JP3561401B2 - State estimation method for manufacturing process - Google Patents

Description

ここでｉ＝１は上部を、ｉ＝２は側壁部を、ｉ＝３は底部を示し、物体とは溶鋼鍋耐火物および盈鍋中のスラグのことである。さらに、
ｑ：単位時間当たりの溶鋼から物体表面（単位面積当たり）への熱移動量
ｄｔ：微少時間
Ｓ：溶鋼と接している物体総面積
Ｈ：溶鋼の熱容量
Ｗ：溶鋼重量
である。
【００４１】
そしてｑは溶鋼と物体表面との境界における伝熱現象を熱伝導として取り扱い、溶鋼鍋の上部、側壁部、底部の各部について次式で表される。
【００４２】
〔数３〕
ｑ_ｉ ＝ λ／ｃ／ρ×（Ｔｎ−Ｔｏ） …（１０）
ここで、
λ：溶鋼と接触している物体の熱伝導率
ｃ：溶鋼と接触している物体の比熱
ρ：溶鋼と接触している物体の密度
Ｔｎ：溶鋼と接触している側（内側）の物体表面温度（℃）
Ｔｏ：溶鋼温度（℃）
である。
【００４３】
Ｔｎはある時刻における溶鋼鍋の内側から外側にいたる物体の厚み方向の複数点（温度計算点）の物体温度を初期値として、厚み方向の温度計算点間隔、熱伝導率、比熱、密度を用いて一次元非定常伝熱差分方程式より微少時間後の物体温度を算出して求められる。この際、厚み方向の温度計算点間隔は物体の厚みより算出される。
【００４４】
また図８中の側壁部（ｉ＝２）、底部（ｉ＝３）のように内側から外側に材質の異なる複数の耐火物および鉄皮の複数の層がある場合には、この層毎に複数の計算点（温度計算点）を設定し、それぞれの層の厚み、熱伝導率、比熱、密度、層間の境界条件を用いて厚み方向の物体温度を計算する。
さらに外表面においては物体から大気中への放射がありるが、この熱移動量ｑｈは物体と大気との境界における伝熱現象を熱放射として取り扱い、次式で表される。
【００４５】
〔数４〕
ｑｈ_ｉ ＝ σ×ε×｛Ｔｇ^４ −Ｔｏ^４ ｝ …（１１）
ここで、
σ：ステファン・ボルツマン定数
ε：放射率
Ｔｇ：外側の物体表面温度（℃）
Ｔｏ：大気温度（℃）
である。
【００４６】
一方図８中の空鍋放冷時（ａ）、空鍋予熱時（ｂ）のように溶鋼鍋内に溶鋼がない場合には、上部（ｉ＝１）については伝熱計算を実行しない。
側壁部（ｉ＝２）および底部（ｉ＝３）については内側の境界条件を、空鍋放冷時（ａ）は大気への熱放射として（１１）式をそのまま使用して、空鍋予熱時（ｂ）は予熱ガスからの熱輻射として（１１）式のＴｏをＴｙ（予熱ガスの温度（℃））に置き換えて、温度分布を算出する。
【００４７】
図９のように鍋蓋を掛ける場合には、鍋蓋について厚み方向の温度分布を追加して算出する。その際の外側境界条件としては（１１）式を用いる。内側境界条件としては空鍋放冷時（ａ）は鍋底部の内側表面（底内面）への熱放射として（１１）式のＴｏをＴｔｎ（鍋底部の内側表面温度（℃））に置き換えて、盈鍋時（ｃ）はスラグの外側表面との熱放射として（１１）式のＴｏをＴｓｇ（スラグ外側表面温度（℃））に置き換えて微少時間後の温度分布および溶鋼から溶鋼鍋への熱移動量を算出する。なお、空鍋予熱時（ｂ）は鍋蓋を掛けることができないので図７のｂと同様の計算を行う。
【００４８】
そして、データ記憶部１に記憶されている各工程の予定時刻に基づき降下温度を計算する。
例えば、転炉吹錬開始時刻（ｔ１）、転炉出鋼開始時刻（ｔ２）、転炉出鋼終了時刻（ｔ３）、転炉炉裏作業終了時刻（ｔ４）に基づき、転炉出鋼中の溶鋼降下温度△ＴｎｓＬ２を以下のように算出する。なお、現時刻からチャージＡで使用するまではその鍋は使用されないものとする。
【００４９】
現時刻で溶鋼容器温度分布記憶部４６に記憶されているチャージＡを処理する鍋ナンバーの溶鋼鍋の厚み方向の温度分布を初期値として、予熱有無および鍋蓋有無に応じて図８（ａ）、（ｂ）もしくは図９（ａ）、（ｂ）に示す境界条件に従って１０秒毎に温度分布を算出していき、チャージＡの転炉出鋼開始時刻（ｔ２）における厚み方向の温度分布を算出する。
【００５０】
さらにこの時刻（ｔ２）における温度分布を初期値として、図８（ｃ）に示す境界条件に従って２．５秒単位で温度分布を算出して（９）、（１０）式より溶鋼降下温度△Ｔを算出し、転炉出鋼終了時刻（ｔ３）までの積算値を△ＴｎｓＬ２とする。
同様な計算により、炉裏作業中の溶鋼降下温度△ＴｎｓＬ３、転炉炉裏作業終了から２次精錬処理開始までの溶鋼降下温度△ＴｎｓＬＲが算出される。
【００５１】
なお、現時刻からチャージＡで使用されるまでの間に同一ナンバーの鍋を使用するチャージがある場合には、そのチャージについて時刻ｔ２、ｔ３、ｔ４における温度分布、チャージＡについて時刻ｔ２での温度分布に基づいて、チャージＡについての△ＴｎｓＬ２、△ＴｎｓＬ３、△ＴｎｓＬＲが算出される。
上記の計算方法は、転炉、２次精錬処理槽、タンディッシュについても同様であであって、データ記憶部１の転炉工程の予定時刻、すなわち転炉炉吹錬開始時刻（ｔ１）、転炉出鋼開始時刻（ｔ２）、転炉出鋼終了時刻（ｔ３）より、出鋼中の転炉炉体に起因する溶鋼降下温度△ＴｌＬ２は以下のように算出される。
【００５２】
即ち図１０に示すように転炉炉体については伝熱現象を直胴部および底部に分解し、空釜放冷時（ａ）、空釜予熱時（ｂ）および盈釜（転炉炉体内に溶鋼が存在する溶銑装入時〜出鋼完了時）時（ｃ）の境界条件に従って、チャージＡを処理する転炉炉体と同一の転炉炉体の現時刻において溶鋼容器温度分布記憶部４６に記憶されている転炉炉体の厚み方向の温度分布を初期値として、予熱有無に応じて図１０（ａ）および（ｂ）に示す境界条件に従って微少時間１０秒単位で時刻ｔ１における厚み方向の温度分布を算出する。
【００５３】
さらに図１０（ｃ）の境界条件に従って微少時間２．５秒単位で時刻ｔ２における厚み方向の温度分布を算出する。そしてこの温度分布を初期値として図１０（ｃ）の境界条件に従って微少時間２．５秒単位で温度分布を算出して、（９）、（１０）式より溶鋼降下温度△Ｔを算出する。この△Ｔのｔ３までの積算値が、転炉出鋼中の転炉炉体に起因する溶鋼降下温度△ＴｌＬ２となる。
【００５４】
データ記憶部１に記憶されている２次精錬工程の予定時刻、すなわち２次精錬処理開始時刻（ｔ５）、２次精錬処理終了時刻（ｔ６）から、２次精錬処理中の２次精錬処理槽に起因する溶鋼温度降下量△ＴｒＲ１は以下のように算出される。
２次精錬処理槽については、図１１に示すように伝熱現象を下部槽および浸漬管に分解し、空槽放冷時（ａ）、空槽予熱時（ｂ）、盈槽（２次精錬処理槽内に溶鋼が存在する２次精錬処理開始〜２次精錬処理終了まで）時（ｃ）の境界条件に従って、２次精錬処理中の２次精錬処理槽に起因する溶鋼温度降下量△ＴｒＲ１を算出する。
【００５５】
データ記憶部４１に記憶されている連鋳工程の予定時刻、すなわち連鋳注入開始時刻（ｔ７）、鋳込代表温度測温時刻（ｔ８）、連鋳注入終了時刻（ｔ９）から、連鋳開始から鋳込代表温度測温間のタンディッシュに起因する溶鋼降下温度△ＴｃＣ２は以下のように算出される。タンディッシュについては、図１２に示すように伝熱現象を側壁部のみとし、空タンディッシュ放冷時（ａ）、空タンディッシュ予熱時（ｂ）、盈タンディッシュ（タンディッシュ内に溶鋼が存在する連々鋳先頭鍋の注入開始時〜連々鋳最終鍋の注入終了時）時（ｃ）の境界条件に従って連鋳注入開始からタンディッシュ代表温度測温までのタンディッシュに起因する溶鋼降下温度△ＴｃＣ２を算出する。
【００５６】
また、転炉炉体については転炉プロコン５３より上位計算機５１を経由してチャージＡの転炉出鋼終了を、２次精錬処理槽については２次精錬プロコン５４より上位計算機５１を経由してチャージＡの２次精錬処理終了を、タンディッシュについては連鋳プロコン５５より上位計算機１１を経由して連々鋳最終鍋の連鋳注入終了を受信した場合には、現時刻において溶鋼容器温度分布記憶部６に記憶されている同一容器の温度分布を初期値として、転炉炉体についてはチャージＡの転炉出鋼終了時、２次精錬処理槽についてはチャージＡの２次精錬処理終了時、タンディッシュについては連々鋳最終鍋の注入終了時まで温度分布を計算して、溶鋼容器温度分布記憶部６の温度分布を更新して記憶し、次回同一容器が使用される際の温度分布の初期値として用いる。
【００５７】
上述のように算出された降下温度、△ＴｎｓＬ２、△ＴｎｓＬ３、△ＴｎｓＬＲ、△ＴｎｓＲ１、△ＴｎｓＲＣ、△ＴｎｓＣ２、△ＴｌＬ２、△ＴｒＲ１、△ＴｃＣ２は各種溶鋼目標温度計算部４７にて使用される。
成分毎投入量算出部４４における算出方法を、転炉出鋼中の投入物を例として説明する。
【００５８】
まず、データ記憶部４１に記憶されているチャージＡに使用する複数種の副原料毎の投入量に、データ記憶部１に記憶されている各副原料毎のＣ、Ｓｉ、Ｍｎ、Ａｌ、ＣａＯ、その他各成分毎の含有率を乗じて、各副原料毎の各成分の含有量を算出する。次に製造する鋼種に含まれるべき各成分の含有量に基づいて各副原料毎に添加すべき各成分量を算出し、各成分毎に合計することによってＣ成分投入量ＭＬｃ、Ｓｉ成分投入量ＭＬｓｉ、Ｍｎ成分投入量ＭＬｍｎ、Ａｌ成分投入量ＭＬａｌ、ＣａＯ分投入量ＭＬｃａｏ、その他投入量ＭＬｅｔｃを算出し、各種溶鋼目標温度算出部４７で用いる。
【００５９】
同様に２次精錬処理中の投入物についても、各成分毎の成分投入量ＭＲｃ、ＭＲｓｉ、ＭＲｍｎ、ＭＲａｌ、ＭＲｃａｏ、ＭＲｅｔｃが算出され、各種溶鋼目標温度算出部４７で使用される。
また、溶鋼目標温度算出モデルにより、チャージＡの転炉での吹錬開始までは鋳込目標温度を起点に、２次精錬処理終了目標温度、２次精錬処理開始目標温度、鍋上目標温度、吹止目標温度までを算出し、転炉はこの吹止目標温度を目標に吹錬する。
【００６０】
その後の出鋼、炉裏作業を経て２次精錬処理開始まではチャージＡの処理の進捗実績を反映して経過時間算出部４２、溶鋼温度降下量算出部４３、成分毎投入量算出部４４、鋼種補正値算出部４５で各項目の数値を再度算出した後、鋳込目標温度を起点として２次精錬処理終了目標温度を再度算出して目標温度の高精度化を図っている。
【００６１】
最後に、ＮＮ４７１〜４７４の重み係数、しきい値および鋼種補正値Ｈａ〜Ｈｄの決定方法について説明する。
まず、温度管理計算機５２が上位計算機５１より受信した過去の操業条件の各データをデータ記憶部１に記憶する。次に、経過時間算出部４２にて実績時刻ｔ１〜ｔ９からｔＬ２、ｔＬ３、ｔＬＲ、ｔＲ１、ｔＲＣ、ｔＣ２を算出し、溶鋼温度降下量算出部４３にて使用した各溶鋼容器のナンバーや実績時刻ｔ１〜ｔ９から温度降下△ＴｎｓＬ２、△ＴｎｓＬ３、△ＴｎｓＬＲ、△ＴｎｓＲ１、△ＴｎｓＲＣ、△ＴｎｓＣ２、△ＴｌＬ２、△ＴｒＲ１、△ＴｃＣ２を算出する。
【００６２】
さらに、成分毎投入量算出部４４にて成分毎投入量ＭＬｃ、ＭＬｓｉ、ＭＬｍｎ、ＭＬａｌ、ＭＬｃａｏ、ＭＬｅｔｃ、ＭＲｃ、ＭＲｓｉ、ＭＲｍｎ、ＭＲａｌ、ＭＲｅｔｃを算出する。
そしてＮＮ４７１については、上記各種データの中から前記したＮＮ４７１の入力状態量を入力層に入力するが、鋳込目標温度については実測された実績の鋳込代表温度を入力する。さらに出力層には実測された２次精錬処理終了温度を入力して学習することにより、重み係数およびしきい値を決定して保存する。
【００６３】
ＮＮ４７１の重み係数およびしきい値の学習に用いた全チャージを複数の鋼種グループに分類し、この鋼種グループ毎の２次精錬処理終了温度の推定温度と実測温度の差の平均値が、鋼種補正部４７５の鋼種グループ毎の鋼種補正値Ｈａとして保存する。今回は全チャージを４６種類に分類して鋼種補正値を算出している。
【００６４】
ＮＮ４７２、４７３、４７４についてもＮＮ４７１と同様に重み係数、しきい値、および鋼種補正係数Ｈｂ、Ｈｃ、Ｈｄを決定して保存する。そして何れも４６種類に分類して鋼種補正値Ｈｂ、Ｈｃ、Ｈｄを決定している。
図１３は本発明に係る温度推定システムの効果の説明図であって、（イ）は全チャージについてＮＮから出力された推定温度と実測温度の差を横軸に、チャージ数を縦軸にした度数分布図である。
【００６５】
（ロ）は全チャージを鋼種により４グループに分類して度数分布図を取り直したもので、この鋼種グループ毎の平均値をいわゆる鋼種により異なる温度降下傾向と見なして、ＮＮから出力された推定温度をこの平均値により補正する。
（ハ）は全チャージを補正した結果の度数分布であり、（イ）と比較して推定精度が大きく向上していることがわかる。
【００６６】
【発明の効果】
第１の発明に係る状態量推定方法によれば、各工程のＮＮモデルを直列接続して製造プロセスをモデル化することにより、各ＮＮの入力状態量数は減少し学習が容易となるとともに推定精度を向上することが可能となる。
第２の発明に係る状態量推定方法によれば、各工程のモデルの一部として数式モデルを使用することにより、各ＮＮの入力状態量数を一層減少することが可能となる。
【００６７】
第３の発明に係る状態量推定方法によれば、製品の種類に応じて予めデーターベース化することが可能な補正量を使用して推定された状態量を補正することにより推定精度を一層向上することが可能となる。
第４の発明に係る状態量推定方法によれば、製鋼プロセスにおいて予め定められた目標鋳込溶鋼温度に基づいて転炉における吹止溶鋼温度を精度よく推定することが可能となる。
【図面の簡単な説明】
【図１】３層ＮＮの構成図である。
【図２】製鋼プロセスの流れ図である。
【図３】溶鋼温度の変化を示すグラフである。
【図４】溶鋼温度推定システムの機能図（１／２）である。
【図５】溶鋼温度推定システムの機能図（２／２）である。
【図６】溶鋼温度推定システムのハードウエア構成図である。
【図７】経過時間算出方法の説明図である。
【図８】溶鋼鍋（鍋蓋無）の伝熱現象説明図である。
【図９】溶鋼鍋（鍋蓋有）の伝熱現象説明図である。
【図１０】転炉の伝熱現象説明図である。
【図１１】２次精錬処理槽の伝熱現象説明図である。
【図１２】ダンディッシュの伝熱現象説明図である。
【図１３】本発明の効果の説明図である。
【符号の説明】
４１…データ記憶部
４２…経過時間算出部
４３…溶鋼容器に起因する溶鋼降下温度算出部
４４…成分毎投入量算出部
４５…鋼種補正値算出部
４６…溶鋼容器温度分布記憶部
４７…溶鋼温度算出部
４７１…第２搬送工程ネットワークモデル
４７２…２次精錬工程ネットワークモデル
４７３…第１搬送工程ネットワークモデル
４７４…出鋼ネットワークモデル
４７５、４７６、４７７、４７８…鋼種補正部[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method for estimating a state quantity of a manufacturing process, and more particularly to a method for estimating a state quantity of a manufacturing process using a neural network (hereinafter referred to as NN).
[0002]
[Prior art]
In order to manufacture a product of a predetermined quality in a process of manufacturing a product, it is important to accurately measure various state quantities during the manufacturing process.
For example, in the process of producing steel such as billets or slabs by continuous casting through molten refining of molten steel taken out of a converter, the quality of the product is determined by the casting temperature of the last step, the continuous casting step, It is most important to control the casting temperature within a specified value.
[0003]
Therefore, taking into account the heat removal by the refractory lining of the molten steel pot, which is one of the molten steel vessels, and the amount of temperature drop of the molten steel caused by heat radiation during the movement of the molten steel pot, it is a pre-process of the continuous casting process. Determine the temperature of molten steel in the next refining process, then determine the temperature of the molten steel in the converter process, which is the previous process, based on the temperature of the molten steel in the secondary refining process, and control the temperature of the molten steel to the predetermined target temperature. By doing so, it becomes possible to control the casting temperature within a specified value.
[0004]
Therefore, as disclosed in Japanese Patent Application Laid-Open No. 3-161161, for example, the temperature of the front and back surfaces of the refractory lining of the molten steel pot that has been emptied by dispensing the molten steel, the specific gravity and weight of the refractory, and the heat radiation correction coefficient obtained from experiments. Predict the amount of molten steel temperature drop based on the amount of heat stored in a container such as a molten steel pot using a model formula that uses, and measure the actual molten steel temperature at least twice, and estimate the refractory based on the measured value based on the model formula. There is also a method in which the error of the heat storage amount is corrected, the molten steel temperature drop is re-predicted, and the tapping temperature is set based on the predicted molten steel temperature drop.
[0005]
However, since there are factors such as the weight of refractories used in the model formula for estimating the amount of heat stored in the molten steel pot, which change due to erosion during use, it is extremely burdensome to create a simple model that takes these factors into account. It becomes. Further, even if a model can be produced, there are many practical problems such as a large amount of labor is required for maintenance for maintaining accuracy.
[0006]
In order to solve such a problem, a method of modeling a process using an NN and estimating a state quantity using the NN model has already been proposed.
FIG. 1 is a configuration diagram of a three-layer NN, which is composed of an input layer 11, an intermediate layer 12, and an output layer 13, and information is transmitted from the input layer 11 to the intermediate layer 12 and from the intermediate layer 12 via weighting coefficients. The signal is transmitted to the input layer 13. The output information of the neurons included in the intermediate layer 12 is obtained as a function of the input information and the threshold.
[0007]
Note that the coupling coefficient and the threshold can be determined by learning.
[0008]
[Problems to be solved by the invention]
However, when applying the NN model to the steelmaking process and estimating the shutoff temperature from the target pouring temperature, the following problems occur.
That is, if a large NN model is used in which the input state quantity to the input layer and the output state quantity from the output are each 50 or more, not only the learning time becomes longer, but also the coupling coefficient and threshold value are determined by learning. In some cases, it is necessary to narrow the input / output state quantity as much as possible.
[0009]
However, as a result of narrowing down the number of input / output state quantities, the actual operation progress was not reflected, the heat storage status of the molten steel pot could not be expressed in detail, and the drop temperature differing depending on the steel type could not be expressed, etc. If the progress of the process is greatly different from the initial plan, there is a problem that the accuracy of estimating the blowoff temperature is reduced when the use of the molten steel ladle is special.
[0010]
Furthermore, although the NN model can be set accurately once, it is vulnerable to changes in conditions that occur in the subsequent steelmaking process, and cannot follow accurately. In this case, it is necessary to learn again to obtain the coupling coefficient of the NN again, and it is not possible to avoid a long-term decrease in estimation accuracy until the learning data is accumulated.
[0011]
For this reason, the setting of the temperature of the molten steel at the blow stop, the temperature of the molten steel at the inlet on the ladle, the temperature of the molten steel at the inlet of the secondary refining process, and the temperature of the molten steel at the outlet of the secondary refining process is actually performed based on the intuition and experience of a skilled operator. In addition, there is a problem that sufficient accuracy cannot be obtained due to individual differences between skilled operators, and that it is necessary to constantly train and secure excellent skilled operators.
[0012]
The present invention has been made in view of the above problems, and has as its object to provide a method for estimating a state quantity of a manufacturing process that can accurately estimate a state quantity of a manufacturing process using NNs.
[0013]
[Means for Solving the Problems]
A method of estimating a state quantity of a manufacturing process according to a first invention is a manufacturing process of a product including a first step, a last step, and at least one intermediate step disposed between the first step and the last step. A state quantity estimating method of a manufacturing process for estimating an output state quantity of a first step based on an input state quantity of a final step,
An output state quantity measuring step of measuring a measurement state quantity that is a measurable output state quantity among output state quantities that are state quantities after processing in each process,
The output state quantity measured in the output state quantity measurement step and the output state quantity other than the measured output state quantity determined based on the input state quantity which is the state quantity before processing in the downstream process of each process are input. And an input state quantity estimating step of estimating an input state quantity of the process based on the NN of the process. The manufacturing process is repeated from the final process to the earliest process for each process, and the measured output state of the earliest process is repeated. The output state quantity other than the quantity is estimated based on the input state quantity of the downstream process.
[0014]
In the state quantity estimating method of the manufacturing process according to the first invention, the state quantity of each step is estimated by using the NN model for each step included in the manufacturing process, so that the state quantity of the manufacturing process is reduced. Presumed.
In the method for estimating a state quantity of a manufacturing process according to a second invention, the input state quantity estimating step includes, based on a mathematical model, an output other than the measured output state quantity based on the output state quantity measured in the output state quantity measuring step. An output state quantity calculating step of calculating a part of the state quantity,
The output state quantity measured in the output state quantity measurement step, the calculated output state quantity calculated in the output state quantity calculation step, and the measured output state quantity and calculation determined based on the input state quantity in the downstream process of each process An output state quantity other than the output state quantity is input, and the input state quantity of the process is estimated based on the NN model of the process.
[0015]
In the state quantity estimating method for the manufacturing process according to the second invention, a part of the state quantity is calculated by a mathematical model.
In the method of estimating a state quantity of a manufacturing process according to a third aspect, the input state quantity estimating step is a step of correcting an input state quantity in a downstream process of each step by a predetermined correction for each type of a product manufactured in the manufacturing process. The output state quantity other than the measured output state quantity and the calculated output state quantity is determined by correcting with the value.
[0016]
In the state quantity estimating method for the manufacturing process according to the third invention, the state quantity estimated by the NN model is corrected by a correction value predetermined for each type of product manufactured in the manufacturing process.
A method for estimating a state quantity of a manufacturing process according to a fourth invention is a method of subjecting molten steel blown in a converter to a secondary refining process in a secondary refining process and a target molten steel temperature before the start of casting in a process of casting in a casting process. A state quantity estimating method of a manufacturing process for estimating a blow-end molten steel temperature that is a molten steel temperature at the time of converter blow-off based on a target pouring temperature that is
The time from the start of continuous casting calculated using the mathematical model to the start of the temperature measurement of the representative casting temperature, the processing time of the second transporting step, the temperature drop within each time calculated using the mathematical model, and the manufacturing A secondary refining process outlet molten steel temperature estimating step of estimating a secondary refining process exit molten steel temperature using a second transport process neural network model based on a target casting molten steel temperature predetermined according to a steel type to be subjected to
A secondary refining process outlet molten steel temperature correction stage that corrects the secondary refining process outlet molten steel temperature estimated in the secondary refining process outlet molten steel temperature estimation stage in accordance with a steel type to be manufactured;
The processing time of the secondary refining process calculated using the mathematical model, the temperature drop during the secondary refining process calculated using the mathematical model, and the secondary corrected in the secondary refining process outlet molten steel temperature correction stage Estimating a secondary refining process inlet molten steel temperature using the secondary refining process neural network model based on the refining process outlet molten steel temperature;
A secondary refining process inlet molten steel temperature correction stage for correcting the secondary refining process inlet molten steel temperature estimated in the secondary refining process inlet molten steel temperature estimation stage according to a steel type to be manufactured;
The time from the end of converter tapping to the end of the hearth operation calculated using the mathematical model, the time from the end of the hearth operation to the start of secondary refining, and the drop temperature for each time calculated using the mathematical model And estimating the molten steel temperature on the pot using the neural network model for the first transporting process based on the molten steel temperature at the secondary refining process inlet corrected at the stage of correcting the molten steel temperature at the secondary refining process inlet Stages and
A molten steel temperature on the pot to correct the molten steel temperature on the pot estimated in the molten steel temperature on the pot according to the steel type to be manufactured,
The time from the start of converter tapping to the end of tapping, calculated using the mathematical model, the temperature drop during that time, calculated using the mathematical model, and the pot corrected in the pot molten steel temperature correction stage Estimating the blow-end molten steel temperature using the neural network model for tapping process based on the upper molten steel temperature,
A temperature correction step for correcting the temperature of the molten steel at the blow stop estimated at the temperature estimation step of the molten steel at the blow stop according to the type of steel to be manufactured.
[0017]
In the state quantity estimating method for the manufacturing process according to the fourth invention, the molten steel temperature is estimated from the target pouring temperature by sequentially estimating the molten steel temperature by going backward in the steelmaking process.
First, the temperature drop phenomenon of the molten steel in each step of the steelmaking process is decomposed into those caused by the elapsed time, the molten steel container such as a molten steel pot, the input material such as an alloy, etc. Be detailed and accurate.
[0018]
That is, the molten steel temperature drop due to the elapsed time needs to accurately estimate the plan data. In addition, since this plan data is to be used to estimate the heat storage state of the molten steel vessel, it is an indispensable element for accurately predicting the molten steel temperature drop. Accordingly, detailed division of the processing time of each process according to the processing category, accurate estimation of the transport time between processes, etc. are performed to edit accurate plan data, and the progress of each process is monitored to To re-predict molten steel temperature drop. In this way, the elapsed time can be accurately estimated in the same manner with respect to a change in operation conditions such as disturbance due to an operation abnormality in each process.
[0019]
The temperature drop of molten steel caused by a molten steel vessel such as a molten steel ladle is based on the temperature distribution in the thickness direction of the refractory lining etc. at the end of the last use of each piece. Based on the boundary conditions, the temperature distribution in the thickness direction of the lining refractory etc. is calculated from the material and thickness of the refractory using the unsteady heat transfer difference equation, and the heat is stored. From the change in the heat storage condition, the temperature drop of the molten steel caused by the molten steel container in the process is obtained. This makes it possible to express the dominant unsteadiness in the steelmaking process, and extrapolation is possible, so that it is possible to accurately estimate the same when operating conditions change, such as changing the material and thickness of the refractory or installing a pot lid. it can.
[0020]
For the temperature change of molten steel caused by heat, endotherm, and latent heat caused by the input materials for manufacturing alloys, etc., instead of using the input amount, multiply the input amount by the component ratio for the component related to the temperature drop. The calculated input amount for each component is used. As a result, the yield of each component of the input material can be expressed, and extrapolation is possible, so that it is possible to accurately estimate the operating conditions such as a change in the alloy component ratio due to a change in the alloy brand.
[0021]
Then, the temperature drop phenomenon of molten steel is reconstructed by the learning function of the hierarchical NN, but the inverse problem is solved without significantly increasing the calculation load by using the NN, and the steelmaking is performed by the nonlinearity of the constituent neurons. It is possible to express the dominant nonlinearity in the process, and to optimize the influence of the molten steel temperature drop component of the input layer, which is estimated more accurately and in detail, and to obtain a more accurate molten steel temperature than when only NN is used. It is possible to express the phenomenon of descent. Further, since extrapolation is possible in the molten steel drop temperature calculation unit, even when the operating conditions change, it is possible to continue to express accurately without the need for re-learning.
[0022]
Further, the molten steel temperature drop phenomenon due to the steel type or the like that could not be expressed so far is classified and corrected and output, so that the molten steel temperature drop phenomenon can be more accurately expressed.
Such a model for predicting the temperature drop of molten steel is very effective in predicting the phenomenon of temperature drop of molten steel in a steelmaking process in which unsteadiness and nonlinearity are dominant and operating conditions are frequently changed.
[0023]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, a steelmaking process according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 2 is a flowchart showing the outline of the steelmaking process. The molten steel that has been blown in the converter 21 is once taken out to the molten steel pot 22 and then transferred to the secondary refining factory 23 by the molten steel pot 22. . The molten steel subjected to the secondary refining is supplied to the continuous casting machine 24 by the molten steel pot 22.
[0024]
In the above process, the quality of the ingot cast by the continuous casting machine 24 is determined by the casting steel temperature, which is the temperature of the molten steel before the casting step in the continuous casting machine 24 is started. Is required to be controlled to a predetermined target temperature.
FIG. 3 is a graph showing a change in molten steel temperature, in which the vertical axis represents molten steel temperature and the horizontal axis represents time. That is, during the blowing in the converter 21 in the converter process, the molten steel temperature rises, and the blow-off temperature, which is the temperature of the molten steel at the end of the blowing, increases the molten steel temperature into the molten steel pot 22 in the steelmaking process and the secondary refining plant. Through the secondary refining step and the continuous casting step at 23, the temperature is lowered to the molten steel temperature.
[0025]
Therefore, it is possible to determine the blow-off temperature by adding the temperature drop due to heat removal in each step to the target pouring temperature.
Therefore, in the present invention, in order to estimate the inlet molten steel temperature of each step based on the outlet molten steel temperature and other state quantities of each step, the tapping step, the first transporting step, the secondary refining step, and the second transporting step are performed. Are modeled using NN, respectively.
[0026]
However, in the present invention, the following measures are used to reduce the dimension of the NN model as much as possible in order to ensure the convergence of the learning and to reduce the time required for the learning.
(1) The processing time in each step is determined based on the start time and end time of each step.
[0027]
(2) The heat removal characteristic of the molten steel container is determined based on a mathematical model representing a temperature change of the molten steel container.
(3) In the converter and secondary refining, the temperature drop due to the introduction of various inputs and the temperature drop correction based on the type of molten steel are determined using a database that has collected actual measurement data.
[0028]
4 and 5 are functional diagrams of a molten steel temperature estimating system to which the state quantity estimating method according to the present invention is applied, and include the following parts.
41: Data storage unit for storing operating conditions in converter, secondary refining, continuous casting, molten steel ladle, etc.
42 ... Elapsed time calculation unit that calculates the elapsed time from the processing start time and the processing end time of each process of the converter, secondary refining, and continuous casting using a mathematical model
43 ... Calculation part of molten steel temperature drop in molten steel container
44: Component input amount calculation unit for calculating the component input amount of additives to be added for the purpose of component adjustment and the like during converter tapping and secondary refining processing
45 ... Steel type correction value calculation unit
46: Molten steel container temperature distribution storage
Each of the calculation units 42 to 46 executes a calculation based on the operation condition data in the data storage unit 41.
[0029]
The reference numeral 47 designates NNs 471 to 474 which use the calculation results of the calculation units 42 to 45 and the data stored in the data storage unit 41 as input data, and correct the output results by the calculation results of the steel type correction value calculation unit 45. This is a molten steel temperature calculation unit composed of steel type correction units 475 to 478.
In addition, among NN, 471 is a 2nd conveyance process, 472 is a secondary refining process, 473 is a 1st conveyance process, and 474 is NN which shows a steel tapping process.
[0030]
That is, from the target pouring temperature given as data from the data storage unit 41, the molten steel temperature at the end of the secondary refining using the NN model 471 in the second transporting process, and from this temperature, the NN model 472 in the secondary refining process is used. From the temperature of the molten steel at the start of the secondary refining process, the temperature of the molten steel tapping using the NN model 473 in the first transport process is estimated from this temperature, and the temperature of the blow-off stop using the NN model 474 of the tapping process from this temperature. I do.
[0031]
FIG. 6 shows an example of the configuration of the molten steel temperature estimation system. Reference numeral 51 denotes a higher-level computer which edits operating conditions in a converter, secondary refining, continuous casting, a molten steel ladle, and transmits the edited data to another computer.
Reference numeral 52 denotes a data storage unit 41, an elapsed time calculation unit 42, a molten steel drop temperature calculation unit 43, a component-based input amount calculation unit 44, a steel type correction value calculation unit 45, a molten steel container temperature distribution storage unit 46, a molten steel temperature shown in FIG. The temperature management computer configured by the calculation unit 47 calculates the molten steel temperature and sends it back to the host computer 51.
[0032]
Reference numeral 53 denotes a converter process control (converter process control computer) for controlling the converter, 54 denotes a secondary refining process control for controlling the secondary refining process, and 55 denotes a continuous casting process control for controlling the continuous casting machine. Reference numeral 55 denotes a target molten steel temperature calculated by the temperature management computer 52 and received via the higher-level computer 51, and a target temperature such as a molten steel temperature and an input material input amount of each step based on target values such as a molten steel component from the higher-level computer 51. Perform control.
[0033]
Hereinafter, the operation of the molten steel temperature estimation system will be described.
First, the operating condition data for charge A blown by the converter and scheduled to be processed in the continuous casting process directly or through the secondary refining process, transmitted from the host computer 51, is stored in the data management unit 52 by the temperature management computer 52. 1 is stored.
Based on the data in the data storage unit 1 and the calculation results of the calculation units 42 to 45, the molten steel temperature calculation unit 7 calculates various molten steel temperatures.
[0034]
That is, the NN471 and the steel type correcting unit 475 calculate the molten steel temperature at the end of the secondary refining process from the casting target temperature molten steel and the like, and the NN472 and the steel type correcting unit 476 use the NN472 and the steel type correcting unit 476 to calculate the molten steel temperature at the end of the secondary refining process. NN473 and the steel type correction unit 477 calculate the tapping process molten steel temperature from the secondary refining start molten steel temperature and the like, and the NN474 and the steel type correction unit 478 calculate the blow-off molten steel temperature from the tapping process molten steel temperature and the like. I do.
[0035]
The input state quantity and output state quantity of each NN are shown below. It should be noted that the input state quantity and the output state quantity mean the state quantity input to and output from the NN, but do not mean the state quantity on the entrance side and the state quantity on the exit side of the process.
(1) NN471 of the second transport process
・ Input state quantity
D0: Casting target temperature (TD target temperature)
D1: Empty pot time
D2: Necessity of furnace hearth slag reforming
tRC: elapsed time from the end of secondary refining process to the start of continuous casting
tC2: Elapsed time from start of continuous casting to representative temperature measurement of casting
ΔTnsRC: Temperature of molten steel falling from molten steel ladle and slag from the end of secondary refining to the start of continuous casting
ΔTnsC2: Temperature drop of molten steel from molten steel pot and slag from start of continuous casting to representative temperature measurement of casting
ΔTcC2: Temperature drop of molten steel by dandishes from start of continuous casting to representative temperature measurement of casting
・ Output state quantity
Molten steel temperature at secondary refining process outlet
(2) NN472 of the secondary refining process
・ Input state quantity
D0: Temperature of molten steel at secondary refining process outlet
D1: Empty pot time
D2: Necessity of furnace hearth slag reforming
D3: Secondary refining OB amount
tR1: processing time from the start of the secondary refining process to the end of the secondary refining process
ΔTnsR1: Temperature of molten steel drop by molten steel pot and slag from start of secondary refining process to end of secondary refining process
ΔTrR1: Temperature drop of molten steel in the secondary refining tank from the start of secondary refining to the end of secondary refining
MRc: C (carbon) input amount from the start of the secondary refining process to the end of the secondary refining process
MRsi: Si (silicon) input amount from the start of the secondary refining process to the end of the secondary refining process
MRmn: Mn (manganese) input amount from the start of the secondary refining process to the end of the secondary refining process
MRal: Amount of Al (aluminum) charged from the start of the secondary refining process to the end of the secondary refining process
MRetc: Input amount of other additives from the start of the secondary refining process to the end of the secondary refining process
・ Output state quantity
Secondary refining start molten steel temperature
(3) NN473 of the first transport step
・ Input state quantity
D0: Secondary refining start molten steel temperature
D1: Empty pot time
D2: Necessity of furnace hearth slag reforming
D4: Amount of metal with steel ladle
D5: Hot spray amount of molten steel pot
tL3: so-called hearth work time from converter tapping to completion of converter hearth work
tLR: Elapsed time from the end of the converter hearth work to the start of the secondary refining process
ΔTnsL3: Temperature drop of molten steel due to molten steel pot and slag during hearth work
ΔTnsLR: Temperature of molten steel falling from the molten steel pot and slag from the end of the converter back work to the start of the secondary refining process
・ Output state quantity
Molten steel temperature on pot
(4) NN474 in the tapping process
・ Input state quantity
D0: Molten steel temperature on pot
D4: Amount of metal with steel ladle
D5: Hot spray amount of molten steel pot
D6: Number of use of converter furnace body
D7: Target value of C component when the converter is shut off
tL2: so-called tapping time from the start of converter tapping to the end of converter tapping
ΔTnsL2: Temperature drop of molten steel by molten steel ladle and slag during tapping
ΔTlL2: Temperature drop of molten steel by converter furnace during tapping
MLc: C input during tapping
MLsi: Si input during tapping
MLmn: Mn input during tapping
MLal: Al input amount during tapping
MLcao: CaO input during tapping
MLetc: Input of other additives during tapping
・ Output state quantity
Converter shutoff temperature
Note that the secondary refining end molten steel temperature is obtained by correcting the output of NN471 by the steel type correction value calculation unit 475, and the secondary refining start molten steel temperature is obtained by correcting the output of NN472 by the steel type correction value calculation unit 476. The molten steel temperature on the converter pan is obtained by correcting the output of NN473 by the steel type correction value calculation unit 477, and the blow-off temperature is obtained by correcting NN474 by the steel type correction value calculation unit 478.
[0036]
Here, the steel type correction value is used to correct the estimated temperature for each process according to the steel type in order to improve the accuracy of estimating the temperature, and is stored in a database of the actual measured values of the molten steel temperature in each process using the steel type as a parameter. It is calculated based on:
Next, an elapsed time calculation method in the elapsed time calculation unit 42 will be described with reference to FIG. 7 which is an explanatory diagram of the elapsed time calculation method. Scheduled time of each step of charge A stored in the data storage unit 41, ie, converter furnace blowing start time (t1), converter tapping start time (t2), converter tapping end time (t3), Furnace hearth work end time (t4), secondary refining process start time (t5), secondary refining process end time (t6), continuous casting pouring start time (t7), TD representative temperature measurement time (t8), continuous casting From the pouring end time (t9), the blowing time (tL1), tapping time (tL2), hearth work time (tL3) in the converter process, and the time (tLR) from the end of the converter furnace work to the start of the secondary refining process ) The refining time in the secondary refining process (tR1), the time from the end of the secondary refining process to the start of continuous casting (tRC), and the time from the start of continuous casting to the representative temperature measurement of casting (tC2) are represented by the following formulas. Calculated from
[0037]
[Equation 1]
tL1 = t2-t1 (1)
tL2 = t3-t2 (2)
tL3 = t4-t3 (3)
tLR = t5-t4 (4)
tR1 = t6-t5 (5)
tRC = t7−t6 (6)
tC1 = t9−t7 (7)
tC2 = t8−t7 (8)
Of these elapsed times, tL2, tL3, tLR, tR1, tRC, and tC2 are used by the molten steel temperature calculation unit 47.
[0038]
Next, the method of calculating the temperature drop of the molten steel caused by the molten steel vessel in the molten steel drop temperature calculating section 43 is described as follows. The molten steel drop temperature caused by the following will be described with reference to FIG. That is, the heat transfer phenomenon is calculated by disassembling the molten steel pot into an upper portion, a side wall portion, and a bottom portion.
Calculation conditions such as the number of layers from the inner side (closer side to molten steel) to the outside and their materials, temperature calculation points in the thickness direction, etc., and when the pan is allowed to cool (a), when the pan is pre-heated (b), and when the pot is lit The boundary conditions in (c) are determined for the top, side wall, and bottom.
[0039]
When there is molten steel in the molten steel pot as shown in FIG. 8 (c), the temperature drop ΔT of the molten steel due to the heat transfer from the molten steel to the object in contact with the molten steel is determined as the total of each part. You.
[0040]

Here, i = 1 indicates the upper portion, i = 2 indicates the side wall portion, and i = 3 indicates the bottom portion. The objects are the refractory in the molten steel pot and the slag in the elongating pot. further,
q: Heat transfer from molten steel to the object surface (per unit area) per unit time
dt: minute time
S: Total area of object in contact with molten steel
H: Heat capacity of molten steel
W: Weight of molten steel
It is.
[0041]
Then, q treats the heat transfer phenomenon at the boundary between the molten steel and the surface of the object as heat conduction, and is expressed by the following equation for each of the top, side wall, and bottom of the molten steel pot.
[0042]
[Equation 3]
q _i = Λ / c / ρ × (Tn−To) (10)
here,
λ: Thermal conductivity of the object in contact with molten steel
c: Specific heat of the object in contact with molten steel
ρ: Density of the object in contact with molten steel
Tn: Body surface temperature on the side (inside) in contact with molten steel (° C)
To: molten steel temperature (° C)
It is.
[0043]
Tn is defined as the initial value of the object temperature at a plurality of points (temperature calculation points) in the thickness direction of the object from the inside to the outside of the molten steel pot at a certain time, using the temperature calculation point interval in the thickness direction, thermal conductivity, specific heat, and density. The object temperature after a minute time is calculated from the one-dimensional unsteady heat transfer difference equation. At this time, the temperature calculation point interval in the thickness direction is calculated from the thickness of the object.
[0044]
Further, when there are a plurality of layers of refractory and steel having different materials from the inside to the outside such as a side wall (i = 2) and a bottom (i = 3) in FIG. A plurality of calculation points (temperature calculation points) are set, and the object temperature in the thickness direction is calculated using the thickness of each layer, thermal conductivity, specific heat, density, and boundary conditions between layers.
Further, radiation from the object to the atmosphere is present on the outer surface, and the heat transfer amount qh treats the heat transfer phenomenon at the boundary between the object and the atmosphere as heat radiation and is expressed by the following equation.
[0045]
[Equation 4]
qh _i = Σ × ε × ｛Tg ⁴ -To ⁴ ｝… (11)
here,
σ: Stefan-Boltzmann constant
ε: emissivity
Tg: outside object surface temperature (° C)
To: Atmospheric temperature (° C)
It is.
[0046]
On the other hand, when there is no molten steel in the molten steel pot as in the case of cooling the empty pot (a) and the preheating of the empty pot (b) in FIG. 8, the heat transfer calculation is not executed for the upper part (i = 1).
Using the inner boundary conditions for the side wall (i = 2) and the bottom (i = 3), and using the equation (11) as the heat radiation to the atmosphere when the empty pan is allowed to cool (a), preheating the empty pan At the time (b), the temperature distribution is calculated by replacing To in equation (11) with Ty (temperature (° C.) of the preheating gas) as heat radiation from the preheating gas.
[0047]
When a pot lid is hung as shown in FIG. 9, the temperature distribution in the thickness direction is added to the pot lid and calculated. Equation (11) is used as the outer boundary condition at that time. As the inner boundary condition, when the empty pot is left to cool (a), To in Equation (11) is replaced with Ttn (the inner surface temperature of the pot bottom (° C.)) as heat radiation to the inner surface (bottom inner surface) of the pot bottom. At the time of the slag, (c) replaces To in equation (11) with Tsg (slag outer surface temperature (° C.)) as heat radiation to the outer surface of the slag, and changes the temperature distribution after a short time and changes the molten steel to the molten steel pot. Calculate the heat transfer amount. In addition, during the preheating of the empty pot (b), the same calculation as in FIG.
[0048]
Then, a temperature drop is calculated based on the scheduled time of each process stored in the data storage unit 1.
For example, based on the converter blowing start time (t1), the converter tapping start time (t2), the converter tapping end time (t3), and the converter backside work end time (t4), the converter tapping is performed. Is calculated as follows. It is assumed that the pot is not used from the current time until it is used for Charge A.
[0049]
The temperature distribution in the thickness direction of the molten steel pot of the pot number for processing the charge A stored in the molten steel container temperature distribution storage unit 46 at the current time is set as an initial value, and FIG. , (B) or the temperature distribution is calculated every 10 seconds according to the boundary conditions shown in FIGS. 9 (a) and 9 (b). calculate.
[0050]
Further, using the temperature distribution at this time (t2) as an initial value, the temperature distribution is calculated in units of 2.5 seconds in accordance with the boundary conditions shown in FIG. 8C, and the molten steel drop temperature ΔT is obtained from the equations (9) and (10). Is calculated, and the integrated value up to the converter tapping end time (t3) is set to ΔTnsL2.
By the same calculation, the molten steel drop temperature ΔTnsL3 during the hearth work and the molten steel fall temperature ΔTnsLR from the end of the converter hearth work to the start of the secondary refining process are calculated.
[0051]
If there is a charge using the same number of pots from the current time to the time of use at charge A, the temperature distribution at time t2, t3, and t4 for that charge, and the temperature at charge t at time t2 for charge A Based on the distribution, ΔTnsL2, ΔTnsL3, and ΔTnsLR for charge A are calculated.
The above calculation method is the same for the converter, the secondary refining tank, and the tundish. The scheduled time of the converter process in the data storage unit 1, that is, the converter furnace blowing start time (t1), From the converter tapping start time (t2) and the converter tapping end time (t3), the molten steel drop temperature ΔTlL2 caused by the converter body during tapping is calculated as follows.
[0052]
That is, as shown in FIG. 10, the heat transfer phenomenon of the converter furnace body is decomposed into a straight body portion and a bottom portion. The molten steel vessel temperature distribution storage unit at the current time of the converter furnace body that is the same as the converter furnace body that processes the charge A, according to the boundary condition at the time of charging the molten iron and the completion of tapping when the molten steel exists at the time (c). Using the temperature distribution in the thickness direction of the converter body stored in the converter 46 as an initial value, the thickness at the time t1 in units of a minute 10 seconds in accordance with the boundary conditions shown in FIGS. Calculate the temperature distribution in the direction.
[0053]
Further, the temperature distribution in the thickness direction at time t2 is calculated in units of minute time 2.5 seconds in accordance with the boundary condition of FIG. 10C. Then, using this temperature distribution as an initial value, the temperature distribution is calculated in units of a minute time of 2.5 seconds in accordance with the boundary condition of FIG. The integrated value of this △ T up to t3 is the molten steel falling temperature △ TlL2 caused by the converter body during converter tapping.
[0054]
From the scheduled time of the secondary refining process stored in the data storage unit 1, that is, the secondary refining process start time (t5) and the secondary refining process end time (t6), the secondary refining process tank during the secondary refining process Is calculated as follows.
As for the secondary refining treatment tank, as shown in FIG. 11, the heat transfer phenomenon is decomposed into a lower tank and a dip tube, and when the empty tank is allowed to cool (a), when the empty tank is preheated (b), the elongation tank (secondary refining) According to the boundary condition at the time (c) (from the start of the secondary refining process to the end of the secondary refining process in which molten steel is present in the processing tank), the temperature drop of the molten steel caused by the secondary refining process tank during the secondary refining process ΔTrR1 Is calculated.
[0055]
From the scheduled time of the continuous casting process stored in the data storage unit 41, that is, the continuous casting start time (t7), the representative casting temperature measurement time (t8), and the continuous casting end time (t9), the continuous casting is started. Therefore, the molten steel drop temperature ΔTcC2 caused by the tundish during the casting representative temperature measurement is calculated as follows. As for the tundish, as shown in FIG. 12, the heat transfer phenomenon was applied only to the side wall portion, and the tundish was allowed to cool (a), the tundish was pre-heated (b), and the tundish (the molten steel was present in the tundish). (From the start of the casting of the continuous casting top pan to the end of the casting of the continuous casting final pan) and the falling temperature due to the tundish from the start of the continuous casting to the representative temperature measurement of the tundish 時 TcC2 according to the boundary condition of (c). Is calculated.
[0056]
In addition, for the converter furnace body, the completion of tapping of the converter of the charge A from the converter processor 53 via the higher-level computer 51 is performed. For the secondary refining processing tank, the secondary refining process controller 54 transmits the higher-level computer 51 via the higher-level computer 51. When the end of the secondary refining process of the charge A is received and the end of the continuous casting of the final casting pot is continuously received from the continuous casting computer 55 via the host computer 11 for the tundish, the temperature distribution of the molten steel container is stored at the current time. Using the temperature distribution of the same vessel stored in the unit 6 as an initial value, at the time of the end of the tapping of the converter of the charge A for the converter furnace body, and at the end of the secondary refining of the charge A for the secondary refining treatment tank, For the tundish, the temperature distribution is continuously calculated until the end of the casting of the final casting pan, the temperature distribution in the molten steel container temperature distribution storage unit 6 is updated and stored, and the temperature distribution when the same container is used next time is stored. Used as a period value.
[0057]
The temperature drop ΔTnsL2, ΔTnsL3, ΔTnsLR, ΔTnsR1, ΔTnsRC, ΔTnsC2, ΔTlL2, ΔTrR1, ΔTcC2 calculated as described above are used in the various molten steel target temperature calculators 47.
The calculation method in the component-by-component input amount calculation unit 44 will be described by taking the input material during converter tapping as an example.
[0058]
First, the amount of C, Si, Mn, Al, CaO for each of the sub-materials stored in the data storage unit 1 is added to the input amount for each of the plurality of types of sub-materials used for the charge A stored in the data storage unit 41. Then, the content of each component is calculated by multiplying the content of each of the other components by the content of each component. Next, the amount of each component to be added for each auxiliary material is calculated based on the content of each component to be included in the steel type to be manufactured, and the total amount of each component is added to obtain the C component input amount MLc and the Si component input amount MLsi, Mn component input amount MLmn, Al component input amount MLal, CaO component input amount MLcao, and other input amount MLetc are calculated and used in various molten steel target temperature calculation units 47.
[0059]
Similarly, for the input material during the secondary refining process, the component input amounts MRc, MRsi, MRmn, MRal, MRcao, MRetc for each component are calculated and used by the various molten steel target temperature calculation units 47.
Further, according to the molten steel target temperature calculation model, the secondary refining process end target temperature, the secondary refining process start target temperature, the target temperature on the pot, The target temperature is calculated up to the target temperature, and the converter blows to the target temperature.
[0060]
After the tapping and hearth work, until the start of the secondary refining process, the elapsed time calculation unit 42, the molten steel temperature drop amount calculation unit 43, the component-based input amount calculation unit 44, reflecting the progress of the charge A process, After recalculating the numerical value of each item in the steel type correction value calculating section 45, the secondary refining process end target temperature is calculated again with the casting target temperature as a starting point, thereby improving the accuracy of the target temperature.
[0061]
Lastly, a method of determining the weight coefficients, threshold values, and steel type correction values Ha to Hd of NNs 471 to 474 will be described.
First, the temperature management computer 52 stores the data of the past operating conditions received from the host computer 51 in the data storage unit 1. Next, the elapsed time calculation unit 42 calculates tL2, tL3, tLR, tR1, tRC, and tC2 from the actual times t1 to t9, and the number and actual time of each molten steel container used in the molten steel temperature drop amount calculation unit 43. From t1 to t9, temperature drops △ TnsL2, △ TnsL3, △ TnsLR, △ TnsR1, △ TnsRC, △ TnsC2, △ TlL2, △ TrR1, and △ TcC2 are calculated.
[0062]
Further, the component-based input amount calculation unit 44 calculates the component-based input amounts MLc, MLsi, MLmn, MLal, MLcao, MLetc, MRc, MRsi, MRmn, MRal, and MRetc.
For the NN 471, the input state quantity of the NN 471 is input to the input layer from among the various data described above. For the casting target temperature, an actual measured casting representative temperature is input. Further, by inputting and learning the actually measured secondary refining process end temperature in the output layer, the weight coefficient and the threshold value are determined and stored.
[0063]
All charges used for learning the weight coefficient and the threshold value of the NN471 are classified into a plurality of steel type groups, and the average value of the difference between the estimated temperature of the secondary refining process end temperature and the measured temperature for each steel type group is corrected by the steel type correction. The value is stored as a steel type correction value Ha for each steel type group in the section 475. This time, the steel type correction value is calculated by classifying all charges into 46 types.
[0064]
For NNs 472, 473, and 474, weighting factors, threshold values, and steel type correction coefficients Hb, Hc, and Hd are determined and stored in the same manner as NN471. All of them are classified into 46 types and the steel type correction values Hb, Hc, Hd are determined.
FIG. 13 is an explanatory diagram of the effect of the temperature estimation system according to the present invention. FIG. 13A shows the difference between the estimated temperature and the measured temperature output from the NN for all the charges on the horizontal axis, and the number of charges on the vertical axis. It is a frequency distribution diagram.
[0065]
(B) is a diagram in which all charges are classified into four groups according to steel types and frequency distribution charts are re-taken. The average value of each steel type group is regarded as a so-called different temperature drop tendency depending on the steel type, and the estimated temperature output from the NN is calculated. Is corrected by this average value.
(C) is a frequency distribution as a result of correcting all charges, and it can be seen that the estimation accuracy is greatly improved as compared with (a).
[0066]
【The invention's effect】
According to the state quantity estimating method according to the first aspect of the invention, the number of input state quantities of each NN is reduced, and learning is facilitated by estimating the number of input state quantities of each NN by connecting the NN models of the respective steps in series to model the manufacturing process. Accuracy can be improved.
According to the state quantity estimating method according to the second invention, the number of input state quantities of each NN can be further reduced by using a mathematical model as a part of a model of each process.
[0067]
According to the state quantity estimating method according to the third invention, the estimation quantity is further improved by correcting the estimated state quantity using a correction quantity that can be converted into a database in advance according to the type of the product. It is possible to do.
According to the state quantity estimating method according to the fourth invention, it is possible to accurately estimate the blow-end molten steel temperature in the converter based on the target cast molten steel temperature predetermined in the steelmaking process.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a three-layer NN.
FIG. 2 is a flow chart of a steel making process.
FIG. 3 is a graph showing a change in molten steel temperature.
FIG. 4 is a functional diagram (1/2) of a molten steel temperature estimation system.
FIG. 5 is a functional diagram (2/2) of the molten steel temperature estimation system.
FIG. 6 is a hardware configuration diagram of a molten steel temperature estimation system.
FIG. 7 is an explanatory diagram of an elapsed time calculation method.
FIG. 8 is an explanatory diagram of a heat transfer phenomenon of a molten steel pot (without a pot lid).
FIG. 9 is an explanatory diagram of a heat transfer phenomenon of a molten steel pot (with a pot lid).
FIG. 10 is an explanatory diagram of a heat transfer phenomenon of a converter.
FIG. 11 is an explanatory diagram of a heat transfer phenomenon in a secondary refining processing tank.
FIG. 12 is an explanatory diagram of a heat transfer phenomenon of a dandysh.
FIG. 13 is an explanatory diagram of an effect of the present invention.
[Explanation of symbols]
41 Data storage unit
42 ... Elapsed time calculation unit
43: Molten steel drop temperature calculator due to molten steel container
44 ... Input amount calculation unit for each component
45 ... Steel type correction value calculation unit
46: Molten steel container temperature distribution storage
47 ... Steel temperature calculator
471: Second transport process network model
472: Secondary refining process network model
473: First transport process network model
474: Steel tapping network model
475, 476, 477, 478 ... Steel type correction unit

Claims (4)

A first step based on an input state quantity of a last step of a product manufacturing process including a first step, a last step, and at least one intermediate step disposed between the first step and the last step. A method for estimating a state quantity of a manufacturing process for estimating an output state quantity of
An output state quantity measuring step of measuring a measured output state quantity that is a measurable output state quantity among output state quantities that are state quantities after processing in each of the processes,
The output state quantity other than the measured output state quantity determined based on the measured output state quantity measured in the output state quantity measurement step and the input state quantity which is the state quantity before processing in the downstream process of each step, And an input state quantity estimating step of estimating an input state quantity of the process based on a neural network model of the process,
The manufacturing process is repeated from the final process to the earliest process for each process, and an output state amount other than the measured output state amount of the earliest process is estimated based on the input state amount of the downstream process. Method of estimating state quantity of manufacturing process. The input state quantity estimation step includes:
The output state quantity measured in the output state quantity measurement step as an input, including an output state quantity calculation step of calculating a part of the output state quantity other than the measured output state quantity based on a mathematical model,
The measured output state amount measured in the output state amount measuring step, the measured output state amount determined based on the calculated output state amount calculated in the output state amount calculating step and the input state amount of the downstream process of each step. 2. The state quantity estimation of the manufacturing process according to claim 1, wherein the quantity and an output state quantity other than the calculated output state quantity are input and the input state quantity of the step is estimated based on a neural network model of the step. Method. The input state quantity estimation step includes:
The input state quantity in the downstream process of each of the above steps is corrected with a correction value predetermined for each type of product manufactured in the manufacturing process to determine an output state quantity other than the measured output state quantity and the calculated output state quantity. The method for estimating a state quantity of a manufacturing process according to claim 1, wherein the method is performed. The molten steel blown in the converter is subjected to a secondary refining process in a secondary refining process, and a process for casting in the casting process is performed based on a target casting temperature which is a target molten steel temperature before the start of casting. A state quantity estimating method of a manufacturing process for estimating a blow-stop molten steel temperature that is a molten steel temperature,
The time from the start of continuous casting calculated using the mathematical model to the start of the temperature measurement of the representative casting temperature, the processing time of the second transporting step, the temperature drop within each time calculated using the mathematical model, and the manufacturing A secondary refining process outlet molten steel temperature estimating step of estimating a secondary refining process exit molten steel temperature using a second transport process neural network model based on a target casting molten steel temperature predetermined according to a steel type to be subjected to
A secondary refining process outlet molten steel temperature correction step of correcting the secondary refining process outlet molten steel temperature estimated in the secondary refining process outlet molten steel temperature estimation step according to a steel type to be manufactured;
The processing time of the secondary refining process calculated using the mathematical model, the temperature drop during the secondary refining process calculated using the mathematical model, and the temperature corrected in the secondary refining process outlet molten steel temperature correction step. Estimating a secondary refining process inlet molten steel temperature using the secondary refining process neural network model based on the secondary refining process outlet molten steel temperature;
A secondary refining process inlet molten steel temperature correction step of correcting the secondary refining process inlet molten steel temperature estimated in the secondary refining process inlet molten steel temperature estimation step according to a steel type to be manufactured;
The time from the end of converter tapping to the end of the hearth operation calculated using the mathematical model, the time from the end of the hearth operation to the start of secondary refining, and the drop temperature for each time calculated using the mathematical model And estimating the molten steel temperature on the pot using the neural network model for the first transporting process based on the molten steel temperature on the secondary refining process inlet corrected in the molten steel temperature correction step on the secondary refining process. The estimation stage;
The molten steel temperature on the pot to correct the molten steel temperature on the pot estimated in the molten steel temperature on the pot estimation step according to the steel type to be manufactured,
The time from the start of converter tapping to the end of tapping calculated using the mathematical model, the temperature drop during that time calculated using the mathematical model, and the temperature corrected in the ladle-on molten steel temperature correction step. Estimating the blow-end molten steel temperature using a neural network model for tapping process based on the molten steel temperature on the pan,
A blow-end molten steel temperature correction step of correcting the blow-end molten steel temperature estimated in the blow-end molten steel temperature estimation step according to a steel type to be manufactured,
A method for estimating the state quantity of a manufacturing process.

Images