Online fault detection method and system for chemical production based on big data

Abstract

This invention relates to the field of fault detection technology, and more specifically, to a method and system for online fault detection in chemical production based on big data. The method includes: collecting operating data from production equipment; constructing real-time data matrices of different sizes; extracting historical data of the same size as the real-time data matrices to construct a historical data matrix; and calculating the similarity between any real-time data matrix and each historical data matrix of the same size using a sliding window. This invention analyzes potential long- and short-term trends in the data by constructing real-time data matrices of different sizes, calculates the similarity between the real-time data matrix and the historical data matrix using a moving sliding window, estimates the estimated value of the comprehensive control limit of the real-time data based on the control limit values ​​of similar historical data matrices, and accurately determines the fault condition of the equipment based on the difference between the estimated value and the comprehensive control limit of the real-time data.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology. More specifically, this invention relates to a method and system for online fault detection in chemical production based on big data. Background Technology

[0002] Chemical production processes involve multiple stages and equipment, such as reactors, pumps, and valves. The operational status of each stage requires real-time monitoring to promptly detect potential faults. Traditional fault detection methods in chemical production rely on operator experience and subjective judgment, which may fail to detect faults in a timely manner, potentially leading to hazards. With the continuous development of big data technology, technicians are using statistical models, machine learning models, and physical models to construct hybrid models for online fault detection in chemical production.

[0003] Currently, the dynamic characteristics of chemical production processes are typically nonlinear and time-varying, with complex nonlinear relationships existing between parameters such as temperature, pressure, and flow rate. Traditional methods such as PCA and linear regression cannot capture these complex relationships over time, resulting in low accuracy in equipment fault detection. Therefore, this patent addresses the problem of online fault detection in chemical production using fault detection technology. Summary of the Invention

[0004] This invention provides a method and system for online fault detection in chemical production based on big data, aiming to solve the problem that traditional methods such as PCA statistics and linear regression used in related technologies cannot capture the complex relationship of multiple variables over time, resulting in low accuracy in equipment fault detection.

[0005] In a first aspect, the present invention provides an online fault detection method for chemical production based on big data, comprising: collecting operating data of production equipment, constructing real-time data matrices of different sizes, and extracting historical data of the same size as the real-time data matrices to construct a historical data matrix; calculating the similarity between any real-time data matrix and each historical data matrix of the same size using a sliding window, wherein the similarity is used to characterize the degree of difference between the data within the sliding window when traversing the real-time data matrix and the historical data matrix using the sliding window, and selecting the historical data matrix with the highest similarity to the real-time data matrix as the most similar matrix, and calculating the comprehensive control limit of the most similar matrix, the calculation formula being: In the formula, Most similar matrix The comprehensive control limits, Most similar matrix The control limit of the Hotelling statistic. Most similar matrix The squared prediction error control limit, Most similar matrix With the standard F-distribution data matrix Statistic, The matrix size is represented; the average value of the product between the comprehensive control limit of the most similar matrix of different sizes and the similarity with the real-time data matrix is ​​used as the estimated value of the comprehensive control limit of the real-time data. Based on the comparison between the comprehensive control limit of the real-time data and the estimated value, fault detection is performed on the production equipment.

[0006] Furthermore, fault detection of production equipment includes: if the difference between the comprehensive control limit and the estimated value of the real-time data is greater than the alarm threshold, a fault is determined to have occurred in the chemical production process, and an alarm is triggered; if the difference between the comprehensive control limit and the estimated value of the real-time data is less than or equal to the alarm threshold, a fault is determined to have occurred in the production equipment, and an alarm is triggered.

[0007] Furthermore, the estimated value of the comprehensive control limit for real-time data is calculated using the following formula:

[0008] In the formula, It is an estimated value of the comprehensive control limits based on real-time data. Represents real-time data matrix Most similar matrix The degree of similarity, Represents the most similar matrix The comprehensive control limits, This indicates the number of real-time data matrices of different sizes.

[0009] Furthermore, the similarity between any real-time data matrix and each historical data matrix of the same size is calculated using the following formula: In the formula, Represents real-time data matrix With historical data matrix similarity, Represents the normalization function. This indicates the number of times the sliding window has been slid. Indicates the sliding window in the matrix Upper The characteristic standard deviation of the second slip Indicates the sliding window in the matrix Upper The characteristic standard deviation of the second slip Representation matrix In the The second sliding window Line number Column characteristics, Representation matrix In the The second sliding window Line number The column's characteristics, the size of the sliding window is .

[0010] Furthermore, a real-time data matrix of different sizes is constructed, including: firstly, the collected operating data of the production equipment is centrally processed, wherein the operating data includes temperature, pressure, flow rate, liquid level, vibration and current.

[0011] Furthermore, the centralization process for any type of data in the operational data includes: for any type of data, subtracting the mean of the data in that type from any data in that type to obtain the centralized value.

[0012] Furthermore, the empirical value of the alarm threshold is 1.

[0013] In a second aspect, a big data-based online fault detection system for chemical production is also provided, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the big data-based online fault detection method for chemical production described above.

[0014] Beneficial effects: By constructing real-time data matrices of different sizes, we can analyze the potential short- and long-term trends of the data. By using a sliding window to calculate the similarity between the real-time data matrix and the historical data matrix, we can estimate the comprehensive control limit of the real-time data based on the control limit of the similar historical data matrix. Based on the difference between the estimated value and the comprehensive control limit of the real-time data, we can accurately determine the fault status of the equipment. Attached Figure Description

[0015] By referring to the accompanying drawings, several embodiments of the invention are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein:

[0016] Figure 1 This is a flowchart illustrating the estimated value of the comprehensive control limit for calculating real-time data according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] like Figure 1 As shown, S101: Collects operating data of production equipment.

[0020] In one embodiment, sensors are installed at key nodes of chemical production equipment, such as reactors, pumps, valves, and pipelines. The corresponding sensors are used to collect the temperature, pressure, flow rate, liquid level, vibration, and current of different equipment. The collected data is then transmitted to a central data platform for storage and processing through a real-time data acquisition system.

[0021] In one embodiment, the method further includes centralizing the collected operational data of the production equipment. Specifically, for any type of data, the mean of that type of data is subtracted from any data point in that type to obtain the centralized value. For example, if temperature data is collected... , , , ,right The value after centralization is At this point, all values ​​after data centralization processing can be obtained.

[0022] In one embodiment, statistics in the prior art The control limits are calculated based on the premise that the normal data follows a multivariate normal distribution and are fixed values. However, in actual data distributions, due to the influence of environmental noise, the data does not conform to a multivariate normal distribution and exhibits periodic trends, leading to control limit errors. This results in untimely fault identification or false warnings for production equipment. Therefore, this embodiment constructs real-time data matrices of different sizes, analyzes the possible long-term and short-term trends of the data, calculates the similarity between the real-time data matrix and the historical data matrix using a moving sliding window, and predicts the estimated value of the comprehensive control limit of the real-time data based on the control limit values ​​of similar historical data matrices. Based on the difference between the comprehensive value and the estimated value of the real-time data, it accurately determines whether a fault has occurred in the production equipment.

[0023] S102: Calculate the similarity between any real-time data matrix and each historical data matrix of the same size.

[0024] In one embodiment, after obtaining all the values ​​processed by the data center, real-time data matrices of different sizes are constructed. For example, matrices of different sizes are constructed as follows: ,common Real-time data matrices of different sizes The maximum value is the time length from the start of data acquisition to the current time. An example matrix size is... Size is ,matrix Includes the past From time to the present time A sample data set is generated. Then, historical data of the same size as the real-time data matrix is ​​extracted to construct a historical data matrix. Historical data refers to the operational data of the production equipment prior to this data set.

[0025] In one embodiment, a sliding window is used to calculate the similarity between any real-time data matrix and each historical data matrix of the same size. The similarity is used to characterize the degree of difference between the data within the sliding window when the real-time data matrix and the historical data matrix are traversed using the sliding window.

[0026] In one embodiment, the formula for calculating the similarity between any real-time data matrix and each historical data matrix of the same size is: In the formula, Represents real-time data matrix With historical data matrix similarity, Represents the normalization function. This indicates the number of times the sliding window has been slid. Indicates the sliding window in the matrix Upper The characteristic standard deviation of the second slip Indicates the sliding window in the matrix Upper The characteristic standard deviation of the second slip Representation matrix In the The second sliding window Line number Column characteristics, Representation matrix In the The second sliding window Line number The column's characteristics, the size of the sliding window is .

[0027] S103: Select the historical data matrix with the highest similarity to the real-time data matrix as the most similar matrix.

[0028] In one embodiment, for any size real-time data matrix, the historical data matrix with the highest similarity is selected as the most similar matrix to the real-time data matrix. This ultimately yields... The most similar matrix .

[0029] S104: Calculate the comprehensive control limits of the most similar matrix.

[0030] In one embodiment, based on the most similar matrix , and Statistical measure to calculate the comprehensive control limit The calculation formula is: In the formula, Most similar matrix The comprehensive control limits, Most similar matrix The control limit of the Hotelling statistic. Most similar matrix The squared prediction error control limit, Most similar matrix With the standard F-distribution data matrix Statistic, Indicates the size of the matrix. Calculates the most similar matrix. , and The statistical measures are based on existing technology and will not be discussed in detail here.

[0031] In one embodiment, due to the traditional fault identification method, the assumption that the data follows an F-distribution is only used through... or Whether a fault has occurred is determined by whether the corresponding control limits are exceeded. However, data affected by noise can influence the data distribution, making it impossible to determine whether a fault has occurred using a single statistical indicator. Therefore, this patent addresses this issue based on existing technologies. Test the similarity between the most similar matrix and the data matrix with the standard F distribution. Statistics, through , and Statistical calculation of comprehensive control limits By estimating the real-time comprehensive control limit using historical comprehensive control limit values, we obtain an estimated value of the comprehensive control limit based on real-time data.

[0032] S105: Calculate the estimated value of the comprehensive control limit for real-time data.

[0033] In one embodiment, the estimated value of the comprehensive control limit for real-time data is calculated as the average of the products of the comprehensive control limits of the most similar matrices of different sizes and their similarity to the real-time data matrix. The formula for calculating the estimated value is: In the formula, It is an estimated value of the comprehensive control limits based on real-time data. Represents real-time data matrix Most similar matrix The degree of similarity, Represents the most similar matrix The comprehensive control limits, This indicates the number of real-time data matrices of different sizes.

[0034] S106: Based on the comparison results of the comprehensive control limits and the estimated values ​​based on real-time data, fault detection is performed on the production equipment.

[0035] In one embodiment, if the difference between the comprehensive control limit and the estimated value of the real-time data is greater than the alarm threshold, a chemical production malfunction is determined, and an alarm is triggered; if the difference between the comprehensive control limit and the estimated value of the real-time data is less than or equal to the alarm threshold, a production equipment malfunction is determined, and an alarm is triggered. In this embodiment, the empirical value of the alarm threshold is 1. In other embodiments, the empirical value of the alarm threshold is 1.1 or 1.2, etc., and can be adjusted according to the specific implementation.

[0036] By constructing real-time data matrices of different sizes, we can analyze the potential short- and long-term trends of the data. By using a sliding window, we can calculate the similarity between the real-time data matrix and the historical data matrix. By using the control limits of similar historical data matrices, we can estimate the estimated value of the comprehensive control limits of the real-time data. Based on the difference between the estimated value and the comprehensive control limits of the real-time data, we can accurately determine the fault status of the equipment.

[0037] This invention also provides an online fault detection system for chemical production based on big data. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the online fault detection method for chemical production based on big data according to the first aspect of this invention.

[0038] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0039] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0040] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0041] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0042] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for online fault detection in chemical production based on big data, characterized in that, include: Collect operational data from production equipment, construct real-time data matrices of different sizes, and extract historical data of the same size as the real-time data matrices to construct a historical data matrix; A sliding window is used to calculate the similarity between any real-time data matrix and each historical data matrix of the same size. This similarity characterizes the degree of difference between data points within the sliding window when traversing the real-time and historical data matrices. The historical data matrix with the highest similarity to the real-time data matrix is ​​selected as the most similar matrix. A comprehensive control limit for the most similar matrix is ​​then calculated using the following formula: Where, The most similar matrix B n×m The comprehensive control limits, The most similar matrix B n×m The control limit of the Hotelling statistic. The most similar matrix B n×m The squared prediction error control limit, The most similar matrix B n×m The KS statistic of the standard F-distribution data matrix, where n×m represents the matrix size; The average value of the product between the comprehensive control limit of the most similar matrix of different sizes and the similarity with the real-time data matrix is ​​used as the estimated value of the comprehensive control limit of the real-time data. Based on the comparison between the comprehensive control limit of the real-time data and the estimated value, fault detection is performed on the production equipment.

2. The online fault detection method for chemical production based on big data according to claim 1, characterized in that, Fault detection of production equipment includes: If the difference between the comprehensive control limit and the estimated value of the real-time data is greater than the alarm threshold, a fault is determined to have occurred in the chemical production process, and an alarm is triggered.

3. The online fault detection method for chemical production based on big data according to claim 1, characterized in that, The estimated value of the comprehensive control limit for real-time data is calculated using the following formula: In the formula, TS is the estimated value of the comprehensive control limit for real-time data. Represents real-time data matrix A h×m With the most similar matrix B h×m The degree of similarity, B represents the most similar matrix h×m The comprehensive control limit, where o represents the number of real-time data matrices of different sizes.

4. The online fault detection method for chemical production based on big data according to claim 1, characterized in that, The similarity between any real-time data matrix and each historical data matrix of the same size is calculated using the following formula: Where, Represents real-time data matrix A n×m With historical data matrix B n×m The similarity is given by norm(), which represents the normalization function, and (n-2)(m-2) represents the number of times the sliding window slides. This indicates that the sliding window is in matrix A. n×m The characteristic standard deviation of the k-th sliding motion, This indicates that the sliding window is in matrix B. n×m The characteristic standard deviation of the k-th sliding motion, Representing matrix A n×m The feature in the i-th row and j-th column within the k-th sliding window. Representation matrix B n×m The feature in the i-th row and j-th column within the k-th sliding window, with a sliding window size of 3×3.

5. The online fault detection method for chemical production based on big data according to claim 1, characterized in that, Construct real-time data matrices of varying sizes, including: First, the collected operating data of the production equipment is centrally processed, including temperature, pressure, flow rate, liquid level, vibration, and current.

6. The online fault detection method for chemical production based on big data according to claim 5, characterized in that, The centralization process for any type of data in the operational data includes: for any type of data, subtracting the mean of that type of data from any data in that type to obtain the value after centralization.

7. The online fault detection method for chemical production based on big data according to claim 2, characterized in that, The empirical value for the alarm threshold is 1.

8. A big data-based online fault detection system for chemical production, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the big data-based online fault detection method for chemical production as described in any one of claims 1-7.

Images