Method for diagnosing stability of running process of hot rolling mill

Abstract

The invention discloses a hot rolling mill operation process stability diagnosis method, which is characterized in that the definition of the instability of the hot rolling mill operation process is defined, and based on historical working condition parameter data of equipment, a stability monitoring index is extracted from parameters having strong correlation with the quality of a hot rolling product; whether the equipment is unstable or not is judged through a comprehensive anomaly judgment rule based on an SPC criterion, and running process volatility / stability monitoring indexes of all equipment parts are extracted through equipment running control parameters in the equipment; according to the method, judgment of unstable conditions and potential fault cause parts of the hot rolling mill equipment within a recent time range in history is realized, the time range of instability of the hot rolling mill equipment and corresponding abnormal characteristics are given, and a complete instability diagnosis conclusion is obtained. The method is helpful for guiding equipment maintenance and fault location, reducing accidents and improving production quality and efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of industrial equipment diagnosis, and in particular to a method for diagnosing the stability of a hot rolling mill operation process. Background Art

[0002] Hot rolling technology is a metallurgical process that plastically deforms metal billets at high temperatures. It is mainly used to produce steel plates, steel strips and other shaped steel products. The main process steps of hot rolling technology include heating section, rough rolling section, finishing section, cooling section and coiling section, among which the finishing section is the most critical. With the development of information technology, modern hot rolling production lines have adopted advanced automatic control systems and real-time quality monitoring systems to achieve precise control and defect identification, thereby improving product quality and stability.

[0003] The stability of the hot rolling mill equipment operation process is an important performance indicator to ensure its continuous, efficient and stable production of high-quality strip steel products under high temperature conditions, and is an important aspect affecting product quality stability and safe production. During the hot rolling process, the operation status of the mill is affected by many factors, including: roll system stiffness, roll gap control, dynamic response characteristics of the mill, hydraulic system and control system, mechanical design, maintenance and operating conditions, etc.

[0004] At present, the research in this field mainly focuses on the regulation of hot rolling production process parameters, control parameters, hot rolling product quality or specification accuracy. The definition of the stability of the hot rolling mill equipment itself and how to diagnose it during operation are still unclear and need to be studied in depth. Since the hot rolling production line involves many equipment, the causes of equipment instability are also relatively complex. In addition, in the actual production environment, there is an extreme lack of structured data on the causes of equipment failures that can be used for analysis, which makes it difficult to diagnose the stability of the equipment operation process. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for diagnosing the stability of the operation process of a hot rolling mill. Based on business experience and physical logic, the method clarifies the definition of the stability of the operation process of hot rolling equipment and the corresponding monitoring indicators, and combines statistical process control criteria to form a comprehensive judgment rule for abnormal stability monitoring indicators, thereby realizing the judgment of the stability of the operation process and the location of the cause of the fault.

[0006] In order to solve the above technical problems, the hot rolling mill operation process stability diagnosis method of the present invention comprises the following steps: Step 1: Collect the working condition data of the hot rolling mill equipment, including equipment operation data, equipment specification data, and hot rolling material specification data, and synchronously mark the time tag of the data, and continuously collect data for a period of time from the current time point as the original historical data for analysis and diagnosis; Step 2: Data conversion and cleaning: Time alignment: divide the data at the same time point into one data group according to the time tag; Data cleaning: automatically remove abnormal data groups, including data groups with values ​​exceeding the reasonable range, data groups with incorrect timestamp format, data groups that are obviously inconsistent with the actual working conditions, and data groups with missing data items; Step 3: Mark the rolling time label, mark the start and end time label of the complete rolling process of each piece of hot-rolled material, including the start and end time of rolling of each section of the rolling mill; Step 4: extract the original data set of each rolling process, and extract the original data set corresponding to each rolling process according to the rolling time label; Step 5: Taking each rolling section data set as a unit, establish feature engineering, extract characteristic values ​​reflecting the process stability or volatility of each rolling mill's pressing system, bending roll system, looper system and roll shifting system, as well as characteristic values ​​reflecting the stability or volatility of the hot-rolled material quality, and obtain characteristic result data rows as indicators reflecting the stability or volatility of the rolling process and the quality of the hot-rolled materials, and use the rolling start time as the index of the characteristic result data row; Step 6: According to the timestamp alignment, the characteristic values ​​are matched with the equipment specifications of the corresponding rolling process and the hot-rolled material specification data, including the support roll numbers of each section of the rolling mill, the thickness and width of the hot-rolled material, to form a complete characteristic specification data set; Step 7: Continuously collect feature specification data for a period of time from the current time point to form a feature specification data set as feature historical data for analysis and diagnosis; Step 8: retain the characteristic specification data set with the thickness of hot-rolled materials below 3.5 mm, mark the characteristic value grouping labels with the same thickness range and width-to-thickness ratio, mark the characteristic value grouping labels with the same support roll number and continuous time, and divide the characteristic historical data set into multiple working condition grouping characteristic specification data subsets according to the thickness grouping and support roll number grouping labels; Step 9: Taking each working condition grouping characteristic specification data subset as a unit, determine and mark the y characteristic abnormal data points through the comprehensive abnormality judgment rule based on the SPC criterion; Step 10: Calculate the y index, that is, the y characteristic abnormal frequency of each characteristic specification data subset, and obtain a set of y indicator data. The calculation formula is: (1) Among them, y_ano_ratio represents the abnormal frequency of y feature, y_ano_amount represents the number of y feature abnormalities, and y_sample_amount represents the number of y feature samples; Step 11: Perform an over-limit judgment on the abnormal frequency of the y characteristic, and judge the data points exceeding the given threshold as unstable, indicating that the equipment has an unstable state within the time range of the corresponding characteristic specification data subset; Step 12: Calculate the median or mean of each x feature of each feature specification data subset to reflect the system stability or volatility within the time range, including the x features contained in the press-down system, looper system, bending roll system and roll shifting system of each rolling mill; Step 13: Locate the cause of instability. Within the time range when the device is in an unstable state, calculate the x index to judge the x feature mean or median value. Data points exceeding the given threshold are judged as abnormal, indicating that the device has become unstable due to the x feature abnormality within the corresponding time range. Step 14: Sort the causes of instability, and sort the multiple abnormal x features from high to low according to the proportion of their threshold exceeding the limit; Step 15: Based on the determination of the unstable state of the equipment in step 11, the location of the instability cause in step 13, and the sorting of the instability causes in step 14, the time range of the instability of the hot rolling mill equipment and the corresponding abnormal characteristics are given to obtain a complete instability diagnosis conclusion.

[0007] Furthermore, in the step 1, the operating data of the hot rolling mill equipment is collected and obtained from the existing automatic control system of the hot rolling mill equipment or the factory data center by data communication or data copying.

[0008] Furthermore, the y feature and y index respectively refer to defining the alignment feature or its derivative feature at the outlet of the final rolling mill of a single rolling process of a single hot rolling mill equipment as the y feature, and defining the abnormal frequency of the y feature within a period of time as the y index; the x feature and x index respectively refer to defining the statistic that has a significant correlation with the y feature or the y index among the statistics calculated based on the operation / control parameters of the hot rolling mill equipment as the x feature, and defining the mean or median value of the x feature within a period of time as the x index. Since the hot rolling mill operation process stability diagnosis method of the present invention adopts the above-mentioned technical scheme, that is, the method clarifies the definition of the instability of the hot rolling mill operation process, based on the historical operating parameter data of the equipment, the stability monitoring index is extracted from the parameters that have a strong correlation with the quality of the hot-rolled products, and the equipment is judged whether it is unstable through the comprehensive judgment rule based on the SPC criterion, and the operation process volatility / stability monitoring index of each equipment part is extracted from the equipment operation control parameters, so as to realize the judgment of the instability and potential fault cause parts that occurred in the recent period of time in the history of the hot rolling mill equipment, and give the time range and corresponding abnormal characteristics of the instability of the hot rolling mill equipment, so as to obtain a complete instability diagnosis conclusion. The method is helpful to guide equipment maintenance and fault location, reduce accidents, and improve production quality and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flowchart of the method for diagnosing the stability of the hot rolling mill operation process of the present invention. DETAILED DESCRIPTION

[0010] The following will be combined with the attached Figure 1 This method is described in detail.

[0011] 1) Determine the IP address, port, communication protocol and other information, and establish data communication with the existing control system / data center of a certain strip hot rolling production line equipment; 2) Collect the original data of working parameters of finishing mills at all levels from the control system or data center through OPC communication, including operation and control signal parameters: steel bite signal value, roll gap value, oil column value, rolling force, roll shifting force value, roll bending force value, looper contact signal value, looper amount, looper tension, centering, wedge shape, zero adjustment signal value; production specification parameters: steel width, steel thickness, support roll number, strip steel number, rolling start time, etc., involving the pressing system, looper system, roll bending system and roll shifting system of each level of rolling mills, and collect data for 3 months in total; 3) While collecting data, mark the time label of the data sequence and complete the quality check of the original data simultaneously, including judging whether the time index format of the data is correct, calculating the missing rate, number of samples, maximum value, minimum value, and judging whether the standard deviation is zero; 4) Data conversion and cleaning: convert the original data according to the actual situation and parameter units, unify the dimension and magnitude; remove abnormal values ​​that exceed the range; splice, interpolate and fill the original data series of each signal parameter according to the time label, resample, and finally process it into an original data set with a sampling frequency of 1 time / 100ms; customize the standard parameter name table and change the original parameter name to the standard parameter name; 5) Mark the rolling start and end time tags, identify and mark the time points when each stage of the rolling mill starts biting and throwing steel, and extract the original data subset of each rolling process; 6) Feature extraction: for each data subset, further divide it into the head, middle and tail time periods according to the rules, and calculate the statistics (mean, median, maximum value, standard deviation) of the signal parameters in each time period, including the statistics of the alignment, wedge shape, rolling force deviation, bending roll force deviation, shifting roll force deviation, looper amount, stiffness and stiffness deviation, looper adjustment duration and looper stable time ratio in each time period. A data subset generates a row of characteristic value data, and matches the corresponding steel width, thickness, support roll number, and strip steel number, finally forming a complete feature specification data set; 7) Working condition segmentation: combining business experience and physical logic, retain the characteristic specification data set of steel thickness of 3.5mm and below, and group the thickness into [1.6mm, 2.0mm), [2.0mm, 2.4mm), [2.4mm, 3.0mm), [3.0mm, 3.5mm]. Further group the thickness groups according to the mean value of the width-to-thickness ratio, and further group them according to the combination of support roller numbers. The continuous time period under the same group number is divided into a time window, which is defined as a specification sliding window. Finally, mark the grouping labels of the characteristic data according to the thickness specification and time window rules; 8) Determine the y feature, and use the linear combination of the median_middle_median and the median_tail_mean as the final y feature, that is, (2) in, represents the median value of the alignment characteristic in the middle time window of rolling, It represents the mean value of the characteristic value of the degree of centering in the rolling tail time window; 9) Determination of abnormality of feature y: Based on SPC abnormality criteria and physical logic, the feature specification data subset of each specification sliding window is used as the unit to determine and mark whether feature y is abnormal. The mean and standard deviation of the historical feature values ​​for three months are defined as the overall mean (overall center line) and overall standard deviation. The comprehensive abnormality rules of the adopted SPC abnormality criteria are as follows: Rule 1: Group by thickness rule, and perform calculations and judgments of Rules 2 to 9 one by one for the grouped feature data of the same specification sliding window.

[0012] Rule 2: A point is more than 1 standard deviation above the overall mean.

[0013] Rule 3: 9 consecutive points are on the same side of the overall mean line.

[0014] Rule 4: 6 consecutive numbers are all increasing or all decreasing.

[0015] Rule 5: 14 consecutive points alternate above and below the overall mean line.

[0016] Rule 6: 2 out of 3 points are more than 2 overall standard deviations away from the overall mean line (on the same side).

[0017] Rule 7: 4 out of 5 points are greater than 1 standard deviation away from the overall mean line (on the same side).

[0018] Rule 8: 8 consecutive points are greater than 1 standard deviation away from the overall mean line (on either side).

[0019] Rule 9: Based on rules 2 to 8, further, one point exceeds the overall mean line by more than 2 standard deviations.

[0020] Rule 10: Based on rules 1 to 9, further summarize the y feature abnormality labels of each thickness rule grouping in the same specification sliding window, and finally obtain the y feature abnormal points in each specification sliding window; 10) Calculate the y index, that is, the y characteristic abnormal frequency, and obtain a set of y index data; 11) Instability judgment: the abnormal frequency of the y feature is judged to be out of limit, and the data points exceeding the given threshold are judged as unstable. The time range corresponding to the feature data subset is extracted, indicating that the equipment is unstable within the time range of the corresponding specification sliding window; 12) Calculate the x index, and take the statistics reflecting the stability / volatility of the operation / control parameters of each mechanical system of the rolling mill as x features, including stiffness_DS, stiffness_WS, stiffness_absolute deviation, roll gap deviation_tail_standard deviation, roll gap deviation_tail_mean, roll gap deviation_middle_median, roll gap deviation_head_mean, roll gap deviation_head_standard value, looper amount_mean, looper amount stable time proportion, rolling force deviation_middle_standard deviation, rolling force deviation_head_standard deviation, rolling force deviation_head_standard deviation; take the feature specification data subset of each specification sliding window as the unit, calculate the mean of these x features, define it as the final potential cause monitoring indicator, and obtain a set of x index data; 13) Locate the cause of instability. Based on the unstable time range obtained in step 10), the x indicator is judged to be out of limit. Data points exceeding the given threshold are judged as abnormal, indicating that the device has become unstable within the corresponding specification sliding window time range due to the corresponding x feature abnormality. 14) Cause sorting: if there are multiple x feature over-limit anomalies in the same sliding window time range, they are sorted from high to low according to the proportion of exceeding the given threshold; 15) Draw a conclusion. Combining the results of steps 11) and 14), draw a conclusion on whether the operation process of the hot rolling mill equipment is stable or unstable. If instability occurs, continue to give the unstable time range and the corresponding instability cause, which is a complete stability diagnosis conclusion.

[0021] This method clarifies the definition of hot rolling mill operation process stability, based on the equipment's historical operating parameter data, extracts stability monitoring indicators from the parameters that have a strong correlation with the quality of hot-rolled products, and uses the comprehensive judgment rules based on the SPC criterion to determine whether the monitored indicators are unstable, and extracts the operation process volatility / stability monitoring indicators of each equipment system from the equipment operation control parameters, and combines business experience, physical logic and statistical theory to make a stability diagnosis of the hot rolling mill operation process, which conforms to the actual operating conditions of the equipment and has good results. The implementation of this method does not rely on equipment failure cause data, and is particularly suitable for situations where there is a lack of effective structured data on equipment failure causes.

Claims

1. A method for diagnosing the stability of a hot rolling mill operation process, characterized in that The steps include: Step 1: Collect the working condition data of the hot rolling mill equipment, including equipment operation data, equipment specification data, and hot rolling material specification data, and synchronously mark the time tag of the data, and continuously collect data for a period of time from the current time point as the original historical data for analysis and diagnosis; Step 2: Data conversion and cleaning: Time alignment: divide the data at the same time point into one data group according to the time tag; Data cleaning: automatically remove abnormal data groups, including data groups with values ​​exceeding the reasonable range, data groups with incorrect timestamp format, data groups that are obviously inconsistent with the actual working conditions, and data groups with missing data items; Step 3: Mark the rolling time label, mark the start and end time label of the complete rolling process of each piece of hot-rolled material, including the start and end time of rolling of each section of the rolling mill; Step 4: extract the original data set of each rolling process, and extract the original data set corresponding to each rolling process according to the rolling time label; Step 5: Taking each rolling section data set as a unit, establish feature engineering, extract characteristic values ​​reflecting the process stability or volatility of each rolling mill's pressing system, bending roll system, looper system and roll shifting system, as well as characteristic values ​​reflecting the stability or volatility of the hot-rolled material quality, and obtain characteristic result data rows as indicators reflecting the stability or volatility of the rolling process and the quality of the hot-rolled materials, and use the rolling start time as the index of the characteristic result data row; Step 6: According to the timestamp alignment, the characteristic values ​​are matched with the equipment specifications of the corresponding rolling process and the hot-rolled material specification data, including the support roll numbers of each section of the rolling mill, the thickness and width of the hot-rolled material, to form a complete characteristic specification data set; Step 7: Continuously collect feature specification data for a period of time from the current time point to form a feature specification data set as feature historical data for analysis and diagnosis; Step 8: retain the characteristic specification data set with the thickness of hot-rolled materials below 3.5 mm, mark the characteristic value grouping labels with the same thickness range and width-to-thickness ratio, mark the characteristic value grouping labels with the same support roll number and continuous time, and divide the characteristic historical data set into multiple working condition grouping characteristic specification data subsets according to the thickness grouping and support roll number grouping labels; Step 9: Taking each working condition grouping characteristic specification data subset as a unit, determine and mark the y characteristic abnormal data points through the comprehensive abnormality judgment rule based on the SPC criterion; Step 10: Calculate the y index, that is, the y characteristic abnormal frequency of each characteristic specification data subset, and obtain a set of y indicator data. The calculation formula is: （1）， Among them, y_ano_ratio represents the abnormal frequency of y feature, y_ano_amount represents the number of y feature abnormalities, and y_sample_amount represents the number of y feature samples; Step 11: Perform an over-limit judgment on the abnormal frequency of the y characteristic, and judge the data points exceeding the given threshold as unstable, indicating that the equipment has an unstable state within the time range of the corresponding characteristic specification data subset; Step 12: Calculate the median or mean of each x feature of each feature specification data subset to reflect the system stability or volatility within the time range, including the x features contained in the press-down system, looper system, bending roll system and roll shifting system of each rolling mill; Step 13: Locate the cause of instability. Within the time range when the device is in an unstable state, calculate the x index to judge the x feature mean or median value. Data points exceeding the given threshold are judged as abnormal, indicating that the device has become unstable due to the x feature abnormality within the corresponding time range. Step 14: Sort the causes of instability, and sort the multiple abnormal x features from high to low according to the proportion of their threshold exceeding the limit; Step 15: Based on the determination of the unstable state of the equipment in step 11, the location of the instability cause in step 13, and the sorting of the instability causes in step 14, the time range of the instability of the hot rolling mill equipment and the corresponding abnormal characteristics are given to obtain a complete instability diagnosis conclusion.

2. The hot rolling mill operation process stability diagnosis method according to claim 1, characterized in that: In the step 1, the operating data of the hot rolling mill equipment is collected by data communication or data copying from the existing automatic control system of the hot rolling mill equipment or the factory data center.

3. The hot rolling mill operation process stability diagnosis method according to claim 1 or 2, characterized in that: The y feature and y index respectively refer to defining the alignment feature or its derivative feature at the outlet of the final rolling mill of a single rolling process of a single hot rolling mill equipment as the y feature, and defining the abnormal frequency of the y feature within a period of time as the y index; the x feature and x index respectively refer to defining the statistic that has a significant correlation with the y feature or the y index among the statistics calculated based on the operation / control parameters of the hot rolling mill equipment as the x feature, and defining the mean or median value of the x feature within a period of time as the x index.

Images