Feedback control method with furnace separation self-learning and multi-stage laminar cooling

Abstract

The invention belongs to the technical field of hot continuous rolling production, and particularly relates to a furnace-separated self-learning and multi-stage laminar cooling feedback control method, which comprises the following steps of: distinguishing calculated values of heat transfer coefficients of different slabs during cooling setting according to identification codes; independently calculating a coiling temperature CT target value for the plate blanks of different heating furnaces; in the laminar cooling stage, a multi-stage feedback control strategy is adopted to adjust and finely adjust the water volume of a collecting pipe: when a strip steel head heat exchange coefficient learning point reaches a coiling temperature CT detection position, actual coiling temperature CT data of a first group of samples are collected, and the average coiling temperature CT is compared with a coiling temperature CT target value; adjusting the water quantity of the fine adjustment header for the first time according to the difference value; and performing feedback regulation on the next group of samples by using the deviation between the coiling temperature CT data of the group after feedback regulation and the coiling temperature CT target value until the tail part of the strip steel. The method has the advantage that the problem of influence of the heating furnace state on self-learning parameter copying is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of hot rolling production, and in particular relates to a control method with furnace self-learning and multi-stage laminar cooling feedback. Background Art

[0002] When producing high-grade pipeline steel, a low-temperature coiling process is often used to reduce alloy content and control production costs. However, this results in a large amount of intermittent iron oxide scale distributed on the surface of the strip, and even a large amount of water on the surface. This leads to large fluctuations in the surface temperature of the strip detected by the pyrometer, which in turn causes the feedback control system to misadjust and exacerbate the fluctuations in the coiling temperature (CT). At the same time, the steel-burning conditions of different heating furnaces vary significantly. The existing coiling temperature control model (CTC) only uses a single heat transfer coefficient for pre-setting and cannot distinguish between the operating conditions of different heating furnaces. During centralized production scheduling, the cooling setting for continuous steel tapping fluctuates, and the self-learning model has poor continuity. In addition, due to the temperature deviation between the front and rear coils, there is a significant difference in the waiting time, which can cause large deviations across the entire coil during laminar cooling and watering.

[0003] In the prior art, patent application number CN202410422849.1 discloses a control method for improving the temperature hit rate of hot-rolled pipeline steel coiling. It adopts a front-end centralized cooling process without feedback control, optimizes the speed advance rate and lag rate of the laminar cooling area, and achieves a matching of the running speed of the strip in the laminar cooling area with the speed set by the mathematical model. It can avoid the deviation of the pipeline steel coiling temperature from the target value due to the difference in the heating temperature of the slab, improve the prediction accuracy of the mathematical model, improve the stability of the pipeline steel through-strip coiling temperature control, reduce the adverse effects of feedback cooling water on the coiling temperature control, and achieve rapid adjustment during the strip rolling process without affecting other steel grades. The feedback control is relatively rough, and the front-end centralized cooling without feedback is adopted. Although it reduces interference, it sacrifices the dynamic adjustment capability. Xi Xiaoling, Zhou Xinliang, and Wu Shengtian, "Calculation and Optimization of Self-Learning Parameters for a Hot Strip Mill Multi-Heating Furnace Setting Model," Metallurgical Automation, Issue 2, 2014. This paper proposes calculating the self-learning parameters for the hot strip mill setting model for each furnace, using exponential smoothing to smooth the oscillating parameters. Furthermore, different self-learning parameters are optimized using long and short genetic methods, enabling the process control system to adapt well to the heating furnaces. However, this approach does not correlate the cooling heat transfer coefficient. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide a method with furnace self-learning and multi-stage laminar cooling feedback control, which adopts a furnace self-learning mode for the heat transfer coefficient of the strip, and uses multi-stage feedback control to adjust it during the laminar cooling process, so as to improve the CT hit rate of thick-gauge steel plates under low-temperature coiling process conditions.

[0005] To achieve the above object, the present invention is implemented through the following technical solutions:

[0006] A method for controlling furnace self-learning and multi-stage laminar cooling feedback includes the following steps:

[0007] (1) Identify the heating furnace number and slab according to the identification code, and distinguish the calculated values ​​of the heat transfer coefficient of different slabs at the cooling setting; enable the CTC self-learning model to distinguish the slabs according to the identification code;

[0008] (2) Calculate the coiling temperature CT target value for the slabs in different heating furnaces independently and store it in the CTC self-learning model;

[0009] (3) During the laminar cooling stage, a multi-stage feedback control strategy is used to adjust the water volume in the fine-tuning header, specifically including:

[0010] S1: When the learning point of the heat transfer coefficient of the strip head reaches the coiling temperature CT detection position, the actual coiling temperature CT data of the first group of samples are collected, and the average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value. The first water adjustment of the fine-tuning header is performed based on the difference between the two;

[0011] S2: When the sample at the entrance of the fine adjustment area during the first adjustment reaches the coiling temperature CT detection position, the actual coiling temperature CT data of the second group of samples with the same number as the first group of samples are collected. The average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value to determine whether the first feedback adjustment is under-adjusted or over-adjusted, and the second water adjustment is performed on the fine adjustment header;

[0012] By analogy, repeat S1 and S2, or the S2 adjustment process, and use the deviation between the coiling temperature CT data of this group of samples after feedback adjustment and the coiling temperature CT target value to perform feedback adjustment on the next group of samples until the tail of the strip.

[0013] According to the actual coiling temperature CT control curve characteristics of the steel strip, the number of sample groups is set to 5-7.

[0014] The thickness of the steel strip is 15.5-18.5 mm.

[0015] The length of each group of samples is adaptively adjusted according to the speed of the steel belt.

[0016] The mean filtering adopts a sliding window averaging method, and the window width matches the number of sample groups.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] The present application assigns heat exchange self-learning parameters according to different heating furnace working conditions, can significantly distinguish the interference on self-learning trend caused by the difference of different heating furnace working conditions, solves the influence of heating furnace state on self-learning parameter copying. Eliminate the influence of heating furnace capacity difference on CT control, improve the model adaptability. Adopt multi-stage feedback control algorithm, through segmented iteration adjustment and mean filtering algorithm, suppress CT fluctuation, realize high precision control, solve the problem of high or low whole roll temperature between rolls due to different temperature waiting time after learning function is put into operation, balance control precision and production efficiency.

[0019] The present application effectively solves the problems of low CT hit rate and large temperature fluctuation in the low temperature coiling process of 15.5~18.5mm thick steel through the synergistic effect of furnace self-learning and multi-stage feedback, and is suitable for efficient and stable production of hot rolling production line. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 It is a flow chart of the furnace self-learning and multi-stage laminar cooling feedback control method.

[0021] Figure 2 It is a 1# heating furnace CT actual curve graph of 5 sample sections.

[0022] Figure 3 It is a 2# heating furnace CT actual curve graph of 5 sample sections.

[0023] Figure 4 It is a 3# heating furnace CT actual curve graph of 5 sample sections.

[0024] Figure 5 It is a 1# heating furnace CT actual curve graph of 6 sample sections.

[0025] Figure 6 It is a 2# heating furnace CT actual curve graph of 6 sample sections.

[0026] Figure 7 It is a 3# heating furnace CT actual curve graph of 6 sample sections.

[0027] Figure 8 It is a 1# heating furnace CT actual curve graph of 7 sample sections.

[0028] Figure 9 It is a 2# heating furnace CT actual curve graph of 7 sample sections.

[0029] Figure 10 It is a 3# heating furnace CT actual curve graph of 7 sample sections. DETAILED DESCRIPTION

[0030] The present invention will be described in detail below with reference to the accompanying drawings, but it should be noted that the implementation of the present invention is not limited to the following embodiments.

[0031] Example:

[0032] See Figure 1 , with furnace self-learning and multi-stage laminar cooling feedback control method, including the following steps:

[0033] (1) The three walking beam slab heating furnaces (1#, 2#, 3#) and slabs of the 2150 line of a steel plant were independently identified by identification codes, and the calculated values ​​of the heat transfer coefficients of different slabs at cooling settings were distinguished, so that the CTC self-learning model could distinguish the slabs according to the identification codes;

[0034] (2) The CTC self-learning model calculates and stores the heat transfer coefficient of the slabs discharged from each heating furnace when executing the cooling setting based on the identification code and the coiling temperature CT target value, replacing the traditional single cooling heat transfer coefficient and solving the CT fluctuation problem caused by the deviation of the steelmaking capacity of the three heating furnaces.

[0035] (3) During the laminar cooling stage, a multi-stage feedback control strategy is used to adjust the water volume in the fine-tuning header, specifically including:

[0036] First adjustment: When the strip head heat transfer coefficient learning point reaches the coiling temperature CT detection position, the actual coiling temperature CT data of the first group of samples is collected. The average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value. The first water adjustment is made to the fine-tuning header based on the difference between the two values.

[0037] Iterative adjustment: When the samples located at the entrance of the fine-tuning area during the first adjustment reach the coiling temperature CT detection position, the actual coiling temperature CT data of the second group of samples, the same number as the first group of samples, are collected. The average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value to determine whether the first feedback adjustment is under-adjusted or over-adjusted, and a second water adjustment is performed on the fine-tuning header. The mean filtering adopts the sliding window averaging method, and the window width matches the number of sample groups.

[0038] Similarly, during the entire coiling process, after the strip reaches the coiling temperature CT detection position, the above iterative adjustment process is repeated, and the feedback deviation of the previous group of samples is used to adjust the next group until the tail of the strip is cooled.

[0039] Based on the actual coiling temperature CT control curve characteristics of high-grade pipeline steel strip, the number of sample groups is set to 5-7. The length of each sample group is adaptively adjusted based on the steel strip speed to ensure a balance between filtering effectiveness and real-time performance. The thickness of high-grade pipeline steel plate ranges from 15.5 to 18.5 mm.

[0040] By comparing the self-learning effects of 5, 6 and 7 groups of samples (see Table 1-3 and Figure 2-Figure 10 ), verifying the effectiveness of the method. Table 1 shows the heat transfer coefficient and CT hit rate of five samples in the separate furnace self-learning mode, Table 2 shows the heat transfer coefficient and CT hit rate of six samples in the separate furnace self-learning mode, and Table 3 shows the heat transfer coefficient and CT hit rate of six samples in the separate furnace self-learning mode. Figure 2-Figure 4 , are the CT curves of 5 sample sections with different heating furnaces, Figure 5-Figure 7 CT curves of 6 sample sections with different heating furnaces, Figures 8-10 CT curves of 7 sample segments in different heating furnaces.

[0041] Table 1: Heat transfer coefficient and CT hit rate of five sample sections by furnace self-learning

[0042]

[0043] Table 2: Heat transfer coefficient and CT hit rate of six sample sections by furnace self-learning

[0044]

[0045] Table 3: Heat transfer coefficient and CT hit rate of 7 sample sections by furnace self-learning

[0046]

[0047] From the self-learning effects of the above 5, 6, and 7 groups of samples, we can see that:

[0048] In the separate furnace mode, the heat transfer coefficient of each heating furnace is calculated independently (e.g. 0.781~0.955 for furnace 1, s 2 (0.00033~0.004268), 2# furnace 0.80~0.969, s 2 (0.000153~0.002544), 3# furnace 0.749~0.95, s 2 (0.000146-0.003468)). Changes in the heat transfer coefficient of any heating furnace do not affect the heat transfer coefficients of other heating furnaces, significantly reducing the deviation caused by traditional uniform coefficients. Under multi-stage feedback control, the CT hit rate is stable at 70%-80% (for example, the highest was 80% for furnace 1 and 80% for furnace 3 in the seven sample groups), a 10%-15% improvement over traditional single-stage feedback. Figure 2-Figure 10 It shows that the fluctuation amplitude of CT curves of different heating furnaces is reduced, and the stability of low-temperature coiling process is significantly enhanced.

Claims

1. A method for controlling furnace self-learning and multi-stage laminar cooling feedback, characterized in that: The following steps are involved: (1) Identify the heating furnace number and slab according to the identification code, and distinguish the calculated values ​​of the heat transfer coefficient of different slabs at the cooling setting; Enable the CTC self-learning model to distinguish slabs based on identification codes; (2) Calculate the coiling temperature CT target value for the slabs in different heating furnaces independently and store it in the CTC self-learning model; (3) During the laminar cooling stage, a multi-stage feedback control strategy is used to adjust the water volume in the fine-tuning header, specifically including: S1: When the learning point of the heat transfer coefficient of the strip head reaches the coiling temperature CT detection position, the actual coiling temperature CT data of the first group of samples are collected, and the average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value. The first water adjustment of the fine-tuning header is performed based on the difference between the two; S2: When the sample at the entrance of the fine adjustment area during the first adjustment reaches the coiling temperature CT detection position, the actual coiling temperature CT data of the second group of samples with the same number as the first group of samples are collected. The average coiling temperature CT after mean filtering is compared with the coiling temperature CT target value to determine whether the first feedback adjustment is under-adjusted or over-adjusted, and the second water adjustment is performed on the fine adjustment header; Repeat S1 and S2, or the S2 adjustment process, and use the deviation between the coiling temperature CT data of the sample after feedback adjustment and the coiling temperature CT target value to perform feedback adjustment on the next group of samples until the end of the strip; Repeat the water adjustment twice for each group of samples. The data of the strip coiling temperature changing with time is divided into several groups of sample segments. The water volume adjustment is repeated twice for each sample segment. The deviation between the sample coiling temperature CT data of the first group of data after feedback adjustment and the coiling temperature CT target value in the two adjacent groups of sample segments is used to perform feedback adjustment on the latter group of samples until the tail of the strip.

2. The method according to claim 1, characterized in that: According to the actual coiling temperature CT control curve characteristics of the steel strip, the number of sample groups is set to 5-7.

3. The method of controlling the self-learning and multi-stage laminar cooling feedback of furnaces according to claim 2, characterized in that: The thickness of the steel strip is 15.5-18.5 mm.

4. The method of claim 1, wherein the method comprises: The length of each group of samples is adaptively adjusted according to the speed of the steel belt.

5. The method of claim 1, wherein the method comprises: The mean filtering adopts a sliding window averaging method, and the window width matches the number of sample groups.

Images