trends according to the local stability of the data; Integrate the global trend model, the time-adaptive Laplace regularization module, the graph structure regularization module and the adaptive weight mechanism to construct a unified optimization problem; The optimization problem is iteratively solved by using an alternating direction multiplier method until a convergence condition is met, and the interpolated complete data is output.

2. The method for filling and repairing the operation data of an industrial process according to claim 1, characterized in that: The circulant matrix nuclear norm is minimized, including: Converting the portion of observation data into a circulant matrix form; Fast Fourier transform is used to calculate the nuclear norm of the circulant matrix, and global trend modeling and missing data interpolation are achieved by minimizing the nuclear norm.

3. The method for filling and repairing the operation data of an industrial process according to claim 1, characterized in that: The step of introducing a time-adaptive Laplace regularization module to dynamically adjust the local smoothing constraint weight according to the time interval of the data to capture the short-term fluctuation characteristics of some observed data includes: Constructing a Laplace matrix, and determining the adjacent relationship between data according to the Laplace matrix; A time weighting factor is introduced to perform weighted correction on the Laplace regularization term according to the time interval between data points; The time weighting factor is used to adjust the smoothness constraint of the data according to the time interval, and the short-term fluctuation characteristics of part of the observed data are captured in the Laplace matrix.

4. The method for filling and repairing the operation data of an industrial process according to claim 1, characterized in that: The variable similarity weights in the adaptive weighted graph regularization module Defined as: in, is the scale factor to control the similarity, and Represents two different variables, is a normalization constant used to ensure that the sum of all weights is 1.

5. The method for filling and repairing the operation data of an industrial process according to claim 1, characterized in that: The adaptive weight mechanism is used to dynamically adjust the balance parameters γ , specifically including: Calculate the variance in a local time window as a measure of local stationarity for a given time series , in The center is In the time window of It can be calculated as: in, is the mean of the data in this time window; According to the local stationarity metric, adaptive weights are generated through Sigmoid function mapping: in, It's at the time The adaptive weight at and It is an adjustable parameter used to control the shape and range of the mapping function. The value range of the Sigmoid function is between (0, 1); Based on the adaptive weight, the balance parameter γ is dynamically adjusted to achieve a dynamic balance between the global and local trend weights. Assuming that the original fixed γ value is , then the dynamically adjusted γ value The calculation method is: in, Indicates that when more attention is paid to the overall trend The value of .

6. The method for filling and repairing the operation data of an industrial process according to claim 5, characterized in that: The step of integrating the global trend model, the time-adaptive Laplace regularization module, the graph structure regularization module and the adaptive weight mechanism to construct a unified optimization problem includes: The global trend model, the time-adaptive Laplace regularization module, the graph structure regularization module and the adaptive weight mechanism are integrated to construct a unified optimization model, which is expressed as: Among them, the global trend data is , the local trend data is , the variable relationship data is , n represents the number of variables x.

7. The method for filling and repairing the operation data of an industrial process according to claim 6, characterized in that: The step of iteratively solving the optimization problem by using the alternating direction multiplier method until the convergence condition is met and outputting the interpolated complete data comprises: Introducing auxiliary variables ; , , , And the optimization problem is expressed as: The augmented Lagrangian function is constructed as follows: in is the Lagrange multiplier, is the penalty parameter; Update X: Fix the remaining variables and solve the following optimization problem: renew : Fix the remaining variables and solve the following optimization problem: renew : Fix the remaining variables and solve the following optimization problem: renew : Fix the remaining variables and solve the following optimization problem: renew : Fix the remaining variables and solve the following optimization problem: Update the Lagrange multipliers: The above sub-problems are iteratively updated until the convergence conditions are met, and the complete data after interpolation is output.

8. An industrial process operation data filling and repair system, characterized in that: include: A data collection module, used for collecting part of the observation data of the industrial process, wherein the part of the observation data includes missing data; The global trend module is used to construct a data global trend model based on the minimization of the nuclear norm of the circulant matrix, and to interpolate some observation data through the nuclear norm minimization operation of the circulant matrix; The short-term feature module is used to introduce a time-adaptive Laplace regularization module, dynamically adjust the local smoothing constraint weights according to the time interval of the data, and capture the short-term fluctuation characteristics of some observed data; Regularization module, which is used to construct a graph structure regularization module with adaptive weights. It learns the correlation weights between variables through the adjacency matrix and graph structure, so that the interpolation results of related variables tend to be consistent; Adaptive weight module, used to design an adaptive weight mechanism to dynamically adjust the balance parameters between global and local trends according to the local stability of the data; An optimization problem module, used to integrate the global trend model, the time-adaptive Laplace regularization module, the graph structure regularization module and the adaptive weight mechanism to construct a unified optimization problem; The output module is used to iteratively solve the optimization problem by using an alternating direction multiplier method until a convergence condition is met, and output the complete data after interpolation.

9. A computer device, characterized in that: The device comprises: a memory and a processor, and when the processor runs the computer instructions stored in the memory, the processor executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The method comprises instructions, which, when executed on a computer, cause the computer to execute the method according to any one of claims 1 to 7.

Images