CN114896900A - A target tracking system - Google Patents

A target tracking system Download PDF

Info

Publication number
CN114896900A
CN114896900A CN202210831903.9A CN202210831903A CN114896900A CN 114896900 A CN114896900 A CN 114896900A CN 202210831903 A CN202210831903 A CN 202210831903A CN 114896900 A CN114896900 A CN 114896900A
Authority
CN
China
Prior art keywords
furnace
strip
target
variables
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210831903.9A
Other languages
Chinese (zh)
Other versions
CN114896900B (en
Inventor
曹卫华
张永月
宋文硕
贺江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Geosciences
Original Assignee
China University of Geosciences
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Geosciences filed Critical China University of Geosciences
Priority to CN202210831903.9A priority Critical patent/CN114896900B/en
Publication of CN114896900A publication Critical patent/CN114896900A/en
Application granted granted Critical
Publication of CN114896900B publication Critical patent/CN114896900B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • C—CHEMISTRY; METALLURGY
    • C21—METALLURGY OF IRON
    • C21D—MODIFYING THE PHYSICAL STRUCTURE OF FERROUS METALS; GENERAL DEVICES FOR HEAT TREATMENT OF FERROUS OR NON-FERROUS METALS OR ALLOYS; MAKING METAL MALLEABLE, e.g. BY DECARBURISATION OR TEMPERING
    • C21D1/00—General methods or devices for heat treatment, e.g. annealing, hardening, quenching or tempering
    • C21D1/26—Methods of annealing
    • C—CHEMISTRY; METALLURGY
    • C21—METALLURGY OF IRON
    • C21D—MODIFYING THE PHYSICAL STRUCTURE OF FERROUS METALS; GENERAL DEVICES FOR HEAT TREATMENT OF FERROUS OR NON-FERROUS METALS OR ALLOYS; MAKING METAL MALLEABLE, e.g. BY DECARBURISATION OR TEMPERING
    • C21D11/00—Process control or regulation for heat treatments
    • C—CHEMISTRY; METALLURGY
    • C21—METALLURGY OF IRON
    • C21D—MODIFYING THE PHYSICAL STRUCTURE OF FERROUS METALS; GENERAL DEVICES FOR HEAT TREATMENT OF FERROUS OR NON-FERROUS METALS OR ALLOYS; MAKING METAL MALLEABLE, e.g. BY DECARBURISATION OR TEMPERING
    • C21D9/00—Heat treatment, e.g. annealing, hardening, quenching or tempering, adapted for particular articles; Furnaces therefor
    • C21D9/52—Heat treatment, e.g. annealing, hardening, quenching or tempering, adapted for particular articles; Furnaces therefor for wires; for strips ; for rods of unlimited length
    • C21D9/54—Furnaces for treating strips or wire
    • C21D9/56—Continuous furnaces for strip or wire
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Metallurgy (AREA)
  • Thermal Sciences (AREA)
  • Organic Chemistry (AREA)
  • Materials Engineering (AREA)
  • Evolutionary Computation (AREA)
  • Mechanical Engineering (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Control Of Heat Treatment Processes (AREA)

Abstract

本发明提供了一种目标跟踪系统，包括：关键变量获取模块、多元特征获取模块、目标的运行状态确定模块和目标跟踪预测模块，通过将该关键变量分为底层设备层、运行过程层和计划指标层；对多个相关变量进行分析，提取其状态特征，对存在的异常特征进行筛选处理，制定状态跟踪策略，融合多元特征建立目标状态检测模型，实时跟踪在炉体内动态变化的目标。基于跟踪信息，建立目标到各炉段出口等位置的预测模型，预测进出各炉段的时间。通过融合多时空层级参数的序列特征，共同决策并跟踪连续退火过程中目标带钢的实时位置，建立目标带钢到各关键位置的预测模型，为后续过程建模及系统控制奠定基础。

Figure 202210831903

The invention provides a target tracking system, which includes: a key variable acquisition module, a multi-feature acquisition module, a target operation state determination module and a target tracking prediction module. The key variables are divided into a bottom equipment layer, an operation process layer and a plan Index layer: analyzes multiple related variables, extracts their state features, filters existing abnormal features, formulates state tracking strategies, integrates multiple features to establish target state detection models, and tracks dynamically changing targets in the furnace in real time. Based on the tracking information, a prediction model is established from the target to the exit of each furnace section, and the time of entering and leaving each furnace section is predicted. By fusing the sequence features of multi-space-time hierarchical parameters, jointly decide and track the real-time position of the target strip during the continuous annealing process, and establish a prediction model for the target strip to each key position, which lays the foundation for subsequent process modeling and system control.

Figure 202210831903

Description

一种目标跟踪系统a target tracking system

技术领域technical field

本发明涉及退火加热技术领域，尤其涉及一种目标跟踪系统。The invention relates to the technical field of annealing heating, in particular to a target tracking system.

背景技术Background technique

连续退火加热过程是冷轧热镀锌生产线上的一道重要工序，需要将进入退火炉内的多种带钢按照对应的加热要求进行加热。不同炉段中不同带钢的加热需求都不一致，因此需要对炉内的不同加热目标——带钢进行实时跟踪，以便及时进行相应调控。The continuous annealing heating process is an important process in the cold-rolled hot-dip galvanizing production line, and various strip steels entering the annealing furnace need to be heated according to the corresponding heating requirements. The heating requirements of different strips in different furnace sections are inconsistent, so it is necessary to track different heating targets in the furnace - strips in real time, so as to make corresponding adjustments in time.

一般冷轧热镀锌生产线的落成是引进了国外较为成熟的生产线及其系统，在该条件下，并不能及时准确跟踪炉内的不同目标带钢。原因在于：1.虽然每卷带钢在进入生产线前有初始入炉计划，该计划中有每卷带钢的初始长度，但是在进入生产线时，根据不同需求，月牙剪会对带钢进行两侧及前后进行修剪，如会对过渡带钢进行长度上的裁剪，该裁剪的具体长度是不定且未知的，因此在生产线中炉内加热的带钢长度是不确定的，并不等同于初始入炉计划中的带钢长度；2.经过月牙剪修剪过的带钢在进入退火炉前会焊接在一起，也影响了带钢的具体长度；3.生产线中不同种类和规格的带钢组合方式呈现不规律性；4.生产线运行时，整个过程是昼夜连续不间断的高速运行，因此只凭人工观测会耗费大量人力精力；5.由于带钢类型和长度的差异性和运行的高速性，一般生产线设备和现有的控制系统中并没有能够实时检测并跟踪目标带钢的传感器，仅在生产线若干位置离散部署了焊缝检测装置从而大致估算目标带钢位置，并不能满足当对特定目标带钢的较高跟踪要求。Generally, the completion of the cold-rolled hot-dip galvanizing production line is the introduction of relatively mature foreign production lines and systems. Under this condition, it is impossible to accurately track different target strips in the furnace in a timely manner. The reasons are: 1. Although each coil of strip steel has an initial furnace entry plan before entering the production line, and the plan includes the initial length of each coil of strip steel, when entering the production line, according to different needs, the crescent shears will carry out two steps on the strip steel. Trim the side and front and back. If the length of the transition strip is cut, the specific length of the cut is uncertain and unknown. Therefore, the length of the strip heated in the furnace in the production line is uncertain and not equal to the initial length. The length of the strip in the furnace plan; 2. The strip trimmed by the crescent shears will be welded together before entering the annealing furnace, which also affects the specific length of the strip; 3. The combination of different types and specifications of strips in the production line 4. When the production line is running, the whole process is continuous and uninterrupted high-speed operation day and night, so only manual observation will consume a lot of manpower and energy; 5. Due to the difference in strip type and length and the high speed of operation , the general production line equipment and the existing control system do not have sensors that can detect and track the target strip in real time. Only the welding seam detection devices are discretely deployed in several positions of the production line to roughly estimate the target strip position. Higher tracking requirements for the target strip.

发明内容SUMMARY OF THE INVENTION

为了解决上述问题，本发明提供了一种目标跟踪系统，目的是融合多时空层级参数的序列特征，共同决策并跟踪连续退火过程中目标带钢的实时位置，从而建立目标带钢到各关键位置的预测模型，为后续过程建模及系统控制奠定基础。该目标跟踪系统主要包括：In order to solve the above problems, the present invention provides a target tracking system, the purpose is to integrate the sequence features of multi-space-time hierarchical parameters, jointly decide and track the real-time position of the target strip in the continuous annealing process, so as to establish the target strip to each key position Predictive model, which lays the foundation for subsequent process modeling and system control. The target tracking system mainly includes:

关键变量获取模块，用于根据退火炉中连续退火加热过程，提取和目标带钢位置相关的参数作为关键变量，并按照设备及控制类型将这些关键变量分为底层设备层I1、运行过程层I2和计划指标层I3；其中，所述关键变量包括炉计划中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序、入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温；The key variable acquisition module is used to extract parameters related to the target strip position as key variables according to the continuous annealing heating process in the annealing furnace, and divide these key variables into the bottom equipment layer I 1 and the operation process layer according to equipment and control types. I 2 and planning index layer I 3 ; wherein, the key variables include strip type specifications in the furnace plan, strip length in the furnace entry plan, strip entry sequence in the furnace entry plan, and welding seam at the furnace entry Distance, weld seam at PH-NOF junction, weld seam distance at NOF furnace section exit, weld seam distance at RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature;

多元特征获取模块，用于对底层设备层I1和运行过程层I2中多个相关变量进行运行状态分析，提取各变量的数据时序分布特征，结合退火炉生产工艺确定存在的多种异常特征，基于优先级策略和时序分布特征，对异常特征进行筛选处理，得到多元特征；其中，所述多个相关变量包括入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温；The multi-feature acquisition module is used to analyze the operating state of multiple related variables in the underlying equipment layer I 1 and the operation process layer I 2 , extract the data time series distribution characteristics of each variable, and determine the existence of various abnormal characteristics in combination with the annealing furnace production process , based on the priority strategy and time series distribution characteristics, the abnormal characteristics are screened to obtain multivariate characteristics; wherein, the multiple relevant variables include the weld distance at the entrance to the furnace, the weld at the PH-NOF junction, and the NOF furnace section exit. Weld distance, weld distance at RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature;

目标的运行状态确定模块，用于根据现有系统中计划指标层I3变量和多元特征获取模块得到的多元特征，制定渐进明细的状态跟踪策略；基于制定的渐进明细状态跟踪策略条件及约束，融合多元特征建立目标状态检测模型，从而确定目标的运行状态；The operating state determination module of the target is used to formulate a progressively detailed state tracking strategy based on the multivariate features obtained by the planning index layer I 3 variables and the multivariate feature acquisition module in the existing system; based on the formulated progressive detailed state tracking strategy conditions and constraints, Integrate multiple features to establish a target state detection model, so as to determine the running state of the target;

目标跟踪预测模块，用于在时间尺度上将计划指标层I3变量与目标的运行状态确定模块中对某卷钢的运行状态结果进行时间对齐，其中，计划指标层I3变量包括入炉计划中的带钢类型规格、入炉计划中的带钢长度和入炉计划中的带钢入炉顺序，在空间尺度上结合本身炉体的设备信息，从而实时跟踪在炉体内动态变化的目标；基于该目标的跟踪信息，建立目标到各炉段出口位置的预测模型，以预测目标进出各炉段的时间。The target tracking and forecasting module is used to align the running state results of a coil of steel in the time scale with the plan index layer I3 variables and the target operation state determination module, wherein the plan index layer I3 variables include the furnace entry plan The strip type specification in the furnace, the strip length in the furnace entry plan and the strip entry sequence in the furnace entry plan, combined with the equipment information of the furnace body on the spatial scale, so as to track the dynamically changing target in the furnace body in real time; Based on the tracking information of the target, a prediction model from the target to the exit position of each furnace section is established to predict the time when the target enters and exits each furnace section.

进一步地，多元特征获取模块中得到多元特征的具体过程如下：Further, the specific process of obtaining multiple features in the multiple feature acquisition module is as follows:

2.1：选择底层设备层I1相关变量作为集合

Figure 797531DEST_PATH_IMAGE001
和运行过程层I2相关变量作为集合
Figure 213468DEST_PATH_IMAGE002
， 其中，i=入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离或RTF炉段出口处焊 缝距离，j=运行速度、NOF炉段出口板温或RTF炉段出口板温； 2.1: Select the underlying device layer I 1 related variables as a set
Figure 797531DEST_PATH_IMAGE001
and the operating process layer I 2 related variables as a set
Figure 213468DEST_PATH_IMAGE002
, where i=the distance of the weld at the entrance to the furnace, the weld at the PH-NOF junction, the distance of the weld at the exit of the NOF furnace section or the distance of the weld at the exit of the RTF furnace section, j = the running speed, the temperature of the plate at the outlet of the NOF furnace section or RTF furnace section outlet plate temperature;

2.2：对

Figure 808398DEST_PATH_IMAGE003
和
Figure 833729DEST_PATH_IMAGE004
中的变量进行最小采样频率公倍数筛选，最小采样频率公倍数为： 2.2: Yes
Figure 808398DEST_PATH_IMAGE003
and
Figure 833729DEST_PATH_IMAGE004
The variables in are screened by the common multiple of the minimum sampling frequency, and the common multiple of the minimum sampling frequency is:

Figure 830504DEST_PATH_IMAGE005
Figure 830504DEST_PATH_IMAGE005

式中，k=1,2,[

Figure 682922DEST_PATH_IMAGE006
]表示求解
Figure 1034DEST_PATH_IMAGE007
和
Figure 597100DEST_PATH_IMAGE002
中所有变量采样频率的公倍数；得到
Figure 448381DEST_PATH_IMAGE008
后，对
Figure 439078DEST_PATH_IMAGE001
和
Figure 274179DEST_PATH_IMAGE002
中的各变量进行筛选： In the formula, k=1,2,[
Figure 682922DEST_PATH_IMAGE006
] means solving
Figure 1034DEST_PATH_IMAGE007
and
Figure 597100DEST_PATH_IMAGE002
the common multiple of the sampling frequency of all variables in ; get
Figure 448381DEST_PATH_IMAGE008
after, yes
Figure 439078DEST_PATH_IMAGE001
and
Figure 274179DEST_PATH_IMAGE002
to filter each variable in:

Figure 142777DEST_PATH_IMAGE009
Figure 142777DEST_PATH_IMAGE009

Figure 350030DEST_PATH_IMAGE010
和
Figure 544251DEST_PATH_IMAGE011
为筛选后的集合；
Figure 350030DEST_PATH_IMAGE010
and
Figure 544251DEST_PATH_IMAGE011
is the filtered set;

2.3：对筛选后的

Figure 335490DEST_PATH_IMAGE010
和
Figure 506315DEST_PATH_IMAGE011
中各变量的时序分布特征进行探索性分析，得到多元数 据； 2.3: For the filtered
Figure 335490DEST_PATH_IMAGE010
and
Figure 506315DEST_PATH_IMAGE011
Exploratory analysis of the time series distribution characteristics of each variable in the

2.4：对以上经过同采样频率筛选后的多元数据进行预处理，预处理优先级为：运行速度>底层设备层变量>两炉段出口板温；预处理是指：具有不同类型分布特征的变量有不同的异常特征或需要不同的数据处理方法，当运行速度为0或为负时，则生产线停机或现场出现异常，对该时间区间所有变量不作处理。2.4: Preprocess the above multivariate data after screening with the same sampling frequency. The preprocessing priority is: running speed > underlying equipment layer variables > outlet plate temperature of the two furnace sections; preprocessing refers to: variables with different types of distribution characteristics There are different abnormal characteristics or different data processing methods. When the running speed is 0 or negative, the production line stops or there is an abnormality on site, and all variables in this time interval are not processed.

进一步地，目标的运行状态确定模块中，所述渐进明细的状态跟踪策略是指先以 入炉计划中的带钢长度为基准，与运行速度进行相除计算，得到初始线性时间区间，将指计 划指标层I3的变量在时序上进行拉长，与初始线性时间区间对齐，再通过传感器检测到的 某相邻带钢的首尾位置计算出带钢长度，相比于入炉计划中的带钢长度，计算出的带钢长 度更加准确，这是因为入炉计划中的带钢长度为初始值，进入生产线时会根据生产要求和 工艺要求对带钢进行小幅或大幅削减，削减长度未知，因此入炉计划中的带钢长度和实际 入炉的带钢长度并不一致，因此可通过此步骤计算出更准确一步的带钢长度，但是此时还 是存在

Figure 597767DEST_PATH_IMAGE012
区间内的误差，所以最后需要再通过两炉段出口板温检测模型进一步精 细化跟踪目标带钢的位置，以便得到更为精确的带钢长度，其中，
Figure 431731DEST_PATH_IMAGE013
表示
Figure 477310DEST_PATH_IMAGE014
对应的采样 时间间隔，v表示生产线运行速度。 Further, in the target operation state determination module, the progressively detailed state tracking strategy means that the strip length in the furnace entry plan is used as the benchmark, and the operation speed is divided and calculated to obtain an initial linear time interval, which will refer to the plan. The variables of the index layer I 3 are elongated in time sequence to align with the initial linear time interval, and then the length of the strip is calculated by the head and tail positions of an adjacent strip detected by the sensor, which is compared with the strip in the furnace plan. Length, the calculated strip length is more accurate, because the strip length in the furnace entry plan is the initial value, and the strip will be slightly or greatly reduced according to production requirements and process requirements when entering the production line, and the cutting length is unknown, so The strip length in the furnace plan is not consistent with the actual strip length, so a more accurate strip length can be calculated through this step, but at this time there are still
Figure 597767DEST_PATH_IMAGE012
Therefore, it is necessary to further refine and track the position of the target strip through the outlet plate temperature detection model of the two furnace sections in order to obtain a more accurate strip length, among which,
Figure 431731DEST_PATH_IMAGE013
express
Figure 477310DEST_PATH_IMAGE014
The corresponding sampling time interval, v represents the running speed of the production line.

进一步地，目标的运行状态确定模块中，底层设备层变量的状态检测模型的建立过程为：Further, in the target operation state determination module, the establishment process of the state detection model of the underlying device layer variables is as follows:

（1）设每个底层设备层的变量

Figure 953290DEST_PATH_IMAGE015
的数据表示为
Figure 633671DEST_PATH_IMAGE016
，建立三个连续滑动窗口：均值 计算窗口W m 、瞬时变化检测窗口W d 和方差计算窗口W v ，窗口长度分别为m，n和v； (1) Set the variables of each underlying device layer
Figure 953290DEST_PATH_IMAGE015
The data is represented as
Figure 633671DEST_PATH_IMAGE016
, establish three continuous sliding windows: mean calculation window W m , instantaneous change detection window W d and variance calculation window W v , the window lengths are m , n and v respectively;

（2）计算

Figure 238587DEST_PATH_IMAGE015
的W m 均值和W d 均值，表示为M m 和M d ，并计算W v 的均值M v 和方差V计算公式 为： (2) Calculation
Figure 238587DEST_PATH_IMAGE015
The W m mean and W d mean of , denoted as M m and M d , and the mean M v and variance V of W v are calculated as:

Figure 4418DEST_PATH_IMAGE017
Figure 4418DEST_PATH_IMAGE017

式中，k0为第一个采样点，定义开始和结束事件累计和

Figure 284090DEST_PATH_IMAGE018
和
Figure 320441DEST_PATH_IMAGE019
，表达式为 In the formula, k 0 is the first sampling point, which defines the cumulative sum of the start and end events
Figure 284090DEST_PATH_IMAGE018
and
Figure 320441DEST_PATH_IMAGE019
, the expression is

Figure 27366DEST_PATH_IMAGE020
Figure 27366DEST_PATH_IMAGE020

式中，δ为权重参数

Figure 280493DEST_PATH_IMAGE021
为方差阈值，δ越大，在当前
Figure 596811DEST_PATH_IMAGE018
和
Figure 251784DEST_PATH_IMAGE019
统计值中所占的比 值越大，累计能力也越强，反之越小，通过判断
Figure 864030DEST_PATH_IMAGE018
和
Figure 840339DEST_PATH_IMAGE019
的变化情况，即可确定瞬变特征 点的位置。In the formula, δ is the weight parameter
Figure 280493DEST_PATH_IMAGE021
is the variance threshold, the larger the δ, the higher the current
Figure 596811DEST_PATH_IMAGE018
and
Figure 251784DEST_PATH_IMAGE019
The larger the ratio of the statistical value, the stronger the cumulative ability, and vice versa, the smaller
Figure 864030DEST_PATH_IMAGE018
and
Figure 840339DEST_PATH_IMAGE019
The position of the transient feature point can be determined.

进一步地，目标的运行状态确定模块的具体实现过程如下：Further, the specific implementation process of the target operating state determination module is as follows:

3.1：对计划指标

Figure 461813DEST_PATH_IMAGE022
进行初始时间区间线性填充，其中k=1,2,3，分别代表入炉计划 中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序；初始时间区间线 性填充是指以入炉计划中的带钢长度为基准，与运行速度进行相除计算，得到初始线性时 间区间Z0； 3.1: To plan indicators
Figure 461813DEST_PATH_IMAGE022
Perform linear filling of the initial time interval, where k=1, 2, 3, respectively represent the strip type and specification in the furnace entry plan, the strip length in the furnace entry plan, and the strip entry sequence in the furnace entry plan; the initial time Interval linear filling means that the strip length in the furnace entry plan is used as the benchmark, and the operation speed is divided and calculated to obtain the initial linear time interval Z 0 ;

3.2：利用步骤3.1和多元特征获取模块中处理后的变量数据制定渐进明细的状态跟踪策略；3.2: Use step 3.1 and the variable data processed in the multi-feature acquisition module to formulate a progressive and detailed state tracking strategy;

3.3：利用建立的底层设备层的多元状态检测模型，对底层设备层I1的变量状态进行检测，即通过确定瞬变特征点，确定状态发生变化的位置点；3.3: Use the established multi-state detection model of the underlying device layer to detect the variable state of the underlying device layer I 1 , that is, determine the location point where the state changes by determining the transient feature point;

3.4：通过步骤3.3，可进一步将带钢的位置在空间域上从入炉计划中的带钢长度 缩小到计算得到的带钢长度Lr，在时间域上，从Z0缩小到

Figure 971292DEST_PATH_IMAGE023
之间，为了更准确 跟踪目标带钢的位置，通过两炉段出口板温状态检测模型确定具体位置，其建立过程为： 3.4: Through step 3.3, the position of the strip can be further reduced from the strip length in the furnace plan to the calculated strip length L r in the space domain, and from Z 0 to the calculated strip length L r in the time domain.
Figure 971292DEST_PATH_IMAGE023
In between, in order to more accurately track the position of the target strip, the specific position is determined by the detection model of the plate temperature state at the exit of the two furnace sections. The establishment process is as follows:

（1）选取多元特征获取模块中处理后的

Figure 518555DEST_PATH_IMAGE002
中两个变量的数据，任一变量的时序数据 集合为y1,y2,…,yk，计算出数据集的均值作为该数据集的参考值为
Figure 746274DEST_PATH_IMAGE024
，偏差计算为
Figure 640280DEST_PATH_IMAGE025
，i=1,2,…,k； (1) Select the processed data in the multivariate feature acquisition module
Figure 518555DEST_PATH_IMAGE002
The data of two variables in
Figure 746274DEST_PATH_IMAGE024
, the deviation is calculated as
Figure 640280DEST_PATH_IMAGE025
, i=1,2,…,k;

（2）利用以下公式计算出任一变量的时序数据集的标准差，根据稳定状态的数据在所有数据中的出现概率P及标准正态分布表，得到P对应的区间边界±m1σ；m1为正数，判定稳定状态数据分布在以所述参考值为中心的±m1σ范围内；(2) Use the following formula to calculate the standard deviation of the time series data set of any variable, and obtain the interval boundary ±m 1 σ corresponding to P according to the occurrence probability P of the data in the steady state in all the data and the standard normal distribution table; m 1 is a positive number, it is determined that the steady state data is distributed in the range of ±m 1 σ centered on the reference value;

Figure 771310DEST_PATH_IMAGE026
Figure 771310DEST_PATH_IMAGE026

上式中，vi为历史数据序列中第i个数据yi对应的偏差；In the above formula, vi is the deviation corresponding to the i -th data yi in the historical data sequence;

获取实时数据序列，并以连续的s个数据为一组判定实时状态；针对每一组，具体 判定方法为：分别计算该组s个实时数据y1,y2,…,ys对应的偏差，得到s个偏差；将所述s个 偏差求均值，得到均值v；若|v|>nσ，表明该组实时数据属于带钢切换时间点Tr 2，再结合步骤 3.3确定的

Figure 990938DEST_PATH_IMAGE027
或
Figure 440374DEST_PATH_IMAGE028
，取两者间的交集，则为带钢带头和带尾的具体时序位置， 在该时刻，对应的空间位置即为两炉段出口的位置。 Obtain the real-time data sequence, and use the consecutive s data as a group to determine the real-time state; for each group, the specific determination method is: calculate the deviation corresponding to the group of s real-time data y 1 , y 2 ,...,y s respectively , obtain s deviations; average the s deviations to obtain the mean v; if |v|>nσ, it indicates that this group of real-time data belongs to the strip switching time point Tr 2 , and then combined with the determined value in step 3.3
Figure 990938DEST_PATH_IMAGE027
or
Figure 440374DEST_PATH_IMAGE028
, and the intersection between the two is taken as the specific timing position of the strip head and the strip tail. At this moment, the corresponding spatial position is the position of the exits of the two furnace sections.

进一步地，制定状态跟踪策略的过程如下：Further, the process of formulating the state tracking strategy is as follows:

1）：基于底层设备层的多元状态检测模型，检测出某相邻带钢的首尾位置，利用入 炉处焊缝距离

Figure 636607DEST_PATH_IMAGE029
，计算出某卷带钢的首端位置
Figure 120678DEST_PATH_IMAGE030
和尾端位置
Figure 511208DEST_PATH_IMAGE031
，r表示某卷带钢，同时计算其 长度Lr，
Figure 949405DEST_PATH_IMAGE032
； 1): Based on the multi-state detection model of the underlying equipment layer, the position of the head and tail of an adjacent strip is detected, and the distance of the weld at the entrance to the furnace is used.
Figure 636607DEST_PATH_IMAGE029
, calculate the position of the head end of a strip of steel
Figure 120678DEST_PATH_IMAGE030
and tail position
Figure 511208DEST_PATH_IMAGE031
, r represents a strip of steel, and its length L r is calculated at the same time,
Figure 949405DEST_PATH_IMAGE032
;

2）：其次利用Z0进行筛选，若

Figure 185214DEST_PATH_IMAGE033
，则保留结果，转到4）运算，若
Figure 789371DEST_PATH_IMAGE034
，则说明 数据异常，转到3）；2): Secondly, use Z 0 to filter, if
Figure 185214DEST_PATH_IMAGE033
, then keep the result, go to 4) operation, if
Figure 789371DEST_PATH_IMAGE034
, then the data is abnormal, go to 3);

3）：说明用

Figure 849337DEST_PATH_IMAGE035
检测和计算出的
Figure 273365DEST_PATH_IMAGE030
、
Figure 578445DEST_PATH_IMAGE031
和
Figure 538573DEST_PATH_IMAGE036
异常，采用底层设备层的下一个变量，即
Figure 270905DEST_PATH_IMAGE037
进行检测及计算处新的
Figure 916650DEST_PATH_IMAGE030
、
Figure 523956DEST_PATH_IMAGE031
和
Figure 837125DEST_PATH_IMAGE036
，再次进行2），直至i+1=4，若
Figure 474780DEST_PATH_IMAGE038
依然计算出数据异常， 则说明进入生产线中的带钢有误，将错误信息返回系统界面，进行报警，此时需要人工重新 校准生产计划和生产线中的带钢信息； 3): for illustration
Figure 849337DEST_PATH_IMAGE035
detected and calculated
Figure 273365DEST_PATH_IMAGE030
,
Figure 578445DEST_PATH_IMAGE031
and
Figure 538573DEST_PATH_IMAGE036
exception, take the next variable of the underlying device layer, i.e.
Figure 270905DEST_PATH_IMAGE037
Carry out inspection and calculation of new
Figure 916650DEST_PATH_IMAGE030
,
Figure 523956DEST_PATH_IMAGE031
and
Figure 837125DEST_PATH_IMAGE036
, do 2) again, until i+1=4, if
Figure 474780DEST_PATH_IMAGE038
If the calculated data is still abnormal, it means that the strip entering the production line is wrong, and the error information is returned to the system interface to give an alarm. At this time, it is necessary to manually recalibrate the production plan and the strip information in the production line;

4）：计算出的Lr受最小采样频率公倍数

Figure 374865DEST_PATH_IMAGE008
约束，得到一个相比于Z0更准确的范 围，但还是存在
Figure 756168DEST_PATH_IMAGE039
区间内的误差，为了更进一步确定目标带钢的准确位置，基于两 炉段出口板温检测模型，分别检测出NOF段出口带钢的大幅度跳变和RTF段出口带钢的大幅 度跳变位置，进一步精细化跟踪目标带钢的位置。 4): The calculated L r is subject to a common multiple of the minimum sampling frequency
Figure 374865DEST_PATH_IMAGE008
Constraints to get a more accurate range than Z 0 , but still
Figure 756168DEST_PATH_IMAGE039
The error in the interval, in order to further determine the exact position of the target strip, based on the detection model of the plate temperature at the exit of the two furnace sections, the large jump of the exit strip of the NOF section and the large jump of the exit strip of the RTF section were detected respectively. position, and further refine the tracking of the position of the target strip.

进一步地，目标跟踪预测模块中，建立目标到各炉段出口位置的预测模型的具体过程如下：Further, in the target tracking and prediction module, the specific process of establishing the prediction model from the target to the exit position of each furnace section is as follows:

4.1：更新初始线性时间区间Z0和步骤3.4中某卷带钢带头带尾的切换时间区间保 持一致，设为

Figure 687959DEST_PATH_IMAGE040
，即为该卷带钢完全通过炉内某检测点的具体时间，同时与运行速度相乘， 计算得到该卷带钢的距离长度
Figure 496515DEST_PATH_IMAGE041
，同时由步骤3.4可知该卷带钢带头的具体位置
Figure 648010DEST_PATH_IMAGE042
； 4.1: Update the initial linear time interval Z 0 to be consistent with the switching time interval of the head and tail of a strip in step 3.4, set as
Figure 687959DEST_PATH_IMAGE040
, that is, the specific time when the coil of steel completely passes through a certain detection point in the furnace, and at the same time, it is multiplied by the running speed to calculate the distance length of the coiled steel.
Figure 496515DEST_PATH_IMAGE041
, and the specific position of the strip head can be known from step 3.4
Figure 648010DEST_PATH_IMAGE042
;

4.2：炉体信息为固定数值，各关键位置的空间位置都是固定已知值，设某关键位 置的空间位置为

Figure 68890DEST_PATH_IMAGE043
，则建立的目标到各炉段出口位置的预测模型为： 4.2: The furnace body information is a fixed value, and the spatial position of each key position is a fixed known value. Let the spatial position of a key position be
Figure 68890DEST_PATH_IMAGE043
, the established prediction model from the target to the outlet position of each furnace section is:

Figure 622231DEST_PATH_IMAGE044
Figure 622231DEST_PATH_IMAGE044

式中，n表示带钢带头具体位置

Figure 601688DEST_PATH_IMAGE042
到关键位置
Figure 473436DEST_PATH_IMAGE043
期间运行速度变化的次数，
Figure 196541DEST_PATH_IMAGE045
表示 第n次变化的运行速度，
Figure 604389DEST_PATH_IMAGE046
表示以
Figure 256212DEST_PATH_IMAGE045
传送的距离。 In the formula, n represents the specific position of the strip head
Figure 601688DEST_PATH_IMAGE042
to key locations
Figure 473436DEST_PATH_IMAGE043
The number of times the running speed changes during the period,
Figure 196541DEST_PATH_IMAGE045
represents the running speed of the nth change,
Figure 604389DEST_PATH_IMAGE046
means with
Figure 256212DEST_PATH_IMAGE045
distance of transmission.

本发明提供的技术方案带来的有益效果是：节约了人力成本，融合多时空层级参数的序列特征，共同决策并跟踪连续退火过程中目标带钢的实时位置，从而建立目标带钢到各关键位置的预测模型，逐步提高了目标带钢的跟踪精度，为后续过程建模及系统控制奠定基础。The beneficial effects brought by the technical solution provided by the present invention are: saving labor costs, integrating the sequence features of multi-space-time hierarchical parameters, jointly making decisions and tracking the real-time position of the target strip in the continuous annealing process, so as to establish the target strip to each key The prediction model of the position gradually improves the tracking accuracy of the target strip, and lays a foundation for the subsequent process modeling and system control.

附图说明Description of drawings

下面将结合附图及实施例对本发明作进一步说明，附图中：The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

图1是本发明实施例中一种目标跟踪系统的原理图。FIG. 1 is a schematic diagram of a target tracking system in an embodiment of the present invention.

图2是本发明实施例中底层设备层变量的时序分布特征图。FIG. 2 is a time series distribution characteristic diagram of the underlying device layer variables in the embodiment of the present invention.

图3是本发明实施例中运行速度的时序分布特征图。FIG. 3 is a time series distribution characteristic diagram of the running speed in the embodiment of the present invention.

图4是本发明实施例中两炉段出口板温时序分布特征图。FIG. 4 is a characteristic diagram of time series distribution of plate temperature at the outlet of two furnace sections in an embodiment of the present invention.

图5是本发明实施例中非接触式温度检测与接触式温度检测结果图。FIG. 5 is a graph showing the results of non-contact temperature detection and contact temperature detection in an embodiment of the present invention.

图6是本发明实施例中出口板温的时序分布特征图。FIG. 6 is a time series distribution characteristic diagram of the outlet plate temperature in the embodiment of the present invention.

图7是本发明实施例中预处理后的某段NOF板温结果图。FIG. 7 is a result diagram of a certain section of NOF plate temperature after pretreatment in the embodiment of the present invention.

图8是本发明实施例中为

Figure 851142DEST_PATH_IMAGE035
时序分布示意图。 Figure 8 is an embodiment of the present invention for
Figure 851142DEST_PATH_IMAGE035
Schematic diagram of timing distribution.

具体实施方式Detailed ways

请参考图1，图1是本发明实施例中一种目标跟踪系统的原理图，以退火炉连续退火加热过程为实例，采集多空间层和目标跟踪相关的关键变量，这些关键变量在时间尺度上具有不同采样频率及不同滞后性等特性。其中，多空间层包括：底层设备层I1、运行过程层I2和计划指标层I3。这些多空间层的分类是根据变量来源及采集频率差异来确定的，如底层设备层的参数是采集时间单位规律且采集频率为毫秒级或秒级的参数，运行过程层为控制系统部署的传感器采集的过程数据，一般采集时间单位规律且采集频率为秒级，计划指标层为入炉计划中的计划数据，该类数据的时间单位间隔并不规律，一般为分钟级或者小时级。目标跟踪指的是将进入退火生产线中会有连续不断的多卷带钢中的其中一卷或者多卷作为目标，进行实时定位跟踪。关键变量是指在退火生产线的初始生产设备和控制系统中，和目标带钢跟踪相关的变量，包括：入炉计划中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序、入炉处焊缝距离、预热-加热段（后简称“PH-NOF”）结合处焊缝、直燃炉段（后简称“NOF炉段”）出口处焊缝距离、辐射加热炉段（后简称“RTF炉段”）出口处焊缝距离、运行速度、NOF炉段出口板温、RTF炉段出口板温。该目标跟踪系统具体包括：Please refer to FIG. 1. FIG. 1 is a schematic diagram of a target tracking system according to an embodiment of the present invention. Taking the continuous annealing heating process of an annealing furnace as an example, key variables related to multi-space layers and target tracking are collected. These key variables are in the time scale. It has the characteristics of different sampling frequency and different hysteresis. The multi-space layer includes: a bottom equipment layer I 1 , an operation process layer I 2 and a plan index layer I 3 . The classification of these multi-space layers is determined according to the source of variables and the difference in acquisition frequency. For example, the parameters of the underlying device layer are parameters of the acquisition time unit law and the acquisition frequency is at the millisecond or second level, and the operation process layer is the sensor deployed by the control system. For the collected process data, the collection time unit is generally regular and the collection frequency is at the second level. The planning index layer is the planning data in the furnace entry plan. The time unit interval of this type of data is irregular, generally at the minute or hour level. Target tracking refers to taking one or more of the continuous multiple coils of strip steel entering the annealing production line as the target for real-time positioning and tracking. The key variables refer to the variables related to the target strip tracking in the initial production equipment and control system of the annealing production line, including: strip type specifications in the furnace entry plan, strip length in the furnace entry plan, and strip steel in the furnace entry plan. The sequence of entering the furnace, the distance of the welding seam at the entering furnace, the welding seam at the junction of the preheating-heating section (hereinafter referred to as "PH-NOF"), the welding seam at the exit of the direct combustion furnace section (hereinafter referred to as "NOF furnace section") Distance, weld distance at the exit of the radiation heating furnace section (hereinafter referred to as "RTF furnace section"), running speed, NOF furnace section exit plate temperature, RTF furnace section exit plate temperature. The target tracking system specifically includes:

关键变量获取模块，用于根据退火炉中连续退火加热过程，提取和目标带钢位置相关的参数作为关键变量，并按照设备及控制类型将这些关键变量分为底层设备层I1、运行过程层I2和计划指标层I3；其中，所述关键变量包括炉计划中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序、入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温；The key variable acquisition module is used to extract parameters related to the target strip position as key variables according to the continuous annealing heating process in the annealing furnace, and divide these key variables into the bottom equipment layer I 1 and the operation process layer according to equipment and control types. I 2 and planning index layer I 3 ; wherein, the key variables include strip type specifications in the furnace plan, strip length in the furnace entry plan, strip entry sequence in the furnace entry plan, and welding seam at the furnace entry Distance, weld seam at PH-NOF junction, weld seam distance at NOF furnace section exit, weld seam distance at RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature;

多元特征获取模块，用于对底层设备层I1和运行过程层I2中多个相关变量进行运行状态分析，提取其状态特征，结合退火炉生产工艺确定存在的多种异常特征，基于优先级分配和分布类型差异，对异常特征进行筛选处理，得到多元特征；其中，所述多个相关变量包括入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温。得到多元特征的具体过程如下：The multi-feature acquisition module is used to analyze the operating state of multiple related variables in the underlying equipment layer I 1 and the operation process layer I 2 , extract their state features, and determine the existence of a variety of abnormal features in combination with the annealing furnace production process, based on priority According to differences in distribution and distribution types, the abnormal features are screened to obtain multivariate features; wherein, the multiple relevant variables include the weld distance at the entrance to the furnace, the weld at the PH-NOF junction, the weld distance at the NOF furnace section exit, Weld distance at the RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature. The specific process of obtaining multivariate features is as follows:

2.1：选择底层设备层I1相关变量作为集合

Figure 643517DEST_PATH_IMAGE047
和运行过程层I2相关变量作为集合
Figure 162265DEST_PATH_IMAGE048
， 其中，i=入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离或RTF炉段出口处焊 缝距离，j=运行速度、NOF炉段出口板温或RTF炉段出口板温； 2.1: Select the underlying device layer I 1 related variables as a set
Figure 643517DEST_PATH_IMAGE047
and the operating process layer I 2 related variables as a set
Figure 162265DEST_PATH_IMAGE048
, where i=the distance of the weld at the entrance to the furnace, the weld at the PH-NOF junction, the distance of the weld at the exit of the NOF furnace section or the distance of the weld at the exit of the RTF furnace section, j = the running speed, the temperature of the plate at the outlet of the NOF furnace section or RTF furnace section outlet plate temperature;

2.2：对

Figure 483524DEST_PATH_IMAGE003
和
Figure 565750DEST_PATH_IMAGE004
中的变量进行最小采样频率公倍数筛选。底层设备层的变量数据
Figure 663281DEST_PATH_IMAGE003
和 运行过程层的变量数据
Figure 514563DEST_PATH_IMAGE004
都呈现采集规律性，但是采集频率互不相同，因此对数据进行同 采样频率数据处理，为保证数据的真实性，对所有变量进行最小采样频率公倍数筛选，最小 采样频率公倍数为： 2.2: Yes
Figure 483524DEST_PATH_IMAGE003
and
Figure 565750DEST_PATH_IMAGE004
The variables in are filtered by the common multiple of the minimum sampling frequency. Variable data of the underlying device layer
Figure 663281DEST_PATH_IMAGE003
and variable data at the operating process level
Figure 514563DEST_PATH_IMAGE004
They all show the regularity of collection, but the collection frequencies are different from each other. Therefore, the data is processed with the same sampling frequency. In order to ensure the authenticity of the data, all variables are screened by the common multiple of the minimum sampling frequency. The minimum common multiple of the sampling frequency is:

Figure 272303DEST_PATH_IMAGE005
Figure 272303DEST_PATH_IMAGE005

式中，k=1,2,[

Figure 340360DEST_PATH_IMAGE006
]表示求解
Figure 474538DEST_PATH_IMAGE007
和
Figure 180326DEST_PATH_IMAGE002
中所有变量采样频率的公倍数；得到
Figure 610432DEST_PATH_IMAGE008
后，对
Figure 667250DEST_PATH_IMAGE001
和
Figure 339540DEST_PATH_IMAGE002
中的各变量进行筛选： In the formula, k=1,2,[
Figure 340360DEST_PATH_IMAGE006
] means solving
Figure 474538DEST_PATH_IMAGE007
and
Figure 180326DEST_PATH_IMAGE002
the common multiple of the sampling frequency of all variables in ; get
Figure 610432DEST_PATH_IMAGE008
after, yes
Figure 667250DEST_PATH_IMAGE001
and
Figure 339540DEST_PATH_IMAGE002
to filter each variable in:

Figure 663948DEST_PATH_IMAGE009
Figure 663948DEST_PATH_IMAGE009

Figure 763492DEST_PATH_IMAGE010
和
Figure 307605DEST_PATH_IMAGE011
为筛选后的集合；如，若初始集合
Figure 19472DEST_PATH_IMAGE007
中的入炉处焊缝距离变量的采样频率 为500ms，在该采样频率的数据采集数据为1,2,3,4,5,6,7…,而求得
Figure 699852DEST_PATH_IMAGE008
为1s，则
Figure 235875DEST_PATH_IMAGE010
中的 入炉处焊缝距离变量的为1,3,5,7…。
Figure 763492DEST_PATH_IMAGE010
and
Figure 307605DEST_PATH_IMAGE011
is the filtered set; for example, if the initial set
Figure 19472DEST_PATH_IMAGE007
The sampling frequency of the welding seam distance variable in the furnace is 500ms, and the data acquisition data at this sampling frequency are 1, 2, 3, 4, 5, 6, 7..., and the obtained
Figure 699852DEST_PATH_IMAGE008
is 1s, then
Figure 235875DEST_PATH_IMAGE010
The welding seam distance variable in the furnace is 1, 3, 5, 7….

2.3：对筛选后的

Figure 500241DEST_PATH_IMAGE010
和
Figure 514333DEST_PATH_IMAGE011
中各变量的时序分布特征进行探索性分析，可知各变量 的时序分布特征不尽相同。
Figure 49220DEST_PATH_IMAGE010
集合中的各变量时序分布特征如图2所示，呈现出线性递增 特征及大幅度跳变特征；
Figure 257610DEST_PATH_IMAGE011
集合中的速度时序分布特征如图3所示，在时间区间内保持线性 不变性，若发生改变则为发生跳变随后继续保持线性不变性；
Figure 510737DEST_PATH_IMAGE011
集合中的两炉段出口变量 的时序分布特征如图4所示，在时间尺度上呈现缓慢的波动变化特征，即在时间尺度上并不 会保持单一变化趋势和单一值，且相邻时间采样点间的数据并不会发生突变，而是相对平 滑变化。 2.3: For the filtered
Figure 500241DEST_PATH_IMAGE010
and
Figure 514333DEST_PATH_IMAGE011
The exploratory analysis of the time series distribution characteristics of each variable in the data shows that the time series distribution characteristics of each variable are different.
Figure 49220DEST_PATH_IMAGE010
The time series distribution characteristics of each variable in the set are shown in Figure 2, showing linear increasing characteristics and large jump characteristics;
Figure 257610DEST_PATH_IMAGE011
The time series distribution characteristics of velocity in the set are shown in Figure 3, which maintains linear invariance within the time interval, and if it changes, it will jump and then continue to maintain linear invariance;
Figure 510737DEST_PATH_IMAGE011
The time series distribution characteristics of the outlet variables of the two furnace sections in the set are shown in Figure 4, showing slow fluctuation characteristics on the time scale, that is, it does not maintain a single change trend and a single value on the time scale, and adjacent time sampling The data between points does not change suddenly, but changes relatively smoothly.

2.4：对底层设备层和运行过程层的多个变量（入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离，运行速度、NOF炉段出口板温、RTF炉段出口板温）（即对经过同采样频率筛选后的多元数据）进行预处理，预处理优先级：运行速度>底层设备层变量>两炉段出口板温。具有不同类型分布特征的变量有不同的异常特征或需要不同的数据预处理方法：运行速度为0或为负时，则说明生产线停机或现场出现异常，则对该时间区间所有变量不作处理；2.4: For multiple variables of the bottom equipment layer and the operation process layer (weld distance at the furnace entry, weld distance at the PH-NOF junction, weld distance at the exit of the NOF furnace section, weld distance at the RTF furnace section exit, running speed , NOF furnace section outlet plate temperature, RTF furnace section outlet plate temperature) (that is, the multivariate data after filtering with the same sampling frequency) is preprocessed, and the preprocessing priority is: running speed > underlying equipment layer variables > two furnace section outlet plates temperature. Variables with different types of distribution characteristics have different abnormal characteristics or require different data preprocessing methods: when the running speed is 0 or negative, it means that the production line is shut down or there is an abnormality on site, and all variables in this time interval will not be processed;

底层设备层变量和两炉段的异常特征判断方向不同，下面逐一解释：The variables of the underlying equipment layer and the abnormal characteristics of the two furnace sections have different judgment directions, which are explained one by one as follows:

（1）底层设备层变量会出现两种变化特征，需要判断是否为异常特征：第一种为当 某个变量因为检测问题出现负值则为异常特征，则需要丢弃该变量处于负值的时间区数 据；第二种为正常情况下，该类所有变量的变化趋势为线性递增变化，而在某时刻突然某变 量未完全呈现线性递增变化，这种情况就需要进行计算判断。检测到该数据点变化发生改 变的时间点作为起始时间点t0，在t0前，计算出某变量线性递增斜率k0，从t0开始，设置步长 为

Figure 62941DEST_PATH_IMAGE049
，其中n=1,2,3,...,
Figure 482027DEST_PATH_IMAGE050
表示
Figure 828695DEST_PATH_IMAGE008
对应的采样时间间隔，将底层设备层的4个变 量从t0开始的数据组成一个四维特征向量
Figure 834697DEST_PATH_IMAGE051
，其中
Figure 692057DEST_PATH_IMAGE052
表表示炉处焊缝距离，
Figure 467115DEST_PATH_IMAGE053
表示PH-NOF结合处焊缝距离，
Figure 250263DEST_PATH_IMAGE054
表示NOF炉段出口处焊缝距离，
Figure 976517DEST_PATH_IMAGE055
表示RTF炉段出口处焊缝 距离，设置m个步长检测点，及最大检测时间区间为
Figure 136103DEST_PATH_IMAGE056
，在最大检测时间区间内对 每个向量取两两相邻计算每个向量m个步长检测点的斜率变化
Figure 765668DEST_PATH_IMAGE057
，其中l 表示在
Figure 486761DEST_PATH_IMAGE056
时间区间内两两相邻的距离，l=1,2,…,m-1，由于底层设备层各变量检 测点距离固定不变且处于同一运行速度，所以若
Figure 936197DEST_PATH_IMAGE058
，则说明该类变 化特征属于正常变化，不是异常特征；若
Figure 633895DEST_PATH_IMAGE059
中某个斜率变化值与其它 三个不一致，则需修正该单个变量的斜率变化值，与其它三个保持一致；若其中两个不一 致，另外两个一致，则需修正两个不一致的变量斜率变化值与另外两个一致；若两两保持一 致，则取均值赋予所有变量的斜率变化；若各不相同，则舍弃该段时间区的所有变量数值。 (1) There will be two changing characteristics of the underlying device layer variables, and it is necessary to judge whether it is an abnormal characteristic: the first is that when a variable has a negative value due to a detection problem, it is an abnormal characteristic, and the time when the variable is in a negative value needs to be discarded. The second is that under normal circumstances, the change trend of all variables in this category is a linear incremental change, but suddenly a variable does not completely show a linear incremental change at a certain moment. In this case, calculation and judgment are required. The time point at which the change of the data point is detected is taken as the starting time point t 0 . Before t 0 , the linearly increasing slope k 0 of a variable is calculated. Starting from t 0 , the step size is set as
Figure 62941DEST_PATH_IMAGE049
, where n=1,2,3,...,
Figure 482027DEST_PATH_IMAGE050
express
Figure 828695DEST_PATH_IMAGE008
Corresponding sampling time interval, the data of the four variables of the underlying device layer starting from t 0 is formed into a four-dimensional feature vector
Figure 834697DEST_PATH_IMAGE051
,in
Figure 692057DEST_PATH_IMAGE052
The table represents the weld distance at the furnace,
Figure 467115DEST_PATH_IMAGE053
Indicates the weld distance at the PH-NOF junction,
Figure 250263DEST_PATH_IMAGE054
Indicates the weld distance at the outlet of the NOF furnace section,
Figure 976517DEST_PATH_IMAGE055
Indicates the weld distance at the exit of the RTF furnace section, set m step detection points, and the maximum detection time interval is
Figure 136103DEST_PATH_IMAGE056
, in the maximum detection time interval, take each vector adjacent to each other to calculate the slope change of each vector m step detection points
Figure 765668DEST_PATH_IMAGE057
, where l represents in
Figure 486761DEST_PATH_IMAGE056
The distance between two adjacent pairs in the time interval, l=1,2,…,m-1, since the distance between the variable detection points of the underlying equipment layer is fixed and at the same running speed, so if
Figure 936197DEST_PATH_IMAGE058
, it means that this kind of change characteristics belong to normal changes, not abnormal characteristics; if
Figure 633895DEST_PATH_IMAGE059
If a certain slope change value is inconsistent with the other three, the slope change value of the single variable needs to be corrected to be consistent with the other three; if two of them are inconsistent and the other two are consistent, the slopes of the two inconsistent variables need to be corrected The change value is consistent with the other two; if the two are consistent, the mean value is taken to give the slope change of all variables; if they are different, all variable values in this time zone are discarded.

（2）对于NOF炉段出口板温和RTF炉段出口板温，由于工艺需求，只能采用非接触式的红外测温仪进行检测。而非接触式的红外测温仪，受环境因素（对象的温度、空气介质）的影响大，所以如图5所示，（a）为NOF炉段出口板温（非接触式）示意图，（b）为NOF炉段出口板温（接触式）示意图，相比于接触式的测温传感器，会含有较多噪声和离群点。如果不对板温数据进行处理，则后续用该数据进行建模时，无法保证数据模型的准确性（如会产生过拟合问题等），因此需要根据出口板温数据的特点对其进行预处理。(2) For NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature, due to process requirements, only non-contact infrared thermometers can be used for detection. The non-contact infrared thermometer is greatly affected by environmental factors (the temperature of the object, the air medium), so as shown in Figure 5, (a) is a schematic diagram of the outlet plate temperature (non-contact) of the NOF furnace section, ( b) is a schematic diagram of the plate temperature (contact type) at the outlet of the NOF furnace section. Compared with the contact type temperature measurement sensor, it will contain more noise and outliers. If the plate temperature data is not processed, the accuracy of the data model cannot be guaranteed when the data is used for subsequent modeling (such as overfitting problems, etc.), so it needs to be preprocessed according to the characteristics of the exit plate temperature data. .

通过分析两炉段出口板温的时序分布特征，可知该类序列会局部反复出现的离群值和偶尔的“脏数据”，“脏数据”在这里是指由于传感器等设备或加热环境中存在的污染物等原因导致测得的数据出现突变离群值，该数据不能反映生产过程中的实际变化，如图6中的“菱形”表示的数据。于是基于此特征，考虑到实际生产中需对不同工况进行识别，所以不能丢失有效的局部信息，且为提高后续数工作的准确率和效率，需降低数据噪声，提高信噪比，因此需要保留变化趋势并强化有效离群值，同时平滑处理无效“脏数据”。考虑到该类流程工业数据反复出现相同的生产状态，其数据中含有重复的信息，且在时间尺度上呈现连续性分布，在时间维度上，前后数据存在相关性，因此利用相似信息强化对两炉段出口板温的异常特征进行筛选处理，其具体步骤为：By analyzing the time-series distribution characteristics of the plate temperature at the outlet of the two furnace sections, it can be seen that this type of sequence will have local and repeated outliers and occasional "dirty data". The measured data have sudden outliers due to pollutants and other reasons, and the data cannot reflect the actual changes in the production process, such as the data represented by the "diamond" in Figure 6. Therefore, based on this feature, considering that different working conditions need to be identified in actual production, effective local information cannot be lost, and in order to improve the accuracy and efficiency of subsequent data work, it is necessary to reduce data noise and improve the signal-to-noise ratio. Therefore, it is necessary to Preserves trends and reinforces valid outliers, while smoothing out invalid "dirty data". Considering that the same production state appears repeatedly in this type of process industry data, the data contains repeated information, and presents a continuous distribution on the time scale. The abnormal characteristics of the plate temperature at the outlet of the furnace section are screened and processed, and the specific steps are as follows:

1）设置两类固定时间长度的滑动窗，分别是用来限制寻找相关点范围的搜索窗口 和确定去噪点及相似点邻域大小的邻域窗，分别表示为

Figure 882080DEST_PATH_IMAGE060
，
Figure 7031DEST_PATH_IMAGE061
，其 中Ds和ds为决定D和d大小的参数，可以根据生产状态特征和算法速度进行设定； 1) Set up two types of sliding windows with fixed time lengths, which are the search window used to limit the range of finding relevant points and the neighborhood window used to determine the size of the neighborhood of denoising points and similar points, respectively expressed as
Figure 882080DEST_PATH_IMAGE060
,
Figure 7031DEST_PATH_IMAGE061
, where Ds and ds are the parameters that determine the size of D and d, which can be set according to the characteristics of the production state and the speed of the algorithm;

2）对搜索窗口和邻域窗口进行设置，搜说窗口以时间点i为中心，第一邻域窗口以i为中心，第二邻域窗口在搜索窗中滑动，其中以点j为待计算相似性度量点，两点间的相似度用权重因子w(i,j)表示：2) Set the search window and the neighborhood window. The search window is centered on time point i, the first neighborhood window is centered on i, and the second neighborhood window slides in the search window, where point j is the point to be calculated. Similarity measures points, and the similarity between two points is represented by a weight factor w(i,j):

Figure 209342DEST_PATH_IMAGE062
Figure 209342DEST_PATH_IMAGE062

式中，

Figure 946616DEST_PATH_IMAGE063
表示i邻域与j邻域的距离，该值越小，则邻域点与目标点越相 似，权重因子w(i,j)也越大。Z(i)为归一化系数，h为平滑参数； In the formula,
Figure 946616DEST_PATH_IMAGE063
Indicates the distance between the i neighborhood and the j neighborhood. The smaller the value is, the more similar the neighborhood point is to the target point, and the larger the weight factor w(i,j) is. Z(i) is the normalization coefficient, h is the smoothing parameter;

3）另第二邻域窗在搜索窗范围内滑动，遍历所有点，求出搜索窗内所有点与以目标点i为中心的邻域的相似性，由此，v(t)中时间点i处去噪后的数据u(i)为：3) Another second neighborhood window slides within the search window, traverses all points, and finds the similarity between all points in the search window and the neighborhood centered on the target point i, thus, the time point in v(t) The denoised data u(i) at i is:

Figure 550773DEST_PATH_IMAGE064
Figure 550773DEST_PATH_IMAGE064

T是指第二邻域窗在搜索窗中遍历的所有时间点，当两点越相似时，点j在u(i)的计算中占的比重越大；T refers to all the time points traversed by the second neighborhood window in the search window. When the two points are more similar, the greater the proportion of point j in the calculation of u(i);

4）遍历两炉段出口板温某段时间区的每个数据点，根据上述方法对该维度数据进行去噪及平滑处理。以NOF段出口板温某时间区间数据为例，将图6中数据进行预处理，结果如图7所示，可以看出可以保留强化变化趋势同时有效剔除离群值。4) Traverse each data point in a certain time zone of the outlet plate temperature of the two furnace sections, and perform denoising and smoothing processing on the dimension data according to the above method. Taking the data in a certain time interval of the outlet plate temperature of the NOF section as an example, the data in Figure 6 is preprocessed, and the result is shown in Figure 7. It can be seen that the trend of strengthening changes can be retained and outliers can be effectively eliminated.

目标的运行状态确定模块，用于提取根据现有系统中计划指标层I3的变量（一般 在现有生产线中，计划指标层变量有入炉计划中的带钢类型规格、入炉计划中的带钢长度、 入炉计划中的带钢入炉顺序）和多元特征获取模块中得到的多元特征，制定渐进明细的状 态跟踪策略。基于制定的策略条件及约束，融合多元特征建立目标状态检测模型，从而确定 目标的运行状态；所述渐进明细的状态跟踪策略是指先以入炉计划中的带钢长度为基准， 与运行速度进行相除计算，得到初始线性时间区间，将指计划指标层I3的变量在时序上进 行拉长，与初始线性时间区间对齐，再通过传感器检测到的某相邻带钢的首尾位置计算出 带钢长度，相比于入炉计划中的带钢长度，计算出的带钢长度更加准确（入炉计划中的带钢 长度为初始值，进入生产线时会根据生产要求和工艺要求对带钢进行小幅或大幅削减，削 减长度未知，因此入炉计划中的带钢长度和实际入炉的带钢长度一般并不一致，且一般情 况下相差较大），因此可通过此步骤计算出更准确一步的带钢长度，此时还是存在

Figure 846625DEST_PATH_IMAGE039
区间内的误差，最后再通过两炉段出口板温检测模型进一步精细化跟踪目标带 钢的位置，其中，
Figure 34767DEST_PATH_IMAGE013
表示
Figure 339847DEST_PATH_IMAGE014
对应的采样时间间隔，v表示生产线运行速度。目标的运行状 态确定模块的具体实现过程如下： The operating state determination module of the target is used to extract the variables according to the planning index layer I 3 in the existing system (generally in the existing production line, the planning index layer variables include the strip type specifications in the furnace entry plan, and the parameters in the furnace entry plan. Strip length, strip charging sequence in the charging schedule) and multi-features obtained in the multi-feature acquisition module to develop a progressively detailed status tracking strategy. Based on the formulated strategy conditions and constraints, a target state detection model is established by integrating multiple features, so as to determine the operating state of the target; the progressive and detailed state tracking strategy refers to the first step based on the strip length in the furnace plan, and the running speed. Divide the calculation to obtain the initial linear time interval, and lengthen the variables of the planning index layer I 3 in time series to align with the initial linear time interval, and then calculate the strip through the head and tail positions of an adjacent strip detected by the sensor. Compared with the strip length in the furnace entry plan, the calculated strip length is more accurate (the strip length in the furnace entry plan is the initial value, and the strip steel will be processed according to production requirements and process requirements when entering the production line. Slightly or substantially cut, the cut length is unknown, so the strip length in the furnace plan and the actual strip length are generally not consistent, and generally the difference is large), so this step can be used to calculate a more accurate one step. Strip length, which still exists at this time
Figure 846625DEST_PATH_IMAGE039
The error in the interval is calculated, and finally, the position of the target strip is further refined and tracked through the outlet plate temperature detection model of the two furnace sections. Among them,
Figure 34767DEST_PATH_IMAGE013
express
Figure 339847DEST_PATH_IMAGE014
The corresponding sampling time interval, v represents the running speed of the production line. The specific implementation process of the target operating state determination module is as follows:

3.1：对计划指标

Figure 34396DEST_PATH_IMAGE022
进行初始时间区间线性填充，其中k=1,2,3，分别代表入炉计划 中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序。初始时间区间线 性填充是指以入炉计划中的带钢长度为基准，与运行速度进行相除计算，得到初始线性时 间区间Z0，注：对于某一类带钢的各个变量
Figure 766728DEST_PATH_IMAGE022
来说，具有统一的Z0； 3.1: To plan indicators
Figure 34396DEST_PATH_IMAGE022
Perform linear filling of the initial time interval, where k=1, 2, 3, respectively represent the strip type specification in the furnace entry plan, the strip length in the furnace entry plan, and the strip entry sequence in the furnace entry plan. The linear filling of the initial time interval refers to dividing the length of the strip in the furnace plan as the benchmark and dividing the running speed to obtain the initial linear time interval Z 0 . Note: For each variable of a certain type of strip
Figure 766728DEST_PATH_IMAGE022
, with a uniform Z 0 ;

3.2：利用步骤3.1和多元特征获取模块中处理后的变量数据制定渐进明细的状态跟踪策略。具体是指：3.2: Use step 3.1 and the variable data processed in the multivariate feature acquisition module to formulate a progressive and detailed state tracking strategy. Specifically:

Step1：基于底层设备层的多元状态检测模型，检测出某相邻带钢的首尾位置，利 用入炉处焊缝距离

Figure 678052DEST_PATH_IMAGE029
，计算出某卷带钢的首端位置
Figure 308795DEST_PATH_IMAGE030
和尾端位置
Figure 356386DEST_PATH_IMAGE031
，r表示某卷带钢，同时计 算其长度Lr，
Figure 259620DEST_PATH_IMAGE032
； Step1: Based on the multi-state detection model of the underlying equipment layer, the head and tail positions of a certain adjacent strip steel are detected, and the distance of the weld at the entrance to the furnace is used.
Figure 678052DEST_PATH_IMAGE029
, calculate the position of the head end of a strip of steel
Figure 308795DEST_PATH_IMAGE030
and tail position
Figure 356386DEST_PATH_IMAGE031
, r represents a strip of steel, and its length L r is calculated at the same time,
Figure 259620DEST_PATH_IMAGE032
;

Step2：其次利用Z0进行筛选，若

Figure 159705DEST_PATH_IMAGE033
，则保留结果，转到Step4运算，若
Figure 806587DEST_PATH_IMAGE034
， 则说明数据异常，转到Step3； Step2: Secondly, use Z 0 to filter, if
Figure 159705DEST_PATH_IMAGE033
, then keep the result, go to Step4 operation, if
Figure 806587DEST_PATH_IMAGE034
, it means that the data is abnormal, go to Step3;

Step3：说明用

Figure 974263DEST_PATH_IMAGE029
检测和计算出的
Figure 546934DEST_PATH_IMAGE030
、
Figure 432850DEST_PATH_IMAGE031
和
Figure 617844DEST_PATH_IMAGE036
异常，采用底层设备层的下一个变 量，即
Figure 407070DEST_PATH_IMAGE037
进行检测及计算处新的
Figure 652107DEST_PATH_IMAGE030
、
Figure 25319DEST_PATH_IMAGE031
和
Figure 246960DEST_PATH_IMAGE036
，再次进行Step2，直至i+1=4，若
Figure 389228DEST_PATH_IMAGE038
依然计算出 数据异常，则说明进入生产线中的带钢有误，将错误信息返回系统界面，进行报警，此时需 要人工重新校准生产计划和生产线中的带钢信息； Step3: Instructions
Figure 974263DEST_PATH_IMAGE029
detected and calculated
Figure 546934DEST_PATH_IMAGE030
,
Figure 432850DEST_PATH_IMAGE031
and
Figure 617844DEST_PATH_IMAGE036
exception, take the next variable of the underlying device layer, i.e.
Figure 407070DEST_PATH_IMAGE037
Carry out inspection and calculation of new
Figure 652107DEST_PATH_IMAGE030
,
Figure 25319DEST_PATH_IMAGE031
and
Figure 246960DEST_PATH_IMAGE036
, perform Step2 again until i+1=4, if
Figure 389228DEST_PATH_IMAGE038
If the calculated data is still abnormal, it means that the strip entering the production line is wrong, and the error information is returned to the system interface to alarm. At this time, it is necessary to manually recalibrate the production plan and the strip information in the production line;

Step4：计算出的Lr受最小采样频率公倍数

Figure 805166DEST_PATH_IMAGE008
约束，得到一个相比于Z0更准确的 范围，但还是存在
Figure 901560DEST_PATH_IMAGE039
区间内的误差，由于生产线运行速度很高，因此需要更进一步 确定目标带钢的准确位置，基于两炉段出口板温检测模型，分别检测出NOF段出口带钢的大 幅度跳变和RTF段出口带钢的大幅度跳变位置，进一步精细化跟踪目标带钢的位置。 Step4: The calculated L r is subject to the common multiple of the minimum sampling frequency
Figure 805166DEST_PATH_IMAGE008
Constraints to get a more accurate range than Z 0 , but still
Figure 901560DEST_PATH_IMAGE039
The error in the interval, due to the high running speed of the production line, it is necessary to further determine the exact position of the target strip. Based on the detection model of the plate temperature at the exit of the two furnace sections, the large jump of the exit strip in the NOF section and the RTF section are detected respectively. The large jump position of the exit strip further refines and tracks the position of the target strip.

3.3：建立底层设备层的多元状态检测模型，用于对底层设备层

Figure 693936DEST_PATH_IMAGE065
的变量状态进行 检测，该检测是检测大幅跳变趋势，即通过确定瞬变特征点，确定状态发生变化的位置点。 通过分析底层设备层变量的时序分布特征可知，一卷带钢的
Figure 690711DEST_PATH_IMAGE066
变量具有稳定线增趋势，而 当发生带钢切换时，会有大幅度跳变过程，如图8所示为
Figure 776085DEST_PATH_IMAGE066
时序分布示意图，说明该卷带钢 在
Figure 123890DEST_PATH_IMAGE067
期间带头进入该检测点，即带钢首部位置，带钢在
Figure 955842DEST_PATH_IMAGE068
期间带尾进入该检测 点，即带钢尾部位置，该卷带钢在时序区间上的分布则在
Figure 72702DEST_PATH_IMAGE023
之间，因此需要检 测出每卷带钢的
Figure 830443DEST_PATH_IMAGE066
在时序上的瞬时变化特征点。建立底层设备层的多元状态检测模型步骤 如下： 3.3: Establish a multi-state detection model of the underlying device layer, which is used to detect the underlying device layer
Figure 693936DEST_PATH_IMAGE065
The variable state is detected, and the detection is to detect a large jump trend, that is, by determining the transient feature points, determine the position point where the state changes. By analyzing the time series distribution characteristics of the underlying equipment layer variables, it can be seen that the
Figure 690711DEST_PATH_IMAGE066
The variable has a stable linear increase trend, and when the strip is switched, there will be a large jump process, as shown in Figure 8:
Figure 776085DEST_PATH_IMAGE066
Schematic diagram of the timing distribution, illustrating that the coil is
Figure 123890DEST_PATH_IMAGE067
During the period, the lead enters the detection point, that is, the position of the head of the strip, and the strip is in the
Figure 955842DEST_PATH_IMAGE068
During the period, the strip tail enters the detection point, that is, the position of the strip tail, and the distribution of the coil strip in the time series is in
Figure 72702DEST_PATH_IMAGE023
between, so it is necessary to detect the
Figure 830443DEST_PATH_IMAGE066
Instantaneous change feature points in time series. The steps to establish the multi-state detection model of the underlying device layer are as follows:

设

Figure 164079DEST_PATH_IMAGE015
的时间序列{i(j)},为建立三个连续滑动窗口：均值计算窗口W m 、瞬时变化检 测窗口W d 和方差计算窗口W v ，窗口长度分别为m,n和v。计算
Figure 767098DEST_PATH_IMAGE015
的W m 均值和W d 均值，表示为M m 和M d ，并计算W v 的均值M v 和方差V计算公式为：Assume
Figure 164079DEST_PATH_IMAGE015
For the time series {i(j)}, three continuous sliding windows are established: mean calculation window W m , instantaneous change detection window W d and variance calculation window W v , and the window lengths are m, n and v, respectively. calculate
Figure 767098DEST_PATH_IMAGE015
The W m mean and W d mean of , denoted as M m and M d , and the mean M v and variance V of W v are calculated as:

Figure 4045DEST_PATH_IMAGE069
Figure 4045DEST_PATH_IMAGE069

式中，k0为第一个采样点。定义开始和结束事件累计和

Figure 168572DEST_PATH_IMAGE018
和
Figure 490969DEST_PATH_IMAGE019
，表达式为： In the formula, k 0 is the first sampling point. Define start and end event cumulative sums
Figure 168572DEST_PATH_IMAGE018
and
Figure 490969DEST_PATH_IMAGE019
, the expression is:

Figure 897680DEST_PATH_IMAGE020
Figure 897680DEST_PATH_IMAGE020

式中，δ为权重参数，

Figure 487668DEST_PATH_IMAGE021
为方差阈值，δ越大，在当前
Figure 587211DEST_PATH_IMAGE018
和
Figure 131324DEST_PATH_IMAGE019
统计值中所占的 比值越大，累计能力也越强，反之越小，通过判断
Figure 108770DEST_PATH_IMAGE018
和
Figure 54729DEST_PATH_IMAGE019
的变化情况，即可确定瞬变特 征点的位置。 In the formula, δ is the weight parameter,
Figure 487668DEST_PATH_IMAGE021
is the variance threshold, the larger the δ, the higher the current
Figure 587211DEST_PATH_IMAGE018
and
Figure 131324DEST_PATH_IMAGE019
The larger the ratio of the statistical value, the stronger the cumulative ability, and vice versa, the smaller
Figure 108770DEST_PATH_IMAGE018
and
Figure 54729DEST_PATH_IMAGE019
The position of the transient feature point can be determined.

3.4：两炉段出口板温状态检测模型。以上步骤已经带钢的位置在空间域上从计划 表中的带钢长度缩小到Lr，在时间域上，从Z0缩小到

Figure 325173DEST_PATH_IMAGE023
之间，其中，带钢切换 时，带头切换检测出两个状态变化时间点
Figure 855119DEST_PATH_IMAGE070
，带尾切换检测出两个状态变化时间点
Figure 603632DEST_PATH_IMAGE068
，这一卷带钢完整长度在炉内的时间范围即为
Figure 669677DEST_PATH_IMAGE023
，此时，仍存在偏差， 为了更准确跟踪目标带钢的位置，通过两炉段出口板温状态检测模型确定具体位置，两炉 出口板温状态检测模型建立过程为： 3.4: The detection model of the plate temperature state at the exit of the two furnace sections. The above steps have reduced the position of the strip from the strip length in the schedule to L r in the space domain, and from Z 0 to L r in the time domain
Figure 325173DEST_PATH_IMAGE023
between, among which, when the strip is switched, the lead switch detects two state change time points
Figure 855119DEST_PATH_IMAGE070
, the tail switching detects two state change time points
Figure 603632DEST_PATH_IMAGE068
, the time range of the complete length of this coil in the furnace is
Figure 669677DEST_PATH_IMAGE023
, at this time, there is still a deviation. In order to more accurately track the position of the target strip, the specific position is determined by the detection model of the plate temperature state at the exit of the two furnaces. The process of establishing the detection model of the plate temperature state at the exit of the two furnaces is as follows:

（1）选取多元特征获取模块处理后的

Figure 612487DEST_PATH_IMAGE002
中两个变量的数据，任一变量的时序数据集 合为y1,y2,…,yk，计算出数据集的均值作为该数据集的参考值为
Figure 131193DEST_PATH_IMAGE071
，偏差计算为
Figure 683397DEST_PATH_IMAGE025
，i=1,2,…,k； (1) Select the multi-feature acquisition module after processing
Figure 612487DEST_PATH_IMAGE002
The data of two variables in
Figure 131193DEST_PATH_IMAGE071
, the deviation is calculated as
Figure 683397DEST_PATH_IMAGE025
, i=1,2,…,k;

（2）利用以下公式计算出任一变量的时序数据集的标准差，根据稳定状态的数据在所有数据中的出现概率P及标准正态分布表，得到P对应的区间边界±m1σ；m1为正数，判定稳定状态数据分布在以所述参考值为中心的±m1σ范围内（本实施例中m1>3）；(2) Use the following formula to calculate the standard deviation of the time series data set of any variable, and obtain the interval boundary ±m 1 σ corresponding to P according to the occurrence probability P of the data in the steady state in all the data and the standard normal distribution table; m 1 is a positive number, and it is determined that the steady state data is distributed within the range of ±m 1 σ centered on the reference value (m 1 >3 in this embodiment);

Figure 571325DEST_PATH_IMAGE026
Figure 571325DEST_PATH_IMAGE026

上式中，vi为所述历史数据序列中第i个数据yi对应的偏差；In the above formula, vi is the deviation corresponding to the i -th data yi in the historical data sequence;

（3）获取实时数据序列，并以连续的s个数据为一组判定实时状态；针对每一组，具 体判定方法为：分别计算该组s个实时数据y1,y2,…,ys对应的偏差，得到s个偏差；将所述s 个偏差求均值，得到均值v；若|v|>nσ，表明该组实时数据属于带钢切换时间点

Figure 449152DEST_PATH_IMAGE072
，再结合 步骤S3.3确定的
Figure 189575DEST_PATH_IMAGE027
或
Figure 843672DEST_PATH_IMAGE028
，取两者间的交集，则为带钢带头和带尾的具体时序 位置，在该时刻，对应的空间位置为两炉段出口的位置。 (3) Obtain the real-time data sequence, and determine the real-time state by taking the continuous s data as a group; for each group, the specific determination method is: calculate the group of s real-time data y 1 , y 2 ,...,y s respectively Corresponding deviations, s deviations are obtained; the s deviations are averaged to obtain the mean value v; if |v|>nσ, it indicates that this group of real-time data belongs to the strip switching time point
Figure 449152DEST_PATH_IMAGE072
, and then combined with step S3.3 determined
Figure 189575DEST_PATH_IMAGE027
or
Figure 843672DEST_PATH_IMAGE028
, and the intersection between the two is taken as the specific timing position of the strip head and the strip tail. At this moment, the corresponding spatial position is the position of the exit of the two furnace sections.

目标跟踪预测模块，用于在时间尺度上将计划指标层

Figure 618730DEST_PATH_IMAGE073
中的变量信息（具体的带钢 编号、类型、顺序）与目标的运行状态确定模块中对某卷钢的运行状态结果进行时间对齐， 在空间尺度上结合本身炉体的设备信息，从而实时跟踪在炉体内动态变化的目标。基于该 目标的跟踪信息，建立目标到各炉段出口等关键位置的预测模型，预测进出各炉段的时间， 为后续过程建模及系统控制奠定基础。建立目标到各炉段出口位置的预测模型的具体过程 如下： target tracking forecasting module for planning indicator layers on time scales
Figure 618730DEST_PATH_IMAGE073
The variable information (specific strip number, type, sequence) in the target operating state determination module is time-aligned with the operating state results of a certain coil, and the equipment information of its own furnace body is combined on the spatial scale to track real-time tracking. Dynamically changing targets within the furnace body. Based on the tracking information of the target, a prediction model is established from the target to key positions such as the exit of each furnace section, and the time of entering and leaving each furnace section is predicted, which lays the foundation for subsequent process modeling and system control. The specific process of establishing the prediction model from the target to the exit position of each furnace section is as follows:

4.1：更新初始线性时间区间Z0和步骤3.4中某卷带钢带头带尾的切换时间区间保 持一致，设为

Figure 923851DEST_PATH_IMAGE040
，即为该卷带钢完全通过炉内某检测点的具体时间，同时与运行速度相乘， 计算得到该卷带钢的距离长度
Figure 151570DEST_PATH_IMAGE041
，同时由步骤3.4可知该卷带钢带头的具体位置
Figure 311156DEST_PATH_IMAGE042
； 4.1: Update the initial linear time interval Z 0 to be consistent with the switching time interval of the head and tail of a strip in step 3.4, set as
Figure 923851DEST_PATH_IMAGE040
, that is, the specific time when the coil of steel completely passes through a certain detection point in the furnace, and at the same time, it is multiplied by the running speed to calculate the distance length of the coiled steel.
Figure 151570DEST_PATH_IMAGE041
, and the specific position of the strip head can be known from step 3.4
Figure 311156DEST_PATH_IMAGE042
;

4.2：炉体信息为固定数值，各关键位置的空间位置都是固定已知值。设某关键位 置的空间位置为

Figure 707764DEST_PATH_IMAGE043
，则预测经过该关键位置的具体时间为： 4.2: The furnace body information is a fixed value, and the spatial position of each key position is a fixed and known value. Let the spatial position of a key position be
Figure 707764DEST_PATH_IMAGE043
, then the specific time predicted to pass the key position is:

Figure 661814DEST_PATH_IMAGE044
Figure 661814DEST_PATH_IMAGE044

式中，n表示带钢带头具体位置

Figure 376829DEST_PATH_IMAGE042
到关键位置
Figure 838641DEST_PATH_IMAGE043
期间运行速度变化的次数，该数据 是可以通过生产线中的速度传感器检测得到的，
Figure 322712DEST_PATH_IMAGE045
表示第n次变化的运行速度，
Figure 949128DEST_PATH_IMAGE046
表示以
Figure 151439DEST_PATH_IMAGE045
传送的距离，其中满足条件
Figure 652827DEST_PATH_IMAGE074
。 In the formula, n represents the specific position of the strip head
Figure 376829DEST_PATH_IMAGE042
to key locations
Figure 838641DEST_PATH_IMAGE043
The number of times the running speed changes during the period, this data can be detected by the speed sensor in the production line,
Figure 322712DEST_PATH_IMAGE045
represents the running speed of the nth change,
Figure 949128DEST_PATH_IMAGE046
means with
Figure 151439DEST_PATH_IMAGE045
The distance transmitted, where the condition is met
Figure 652827DEST_PATH_IMAGE074
.

本发明提供的技术方案带来的有益效果是：融合多时空层级参数的序列特征，共同决策并跟踪连续退火过程中目标带钢的实时位置，从而建立目标带钢到各关键位置的预测模型，为后续过程建模及系统控制奠定基础。The beneficial effects brought by the technical solution provided by the present invention are: integrating the sequence features of multi-temporal and spatial hierarchical parameters, jointly making decisions and tracking the real-time position of the target strip in the continuous annealing process, so as to establish a prediction model for the target strip to each key position, Lay the foundation for subsequent process modeling and system control.

以上所述仅为本发明的较佳实施例，并不用以限制本发明，凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (7)

1.一种目标跟踪系统，其特征在于：包括：1. a target tracking system, is characterized in that: comprise: 关键变量获取模块，用于根据退火炉中连续退火加热过程，提取和目标带钢位置相关的参数作为关键变量，并按照设备及控制类型将这些关键变量分为底层设备层I1、运行过程层I2和计划指标层I3；其中，所述关键变量包括炉计划中的带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序、入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温；The key variable acquisition module is used to extract parameters related to the target strip position as key variables according to the continuous annealing heating process in the annealing furnace, and divide these key variables into the bottom equipment layer I 1 and the operation process layer according to equipment and control types. I 2 and planning index layer I 3 ; wherein, the key variables include strip type specifications in the furnace plan, strip length in the furnace entry plan, strip entry sequence in the furnace entry plan, and welding seam at the furnace entry Distance, weld seam at PH-NOF junction, weld seam distance at NOF furnace section exit, weld seam distance at RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature; 多元特征获取模块，用于对底层设备层I1和运行过程层I2中多个相关变量进行运行状态分析，提取各变量的数据时序分布特征，结合退火炉生产工艺确定存在的多种异常特征，基于优先级策略和时序分布特征，对异常特征进行筛选处理，得到多元特征；其中，所述多个相关变量包括入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离、RTF炉段出口处焊缝距离、运行速度、NOF炉段出口板温和RTF炉段出口板温；The multi-feature acquisition module is used to analyze the operating state of multiple related variables in the underlying equipment layer I 1 and the operation process layer I 2 , extract the data time series distribution characteristics of each variable, and determine the existence of various abnormal characteristics in combination with the annealing furnace production process , based on the priority strategy and time series distribution characteristics, the abnormal characteristics are screened to obtain multivariate characteristics; wherein, the multiple relevant variables include the weld distance at the entrance to the furnace, the weld at the PH-NOF junction, and the NOF furnace section exit. Weld distance, weld distance at RTF furnace section exit, running speed, NOF furnace section outlet plate temperature and RTF furnace section outlet plate temperature; 目标的运行状态确定模块，用于根据现有系统中计划指标层I3变量和多元特征获取模块得到的多元特征，制定渐进明细的状态跟踪策略；基于制定的渐进明细状态跟踪策略条件及约束，融合多元特征建立目标状态检测模型，从而确定目标的运行状态；The operating state determination module of the target is used to formulate a progressively detailed state tracking strategy based on the multivariate features obtained by the planning index layer I 3 variables and the multivariate feature acquisition module in the existing system; based on the formulated progressive detailed state tracking strategy conditions and constraints, Integrate multiple features to establish a target state detection model, so as to determine the running state of the target; 目标跟踪预测模块，用于在时间尺度上将计划指标层I3变量与目标的运行状态确定模块中对某卷钢的运行状态结果进行时间对齐，其中，计划指标层I3变量包括入炉计划中的带钢类型规格、入炉计划中的带钢长度和入炉计划中的带钢入炉顺序，在空间尺度上结合本身炉体的设备信息，从而实时跟踪在炉体内动态变化的目标；基于该目标的跟踪信息，建立目标到各炉段出口位置的预测模型，以预测目标进出各炉段的时间。The target tracking and forecasting module is used to align the running state results of a coil of steel in the time scale with the plan index layer I3 variables and the target operation state determination module, wherein the plan index layer I3 variables include the furnace entry plan The strip type specification in the furnace, the strip length in the furnace entry plan and the strip entry sequence in the furnace entry plan, combined with the equipment information of the furnace body on the spatial scale, so as to track the dynamically changing target in the furnace body in real time; Based on the tracking information of the target, a prediction model from the target to the exit position of each furnace section is established to predict the time when the target enters and exits each furnace section. 2.如权利要求1所述的一种目标跟踪系统，其特征在于：多元特征获取模块中得到多元特征的具体过程如下：2. a kind of target tracking system as claimed in claim 1, is characterized in that: the concrete process that obtains multi-features in multi-feature acquisition module is as follows: 2.1：选择底层设备层I1相关变量作为集合
Figure 228113DEST_PATH_IMAGE001
和运行过程层I2相关变量作为集合
Figure 482376DEST_PATH_IMAGE002
，其 中，i=入炉处焊缝距离、PH-NOF结合处焊缝、NOF炉段出口处焊缝距离或RTF炉段出口处焊缝 距离，j=运行速度、NOF炉段出口板温或RTF炉段出口板温； 2.1: Select the underlying device layer I 1 related variables as a set
Figure 228113DEST_PATH_IMAGE001
and the operating process layer I 2 related variables as a set
Figure 482376DEST_PATH_IMAGE002
, where i=the distance of the weld at the entrance to the furnace, the weld at the PH-NOF junction, the distance of the weld at the exit of the NOF furnace section or the distance of the weld at the exit of the RTF furnace section, j = the running speed, the temperature of the plate at the outlet of the NOF furnace section or RTF furnace section outlet plate temperature; 2.2：对
Figure 860530DEST_PATH_IMAGE003
和
Figure 542047DEST_PATH_IMAGE004
中的变量进行最小采样频率公倍数筛选，最小采样频率公倍数为： 2.2: Yes
Figure 860530DEST_PATH_IMAGE003
and
Figure 542047DEST_PATH_IMAGE004
The variables in are screened by the common multiple of the minimum sampling frequency, and the common multiple of the minimum sampling frequency is:
Figure 402556DEST_PATH_IMAGE005
Figure 402556DEST_PATH_IMAGE005
 式中，k=1,2,[
Figure 427887DEST_PATH_IMAGE006
]表示求解
Figure 424662DEST_PATH_IMAGE007
和
Figure 11501DEST_PATH_IMAGE002
中所有变量采样频率的公倍数；得到
Figure 595192DEST_PATH_IMAGE008
后，对
Figure 925679DEST_PATH_IMAGE001
和
Figure 776960DEST_PATH_IMAGE002
中的各变量进行筛选： In the formula, k=1,2,[
Figure 427887DEST_PATH_IMAGE006
] means solving
Figure 424662DEST_PATH_IMAGE007
and
Figure 11501DEST_PATH_IMAGE002
the common multiple of the sampling frequency of all variables in ; get
Figure 595192DEST_PATH_IMAGE008
after, yes
Figure 925679DEST_PATH_IMAGE001
and
Figure 776960DEST_PATH_IMAGE002
to filter each variable in:
Figure 298815DEST_PATH_IMAGE009
Figure 298815DEST_PATH_IMAGE009
Figure 602758DEST_PATH_IMAGE010
和
Figure 471356DEST_PATH_IMAGE011
为筛选后的集合；
Figure 602758DEST_PATH_IMAGE010
and
Figure 471356DEST_PATH_IMAGE011
is the filtered set; 2.3：对筛选后的
Figure 209768DEST_PATH_IMAGE010
和
Figure 138409DEST_PATH_IMAGE011
中各变量的时序分布特征进行探索性分析，得到多元数据； 2.3: For the filtered
Figure 209768DEST_PATH_IMAGE010
and
Figure 138409DEST_PATH_IMAGE011
Exploratory analysis of the time series distribution characteristics of each variable in the 2.4：对以上经过同采样频率筛选后的多元数据进行预处理，预处理优先级为：运行速度>底层设备层变量>两炉段出口板温；预处理是指：具有不同类型分布特征的变量有不同的异常特征或需要不同的数据处理方法，当运行速度为0或为负时，则生产线停机或现场出现异常，对该时间区间所有变量不作处理。2.4: Preprocess the above multivariate data after screening with the same sampling frequency. The preprocessing priority is: running speed > underlying equipment layer variables > outlet plate temperature of the two furnace sections; preprocessing refers to: variables with different types of distribution characteristics There are different abnormal characteristics or different data processing methods. When the running speed is 0 or negative, the production line stops or there is an abnormality on site, and all variables in this time interval are not processed. 3.如权利要求1所述的一种目标跟踪系统，其特征在于：目标的运行状态确定模块中， 所述渐进明细的状态跟踪策略是指先以入炉计划中的带钢长度为基准，与运行速度进行相 除计算，得到初始线性时间区间，将指计划指标层I3的变量在时序上进行拉长，与初始线性 时间区间对齐，再通过传感器检测到的某相邻带钢的首尾位置计算出带钢长度，相比于入 炉计划中的带钢长度，计算出的带钢长度更加准确，这是因为入炉计划中的带钢长度为初 始值，进入生产线时会根据生产要求和工艺要求对带钢进行小幅或大幅削减，削减长度未 知，因此入炉计划中的带钢长度和实际入炉的带钢长度并不一致，因此可通过此步骤计算 出更准确一步的带钢长度，但是此时还是存在
Figure 195227DEST_PATH_IMAGE012
区间内的误差，所以最后需要再 通过两炉段出口板温检测模型进一步精细化跟踪目标带钢的位置，以便得到更为精确的带 钢长度，其中，
Figure 366052DEST_PATH_IMAGE013
表示
Figure 191925DEST_PATH_IMAGE014
对应的采样时间间隔，v表示生产线运行速度。 3. A target tracking system according to claim 1, characterized in that: in the target operating state determination module, the progressively detailed state tracking strategy refers to first taking the strip length in the furnace entry plan as a benchmark, and The running speed is divided and calculated to obtain the initial linear time interval, and the variables that refer to the plan index layer I 3 are elongated in time sequence to align with the initial linear time interval, and then the head and tail positions of an adjacent strip detected by the sensor are used. The calculated strip length is more accurate than the strip length in the furnace entry plan. This is because the strip length in the furnace entry plan is the initial value, and when it enters the production line, it will be based on production requirements and The process requires a small or large reduction of the strip, and the cutting length is unknown, so the strip length in the furnace plan and the actual strip length are not consistent, so a more accurate strip length can be calculated through this step. But it still exists
Figure 195227DEST_PATH_IMAGE012
Therefore, it is necessary to further refine and track the position of the target strip through the outlet plate temperature detection model of the two furnace sections in order to obtain a more accurate strip length, among which,
Figure 366052DEST_PATH_IMAGE013
express
Figure 191925DEST_PATH_IMAGE014
The corresponding sampling time interval, v represents the running speed of the production line. 4.如权利要求1所述的一种目标跟踪系统，其特征在于：目标的运行状态确定模块中，底层设备层变量的状态检测模型的建立过程为：4. a kind of target tracking system as claimed in claim 1 is characterized in that: in the running state determination module of the target, the establishment process of the state detection model of the underlying equipment layer variable is: （1）设每个底层设备层的变量
Figure 25889DEST_PATH_IMAGE015
的数据表示为
Figure 337047DEST_PATH_IMAGE016
，建立三个连续滑动窗口：均值计算 窗口W m 、瞬时变化检测窗口W d 和方差计算窗口W v ，窗口长度分别为m，n和v； (1) Set the variables of each underlying device layer
Figure 25889DEST_PATH_IMAGE015
The data is represented as
Figure 337047DEST_PATH_IMAGE016
, establish three continuous sliding windows: mean calculation window W m , instantaneous change detection window W d and variance calculation window W v , the window lengths are m , n and v respectively; （2）计算
Figure 547449DEST_PATH_IMAGE015
的W m 均值和W d 均值，表示为M m 和M d ，并计算W v 的均值M v 和方差V计算公式为： (2) Calculation
Figure 547449DEST_PATH_IMAGE015
The W m mean and W d mean of , denoted as M m and M d , and the mean M v and variance V of W v are calculated as:
Figure 227829DEST_PATH_IMAGE017
Figure 227829DEST_PATH_IMAGE017
 式中，k0为第一个采样点，定义开始和结束事件累计和
Figure 262387DEST_PATH_IMAGE018
和
Figure 28218DEST_PATH_IMAGE019
，表达式为 In the formula, k 0 is the first sampling point, which defines the cumulative sum of the start and end events
Figure 262387DEST_PATH_IMAGE018
and
Figure 28218DEST_PATH_IMAGE019
, the expression is
Figure 809355DEST_PATH_IMAGE020
Figure 809355DEST_PATH_IMAGE020
 式中，δ为权重参数
Figure 609820DEST_PATH_IMAGE021
为方差阈值，δ越大，在当前
Figure 785587DEST_PATH_IMAGE018
和
Figure 802828DEST_PATH_IMAGE019
统计值中所占的比值越 大，累计能力也越强，反之越小，通过判断
Figure 620611DEST_PATH_IMAGE018
和
Figure 275583DEST_PATH_IMAGE019
的变化情况，即可确定瞬变特征点的 位置。 In the formula, δ is the weight parameter
Figure 609820DEST_PATH_IMAGE021
is the variance threshold, the larger the δ, the higher the current
Figure 785587DEST_PATH_IMAGE018
and
Figure 802828DEST_PATH_IMAGE019
The larger the ratio of the statistical value, the stronger the cumulative ability, and vice versa, the smaller
Figure 620611DEST_PATH_IMAGE018
and
Figure 275583DEST_PATH_IMAGE019
The position of the transient feature point can be determined. 5.如权利要求4所述的一种目标跟踪系统，其特征在于：目标的运行状态确定模块的具体实现过程如下：5. a kind of target tracking system as claimed in claim 4 is characterized in that: the concrete realization process of the running state determination module of target is as follows: 3.1：对计划指标
Figure 654874DEST_PATH_IMAGE022
进行初始时间区间线性填充，其中k=1,2,3，分别代表入炉计划中的 带钢类型规格、入炉计划中的带钢长度、入炉计划中的带钢入炉顺序；初始时间区间线性填 充是指以入炉计划中的带钢长度为基准，与运行速度进行相除计算，得到初始线性时间区 间Z0； 3.1: To plan indicators
Figure 654874DEST_PATH_IMAGE022
Perform linear filling of the initial time interval, where k=1, 2, 3, respectively represent the strip type and specification in the furnace entry plan, the strip length in the furnace entry plan, and the strip entry sequence in the furnace entry plan; the initial time Interval linear filling means that the strip length in the furnace entry plan is used as the benchmark, and the operation speed is divided and calculated to obtain the initial linear time interval Z 0 ; 3.2：利用步骤3.1和多元特征获取模块中处理后的变量数据制定渐进明细的状态跟踪策略；3.2: Use step 3.1 and the variable data processed in the multi-feature acquisition module to formulate a progressive and detailed state tracking strategy; 3.3：利用建立的底层设备层的多元状态检测模型，对底层设备层I1的变量状态进行检测，即通过确定瞬变特征点，确定状态发生变化的位置点；3.3: Use the established multi-state detection model of the underlying device layer to detect the variable state of the underlying device layer I 1 , that is, determine the location point where the state changes by determining the transient feature point; 3.4：通过步骤3.3，可进一步将带钢的位置在空间域上从入炉计划中的带钢长度缩小 到计算得到的带钢长度Lr，在时间域上，从Z0缩小到
Figure 395297DEST_PATH_IMAGE023
之间，为了更准确跟踪 目标带钢的位置，通过两炉段出口板温状态检测模型确定具体位置，其建立过程为： 3.4: Through step 3.3, the position of the strip can be further reduced from the strip length in the furnace plan to the calculated strip length L r in the space domain, and from Z 0 to the calculated strip length L r in the time domain.
Figure 395297DEST_PATH_IMAGE023
In between, in order to more accurately track the position of the target strip, the specific position is determined by the detection model of the plate temperature state at the exit of the two furnace sections. The establishment process is as follows: （1）选取多元特征获取模块中处理后的
Figure 751192DEST_PATH_IMAGE002
中两个变量的数据，任一变量的时序数据集合 为y1,y2,…,yk，计算出数据集的均值作为该数据集的参考值为
Figure 759206DEST_PATH_IMAGE024
，偏差计算为
Figure 807934DEST_PATH_IMAGE025
， i=1,2,…,k； (1) Select the processed data in the multivariate feature acquisition module
Figure 751192DEST_PATH_IMAGE002
The data of two variables in
Figure 759206DEST_PATH_IMAGE024
, the deviation is calculated as
Figure 807934DEST_PATH_IMAGE025
, i=1,2,…,k; （2）利用以下公式计算出任一变量的时序数据集的标准差，根据稳定状态的数据在所有数据中的出现概率P及标准正态分布表，得到P对应的区间边界±m1σ；m1为正数，判定稳定状态数据分布在以所述参考值为中心的±m1σ范围内；(2) Use the following formula to calculate the standard deviation of the time series data set of any variable, and obtain the interval boundary ±m 1 σ corresponding to P according to the occurrence probability P of the data in the steady state in all the data and the standard normal distribution table; m 1 is a positive number, it is determined that the steady state data is distributed in the range of ±m 1 σ centered on the reference value;
Figure 301232DEST_PATH_IMAGE026
Figure 301232DEST_PATH_IMAGE026
 上式中，vi为历史数据序列中第i个数据yi对应的偏差；In the above formula, vi is the deviation corresponding to the i -th data yi in the historical data sequence; 获取实时数据序列，并以连续的s个数据为一组判定实时状态；针对每一组，具体判定 方法为：分别计算该组s个实时数据y1,y2,…,ys对应的偏差，得到s个偏差；将所述s个偏差 求均值，得到均值v；若|v|>nσ，表明该组实时数据属于带钢切换时间点Tr 2，再结合步骤3.3 确定的
Figure 696703DEST_PATH_IMAGE027
或
Figure 326268DEST_PATH_IMAGE028
，取两者间的交集，则为带钢带头和带尾的具体时序位置，在该 时刻，对应的空间位置即为两炉段出口的位置。 Obtain the real-time data sequence, and use the consecutive s data as a group to determine the real-time state; for each group, the specific determination method is: calculate the deviation corresponding to the group of s real-time data y 1 , y 2 ,...,y s respectively , obtain s deviations; average the s deviations to obtain the mean v; if |v|>nσ, it indicates that this group of real-time data belongs to the strip switching time point Tr 2 , and then combined with the determined value in step 3.3
Figure 696703DEST_PATH_IMAGE027
or
Figure 326268DEST_PATH_IMAGE028
, and the intersection between the two is taken as the specific timing position of the strip head and the strip tail. At this moment, the corresponding spatial position is the position of the exit of the two furnace sections. 6.如权利要求5所述的一种目标跟踪系统，其特征在于：制定状态跟踪策略的过程如下：6. a kind of target tracking system as claimed in claim 5 is characterized in that: the process of formulating state tracking strategy is as follows: 1）：基于底层设备层的多元状态检测模型，检测出某相邻带钢的首尾位置，利用入炉处 焊缝距离
Figure 280317DEST_PATH_IMAGE029
，计算出某卷带钢的首端位置
Figure 595384DEST_PATH_IMAGE030
和尾端位置
Figure 27503DEST_PATH_IMAGE031
，r表示某卷带钢，同时计算其长度 Lr，
Figure 777153DEST_PATH_IMAGE032
； 1): Based on the multi-state detection model of the underlying equipment layer, the position of the head and tail of an adjacent strip is detected, and the distance of the weld at the entrance to the furnace is used.
Figure 280317DEST_PATH_IMAGE029
, calculate the position of the head end of a strip of steel
Figure 595384DEST_PATH_IMAGE030
and tail position
Figure 27503DEST_PATH_IMAGE031
, r represents a strip of steel, and its length L r is calculated at the same time,
Figure 777153DEST_PATH_IMAGE032
; 2）：其次利用Z0进行筛选，若
Figure 403569DEST_PATH_IMAGE033
，则保留结果，转到4）运算，若
Figure 605880DEST_PATH_IMAGE034
，则说明数据 异常，转到3）； 2): Secondly, use Z 0 to filter, if
Figure 403569DEST_PATH_IMAGE033
, then keep the result, go to 4) operation, if
Figure 605880DEST_PATH_IMAGE034
, then the data is abnormal, go to 3); 3）：说明用
Figure 107268DEST_PATH_IMAGE035
检测和计算出的
Figure 209960DEST_PATH_IMAGE030
、
Figure 505812DEST_PATH_IMAGE031
和
Figure 195420DEST_PATH_IMAGE036
异常，采用底层设备层的下一个变量，即
Figure 470805DEST_PATH_IMAGE037
进 行检测及计算处新的
Figure 195048DEST_PATH_IMAGE030
、
Figure 425916DEST_PATH_IMAGE031
和
Figure 337240DEST_PATH_IMAGE036
，再次进行2），直至i+1=4，若
Figure 914852DEST_PATH_IMAGE038
依然计算出数据异常，则说明 进入生产线中的带钢有误，将错误信息返回系统界面，进行报警，此时需要人工重新校准生 产计划和生产线中的带钢信息； 3): for illustration
Figure 107268DEST_PATH_IMAGE035
detected and calculated
Figure 209960DEST_PATH_IMAGE030
,
Figure 505812DEST_PATH_IMAGE031
and
Figure 195420DEST_PATH_IMAGE036
exception, take the next variable of the underlying device layer, i.e.
Figure 470805DEST_PATH_IMAGE037
Carry out inspection and calculation of new
Figure 195048DEST_PATH_IMAGE030
,
Figure 425916DEST_PATH_IMAGE031
and
Figure 337240DEST_PATH_IMAGE036
, do 2) again, until i+1=4, if
Figure 914852DEST_PATH_IMAGE038
If the calculated data is still abnormal, it means that the strip entering the production line is wrong, and the error information is returned to the system interface to alarm. At this time, it is necessary to manually recalibrate the production plan and the strip information in the production line; 4）：计算出的Lr受最小采样频率公倍数
Figure 995065DEST_PATH_IMAGE008
约束，得到一个相比于Z0更准确的范围，但 还是存在
Figure 632720DEST_PATH_IMAGE039
区间内的误差，为了更进一步确定目标带钢的准确位置，基于两炉段 出口板温检测模型，分别检测出NOF段出口带钢的大幅度跳变和RTF段出口带钢的大幅度跳 变位置，进一步精细化跟踪目标带钢的位置。 4): The calculated L r is subject to a common multiple of the minimum sampling frequency
Figure 995065DEST_PATH_IMAGE008
Constraints to get a more accurate range than Z 0 , but still
Figure 632720DEST_PATH_IMAGE039
The error in the interval, in order to further determine the exact position of the target strip, based on the detection model of the plate temperature at the exit of the two furnace sections, the large jump of the exit strip of the NOF section and the large jump of the exit strip of the RTF section were detected respectively. position, and further refine the tracking of the position of the target strip. 7.如权利要求6所述的一种目标跟踪系统，其特征在于：目标跟踪预测模块中，建立目标到各炉段出口位置的预测模型的具体过程如下：7. A kind of target tracking system as claimed in claim 6 is characterized in that: in the target tracking prediction module, the concrete process of establishing the prediction model of the target to the exit position of each furnace section is as follows: 4.1：更新初始线性时间区间Z0和步骤3.4中某卷带钢带头带尾的切换时间区间保持一 致，设为
Figure 296920DEST_PATH_IMAGE040
，即为该卷带钢完全通过炉内某检测点的具体时间，同时与运行速度相乘，计算 得到该卷带钢的距离长度
Figure 176758DEST_PATH_IMAGE041
，同时由步骤3.4可知该卷带钢带头的具体位置
Figure 610013DEST_PATH_IMAGE042
； 4.1: Update the initial linear time interval Z 0 to be consistent with the switching time interval of the head and tail of a strip in step 3.4, set as
Figure 296920DEST_PATH_IMAGE040
, that is, the specific time when the coil of steel completely passes through a certain detection point in the furnace, and at the same time, it is multiplied by the running speed to calculate the distance length of the coiled steel.
Figure 176758DEST_PATH_IMAGE041
, and the specific position of the strip head can be known from step 3.4
Figure 610013DEST_PATH_IMAGE042
; 4.2：炉体信息为固定数值，各关键位置的空间位置都是固定已知值，设某关键位置的 空间位置为
Figure 418569DEST_PATH_IMAGE043
，则建立的目标到各炉段出口位置的预测模型为： 4.2: The furnace body information is a fixed value, and the spatial position of each key position is a fixed known value. Let the spatial position of a key position be
Figure 418569DEST_PATH_IMAGE043
, the established prediction model from the target to the outlet position of each furnace section is:
Figure 540371DEST_PATH_IMAGE044
Figure 540371DEST_PATH_IMAGE044
 式中，n表示带钢带头具体位置
Figure 990944DEST_PATH_IMAGE042
到关键位置
Figure 13127DEST_PATH_IMAGE043
期间运行速度变化的次数，
Figure 756698DEST_PATH_IMAGE045
表示第n次 变化的运行速度，
Figure 395490DEST_PATH_IMAGE046
表示以
Figure 118595DEST_PATH_IMAGE045
传送的距离。 In the formula, n represents the specific position of the strip head
Figure 990944DEST_PATH_IMAGE042
to key locations
Figure 13127DEST_PATH_IMAGE043
The number of times the running speed changes during the period,
Figure 756698DEST_PATH_IMAGE045
represents the running speed of the nth change,
Figure 395490DEST_PATH_IMAGE046
means with
Figure 118595DEST_PATH_IMAGE045
distance of transmission.
CN202210831903.9A 2022-07-15 2022-07-15 a target tracking system Active CN114896900B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210831903.9A CN114896900B (en) 2022-07-15 2022-07-15 a target tracking system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210831903.9A CN114896900B (en) 2022-07-15 2022-07-15 a target tracking system

Publications (2)

Publication Number Publication Date
CN114896900A true CN114896900A (en) 2022-08-12
CN114896900B CN114896900B (en) 2022-09-30

Family

ID=82730151

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210831903.9A Active CN114896900B (en) 2022-07-15 2022-07-15 a target tracking system

Country Status (1)

Country Link
CN (1) CN114896900B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116505871A (en) * 2023-04-28 2023-07-28 江苏中信博新能源科技股份有限公司 Anomaly detection method, device and system for a tracker in a photovoltaic power generation system
CN117070740A (en) * 2023-07-18 2023-11-17 中冶京诚工程技术有限公司 Method and device for tracking and calibrating billet of heating furnace

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5558881A (en) * 1978-10-23 1980-05-01 Ibm Page request system using i*o adapter
CN1704487A (en) * 2004-05-28 2005-12-07 上海宝信软件股份有限公司 Process for controlling temperature of strip steel
CN102886383A (en) * 2011-07-22 2013-01-23 宝山钢铁股份有限公司 Method for controlling mechanical properties of cold-rolling strip steel in online manner
US20180347018A1 (en) * 2014-11-18 2018-12-06 Salzgitter Flachstahl Gmbh High-strength air-hardening multiphase steel having excellent processing properties, and method for manufacturing a strip of said steel
CN111625934A (en) * 2020-04-30 2020-09-04 中国地质大学（武汉） Multi-mode identification method for annealing heating process based on D-S evidence theory

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5558881A (en) * 1978-10-23 1980-05-01 Ibm Page request system using i*o adapter
CN1704487A (en) * 2004-05-28 2005-12-07 上海宝信软件股份有限公司 Process for controlling temperature of strip steel
CN102886383A (en) * 2011-07-22 2013-01-23 宝山钢铁股份有限公司 Method for controlling mechanical properties of cold-rolling strip steel in online manner
US20180347018A1 (en) * 2014-11-18 2018-12-06 Salzgitter Flachstahl Gmbh High-strength air-hardening multiphase steel having excellent processing properties, and method for manufacturing a strip of said steel
CN111625934A (en) * 2020-04-30 2020-09-04 中国地质大学（武汉） Multi-mode identification method for annealing heating process based on D-S evidence theory

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
YONGYUE ZHANG 等: "A Dynamic Data-driven Model for Predicting Strip Temperature in Continuous Annealing Line Heating Process", 《2018 37TH CHINESE CONTROL CONFERENCE (CCC)》 *
李建平 等: "两段探测目标的传感器任务调度问题0－1规划模型及算法", 《国防科技大学学报》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116505871A (en) * 2023-04-28 2023-07-28 江苏中信博新能源科技股份有限公司 Anomaly detection method, device and system for a tracker in a photovoltaic power generation system
CN117070740A (en) * 2023-07-18 2023-11-17 中冶京诚工程技术有限公司 Method and device for tracking and calibrating billet of heating furnace

Also Published As

Publication number Publication date
CN114896900B (en) 2022-09-30

Similar Documents

Publication Publication Date Title
CN114896900B (en) a target tracking system
CN107146004B (en) A kind of slag milling system health status identifying system and method based on data mining
CN109396576B (en) Deep learning-based electric spark machining stability and energy consumption state optimization decision system and decision method
US11105758B2 (en) Prediction method for mold breakout based on feature vectors and hierarchical clustering
CN106980729B (en) A kind of continuous casting breakout prediction method based on mixed model
CN116644373B (en) Automobile flow data analysis management system based on artificial intelligence
CN101739814A (en) SCATS coil data-based traffic state online quantitative evaluation and prediction method
CN108469180A (en) The method for building up of sintering end point forecasting system based on big data and machine learning
CN108436050B (en) A Method for Predicting Mold Breakout in Continuous Casting Using Spatial Density Clustering DBSCAN
CN107239066B (en) A kind of Vertical Mill operation closed-loop control device and method based on data mining
CN106990763A (en) A kind of Vertical Mill operation regulator control system and method based on data mining
CN109647899B (en) A method for predicting power consumption of multi-specification rolling stock in the process of hot strip finishing rolling
CN102592451B (en) Method for detecting road traffic incident based on double-section annular coil detector
CN110210730A (en) A kind of regional complex energy resource system energy efficiency state evaluation method
CN116227752B (en) Park facility management system based on Internet of things
Zhang et al. A study on the method for cleaning and repairing the probe vehicle data
CN111259307A (en) Method for predicting bulging deformation of continuous casting billet by using Hilbert-Huang transform
CN118070683B (en) Logistics and transportation big data collection method and system based on digital twin
CN110310491B (en) Short-distance double-node geomagnetic vehicle speed detection system and detection method
CN106845817A (en) Online strengthening learns transaction system and method
CN113159374A (en) Data-driven urban traffic flow rate mode identification and real-time prediction early warning method
CN117422310A (en) Electric control decision optimization method for all-steel inner liner linkage line group
CN103834758A (en) Blast furnace materiel level real-time detection method with continuous high precision
CN119337300B (en) A CNC machine tool health monitoring method, system and program product based on multi-source data fusion
CN107357994A (en) A kind of aero-engine performance decline mode excavation method stage by stage

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant