KR20180072050A - A method and device for analyzing operational abnormality of rolling process - Google Patents

Abstract

A method of analyzing operational abnormality of a rolling process according to the present invention includes: a data collection step of collecting time series data indicating an operational pattern generated in the rolling process, control set values, and major process variables; a segmentation step of segmenting the time series data corresponding to one material into a plurality of segments; a statistical information generation step of generating statistical summary information on the segmented time series data; an abnormality segment identification step of distinguishing the segments into an abnormal segment group and a normal segment group based on the statistical summary information; a segment sampling step of sampling a plurality of abnormal segments and normal segments from the abnormal segment group and the normal segment group, respectively; a key factor derivation step of deriving at least one key factor affecting the abnormal segment, based on the statistical summary information in the sampled segment; and an operational pattern modeling step of generating an operational pattern model by using the key factor as a key parameter. Accordingly, a cause of operational abnormality of equipment in the rolling process is promptly and efficiently analyzed.

Description

Technical Field [0001] The present invention relates to a method and an apparatus for analyzing an operational abnormality of a rolling process,

The present invention relates to a method and an apparatus for analyzing an operational disorder of a rolling process, and more particularly, to a method and an apparatus for analyzing an operational disorder of a rolling process by segmenting a plurality of time series data generated in a rolling process, And more particularly, to a method and apparatus for analyzing an operating disorder of a rolling process capable of effectively analyzing a rolling operation and providing a driver with a guide on how to operate the facility.

Typical steelmaking processes consist of a steelmaking process to produce molten iron, a steelmaking process to remove impurities from molten iron, a casting process to transform solid iron into solid, and a rolling process to make iron into steel or wire. Among them, the rolling process refers to a process of stretching or thinning slab, bloom, billet, and the like produced by the performance process by applying a continuous force while passing through a plurality of rotating rolls. The products produced in this rolling process vary in size, grade and size, and accordingly, the operating methods of the equipment for producing the products in the rolling process are also varied.

The quality of the product produced in the rolling process depends largely on how to control the processing equipment. In order to produce a product of uniform quality, the real-time management Is important. While a significant portion of the current plant maneuvering process is automated, the key processes for determining quality, such as material heating time, rolling load distribution, and temperature, are still dependent on the driver's experience and these variables, The driver's skill level, and the driver's condition. As a result, this is a variation factor in the production process and adversely affects the uniformity of product quality.

In particular, the driving difficulties such as meandering, breaking, overlapping, and foreign objects caused by the wrong facility operation of the driver occur only in some part of the whole length of the product, but when products occur, The effect on the productivity of the product is considerably serious. In order to solve this problem, it is necessary to construct a system capable of guiding the operator to analyze the operation pattern of the operator in real time so as not to perform the erroneous operation. However, as described above, There are limitations in guiding the driver 's operation method with the uniformized driver' s guide system due to different driving methods. In addition, changes in the equipment such as the roll replacement occur from time to time, and accordingly, there is a problem that the driver's guide system needs to be changed.

In spite of these problems, it is essential to formulate the driver's operation pattern and reduce the operating deviation even by manual operation in order to produce a uniform quality product in the rolling process. However, since the time series data of the driver's operation pattern is generated at intervals of 10 to 50 ms and the operating parameters used in one rolling process are tens of thousands of thousands, a single (24 hours) The amount of data is so large that it reaches several gigabytes (GB). Therefore, there is a limitation in visually identifying and analyzing time series data of a single material on a time basis using a dedicated tool.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide a rolling process capable of efficiently analyzing the cause of abnormal operation of a facility by using a plurality of time series data generated in the rolling process, And to provide a method and apparatus for analyzing the cause of abnormal operation of the facility.

Another object of the present invention is to provide an apparatus and method for analyzing an abnormal operation cause of equipment capable of providing a real time operation guide to a facility driver by modeling an operation pattern based on analyzed data and monitoring the time series data in real time according to modeling I would like to.

Other objects of the present invention will become more apparent from the following detailed description and drawings.

A method for analyzing an operational fault in a rolling process according to an embodiment of the present invention includes collecting time series data, control setting values, and main process variables indicating operation patterns generated in a rolling process; A segmentation step of segmenting the time series data pertaining to the one material into a plurality of segments; A statistical information generating step of generating statistical summary information on the segmented time series data; A disturbance section identification step of distinguishing the plurality of sections as the obstacle section group and the normal section group based on the statistical summary information; A section sampling step of sampling a plurality of obstacle sections and a normal section from the obstacle section group and the normal section group; Deriving at least one key factor affecting the impairment interval based on the statistical summary information of the sampled interval; And an operation pattern modeling step of generating an operation pattern model using the key factor as a main variable.

The method for analyzing the operational failure of the rolling process may further include a step of providing a driver's guide to the driver of the rolling process by monitoring the time series data generated from the rolling process in real time, In the providing step, the real time statistical summary information is generated after sectioning the monitored time series data, and the real time statistical summary information is applied to the operation pattern model to determine whether the current operation pattern is a failure operation pattern in real time .

The overlapping material number processing step may further include a duplicate material number processing step of processing a section corresponding to a duplicate material number in the sampled section, wherein the overlapping material number processing step includes: And if the normal section data includes the normal section data, the normal section data can be deleted.

The duplicate material number processing step may integrate and summarize the statistical summary information of each normal section when all the sections belonging to the overlapped material number are in the normal section.

A computer-readable storage medium, according to another embodiment of the present invention, stores program instructions that, when executed by a computing device, cause the computing device to perform the method of analyzing the operational failure of the rolling process described above.

An apparatus for analyzing an operational disorder in a rolling process according to another embodiment of the present invention includes a data collection module for collecting time series data, control setting values, and main process variables indicating operation patterns generated in a rolling process; A segmentation and statistical information generation module for segmenting time series data belonging to the one material into a plurality of segments and generating statistical summary information for the segmented time series data; A disturbance interval determination module that distinguishes the plurality of intervals as the obstacle interval group and the normal interval group based on the statistical summary information; A section sampling module for sampling a plurality of obstacle sections and a normal section from the obstacle section group and the normal section group; Deriving at least one key factor affecting the impairment interval based on the statistical summary information of the sampled interval; And an operation pattern modeling module for generating an operation pattern model with the key factor as a main variable.

The apparatus for analyzing the operational trouble of the rolling process may further include a driver guide providing module for providing a driver's guide to the driver of the rolling process by monitoring the time series data generated from the rolling process in real time, The providing module may divide the monitored time series data and generate real-time statistical summary information and apply the real-time statistical summary information to the operation pattern model to determine in real time whether the current operation pattern is a faulty operation pattern .

The section sampling module may further include a duplicate material number processing module for processing a section corresponding to a duplicate material number in the sampled section, and the duplicate material number processing module may further include: And if the normal section data includes the normal section data, the normal section data can be deleted.

The duplicate material number processing module may integrate the statistical summary information of each normal section when all the sections belonging to the overlapped material number are in the normal section.

According to the present invention, it is possible to quickly and efficiently analyze the cause of the operational trouble of the facility in the rolling process by comparing a large amount of time series data using the segmentation and statistical information. Also, by providing a real-time driving guide to the driver based on the analysis, the production efficiency of the product can be improved by minimizing the occurrence of the driving disorder.

FIG. 1 is a flowchart illustrating a method for analyzing an operational failure of a rolling process according to an embodiment of the present invention. Referring to FIG.
FIG. 2 is a flowchart illustrating a method of generating a data set for classification analysis in a method for analyzing an operational disorder in a rolling process according to an exemplary embodiment of the present invention. Referring to FIG.
3 is a conceptual diagram illustrating a method of generating a data set for classification analysis in a method for analyzing an operational disorder in a rolling process according to an embodiment of the present invention.
FIG. 4 is a block diagram illustrating a structure of an apparatus for analyzing an operational trouble in a rolling process according to an embodiment of the present invention. Referring to FIG.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The embodiments of the present invention can be modified in various forms, and the scope of the present invention should not be construed as being limited to the embodiments described below. The embodiments are provided to explain the present invention to a person having ordinary skill in the art to which the present invention belongs. Accordingly, the shape of each element shown in the drawings may be exaggerated to emphasize a clearer description.

FIG. 1 is a flowchart illustrating a method for analyzing an operational failure of a rolling process according to an embodiment of the present invention. Referring to FIG. Referring to FIG. 1, the method 100 for analyzing an operational failure of a rolling process of the present invention starts with collecting time series data having control pattern information, control set values, and key process variables from a rolling facility (S110) . The time series data is micro data representing the operation pattern for each material. It is the data of each roll in the rolling process, the rotational speed of each roll, the interval between symmetrical pair of rolls, torque, And data about any suitable variables that can be manipulated by the driver in the rolling process. The time series data is linked with a unique material number of the material from which the time series data is generated, and can be generated in units of 10 to 50 ms as described above. The control set values are macro data representing the physical characteristics or specifications of the product to be produced, including, but not limited to, thickness, width, length, composition, and rolling temperature of the material, And may include any suitable setting values that can be made. The main process variables are important variables that determine the properties of the material, and may include, for example, furnace temperatures.

Next, the time series data is segmented to generate statistical summary information (S120). The products produced in the rolling process are different in length, thickness, and width, and the division into time series data is performed by dividing the entire length of the material by n. That is, even if the lengths of the materials are different, since the same number of intervals are progressed, the length of one section may be different depending on the material. Therefore, n intervals are formed for each material, and the average (μ) and standard deviation (σ) of the time series data are obtained for each interval, which is called statistical summary information.

When segmentation of time series data and statistical summary information of each segment are generated, it is discriminated whether or not each segment is a failure segment (S130). The obstacle section means a section of the material produced in the rolling process in which the physical characteristics of the section are out of a predetermined value. There are three methods of determining whether the section is a failure section.

First, in the case of meandering or breaking, if the data generated in the measurement center such as the width meter or the load cell is out of the predetermined range, it is judged that the meander or breakage occurs, and this interval can be judged as the disturbance interval. Secondly, when the time of occurrence of the target driver's impairment is known, the obstacle section can be manually determined by searching and inputting the section corresponding to the time. Third, if there is no specific definition of the drift impairment interval, dependent variables according to independent variables such as job setting values are analyzed by using PCA, spectral clustering, It is possible to determine whether the corresponding interval is a failure interval according to whether there is a difference.

If it is determined in step S130 that the corresponding interval is a failure interval, the corresponding interval is set as a failure interval (S140). If it is determined in step S130 that the corresponding interval is not a failure interval, the corresponding interval is set as a normal interval (S150). It is determined whether the corresponding section is the last section of the entire material (S160). If the section is not the last section (NO in S160), steps S130 through S150 are repeated.

If the corresponding section is the last section (YES in S160), that is, when the failure section is set for all the materials, a data set for classification analysis is generated through the section sampling (S170). A method of generating a data set for classification analysis will be described in detail below with reference to FIGS. 2 and 3. FIG.

FIG. 2 is a flowchart showing a method of generating a data set (hereinafter referred to as a final data set) for classification analysis in the method for analyzing the operational trouble of the rolling process according to an embodiment of the present invention. Referring to FIG. 2, a method of generating a final data set starts with sampling a normal interval and a disabled interval to perform a period sampling to generate a normal interval group and a disabled interval interval (S171). Since the number of disability intervals is generally much smaller than the number of normal intervals, the disability interval group samples the entire disability interval, and the normal group is the whole Some of the normal sections are randomly sampled (for example, random non-restored sampling).

In general, the number of materials produced per day in one rolling mill is about 700, which means that about 20,000 materials are produced per month. The number of defects is only ten or fewer, and the number of obstacle sections is much smaller than the number of normal sections. Therefore, as described above, in the interval sampling, the obstacle interval group samples the entire obstacle interval, and the normal group randomly samples some of the entire normal interval.

3 is a conceptual diagram illustrating a method of generating a data set for classification analysis in a method for analyzing an operational disorder in a rolling process according to an embodiment of the present invention. As can be seen from FIG. 3 (a), when n pieces of sections are m sectioned, m × n sections are generated, and some of them (section k of the i th material) Section. Therefore, the entire obstacle section is sampled into the obstacle section group (A) among the m × n sections, and some normal sections are non-restored randomly sampled into the normal section group (B). As can be seen from FIG. 3 (b), each sampled group is represented by a set, and each interval in the set is divided into a material number (a, b) to which the corresponding interval belongs, μ) and the standard deviation (σ) of the time series data of the corresponding interval. Here, a represents the material number including the normal section, and b represents the material number including the failure section.

When the normal interval group A and the obstacle interval group B are sampled, the sampled intervals are compared one by one to determine whether or not the material numbers of the two intervals are overlapped (S172). If the material numbers of the two sections overlap (YES in S172), it is determined whether one of the two sections is a failure section (S173). If it is determined that one of the two sections is a failure section (YES in S173), the data corresponding to the normal section is deleted (S174). That is, the section corresponding to the material number having the obstacle section should be excluded from the normal section group. This is because the driver who produces the material containing the obstacle section is likely to operate similarly to the obstacle section other than the obstacle section, and therefore, even if it is in the normal section, it is likely to include the abnormal parameter value. The reliability of the data can be improved by excluding the section corresponding to the material including the failure section from among the sampled normal sections.

If one of the two sections is not a failure section, that is, if both sections are in the normal section (NO in S173), the two sections are integrated into one section by shipment of the mean and standard deviation corresponding to the two sections (S175). This is because, when a plurality of sections belonging to the normal section material number corresponding to the same material number are included in a plurality of sections, a weighting may be given to the operation pattern used for producing the material.

As described above, when there are a plurality of sections belonging to the same material number based on the sampled material number, except for the section corresponding to the material number to which the obstacle section belongs from the normal section group, two or more normal sections are sampled In this case, the reliability of the data for classification analysis is increased by shipment of the mean and the standard deviation for the interval.

As described above, when a normal interval group A and a disabled interval group B are constructed, a union of them is generated and a final data set for classification analysis is generated by merging the data based on the basic information of the material and the material number S176). The basic information of the material may include any suitable information that can classify the material according to the characteristics to be produced, including, but not limited to, heating temperature, width, length, thickness, The produced material has its own unique material number, and a plurality of material numbers can be classified into one material. Accordingly, the final data set for the classification analysis is completed by integrating the sections generated in the above manner on the basis of the material based on the material basic information and the material number.

When the data set for the classification analysis is completed as described above, the main variables are derived through classification analysis (S180). For example, a classification method such as Decision Tree, Boosting, and Random Forest is applied to the intervals included in the final data set. By listing the variables in order of importance and visualizing them, we can derive the key factors that affect the operation pattern. Detailed models used in the classification analysis are already known, so a detailed description thereof will be omitted.

Next, when the key parameters are derived as the main variable, the result according to the operation pattern is modeled through the derived key factors (S190). The operation pattern model is created by performing linear / non-linear modeling or neural network-based modeling using the derived core factors as the main variables.

After the operation pattern model is created by modeling, the driver's guide is provided to the driver by monitoring data generated in the subsequent rolling process in real time (S200). Specifically, in order to provide the driver's guide, the time series data generated in real time in the rolling process is segmented in the same manner as that performed in the segmentation and statistical summary information generation step (S120), and the statistical summary information is generated. Then, it is determined whether the current operation pattern is the normal operation pattern or the operation pattern causing the fault (fault occurrence operation pattern) by using the statistical summary information for the relevant section as the input of the operation pattern model, It is possible to provide an appropriate guide to the driver when it corresponds to the generated operation pattern.

As described above, according to the present invention, a large amount of time series data generated in the rolling process is sectioned, the fault section is identified by using the statistical summary information for each section, and the operation pattern It is possible to quickly and efficiently analyze the cause of the driver 's obstacle by analyzing and modeling the driver' s obstacle. Also, it is possible to minimize the occurrence of the driver 's trouble by providing the driver with the driver' s guide in real time based on the analysis.

Next, referring to FIG. 4, an apparatus for analyzing an operational disorder in a rolling process according to an embodiment of the present invention will be described. FIG. 4 is a block diagram illustrating a structure of an apparatus for analyzing an operational trouble in a rolling process according to an embodiment of the present invention. Referring to FIG.

4, the apparatus for analyzing the operational trouble in the rolling process of the present invention includes a data collection module 401, a segmentation and statistical information generation module 402, a failure interval determination module 403, A key factor derivation module 405, an operation pattern modeling module 406, and an operator guide providing module 407. The key factor derivation module 405, Those skilled in the art will appreciate that a plurality of modules may be integrated into one functional module or one module may be integrated into a plurality of modules As shown in FIG.

The data collection module 401 is a module for collecting time series data, control set values, and main process variables having operation pattern information from the rolling facilities. The time series data is micro data representing the operation pattern for each material. It is the data of each roll in the rolling process, the rotational speed of each roll, the interval between symmetrical pair of rolls, torque, And the control set value is macro data showing the physical characteristics or specifications of the product to be produced, and includes the thickness, width, length, constituent components and rolling temperature of the material, For example, the heating furnace temperature.

The segmentation and statistical information generation module 402 divides the time series data and generates statistical summary information. The segmentation and statistical information generation module 402 performs segmentation of the material by dividing the entire length of the material by n, (Μ) and standard deviation (σ) of the time series data with respect to the interval are generated as statistical summary information.

The failure section determination module 403 is a module that determines whether a section generated by the sectioning and statistical information generation module 402 belongs to a failure section or a normal section, The detailed judgment is as described above, and a description thereof will be omitted.

Next, the classification analysis data set generation module 404 is a module for generating a data set for classification analysis through interval sampling. The classification analysis data set generation module 404 includes a section sampling module 408, a redundant material processing module 409, Module 410, as shown in FIG.

The section sampling module 408 generates a normal interval group A and a disabled interval group B by sampling a normal interval and a disabled interval. Specifically, the interval sampling is a group of a discrete interval A Sampling the entire impairment interval and non-restoring random sampling of some normal intervals with the normal interval population (B).

The duplicate material processing module 409 processes a duplicate material number if there is a duplicate material number based on the material number of the sampled section. If one of the sections corresponding to the duplicate material number corresponds to the failure section, The normal section corresponding to the material number is deleted, and when all the sections corresponding to the duplicate material number are normal sections, the overlapping sections are integrated into one section by shipment of the mean and standard deviation corresponding to each section.

The final data generation module 410 is a module for generating a final data set for classification analysis by integrating the sections generated in the above manner on the basis of the material basis information and the material number.

The core factor derivation module 405 is a module for deriving core factors that affect the operation pattern by performing a classification analysis on the section data generated by the final data generation module. The module 405 may use a classification analysis method such as a Decision Tree, Boosting, and Random Forest to derive a key factor.

The operation pattern modeling module 406 is a module for generating a model of an operation pattern using the derived core factors as a main variable and can be a linear / non- .

The driver guide providing module 407 is a module for providing a driver's guide to the driver by monitoring the data generated in the rolling process after the driving pattern model is created by modeling in real time. Specifically, the driver guide providing module 407 divides the time series data generated in real time, generates statistical summary information, and outputs statistical summary information on the corresponding section as an input of the operation pattern model generated in the operation pattern modeling module It is determined whether the current operation pattern is a normal operation pattern or an operation pattern causing a failure (failure occurrence operation pattern), and provides an appropriate guide to the driver when the operation pattern corresponds to the failure occurrence operation pattern.

As described above, the apparatus for analyzing the operational trouble in the rolling process of the present invention divides a large amount of time series data generated in the rolling process, identifies the normal section and the fault section using the statistical summary information for each section, By providing the driver's guide in real time by searching and learning the contributing driving patterns and the key factors of the fault occurrence through the data analysis method, it is possible to minimize the occurrence of the driving disorder.

401: Data Acquisition Module
402: Interval and statistical information generation module
403: Failure section determination module
404: Classification Analysis Data Set Generation Module
405: Core factor derivation module
406: Operation pattern modeling module
407: Driver's Guide Module

Claims (9)

A data collecting step of collecting time series data, control set values and main process variables indicating the operation pattern generated in the rolling process;
A segmentation step of segmenting the time series data pertaining to the one material into a plurality of segments;
A statistical information generating step of generating statistical summary information on the segmented time series data;
A disturbance section identification step of distinguishing the plurality of sections as the obstacle section group and the normal section group based on the statistical summary information;
A section sampling step of sampling a plurality of obstacle sections and a normal section from the obstacle section group and the normal section group;
Deriving at least one key factor affecting the impairment interval based on the statistical summary information of the sampled interval; And
And an operation pattern modeling step of generating an operation pattern model with the key factor as a main variable. The method according to claim 1,
Further comprising the step of providing a driver's guide to the driver of the rolling process by monitoring the time series data generated from the rolling process in real time,
The driver guide providing step may include generating real-time statistical summary information after segmenting the monitored time series data and applying the real-time statistical summary information to the operation pattern model to determine whether the current operation pattern is a faulty driving pattern in real time The method comprising the steps of: The method according to claim 1,
In the section sampling step,
Further comprising a duplicate material number processing step of processing a section corresponding to a duplicate material number in the sampled section,
Wherein the duplicate material number processing step deletes the normal section data when the section belonging to the overlapped material number includes the failure section and the normal section. 3. The method of claim 2,
Wherein the duplicate material number processing step integrates the statistical summary information of each normal section when all the sections belonging to the overlapped material number are in the normal section, and then shipment the property summary information of the rolling section in the rolling process. 15. A computer-readable storage medium having stored thereon program instructions for causing a computing device to perform a method for analyzing a driver obstacle according to any one of claims 1 to 4, when executed by a computing device. A data collection module for collecting time series data, control set values, and main process variables indicating the operation pattern generated in the rolling process;
A segmentation and statistical information generation module for segmenting time series data belonging to the one material into a plurality of segments and generating statistical summary information for the segmented time series data;
A disturbance interval determination module that distinguishes the plurality of intervals as the obstacle interval group and the normal interval group based on the statistical summary information;
A section sampling module for sampling a plurality of obstacle sections and a normal section from the obstacle section group and the normal section group;
Deriving at least one key factor affecting the impairment interval based on the statistical summary information of the sampled interval; And
And an operation pattern modeling module for generating an operation pattern model with the key factor as a main variable. The method according to claim 6,
And a driver guide providing module for providing a driver's guide to the driver of the rolling process by monitoring the time series data generated from the rolling process in real time,
The driver guide providing module generates real-time statistical summary information after segmenting the monitored time series data and applies the real-time statistical summary information to the operation pattern model to determine whether the current operation pattern is a faulty driving pattern in real time Wherein the apparatus is operable to analyze the operational failure of the rolling process. The method according to claim 6,
The section sampling module includes:
Further comprising a duplicate material number processing module for processing a section corresponding to a duplicate material number in the sampled section,
Wherein the duplicate material number processing module deletes the normal section data when the section belonging to the overlapped material number includes the failure section and the normal section. 9. The method of claim 8,
Wherein the duplicate material number processing module integrates the statistical summary information of each normal section when all the sections belonging to the overlapped material number are in the normal section, and outputs the combined summary information.

Images