TW202211091A - Blast furnace condition determination device, operation method of blast furnace, and manufacturing method for molten iron - Google Patents
Blast furnace condition determination device, operation method of blast furnace, and manufacturing method for molten iron Download PDFInfo
- Publication number
- TW202211091A TW202211091A TW110130616A TW110130616A TW202211091A TW 202211091 A TW202211091 A TW 202211091A TW 110130616 A TW110130616 A TW 110130616A TW 110130616 A TW110130616 A TW 110130616A TW 202211091 A TW202211091 A TW 202211091A
- Authority
- TW
- Taiwan
- Prior art keywords
- blast furnace
- temperature distribution
- learning model
- determination device
- unit
- Prior art date
Links
Images
Classifications
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B5/00—Making pig-iron in the blast furnace
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B7/00—Blast furnaces
- C21B7/24—Test rods or other checking devices
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Manufacturing & Machinery (AREA)
- Materials Engineering (AREA)
- Metallurgy (AREA)
- Organic Chemistry (AREA)
- Manufacture Of Iron (AREA)
- Blast Furnaces (AREA)
- Waste-Gas Treatment And Other Accessory Devices For Furnaces (AREA)
Abstract
Description
本發明係關於高爐爐況狀態判定裝置、高爐之作業方法以及鐵水之製造方法。The present invention relates to a blast furnace state determination device, a blast furnace operation method, and a molten iron manufacturing method.
近年,會有在低焦炭比(CR)及高粉煤吹入(PCI)流量下進行高爐作業的情況。在這樣的作業環境下,儘早且準確地掌握高爐的爐況是重要的。因為高爐呈圓筒形,當發生了原料裝入之圓周方向的偏差、反應之圓周方向的偏差等的情況,有出鐵狀態在圓周方向變得不均一而使作業狀態惡化的疑慮。因此,特別是對圓周方向的偏差儘早且準確偵測是重要。In recent years, blast furnace operations have been performed at low coke ratio (CR) and high pulverized coal injection (PCI) flow rates. Under such an operating environment, it is important to grasp the furnace conditions of the blast furnace as early as possible and accurately. Since the blast furnace has a cylindrical shape, if there is a deviation in the circumferential direction of the charging of raw materials, deviation in the circumferential direction of the reaction, etc., there is a possibility that the tapping state will become uneven in the circumferential direction, and the working state may be deteriorated. Therefore, it is important to detect the deviation in the circumferential direction early and accurately.
作為偵測圓周方向的偏差的方法之一,係計測原料裝入面之正上方的溫度分布之方法。已知當發生了原料裝入的偏差及反應的偏差之至少一方的情況,會顯現溫度分布的偏差。以往,原料裝入面之正上方的溫度分布,大多是藉由插入爐內的熱電偶進行計測。為了避免妨礙原料裝入作業,熱電偶通常是從4個方向插入,藉此計測各熱電偶的半徑方向上之複數點的溫度。但因為僅限於4個方向,難以掌握整體的溫度分布。又具有耐久性的熱電偶大多響應時間較長。因此,要對應於利用高速迴旋的原料滑槽之原料裝入來掌握氣體的溫度變化是困難的。As one of the methods of detecting the deviation in the circumferential direction, it is a method of measuring the temperature distribution just above the raw material charging surface. It is known that when at least one of the variation in the charging of the raw materials and the variation in the reaction occurs, the variation in the temperature distribution occurs. Conventionally, the temperature distribution just above the raw material charging surface was often measured by a thermocouple inserted into the furnace. In order to avoid interfering with the work of loading raw materials, thermocouples are usually inserted from four directions, and the temperature of plural points in the radial direction of each thermocouple is measured. However, since it is limited to four directions, it is difficult to grasp the overall temperature distribution. Most of the thermocouples with durability have a longer response time. Therefore, it is difficult to grasp the temperature change of the gas in response to the charging of the raw material by the high-speed rotating raw material chute.
在此,因為近年的技術提升,以較高的取樣周期收集原料裝入面之正上方的溫度分布變成可能。藉此,要將原料裝入面之正上方的溫度分布大致即時地可視化、以及要掌握溫度分布之偏差等的異常變成可能。然而,要始終派人監視溫度分布的畫面等之可視化資訊是困難的。又當人為判斷是異常的情況,其判斷基準是屬於個人的,要始終進行適切的判斷也是困難的。針對此,專利文獻1揭示出,藉由將溫度的不均一指標化來進行適切的判定之高爐爐況狀態判定裝置。 [先前技術文獻] [專利文獻]Here, due to recent technological improvements, it has become possible to collect the temperature distribution just above the raw material charging surface at a relatively high sampling cycle. Thereby, it becomes possible to visualize the temperature distribution just above the raw material charging surface in real time, and to grasp the abnormality such as the deviation of the temperature distribution. However, it is difficult to always send people to monitor the visual information such as the temperature distribution screen. Also, when human judgment is abnormal, the standard of judgment belongs to the individual, and it is difficult to always make appropriate judgments. In view of this, Patent Document 1 discloses a blast furnace state determination device that performs appropriate determination by indexing the nonuniformity of temperature. [Prior Art Literature] [Patent Literature]
專利文獻1:日本特開2018-165399號公報Patent Document 1: Japanese Patent Laid-Open No. 2018-165399
[發明所欲解決之問題][Problems to be Solved by Invention]
在此,當判定為高爐爐況狀態異常的情況,宜為了消除異常而控制高爐的作業。然而,用於消除異常之高爐的控制內容會依溫度分布的紊亂狀態而有差異。因此要求一種技術,除了偵測異常以外,不須人為的介入就能將溫度分布的紊亂狀態進行判定並分類。Here, when it is determined that the state of the blast furnace is abnormal, it is appropriate to control the operation of the blast furnace in order to eliminate the abnormality. However, the control content of the blast furnace for eliminating the abnormality varies depending on the disordered state of the temperature distribution. Therefore, there is a need for a technology that can determine and classify the disordered state of temperature distribution without human intervention in addition to detecting anomalies.
為了解決以上的問題之本發明的目的是為了提供一種高爐爐況狀態判定裝置,可將高爐之溫度分布的紊亂狀態自動判定並分類。又本發明的其他目的是為了提供一種高爐之作業方法以及鐵水之製造方法,能根據高爐爐況狀態判定裝置之判定和分類的結果來消除高爐之溫度分布的紊亂,而將作業的自動化比率提高。 [解決問題之技術手段]An object of the present invention to solve the above problems is to provide a blast furnace furnace condition state determination device that can automatically determine and classify the disturbance state of the temperature distribution of the blast furnace. Another object of the present invention is to provide a blast furnace operation method and a molten iron manufacturing method, which can eliminate the disturbance of the temperature distribution of the blast furnace according to the results of the determination and classification of the blast furnace state determination device, and reduce the automation ratio of the operation. improve. [Technical means to solve problems]
本發明的一實施形態之高爐爐況狀態判定裝置,係判定高爐的爐況狀態之高爐爐況狀態判定裝置,其係具備溫度分布計測部、儲存部、判定部;前述溫度分布計測部,係計測前述高爐中之原料裝入面之正上方的溫度分布;前述儲存部,係根據所計測之前述溫度分布的資料,儲存以對於前述溫度分布的模式輸出複數個類型的方式進行學習後的學習模型;前述判定部,係使用所儲存之前述學習模型,判定前述溫度分布計測部所計測之前述溫度分布是前述複數個類型之哪一個。A blast furnace state determination device according to an embodiment of the present invention is a blast furnace state determination device for determining the furnace state of a blast furnace, and includes a temperature distribution measurement unit, a storage unit, and a determination unit; the temperature distribution measurement unit is a Measure the temperature distribution just above the raw material charging surface in the blast furnace; the storage unit stores the learning after learning by outputting a plurality of patterns for the temperature distribution based on the measured data of the temperature distribution The model; the determining unit uses the stored learning model to determine which of the plurality of types the temperature distribution measured by the temperature distribution measuring unit is.
本發明的一實施形態之高爐之作業方法,係因應藉由上述高爐爐況狀態判定裝置所判定之高爐的爐況狀態來變更作業條件。The blast furnace operation method according to one embodiment of the present invention changes the operation conditions according to the blast furnace condition state determined by the blast furnace condition state determination device.
本發明的一實施形態之鐵水之製造方法,係使用藉由上述高爐之作業方法進行作業的高爐來製造鐵水。 [發明之效果]The manufacturing method of molten iron which concerns on one Embodiment of this invention manufactures molten iron using the blast furnace operated by the operation method of the said blast furnace. [Effect of invention]
依據本發明,能夠提供可將高爐之溫度分布的紊亂狀態自動判定並分類的高爐爐況狀態判定裝置。又依據本發明能夠提供一種高爐之作業方法及鐵水之製造方法,其可根據高爐爐況狀態判定裝置之判定和分類的結果來將高爐之溫度分布的紊亂消除,藉此將作業的自動化比率提高。According to the present invention, it is possible to provide a blast furnace state determination device capable of automatically determining and classifying the disordered state of the temperature distribution of the blast furnace. Furthermore, according to the present invention, it is possible to provide a blast furnace operation method and a molten iron manufacturing method, which can eliminate the disturbance of the temperature distribution of the blast furnace according to the results of the determination and classification of the blast furnace state determination device, thereby reducing the automation ratio of the operation. improve.
以下說明運用本發明來判定高爐的爐況狀態之高爐爐況狀態判定裝置1、高爐之作業方法以及鐵水之製造方法的一實施形態。Hereinafter, one embodiment of the blast furnace furnace state determination device 1 for determining the furnace state state of a blast furnace, a blast furnace operation method, and a molten iron manufacturing method will be described using the present invention.
(裝置的構成)
如圖1所示般,本實施形態的高爐爐況狀態判定裝置1係具備:溫度分布計測部2、儲存部3、判定部5、顯示部6。又高爐爐況狀態判定裝置1是構成為可與學習模型生成裝置10進行通訊,藉此在其和學習模型生成裝置10之間執行後述學習模型及學習用資料的輸入輸出。在本實施形態中,學習模型生成裝置10具備學習模型生成部4。(Configuration of the device)
As shown in FIG. 1 , the blast furnace state determination device 1 of the present embodiment includes a temperature
溫度分布計測部2係計測位於原料裝入面的正上方之空間的溫度分布,該原料裝入面,是在使用鐵礦石作為原料來生產生鐵之高爐的內部,由從高爐的頂部裝入之原料所構成的層之最上表面。溫度分布計測部2,可以是將高爐中之原料裝入面之正上方的溫度分布高速(例如10秒以內)地計測之各種感測器或裝置。在本實施形態中,溫度分布計測部2係利用音速的溫度相依性之溫度計測裝置。溫度分布計測部2,例如是在高爐爐頂的圓周上設置8個收發機,朝向原料裝入面進行音波的發送接收,根據從音波的發送到接收為止的時間來測定溫度。The temperature
儲存部3,係根據所計測之溫度分布的資料,儲存以對於溫度分布的模式輸出複數個類型的方式進行學習後之學習模型。學習模型,係由學習模型生成部4所生成(所學習),在學習完畢後儲存於儲存部3。又儲存部3係儲存藉由溫度分布計測部2所計測之溫度分布的資料。藉由溫度分布計測部2即時計測高爐中之原料裝入面的正上方之溫度分布,儲存部3蓄積溫度分布的資料。當學習模型生成部4生成學習模型的情況,係使用儲存部3所蓄積之溫度分布的資料來作為學習用資料。又當判定部5判定高爐之爐況狀態的情況,判定部5是透過儲存部3而取得來自溫度分布計測部2之即時之溫度分布的資料。又當判定部5判定高爐之爐況狀態的情況,是藉由判定部5取得儲存部3所儲存之學習模型。The
學習模型生成部4,係根據儲存部3所儲存之溫度分布的資料,來生成輸出溫度分布的模式之複數個類型之學習模型。類型,係將溫度分布的模式分類之集合或群組。在本實施形態所使用之類型的詳細情形隨後敘述。The learning model generation unit 4 generates a plurality of types of learning models for outputting patterns of temperature distribution based on the temperature distribution data stored in the
判定部5,係使用儲存部3所儲存的學習模型(學習完畢模型)來判定高爐的爐況狀態。在本實施形態中,為了偵測高爐之爐況狀態的異常,判定部5係判定溫度分布計測部2所計測的溫度分布是複數個類型之哪一個。在本實施形態中,複數個類型包含:將異常溫度分布的模式分類之至少2個類型。亦即,判定部5在取得異常溫度分布的資料的情況,不是單純地判定異常,而是按照模式來判定被分類成複數個類型之哪一個。在此,本實施形態中之異常溫度分布的模式係至少包含:起因於高爐之圓周方向上之原料裝入的偏差及反應的偏差而可能產生之溫度分布的模式。The
顯示部6係將判定部5的判定結果透過圖像提示操作者。顯示部6可進一步具備輸出聲音的功能,而連同圖像將警報一起朝操作者發送。操作者例如是在高爐進行作業的作業者。藉由高爐爐況狀態判定裝置1所判定之高爐的爐況狀態,可運用在高爐之作業方法中變更作業條件。根據顯示部6所顯示的圖像而得知爐況異常的操作者,為了抑制高爐之圓周方向上之原料裝入的偏差及反應的偏差而變更高爐的作業條件。高爐之作業條件的變更,可包含例如焦炭比的變更。高爐之作業條件的變更,可包含例如送風流量的變更。又例如在高爐之圓周方向上如何將原料裝入之裝入物控制模式事先設定成複數種模式的情況,高爐之作業條件的變更可包含裝入物控制模式的變更。又上述高爐的作業,可作為製造鐵水之製造方法的一部分來執行。在高爐中,作為原料的鐵礦石被熔解、還原而成為生鐵,以鐵水的狀態出鐵。從高爐出鐵後的鐵水,藉由鐵水預備處理工序將硫、磷等的雜質除去。進而在轉爐進行精煉而將碳除去。The
高爐爐況狀態判定裝置1之儲存部3、判定部5及顯示部6,可藉由取得從溫度分布計測部2計測之溫度分布的資料之電腦來實現。電腦係具備:例如記憶體(儲存裝置)、CPU(處理裝置)、硬碟機(HDD)、控制顯示器等的顯示裝置之顯示控制部。作業系統(OS)及用於實施各種處理的應用程式,可儲存於硬碟機,在藉由CPU執行時從硬碟機載入記憶體。因應必要,CPU控制顯示控制部而讓顯示器顯示必要的圖像。又關於處理中的資料,是儲存於記憶體,如果必要的話可儲存於HDD。各種功能,是藉由讓CPU、記憶體等的硬體和OS及必要的應用程式有機地協作而實現。儲存部3可由例如記憶體及硬碟機來實現。判定部5可由例如CPU來實現。又顯示部6可由例如顯示控制部及顯示器來實現。The
學習模型生成裝置10之學習模型生成部4,可由與高爐爐況狀態判定裝置1不同的電腦來實現。學習模型生成部4可由例如CPU來實現。在本實施形態中,學習完畢模型是如上述般儲存於儲存部3,作為其他例子,可儲存於學習模型生成裝置10之記憶體或硬碟機。在此情況,判定部5要判定高爐的爐況狀態時,可對學習模型生成裝置10之記憶體或硬碟機進行存取(access)而讀取學習完畢模型。又在本實施形態中,學習用資料是如上述般儲存於儲存部3,作為其他例子,可儲存於學習模型生成裝置10之記憶體或硬碟機。在此情況,學習模型生成部4可對學習模型生成裝置10之記憶體或硬碟機進行存取而讀取學習用資料。The learning model generating unit 4 of the learning
在此,圖1之高爐爐況狀態判定裝置1的構成只是一例,可以不包含構成要素的一部分。又高爐爐況狀態判定裝置1可以具備其他構成要素。例如,高爐爐況狀態判定裝置1可省略顯示部6而構成。這時,高爐爐況狀態判定裝置1可具備:將判定部5的判定結果經由網路輸出到操作者的終端裝置之通訊部。Here, the configuration of the blast furnace state determination device 1 shown in FIG. 1 is merely an example, and some of the constituent elements may not be included. Further, the blast furnace state determination device 1 may include other components. For example, the blast furnace state determination device 1 may be configured by omitting the
(高爐爐況狀態判定方法) 圖2係顯示高爐爐況狀態判定裝置1所執行之高爐爐況狀態判定方法的大致流程圖。高爐爐況狀態判定方法分成學習、判定及顯示的工序。圖2中,線上(on-line)是表示作為高爐作業的一部分所執行的處理。相反的,離線(off-line)則表示和高爐作業分開執行的處理。(Blast Furnace Condition State Judgment Method) FIG. 2 is a schematic flow chart showing a method for determining the state of a blast furnace state executed by the blast furnace state determination device 1 . The blast furnace state determination method is divided into the steps of learning, determination and display. In FIG. 2 , on-line indicates processing performed as a part of blast furnace operation. In contrast, off-line refers to processing performed separately from the blast furnace operation.
高爐爐況狀態判定裝置1,以離線方式,讓學習模型生成裝置10生成學習模型。例如以來自高爐爐況狀態判定裝置1的指令作為觸發(trigger),首先學習模型生成部4係取得從儲存部3所蓄積之溫度分布的資料中選擇者作為賦予標籤(labeled)圖像資料。學習模型生成部4係使用賦予標籤圖像資料來生成學習模型(步驟S1)。關於學習模型及學習的詳細情形隨後敘述。作為其他例子,學習模型生成裝置10也可以不等待來自高爐爐況狀態判定裝置1的指令就開始學習模型的生成。亦即,只要在判定部5判定爐況狀態之前將學習完畢的模型儲存於儲存部3,能以學習模型生成裝置10為主體而執行步驟S1。The blast furnace state determination device 1 allows the learning
判定部5從儲存部3讀取所儲存的學習模型。又判定部5是透過儲存部3取得來自溫度分布計測部2之即時的溫度分布的資料。判定部5,以線上方式,將即時的溫度分布之圖像資料輸入學習模型,判定高爐的爐況狀態(步驟S2)。在本實施形態中,判定部5,作為爐況狀態的判定,係判定即時的溫度分布是否被分類成異常溫度分布的模式之類型。The
然後,顯示部6將判定部5的判定結果以圖像顯示(步驟S3)。圖像只要是能讓操作者掌握是否被分類成異常溫度分布的模式之類型即可,並不限定於特定形式。根據圖像得知爐況異常的操作者,可將高爐的作業條件變更。Then, the
(學習模型)
已知當原料被正常裝入的情況,亦即原料裝入之偏差不存在的情況,原料裝入面之正上方的溫度分布成為同心圓狀的分布。又已知當反應之偏差不存在的情況也是成為同心圓狀的分布。相反的,當有原料之偏差存在、或原料下降速度不均一而有原料裝入之偏差存在的情況,溫度分布會出現紊亂而無法成為同心圓狀。因此,將溫度分布的紊亂予以圖像化並分類,由高爐爐況狀態判定裝置1判定即時之溫度分布的圖像是屬於哪一類紊亂,藉此可達成不須人為介入之爐況的自動判定。在本實施形態中,判定所使用的學習模型,是以從儲存部3所蓄積之溫度分布的資料中選擇之具有紊亂的溫度分布者作為學習用資料而進行學習,藉此所生成。(Learning Model)
It is known that when the raw materials are normally charged, that is, when there is no variation in the raw material charging, the temperature distribution immediately above the raw material charging surface becomes a concentric distribution. It is also known that when there is no variation in the response, the distribution becomes concentric. Conversely, when there is a variation in the raw materials, or when the falling speed of the raw materials is not uniform and there is a deviation in the charging of the raw materials, the temperature distribution will be disordered and cannot be concentric. Therefore, the disturbance of the temperature distribution is imaged and classified, and the blast furnace condition determination device 1 determines which type of disturbance the image of the immediate temperature distribution belongs to, thereby realizing automatic determination of the furnace condition without human intervention. . In the present embodiment, the learning model to be used for determination is generated by learning a person having a turbulent temperature distribution selected from the temperature distribution data stored in the
作為學習用資料所選擇之溫度分布的資料,當被圖像化的情況,係包含以下所說明之具有異常溫度分布的模式者。以下所說明之異常溫度分布的模式,在高爐作業中有引發異常的可能性是已知的。When the data of the temperature distribution selected as the learning material is imaged, it includes a pattern having an abnormal temperature distribution described below. The pattern of abnormal temperature distribution described below is known to cause abnormality in blast furnace operation.
在第1異常溫度分布的模式,高溫部分偏離原料裝入面之中心。因為高爐呈圓筒形,原料裝入面的形狀為圓形,該模式是高溫部分不包含其中心的模式。高爐,為了讓作業穩定化,通氣是重要的。因此,是以使氣體容易通過高爐之中心部的方式讓原料分布。在正常的情況,原料裝入面之正上方的溫度分布是中心部的溫度較高。第1異常溫度分布的模式,可能因原料裝入之分布的偏差而產生。例如可藉由修正原料裝入的分布,讓高爐的作業正常化。以下,將被分類為第1異常溫度分布的模式之溫度分布的模式所屬之類型稱為第1類型。In the pattern of the first abnormal temperature distribution, the high temperature part is deviated from the center of the raw material charging surface. Since the blast furnace has a cylindrical shape, the shape of the raw material charging surface is circular, and this mode is a mode in which the high temperature portion does not include its center. In blast furnaces, ventilation is important in order to stabilize the operation. Therefore, the raw material is distributed so that the gas can easily pass through the center of the blast furnace. Under normal circumstances, the temperature distribution just above the raw material loading surface is that the temperature in the center is higher. The pattern of the first abnormal temperature distribution may be caused by the variation in the distribution of the charging of the raw materials. For example, the operation of the blast furnace can be normalized by correcting the distribution of the charging of raw materials. Hereinafter, the type to which the pattern of the temperature distribution classified as the pattern of the first abnormal temperature distribution belongs is referred to as the first type.
在第2異常溫度分布的模式,高溫部分從原料裝入面的中心綿延到周邊。在該模式,高溫部分雖包含原料裝入面的中心,但該高溫部分綿延到原料裝入面的周邊(圓周部分),其溫度分布並非同心圓狀。如上述說明般,在正常的情況,因為中心部的溫度較高,中間部(中心部和周邊之間)相對地溫度變低。第2異常溫度分布的模式,應是因原料裝入之偏差及反應之偏差之至少1方所產生的。以下,將被分類為第2異常溫度分布的模式之溫度分布的模式所屬之類型稱為第2類型。In the pattern of the second abnormal temperature distribution, the high temperature portion extends from the center to the periphery of the raw material charging surface. In this mode, the high temperature portion includes the center of the raw material charging surface, but the high temperature portion extends to the periphery (circumferential portion) of the raw material charging surface, and the temperature distribution is not concentric. As described above, in a normal case, since the temperature of the central portion is high, the temperature of the intermediate portion (between the central portion and the periphery) is relatively low. The pattern of the second abnormal temperature distribution should be caused by at least one of the variation in the charging of the raw materials and the variation in the reaction. Hereinafter, the type to which the pattern of the temperature distribution classified as the pattern of the second abnormal temperature distribution belongs is referred to as the second type.
作為學習用資料所選擇之溫度分布的資料,如上述般進行「標籤賦予」。例如,可對於被分類為第1類型之溫度分布的圖像資料賦予標籤「1」,對於被分類為第2類型之溫度分布的圖像資料賦予標籤「2」。The data of the temperature distribution selected as the learning material is "labeled" as described above. For example, the label "1" may be assigned to the image data classified as the temperature distribution of the first type, and the label "2" may be assigned to the image data classified as the temperature distribution of the second type.
(學習) 圖3係顯示學習模型生成部4所進行之學習的流程圖。學習模型生成部4,係取得上述學習用資料,作成輸入資料進行學習而生成學習模型。在本實施形態中,學習模型生成部4,係採用深度學習的一手法、即卷積類神經網路(Convolutional neural network,以下稱為CNN)進行學習。CNN是可辨識圖像之圖像處理的手法。在此,學習模型生成部4之學習手法並不限定於CNN,亦可為可辨識模式圖像之其他手法。(study) FIG. 3 is a flowchart showing the learning performed by the learning model generation unit 4 . The learning model generation unit 4 acquires the above-mentioned learning materials, creates input materials, and performs learning to generate a learning model. In the present embodiment, the learning model generation unit 4 uses a method of deep learning, that is, a convolutional neural network (hereinafter referred to as CNN) to perform learning. CNN is an image processing technique for recognizable images. Here, the learning method of the learning model generation unit 4 is not limited to CNN, and may be other methods for recognizing pattern images.
在此,學習模型生成部4,係將作為學習用資料所選擇之溫度分布使用1個以上的閾值進行多值化後,作為圖像資料來使用。在本實施形態中,學習模型生成部4是使用2值化的圖像資料。如圖3所示般,在作為學習用資料之圖像的例子,閾值以上的溫度的部分對應於高溫部分而用黑色表示。閾值,作為一例是150℃,但並不限定於此。閾值可設定成:正常作業之高爐的原料裝入面之中心部的溫度和中間部的溫度之間的溫度。Here, the learning model generation unit 4 multivalues the temperature distribution selected as the learning data using one or more thresholds, and then uses it as the image data. In the present embodiment, the learning model generation unit 4 uses binarized image data. As shown in FIG. 3 , in an example of an image serving as a learning material, a portion with a temperature equal to or higher than a threshold value is shown in black corresponding to a high temperature portion. The threshold value is 150° C. as an example, but is not limited to this. The threshold value can be set as a temperature between the temperature of the center part and the temperature of the middle part of the raw material charging surface of the blast furnace in normal operation.
學習模型生成部4係將上述圖像資料分割成適當的尺寸(像素大小),對每個分割後的資料取平均,使用該資料作為CNN的輸入資料。學習模型生成部4,可在能掌握溫度分布之模式的範圍內,將分割之圖像尺寸增大。作為一例,學習模型生成部4可將圖像資料在縱方向及橫方向分別分割成10份。相反的,在實現學習模型生成部4之CPU的性能足夠高的情況,當基於CNN之學習可在現實的時間內執行時,學習模型生成部4可以不分割圖像資料而以像素單位作為輸入資料。The learning model generation unit 4 divides the above-mentioned image data into an appropriate size (pixel size), averages each divided data, and uses the data as the input data of the CNN. The learning model generation unit 4 can increase the size of the divided image within the range in which the pattern of the temperature distribution can be grasped. As an example, the learning model generation unit 4 may divide the image data into ten parts in the vertical direction and the horizontal direction, respectively. Conversely, in the case where the performance of the CPU for realizing the learning model generation unit 4 is sufficiently high, when the CNN-based learning can be performed in a realistic time, the learning model generation unit 4 may not divide the image data but use pixel units as input material.
圖3所示的CNN輸入層對應於上述輸入資料。又中間層對應於特徵量。又輸出層對應於溫度分布的模式被分類成複數個類型的哪一個之判定。學習模型生成部4是藉由這樣的手法來生成學習模型。學習模型生成部4可將所生成的學習模型儲存於儲存部3。在此,作為學習用資料所使用的圖像資料,係具有中心對稱的構造之高爐的原料裝入面之溫度分布。因此,以原料裝入面的中心為軸而讓圖像資料旋轉後的資料,也能運用於學習。亦即,學習模型生成部4,藉由使用原始的圖像資料及讓其旋轉後的圖像資料可增加輸入資料的數量,而進行有效率地學習。The CNN input layer shown in Figure 3 corresponds to the above input data. Again, the middle layer corresponds to the feature quantity. In addition, the output layer determines which of a plurality of types the pattern corresponding to the temperature distribution is classified into. The learning model generation unit 4 generates a learning model by such a method. The learning model generation unit 4 may store the generated learning model in the
(異常之判定)
判定部5係使用藉由學習模型生成部4所生成的學習模型來判定高爐的爐況狀態。在判定高爐的爐況狀態的情況,判定部5可從儲存部3讀取學習模型。又判定部5係透過儲存部3來取得來自溫度分布計測部2之即時的溫度分布的資料。判定部5是與學習模型生成部4同樣的,將所取得的溫度分布使用1個以上的閾值進行多值化後,作為圖像資料來使用。又判定部5是與學習模型生成部4同樣的,將圖像資料分割成適當的尺寸,對每個分割後的資料取平均,將該資料輸入學習模型。作為學習模型的輸出是得到判定結果。在本實施形態中,學習模型的輸出是異常溫度分布的模式之複數個類型(對應於第1類型之「1」及對應於第2類型之「2」)。例如即時的溫度分布是高溫部分偏離中心之模式的情況,學習模型輸出「1」。又例如即時的溫度分布是高溫部分從原料裝入面的中心綿延到周邊之模式的情況,學習模型輸出「2」。(judgment of abnormality)
The
顯示部6將判定部5的判定結果透過圖像提示操作者。顯示部6可顯示:例如溫度分布的圖像、和所屬之異常溫度分布的模式。例如當學習模型輸出「1」的情況,顯示部6可顯示溫度分布的圖像和「第1類型」。在此,對於判定為不屬於第1類型也不屬於第2類型的模式,顯示部6可顯示溫度分布的圖像和「第3類型」。The
操作者可根據判定結果進行高爐操作。操作者在例如第1類型(高溫部分偏離中心之模式)的情況,當可用不同的裝入物控制模式對應時,係進行變更裝入物控制模式的操作。操作者在例如第1類型以外的情況,能以使往後的作業穩定的方式實施操作條件的變更。使往後作業穩定之操作條件的變更,例如包括焦炭比的增加、PCI比的減少、送風流量的減少等。 實施例The operator can operate the blast furnace according to the judgment result. For example, in the case of the first type (mode in which the high temperature part is off-center), when a different load control mode is available, the operator performs an operation to change the load control mode. For example, in cases other than the first type, the operator can change the operating conditions so as to stabilize the subsequent work. Changes in operating conditions to stabilize subsequent operations include, for example, an increase in the coke ratio, a decrease in the PCI ratio, and a decrease in the air flow rate. Example
以下,使用實施例,將本發明具體地說明。溫度分布計測部2,係利用上述音速的溫度相依性之溫度計測裝置,且設置在用於生產生鐵之高爐。溫度分布的計測之取樣周期為10秒。Hereinafter, the present invention will be specifically described using examples. The temperature
在本實施例中,針對上述第1異常溫度分布的模式及第2異常溫度分布的模式進行偵測。在此,第1異常溫度分布的模式是高溫部分不包含中心的模式。又在第2異常溫度分布的模式,高溫部分雖包含原料裝入面的中心,但該高溫部分綿延到原料裝入面的周邊,其溫度分布並非同心圓狀。In the present embodiment, detection is performed for the pattern of the first abnormal temperature distribution and the pattern of the second abnormal temperature distribution. Here, the pattern of the first abnormal temperature distribution is a pattern in which the high temperature portion does not include the center. In the second abnormal temperature distribution pattern, although the high temperature portion includes the center of the raw material charging surface, the high temperature portion extends to the periphery of the raw material charging surface, and the temperature distribution is not concentric.
首先,對於這些異常溫度分布的模式,使用CNN進行學習。作為學習用資料,是從儲存部3所蓄積之溫度分布的資料中,用人選出屬於這些異常溫度分布的模式者各50個樣本左右。讓所選擇之溫度分布的資料之圖像資料逐次旋轉30°,將樣本數增加而進行學習。原始樣本和藉由旋轉所獲得的樣本合計數量約7000個。又在本實施例中,為了易於理解異常模式,係對溫度設定閾值而進行二值化。二值化的閾值設定為150℃。但二值化操作並非本發明之必要的處理。亦即,二值化可被省略。First, for these patterns of anomalous temperature distributions, a CNN is used for learning. As the learning data, about 50 samples each belonging to the patterns of these abnormal temperature distributions are selected by human from the data of temperature distribution stored in the
對於上述學習(CNN)所獲得的學習模型,使用和學習用資料不同的資料來驗證判定精度。驗證結果獲得90.3%的判定準確率。圖4例示該驗證的判定結果。For the learning model obtained by the above-mentioned learning (CNN), the determination accuracy is verified using data different from the learning data. The verification results obtained a judgment accuracy of 90.3%. FIG. 4 illustrates the determination result of this verification.
又確認了,可區別既不屬於第1類型也不屬於第2類型之溫度分布的模式(分類為第3類型)。It was also confirmed that the pattern of the temperature distribution which did not belong to either the first type nor the second type (classified as the third type) could be distinguished.
如以上般,基於本發明的高爐爐況狀態判定裝置1,可將高爐之溫度分布的紊亂狀態自動判定並進行分類。又在高爐之作業方法及鐵水之製造方法中,可根據高爐爐況狀態判定裝置1之判定和分類的結果來消除高爐之溫度分布的紊亂,而將作業的自動化比率提高。As described above, according to the blast furnace state determination device 1 of the present invention, the disturbance state of the temperature distribution of the blast furnace can be automatically determined and classified. Furthermore, in the blast furnace operation method and the molten iron manufacturing method, the disturbance of the temperature distribution of the blast furnace can be eliminated according to the results of the determination and classification of the blast furnace state determination device 1, and the automation ratio of the operation can be improved.
以上是根據圖式及實施例來說明本發明,但應了解,只要是所屬技術領域具有通知知識者,要根據本發明進行各種變形及修正是容易的。因此,這些變形及修正也包含於本發明的範圍。例如,各手段、各步驟等所包含之功能等能以合乎邏輯的方式進行再配置,可將複數個手段及步驟等組合成1個或是分割。As mentioned above, although this invention was demonstrated based on drawing and an Example, it should be understood that various deformation|transformation and correction|amendment based on this invention are easy as long as those skilled in the art are informed. Therefore, these deformation|transformation and correction are also included in the scope of the present invention. For example, functions, etc. included in each means, each step, etc. can be rearranged in a logical manner, and a plurality of means, steps, etc. can be combined into one or divided.
在上述實施形態的學習中,作為第1異常溫度分布的模式,係選擇高溫部分不包含中心的圖像資料。亦即,按照高溫部分是否包含原料裝入面的中心點來進行圖像資料的選擇。在此亦可為,不是按照中心點,而是按照具有一定大小之中心部分和高溫部分是否重疊來進行圖像資料的選擇。In the learning of the above-described embodiment, as the pattern of the first abnormal temperature distribution, image data in which the high temperature portion does not include the center are selected. That is, the selection of the image data is performed according to whether or not the high temperature portion includes the center point of the raw material loading surface. Here, the image data may be selected not according to the center point, but according to whether the center part having a certain size and the high temperature part overlap.
又如圖5所示般,高爐爐況狀態判定裝置1可構成為進一步具備學習模型生成部4。亦即,高爐爐況狀態判定裝置1可具備:溫度分布計測部2、儲存部3、學習模型生成部4、判定部5、顯示部6。Also, as shown in FIG. 5 , the blast furnace state determination device 1 may be configured to further include a learning model generation unit 4 . That is, the blast furnace state determination device 1 may include a temperature
1:高爐爐況狀態判定裝置 2:溫度分布計測部 3:儲存部 4:學習模型生成部 5:判定部 6:顯示部 10:學習模型生成裝置1: Blast furnace condition state determination device 2: Temperature distribution measurement section 3: Storage Department 4: Learning model generation part 5: Judgment Department 6: Display part 10: Learning Model Generation Device
[圖1]係顯示本發明的一實施形態之高爐爐況狀態判定裝置的構成例。 [圖2]係顯示本發明的一實施形態之高爐爐況狀態判定裝置所執行的學習、判定及顯示的流程圖。 [圖3]係顯示使用卷積類神經網路之學習的流程圖。 [圖4]係例示判定結果。 [圖5]係顯示變形例的高爐爐況狀態判定裝置的構成例。Fig. 1 shows an example of the configuration of a blast furnace furnace state determination device according to an embodiment of the present invention. Fig. 2 is a flowchart showing learning, determination, and display performed by the blast furnace state determination device according to an embodiment of the present invention. [Fig. 3] is a flowchart showing learning using a convolutional neural network. [ Fig. 4 ] is an example of the determination result. [ Fig. 5] Fig. 5 is a diagram showing a configuration example of a blast furnace state determination device according to a modification.
1:高爐爐況狀態判定裝置 1: Blast furnace condition state determination device
2:溫度分布計測部 2: Temperature distribution measurement section
3:儲存部 3: Storage Department
4:學習模型生成部 4: Learning model generation part
5:判定部 5: Judgment Department
6:顯示部 6: Display part
10:學習模型生成裝置 10: Learning Model Generation Device
Claims (8)
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2020141068A JP7264132B2 (en) | 2020-08-24 | 2020-08-24 | Blast Furnace Status Determining Device, Blast Furnace Operating Method, and Hot Metal Manufacturing Method |
JP2020-141068 | 2020-08-24 |
Publications (1)
Publication Number | Publication Date |
---|---|
TW202211091A true TW202211091A (en) | 2022-03-16 |
Family
ID=80353203
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW110130616A TW202211091A (en) | 2020-08-24 | 2021-08-19 | Blast furnace condition determination device, operation method of blast furnace, and manufacturing method for molten iron |
Country Status (3)
Country | Link |
---|---|
JP (1) | JP7264132B2 (en) |
TW (1) | TW202211091A (en) |
WO (1) | WO2022044816A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI843517B (en) * | 2022-04-22 | 2024-05-21 | 日商Jfe鋼鐵股份有限公司 | Direct-reduction iron melting method, solid iron and method for producing solid iron, civil engineering and construction material and method for producing civil engineering and construction material, and direct-reduction iron melting system |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2512328B2 (en) * | 1989-03-31 | 1996-07-03 | 新日本製鐵株式会社 | Automatic determination system for blast furnace pattern data |
JP2019022251A (en) * | 2017-07-11 | 2019-02-07 | 米沢電気工事株式会社 | Solar cell diagnosis method and solar cell diagnosis system |
JP6870693B2 (en) * | 2018-04-03 | 2021-05-12 | Jfeスチール株式会社 | Blast furnace condition condition determination device, blast furnace operation method, and blast furnace condition condition determination method |
-
2020
- 2020-08-24 JP JP2020141068A patent/JP7264132B2/en active Active
-
2021
- 2021-08-11 WO PCT/JP2021/029680 patent/WO2022044816A1/en active Application Filing
- 2021-08-19 TW TW110130616A patent/TW202211091A/en unknown
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI843517B (en) * | 2022-04-22 | 2024-05-21 | 日商Jfe鋼鐵股份有限公司 | Direct-reduction iron melting method, solid iron and method for producing solid iron, civil engineering and construction material and method for producing civil engineering and construction material, and direct-reduction iron melting system |
Also Published As
Publication number | Publication date |
---|---|
WO2022044816A1 (en) | 2022-03-03 |
JP7264132B2 (en) | 2023-04-25 |
JP2022036716A (en) | 2022-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6119515B2 (en) | Blast furnace tuyere state observation method and blast furnace tuyere state observation apparatus | |
EP3730630B1 (en) | Furnace condition control apparatus and method | |
JP2018119924A (en) | Diagnostic device | |
CN1021833C (en) | Blast furnace operation management method and apparatus | |
JP6897723B2 (en) | Learning model generation method, learning model generation device, blast furnace hot metal temperature control method, blast furnace hot metal temperature control guidance method, and hot metal manufacturing method | |
TW202211091A (en) | Blast furnace condition determination device, operation method of blast furnace, and manufacturing method for molten iron | |
JP2020042757A (en) | Processing device, processing method, processing program and inspection device | |
JP6350159B2 (en) | Blast furnace tuyere state observation method and blast furnace tuyere state observation apparatus | |
Li et al. | A blast furnace coke ratio prediction model based on fuzzy cluster and grid search optimized support vector regression | |
JP2013185234A (en) | Method and apparatus for observing condition of blast furnace tuyere | |
JP5644910B1 (en) | Abnormality detection method and blast furnace operation method | |
CN108759479B (en) | Sintering machine air leakage fault online diagnosis method based on blind source separation unilateral detection method | |
Trofimov | Automated expert systems in blast-furnace process control | |
CN112729559A (en) | Molten steel temperature monitoring system in LF stove | |
Trofimov et al. | About designing an intelligent system for slag detection in oxygen converter steelmaking | |
Matsuzaki et al. | Development of visual evaluation and numerical analysis system of blast furnace | |
WO2021182220A1 (en) | Blast-furnace furnace condition learning method, furnace condition learning device, abnormality detection method, abnormality detection device, and operation method | |
JP2696114B2 (en) | Blast furnace operation management method | |
WO2024036713A1 (en) | Method and apparatus for calculating height of root of cohesive zone of blast furnace, and storage medium | |
JP7549189B2 (en) | Simulation method using a blast furnace mathematical model and blast furnace operation method using the same | |
CN114140379A (en) | Online calculation method for blast furnace center airflow size | |
Ridley et al. | Modelling and flame segmentation for real-time monitoring of rotary kilns | |
He et al. | Angle Diagnosis of Blast Furnace Chute Through Deep Learning of Temporal Images | |
TW202348804A (en) | Method for predicting molten iron temperature in blast furnace, method for training molten iron temperature prediction model for blast furnace, method for operating blast furnace, molten iron temperature prediction device for blast furnace, molten iron temperature prediction system, and terminal device | |
Xu et al. | Blast Furnace Condition Recognizing in the Ironmaking Process Based on Prior Knowledge and Platt Scaling Probability |