Sintering process detection abnormal value elimination and repair method, equipment and storage device

Abstract

The present invention provides a method, device and storage device for eliminating and repairing abnormal values ​​in sintering process detection, which performs differential calculation on the collected historical time series data to obtain first-order differential time series data. Then, the detection sensitivity is set, and the historical time series data and first-order differential time series data of the detection parameters are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order differential time series data. Finally, the newly detected data is judged according to the upper and lower bounds of the fluctuation range. If it exceeds any fluctuation range, the newly detected data is an abnormal value and is directly eliminated. The abnormal value is repaired according to the median of the fluctuation range of the first-order differential time series data and the upper and lower bounds of the fluctuation range of the historical time series data. The beneficial effect of the present invention is to provide effective and accurate data for sintering process modeling and control, which is of great significance to ensuring the production safety of the sintering process.

Description

Technical Field

[0001] The present invention relates to the field of data processing in an iron ore sintering production process, and in particular to a method, device and storage device for eliminating and repairing abnormal values ​​detected in a sintering process. Background Art

[0002] Abnormal data is an outward manifestation of abnormalities in industrial production processes. These process anomalies often lead to low product yields and poor quality. Effective detection of abnormal data provides data support for monitoring and guidance for parameter decision-making in industrial production processes. Abnormal data detection enables production managers and operators to more intuitively understand the operational status of industrial processes, promptly identifying anomalies in production data and providing valuable insights for optimizing and adjusting production processes.

[0003] The sintering process has a complex mechanism. Through a series of processes, a mixture of iron ore, limestone, coke, etc. is manufactured into sintered ore with strong reducing properties, appropriate composition, good permeability, reasonable particle size, and appropriate mechanical strength. There are many parameters in the sintering process. Obtaining accurate and effective process information is the basis for achieving real-time monitoring of the sintering process. Detecting process parameters through sensors is the main way to obtain process information, and then obtaining time series data of the detection parameters. Due to the influence of sensor failure, environmental impact, human factors, etc., outliers often appear in the time series data of detection parameters, which makes it very necessary to eliminate and repair the detection outliers of the sintering process. Therefore, studying the method of eliminating and repairing outliers in the sintering process detection can provide effective and accurate data for sintering process modeling and control, which is of great significance to ensuring the production safety of the sintering process.

[0004] Current methods for removing and repairing outliers in detection primarily consider the statistical properties of the data itself, lacking consideration for data granularity and interpretability, limiting their applicability. The principle of reasonable granularity, which considers both the rationality and specificity of data, has garnered considerable attention and has been widely applied in time series data analysis. Utilizing this principle to address the problem of removing and repairing outliers in sintering process detection is a new trend. Therefore, it is necessary to develop a method for removing and repairing outliers in sintering process detection based on this principle. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides a method, device and storage device for removing and repairing abnormal values ​​in sintering process detection. The method mainly includes:

[0006] S1: Collect historical time series data of a certain detection parameter of the sintering process over a period of time, perform differential calculation on the historical time series data, and obtain first-order differential time series data;

[0007] S2: Using the reasonable granularity principle, the historical time series data and the first-order difference time series data of the detection parameters are processed separately to obtain the fluctuation range of the historical time series data and the first-order difference time series data;

[0008] S3: According to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an outlier and is directly eliminated. The outlier is repaired according to the median of the fluctuation range of the first-order difference time series data and the upper and lower bounds of the fluctuation range of the historical time series data.

[0009] Furthermore, step S1 specifically includes:

[0010] (2-1) The historical time series data of a certain detection parameter of the sintering process is collected over a period of time. The sampling interval of the data is set to 30 seconds to obtain the historical time series data T = {t1, t2, ..., t n}, n is the length of the time series;

[0011] (2-2) Perform differential calculation on the historical time series data to obtain the first-order differential time series data ΔT = {Δt1, Δt2, ..., Δt n-1}, where Δt i =t i+1 -t i ,i=1,2,...,n-1,t i Represents the i-th time series data in the historical time series data T, t i+1 Represents the i+1th time series data in the historical time series data T.

[0012] Furthermore, step S2 specifically includes the following process:

[0013] (3-1) Set the detection sensitivity α∈(0,+∞);

[0014] (3-2) Assume that the fluctuation range of the historical time series data of the detection parameter after processing by the reasonable granularity principle is Ω = [d L ,d H ], where d L Indicates the lower bound of the fluctuation range of T, d H Indicates the upper bound of the fluctuation range of T; suppose the fluctuation range of the processed first-order difference time series data is ΔΩ=[Δd L ,Δd H ], where Δd L Indicates the lower bound of the fluctuation range of ΔT, Δd H Indicates the upper limit of the fluctuation range of ΔT; the reasonable granularity principle uses the following formula to obtain the corresponding fluctuation range,

[0015]

[0016]

[0017] The function card(·) represents the number of elements in the set. The principle of reasonable granularity is to maximize f(Ω) and f(ΔΩ). After multiple calculations, the maximum values ​​of f(Ω) and f(ΔΩ) are obtained. The interval corresponding to the maximum values ​​of f(Ω) and f(ΔΩ) is the final fluctuation interval.

[0018] Furthermore, step S3 specifically includes the following process:

[0019] (4-1) Let the new detection data be t n+1 , then the first-order difference data can be obtained as Δt n , where Δt n =t n+1 -t n , t n The last data point in the historical time series data. The rules for determining whether the newly detected data is an outlier are as follows:

[0020] if Then the newly detected data t n+1 For outliers, remove them;

[0021] if Then the newly detected data t n+1 For outliers, remove them;

[0022] When t n+1 ∈[d L ,d H ] and Δt n ∈[Δd L ,Δd H ], the newly detected data t n+1 If it is a normal value, keep the new test data;

[0023] (4-2) According to the median of the fluctuation interval of the first-order difference time series data and the upper and lower bounds of the fluctuation interval of the historical time series data, the abnormal value is repaired. Let the repair data be t new , then:

[0024]

[0025] If t new <d L , then t new =d L If t new >d H , then t new =d H .

[0026] A storage device stores instructions and data for realizing a method for eliminating and repairing abnormal values ​​in sintering process detection.

[0027] A sintering process detection abnormal value elimination and repair device comprises: a processor and the storage device; the processor loads and executes instructions and data in the storage device to implement a sintering process detection abnormal value elimination and repair method.

[0028] The beneficial effects of the technical solution provided by the present invention are as follows: the present invention first collects the historical time series data of a certain detection parameter of the sintering process for a period of time, performs differential calculation on the historical time series data, and obtains the first-order differential time series data. Then, according to the set detection sensitivity, the historical time series data and the first-order differential time series data of the detection parameter are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order differential time series data. Finally, according to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an outlier and is directly eliminated. The outlier is repaired according to the median of the fluctuation range of the first-order differential time series data and the upper and lower bounds of the fluctuation range of the historical time series data. The reasonable granularity principle can be used to obtain the fluctuation range of the historical time series data and its first-order differential time series data, and the upper and lower bounds of the fluctuation range and the median of the fluctuation range of the first-order differential time series data can be used to repair the outlier. Therefore, the present invention can effectively eliminate and repair abnormal values ​​in the detection of the sintering process, provide effective and accurate data for the modeling and control of the sintering process, and has application value in sintering process status monitoring, safe production and modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0030] Figure 1 The present invention is a flowchart of a method for removing and repairing abnormal values ​​in sintering process detection according to an embodiment of the present invention.

[0031] Figure 2 Schematic diagram of the historical time series data of the sintering endpoint in an embodiment of the present invention.

[0032] Figure 3 Schematic diagram of the historical time series data of the sintering endpoint after elimination and repair in an embodiment of the present invention.

[0033] Figure 4 It is a schematic diagram of the operation of the hardware device in the embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0035] The embodiments of the present invention provide a method, device and storage device for removing and repairing abnormal values ​​detected in a sintering process.

[0036] Please refer to Figure 1 , Figure 1 This is a flow chart of a method for removing and repairing abnormal values ​​in sintering process detection according to an embodiment of the present invention, which specifically includes:

[0037] S1: Collect historical time series data of a certain detection parameter of the sintering process over a period of time, perform differential calculation on the historical time series data, and obtain first-order differential time series data;

[0038] S2: Based on the set detection sensitivity, the historical time series data and first-order difference time series data of the detection parameters are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order difference time series data;

[0039] S3: According to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an outlier and is directly eliminated. The outlier is repaired according to the median of the fluctuation range of the first-order difference time series data and the upper and lower bounds of the fluctuation range of the historical time series data.

[0040] The specific implementation steps are as follows:

[0041] (1) Collect sintering production history data to obtain original sample data

[0042] The historical data of sintering production is stored in the local database of the industrial computer in the operation room in the form of daily reports. Based on the data in the daily reports, taking the sintering end point as an example, one month's historical data is collected with a sampling interval of 5 seconds to form the original sample data.

[0043] (2) Obtaining first-order difference time series data

[0044] 120 data are selected as historical time series data, and the difference calculation is performed on the historical time series data to obtain the first-order difference time series data. Specifically:

[0045] (2-1) The historical time series data of a certain detection parameter of the sintering process is collected over a period of time. The sampling interval of the data is set to 30 seconds to obtain the historical time series data T = {t1, t2, ..., t n}, n is the length of the time series;

[0046] (2-2) Perform differential calculation on the historical time series data to obtain the first-order differential time series data ΔT = {Δt1, Δt2, ..., Δt n-1}, where Δt i =t i+1 -t i ,i=1,2,...,n-1,t i Represents the i-th time series data in the historical time series data T, t i+1 Represents the i+1th time series data in the historical time series data T.

[0047] (3) Obtaining the fluctuation range

[0048] The historical time series data and first-order difference time series data of the detection parameters are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order difference time series data. Specifically:

[0049] (3-1) Set the detection sensitivity α∈(0,+∞), which reflects the sensitivity to abnormal values. In this example, the detection sensitivity is set to 1;

[0050] (3-2) Assume that the fluctuation range of the historical time series data of the detection parameter after processing by the reasonable granularity principle is Ω = [d L ,d H ], where d L Indicates the lower bound of the fluctuation range of T, d H It represents the upper bound of the fluctuation range of T. Assume that the fluctuation range of the first-order difference time series data after processing is ΔΩ=[Δd L ,Δd H ], where Δd L Indicates the lower bound of the fluctuation range of ΔT, Δd H Indicates the upper limit of the fluctuation range of ΔT. The reasonable granularity principle uses the following formula to calculate the corresponding fluctuation range:

[0051]

[0052]

[0053] Here, the function card(·) represents the number of elements in the set. The principle of reasonable granularity is to maximize f(Ω) and f(ΔΩ). Since the upper and lower bounds are derived from the original data, the original data can be sorted from small to large and set as the upper and lower bounds of the fluctuation range in sequence. Multiple calculations can be performed to obtain the maximum values ​​of f(Ω) and f(ΔΩ). The interval corresponding to the maximum values ​​of f(Ω) and f(ΔΩ) is Ω = [d L ,d H ] and ΔΩ=[Δd L ,ΔdH ] is the final result.

[0054] In this implementation, the fluctuation range of the historical time series data is calculated to be [22.2792, 22.9262], and the fluctuation range of the first-order difference time series data is [-0.0508, 0.0319].

[0055] (4) Removal and repair

[0056] According to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an outlier and is directly eliminated. Then, according to the median of the fluctuation range of the first-order difference time series data and the upper and lower bounds of the fluctuation range of the historical time series data, the outlier is repaired. Specifically:

[0057] (4-1) Let the new detection data be t n+1 , then the first-order difference data can be obtained as Δt n , where Δt n =t n+1 -t n , t n The last data point in the historical time series data. The rules for determining whether the newly detected data is an outlier are as follows:

[0058] if Then the newly detected data t n+1 For outliers, remove them;

[0059] if Then the newly detected data t n+1 For outliers, remove them;

[0060] When t n+1 ∈[d L ,d H ] and Δt n ∈[Δd L ,Δd H ], the newly detected data t n+1 If it is a normal value, keep the new test data;

[0061] (4-2) According to the median of the fluctuation interval of the first-order difference time series data and the upper and lower bounds of the fluctuation interval of the historical time series data, the abnormal value is repaired. Let the repair data be t new , then:

[0062]

[0063] If t new <d L , then t new =d L If tnew >d H , then t new =d H .

[0064] If the newly detected data is 21.2752, it is judged as an outlier and the data 21.2752 is directly removed, and the data at that location is patched to 22.6709. Figure 2 The historical time series data and a new data to be tested are shown. The last time series data is the data to be tested, and it can be seen that it has obvious anomalies. After testing and repairing with the method provided by the present invention, the following can be obtained: Figure 3 The results shown show that the abnormal data has been repaired. Figure 2 and Figure 3 It can be seen from the experimental results that the method provided by the present invention can meet the needs of detecting abnormal values, effectively realize the elimination and repair of abnormal values ​​detected in the sintering process, and has application value in sintering process state monitoring and modeling.

[0065] See Figure 4 , Figure 4 4 is a schematic diagram of the working of the hardware device of an embodiment of the present invention, wherein the hardware device specifically comprises: a sintering process detection abnormal value removal and repair device 401, a processor 402 and a storage device 403.

[0066] A device 401 for removing and repairing abnormal values ​​during sintering process detection: The device 401 for removing and repairing abnormal values ​​during sintering process detection implements a method for removing and repairing abnormal values ​​during sintering process detection.

[0067] Processor 402: The processor 402 loads and executes the instructions and data in the storage device 403 to implement the method for eliminating and repairing abnormal values ​​in sintering process detection.

[0068] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the method for eliminating and repairing abnormal values ​​in sintering process detection.

[0069] The beneficial effects of the present invention are as follows: the present invention first collects the historical time series data of a certain detection parameter of the sintering process for a period of time, performs differential calculation on the historical time series data, and obtains the first-order differential time series data. Then, according to the set detection sensitivity, the historical time series data and the first-order differential time series data of the detection parameter are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order differential time series data. Finally, according to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an abnormal value and is directly eliminated. According to the median of the fluctuation range of the first-order differential time series data and the upper and lower bounds of the fluctuation range of the historical time series data, the abnormal value is repaired. The reasonable granularity principle can be used to obtain the fluctuation range of the historical time series data and its first-order differential time series data, and the upper and lower bounds of the fluctuation range and the median of the fluctuation range of the first-order differential time series data can be used to repair the abnormal value. Therefore, the present invention can effectively eliminate and repair abnormal values ​​in the detection of the sintering process, provide effective and accurate data for the modeling and control of the sintering process, and has application value in sintering process status monitoring, safe production and modeling.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for eliminating and repairing abnormal values ​​in sintering process detection, characterized by: include: S1: Collect historical time series data of a certain detection parameter of the sintering process over a period of time, perform differential calculation on the historical time series data, and obtain first-order differential time series data; S2: Based on the set detection sensitivity, the historical time series data and first-order difference time series data of the detection parameters are processed separately using the reasonable granularity principle to obtain the fluctuation range of the historical time series data and the first-order difference time series data; Step S2 specifically includes the following process: (3-1) Set the detection sensitivity α∈(0,+∞); (3-2) Assume that the fluctuation range of the historical time series data of the detection parameter after processing by the reasonable granularity principle is Ω = [d L ,d H ], where d L Indicates the lower bound of the fluctuation range of T, d H Indicates the upper limit of the fluctuation range of T; suppose the fluctuation range of the processed first-order difference time series data is ΔΩ=[Δd L ,Δd H ], where Δd L Indicates the lower bound of the fluctuation range of ΔT, Δd H Indicates the upper limit of the fluctuation range of ΔT; the reasonable granularity principle uses the following formula to obtain the corresponding fluctuation range, The function card(·) represents the number of elements in the set. The principle of reasonable granularity is to maximize f(Ω) and f(ΔΩ). After multiple calculations, the maximum values ​​of f(Ω) and f(ΔΩ) are obtained. The interval corresponding to the maximum values ​​of f(Ω) and f(ΔΩ) is the final fluctuation interval. S3: According to the upper and lower bounds of the fluctuation range, the newly detected data is judged. If it exceeds the fluctuation range, the newly detected data is an outlier and is directly eliminated. The outlier is repaired according to the median of the fluctuation range of the first-order difference time series data and the upper and lower bounds of the fluctuation range of the historical time series data; Step S3 specifically The following processes are included: (4-1) Let the new detection data be t n+1 , then the first-order difference data can be obtained as Δt n , where Δt n =t n+1 -t n , t n The last data point in the historical time series data. The rules for judging whether the newly detected data is an outlier are as follows: if Then the newly detected data t n+1 For outliers, remove them; if Then the newly detected data t n+1 For outliers, remove them; When t n+1 ∈[d L ,d H ] and Δt n ∈[Δd L ,Δd H ], the newly detected data t n+1 If it is a normal value, keep the new test data; (4-2) According to the median of the fluctuation interval of the first-order difference time series data and the upper and lower bounds of the fluctuation interval of the historical time series data, the abnormal value is repaired. Let the repair data be t new , then: If t new <d L , then t new =d L ; if t new >d H , then t new =d H .

2. The method for eliminating and repairing abnormal values ​​in sintering process detection according to claim 1, characterized in that: Step S1 specifically includes: (2-1) The historical time series data of a certain detection parameter of the sintering process is collected over a period of time. The sampling interval of the data is set to 30 seconds to obtain the historical time series data T = {t1, t2, ..., t n }, n is the length of the time series; (2-2) Perform differential calculation on the historical time series data to obtain the first-order differential time series data ΔT = {Δt1, Δt2, ..., Δt n-1 }, where Δt i =t i+1 -t i ,i=1,2,...,n-1,t i Represents the i-th time series data in the historical time series data T, t i+1 Represents the i+1th time series data in the historical time series data T.

3. A storage device, characterized in that: The storage device stores instructions and data for implementing the method for eliminating and repairing abnormal values ​​in sintering process detection according to any one of claims 1 to 2.

4. A device for removing and repairing abnormal values ​​during sintering process detection, characterized by: include: A processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the method for eliminating and repairing abnormal values ​​in sintering process detection according to any one of claims 1 to 2.

Images