JP4031608B2 - Material life evaluation method - Google Patents

Description

【００４０】
次いで、各母材（Ｎｏ．１〜３）とも、２つの母材を突き合わせて該突き合わせ部分を被覆アーク溶接法により溶接し、溶接継手（供試材：Ｎｏ．１〜３）とした。
被覆アーク溶接後の溶接後熱処理条件は、次のとおりである。
母材（Ｎｏ．１〜３）：７１５℃×１５分／ＡＣ
【００４１】
次いで、これらの供試材（Ｎｏ．１〜３）各々に様々な条件で熱処理を施し、寿命評価のための試料とした。
熱処理における加熱温度（最高保持温度）及び加熱時間（前記加熱温度における保持時間）は、各供試材（Ｎｏ．１〜３）とも、次の通りとした。
加熱温度：５５０℃、５７５℃、６００℃、６２５℃、６５０℃の５点
加熱時間：上記加熱時間各々において、３０ｈ（時間）、１００ｈ、３００ｈ
１０００ｈ、３０００ｈ、１００００ｈの６点
これにより、試料点数の合計は、９３点〔＝｛未処理材（１点）＋熱処理材（温度５点×時間６点）｝×供試材（３種類）〕となった。
【００４２】
次いで、熱処理を施した試料（Ｎｏ．１〜３）各々の表面からレプリカを採取し、このレプリカを光学顕微鏡を用いて観察し、その光学顕微鏡組織を撮影するとともに、これらの光学顕微鏡組織像を従来技術で既に説明したラーソン・ミラー・パラメータ法を用いて分類した。なお、パラメータ定数Ｃの値は、各試料（Ｎｏ．１〜３）ともに２０とした。
【００４３】
次いで、この光学顕微鏡組織の画像処理を図１に基づき実施した。
Ａ．原画像（光学顕微鏡組織像）の変換
この画像処理を実施するに先だって、原画像（光学顕微鏡組織像）として、グレースケール２５６階調で、８００ピクセル（画素）×５８０ピクセル（画素）の大きさの画像を用いた。
【００４４】
この原画像に画像処理を施す際の着目点は、次の通りである。
（ａ）白色領域及び黒色領域で表される２相の金属組織を有し、外部からの加熱等の高温での使用による劣化（経時変化）により前記２相間のコントラスト比が低下する。
（ｂ）白色領域では、粒状の炭化物が成長して視認できるようになり、全体として濃度が増加する。
（ｃ）黒色領域では、ラスと称される線状構造が減少する。ただし、ばらつきは大きい。
【００４５】
上記の着目点を念頭に入れて、図１の画像処理を示す流れ図にしたがって「黒領域画像」と「白領域画像」をそれぞれ抽出した。
Ｂ．原画像の２値化
（１）いき値処理
前記変換を施した原画像に対して、２値化いき値決定法として最もよく知られている正準判別法を用いていき値処理を施し、２値化像とした。
【００４６】
例えば、この原画像の濃淡情報を、あるレベルの濃淡度（いき値）で区切り、このレベルより濃い領域（黒色領域）に”１”を、このレベルより淡い領域に”０”を、それぞれ割り当てれば、前記レベルより濃い領域（黒色領域）のみが２値画像として表示される。
低合金ＣｒＭｏ鋼では、フェライト領域が白領域、ベイナイト（またはパーライト）が黒領域となる。
【００４７】
このいき値は、例えば、あるいき値によって画像を２つのクラス（０クラスと１クラス）に分割した場合に、クラス間分散σ_B ²(k)が最大になるいき値ｋを選ぶ。このクラス間分散σ_B ²(k)は次式で表される。
σ_B ²(k)＝ω₀（μ₀−μ_T）²＋ω₁（μ₁−μ_T）²
【００４８】
ここで、
【数１】

【数２】

【数３】

【数４】

【数５】

Ｐ_i＝ｎ_i／Ｎ
ω₀＋ω₁＝１
である。
【００４９】
ただし、
ｉ：濃淡レベル（階調）（ｉ＝１〜Ｌの間の整数）
ｎ_i：レベルｉの画素数
Ｎ：全画素数
Ｐ_i：輝度１〜ｋを０クラス（黒領域）、輝度ｋ＋１〜Ｌを１クラス（白領域） とした場合の、輝度ｉの点が全体に占める面積比率
ω₀：０クラスの面積比率
ω₁：１クラスの面積比率
μ₀：０クラスの平均輝度
μ₁：１クラスの平均輝度
μ_T：画像全体の平均輝度
である。
【００５０】
（２）２値化像の反転
いき値処理で得られた２値化像を反転して黒領域を抽出する。
２値化像を反転する理由は、本実施形態においては、２５６階調のモノクロ画像に変換する際に、黒（0）を無、白（255）を有として処理しているために、いき値処理で得られた２値化像を、白領域（フェライト領域）に注目して処理を進める場合にはこのまま用いればよいが、黒領域（ベイナイト（またはパーライト）領域）に注目して処理を進めるためには、得られた２値化像を反転して黒領域を抽出する必要があるためである。
【００５１】
Ｃ．白領域抽出
（１）微小ラベルの除去
白領域（フェライト領域）内に点在する黒点状の炭化物を除去するために、ある面積いき値以下の領域を消去した。ここでは、面積いき値を２００画素に設定したので、白領域内の２００画素以下の面積の黒点が除去された。
【００５２】
（２）黒領域の膨張・収縮処理
黒領域内の隙間を穴埋めするために、黒領域に対して膨張・収縮処理（クロージング：closing）を施した。
膨張処理は、ｎ×ｎ（ｎは奇数）のフィルタを用意し、２値化後の画像の全ての画素に対して、例えば３×３のフィルタをあて、黒画素の１つ隣の白画素を黒画素化する処理であり、最大値フィルタ処理ともいわれる。
【００５３】
例えば、３×３の画素のうち、中心の画素を囲む８つの画素の中で最も輝度が高いもの（黒い画素）があれば、この画素の輝度を中心の画素の輝度とする。ここでは、２値化後に最大値フィルタをかけることとなるので、輝度は０、２５５のいずれかしか採り得ない。したがって、中心の画素を囲む８つの画素の中に１つでも黒い画素があれば、中心の画素も黒い画素となる。
ここでは、５×５のフィルタを用いて膨張処理（最大値フィルタ処理）を２回実施した。
【００５４】
収縮処理は、膨張処理の逆操作であり、黒画素の部分を薄くはがす機能を有し、最小値フィルタ処理ともいわれる。
ここでは、５×５のフィルタを用いて収縮処理（最小値フィルタ処理）を２回実施した。
【００５５】
この膨張・収縮処理では、最大値フィルタ及び最小値フィルタを用いて各領域内の白点／黒点にあたる隙間を埋めて、１つの領域として抽出した。その理由は、単純な２値化では、黒領域（ベイナイト）の白点（ラス構造）や、白領域（フェライト）の黒点（炭化物粒子）のように、抽出された領域が相互に入り交じっているが、例えば、白領域の黒点は、上述した着目点（ｂ）に当たる重要な情報であるために、この白領域は「１つの白領域」として確実に抽出する必要があるからである。
【００５６】
黒領域に対して膨張・収縮処理を施した結果、黒領域内の小さな白い部分が黒で埋められ、黒領域と白領域との境界領域の凹凸が均されて平になった。
低合金ＣｒＭｏ鋼では、ベイナイト領域は黒領域であるから、ベイナイト領域内の線状構造であるラス構造を消失させることができる。
【００５７】
ただし、「白領域」内の析出物である炭化物粒子の密度が高い場合、膨張処理によって「黒領域」として誤って抽出する可能性が生じるために、膨張処理の前に、析出する炭化物に相当する微小ラベルの除去を実施した。
【００５８】
（３）最大値フィルタ処理
白領域画像を絞り込む（確実に白領域とする）ために、最大値フィルタを用いて黒領域を膨張させた。
【００５９】
（４）輝度反転
境界線を含む黒領域を反転させて白領域とした。
輝度反転を施す前までは、例えば、低合金ＣｒＭｏ鋼では、境界線を含むベイナイト領域が黒領域（画面上では白（輝度２５５））、フェライト領域が白領域（画面上では黒（輝度０））になっている。
次のマスキングでは、白（輝度２５５）の部分（ＯＮの部分）が残ることになるので、黒（輝度０）のフェライト領域を輝度反転して白（輝度２５５）とすることで、該フェライト領域をＯＮとした。
【００６０】
要約すれば、「白領域」の抽出は、単純な２値化で得られた「黒領域」に対し、膨張・収縮処理、最大値フィルタ処理により「黒領域」を最大限に膨張させ、残った部分を「白領域」とすることで完了した。
【００６１】
（５）マスキング
輝度反転した２値画像と原画像をマスキングすることにより、輝度反転して白（輝度２５５）となったフェライト領域のみを抽出した画像とした。
これにより、白領域画像の作成が完了した。
【００６２】
（６）メディアンフィルタ処理
上述した最大値フィルタ処理と同様、５×５のフィルタを用いて、フィルタ内メディアン（中央値）を代表値として（フイルタの中央の画素の新輝度として）置き換えた。
これにより、フィルタ内に突出した輝度の点があったとしても、その影響を緩和することができる。換言すれば、画像中の輝度の急峻な変化が抑制され、なだらかな輝度変化（画像の低周波成分）を取り出すことができる。
【００６３】
（７）差分処理
メディアンフィルタ処理後の画像から、マスキングにより得られた白領域画像を差引き（各画素の輝度の差をとり）、輝度の変化が急峻な部分、つまり析出物と思われる部分を抽出し、高周波成分とする。
【００６４】
（８）再２値化
さらに「白領域」から析出炭化物を抽出するために、再度２値化をおこなった。
低周波成分による悪影響を抑えるために、差分処理により高周波成分のみを残した画像を用いて、正準判別法による２値化をおこなった。
以上により、２値化した炭化物画像を作成することができた。
【００６５】
Ｄ．黒領域抽出
（１）黒領域の膨張→収縮処理
ベイナイト領域（黒領域）内の隙間であるラス線状構造（白領域）を穴埋めするために、黒領域に対して膨張・収縮処理（クロージング：closing）を施した。
ここでは、５×５のフィルタを用いて膨張処理（最大値フィルタ処理）を２回実施した。
その後、５×５のフィルタを用いて収縮処理（最小値フィルタ処理）を２回実施した。
この膨張・収縮処理では、最大値フィルタ及び最小値フィルタを用いて各領域内の白点／黒点にあたる隙間、すなわちラス線状構造（白領域）を埋めて、１つの領域として抽出した。
【００６６】
（２）黒領域の収縮・膨張処理
２相からなる材料では一般に、画像は性質の異なる「黒領域」、「白領域」及び「境界領域」の３つの領域から構成されると考えられている。これらの領域のうち「境界領域」は、単純な２値化では「黒領域」に合わさって抽出されるので、「黒領域」のみを単独で抽出することは難しい。そこで「黒領域」に収縮・膨張処理を施すことにより、「黒領域」から「境界領域」を除去し、黒領域のみを抽出した。
【００６７】
ここでは、ベイナイト領域（黒領域）の境界線である細い線状構造を除去するために、黒領域に対して収縮・膨張処理（オープニング：opening）を施した。
例えば、５×５のフィルタを用いて収縮処理（最小値フィルタ処理）を２回実施し、その後、５×５のフィルタを用いて膨張処理（最大値フィルタ処理）を２回実施した。
この結果、この収縮・膨張処理では、最小値フィルタ及び最大値フィルタを用いて各領域の境界線である細い線状部分が除去され、黒領域のみを抽出することができた。
【００６８】
（３）マスキング
収縮→膨張処理により得られた２値画像では、境界線が除去されたベイナイト領域が黒領域（画面上では白（輝度２５５））、フェライト領域が白領域（画面上では黒）になっているので、この２値画像と原画像のマスキングすることにより、黒領域であるベイナイト領域のみを抽出した画像とした。
これにより、黒領域画像の作成が完了した。
【００６９】
（４）再２値化
さらに、マスキングにより黒領域であるベイナイト領域のみを抽出した画像を用いて、正準判別法による２値化をおこなった。
以上により、２値化したラス構造の画像を作成することができた。
【００７０】
Ｅ．特徴量の計算
上記により得られた白領域画像、黒領域画像、炭化物画像、ラス画像を用いて、次の７つの特徴量を計算した。
（１）「黒領域」と「白領域」のコントラスト比に基づくもの
ａ．領域間クラス分離度
黒領域と白領域との間のクラス間分散を全体の分散で除したものであり、０〜１の間の値を取る。この分離度は、１に近い程クラス間の分離が良い、すなわちコントラスト比が高いことを表している。
【００７１】
領域間クラス分離度η_Aは次式で表される。
η_A＝σ_BA ²／σ_BW ²
だだし、σ_BA ²は黒領域、白領域を２つのクラスとしたときのクラス間分散、σ_BW ²は黒白両領域合わせた全体の輝度分散である。
【００７２】
なお、クラス間分散σ_BA ²は次式で表される。
【数６】

ただし、ｎ_B：黒領域の画素数
ｎ_W：白領域の画素数
μ_B：黒領域の平均輝度
μ_W：白領域の平均輝度
μ_BW：黒白両領域を合わせた領域の平均輝度
である。
【００７３】
ｂ．正準判別クラス分離度
ヒストグラムの正準判別法に基づくクラス間分散を全体の分散で除したものであり、いわゆる大津の方法により輝度ヒストグラム上でいき値を求め、該いき値を境に輝度ヒストグラムの上下それぞれに属する画素群（２つの画素群）をそれぞれのクラスと見なしたもので、０〜１の間の値を取る。この分離度は、１に近い程クラス間の分離が良い、すなわちコントラスト比が高いことを表している。
【００７４】
正準判別クラス分離度η₀は次式で表される。
η₀＝σ_BO ²／σ_T ²
だだし、σ_BO ²は画像を輝度いき値ｋで２つのクラスに分割したときのクラス間分散、σ_T ²は画像全体の輝度分散である。
【００７５】
なお、クラス間分散σ_BO ²は次式で表される。
【数７】

ただし、ｎ_BO：輝度が大津いき値以下の領域の画素数
ｎ_WO：輝度が大津いき値より大きい領域の画素数
Ｎ（＝ｎ_BO＋ｎ_WO）：画像全体の画素数
μ_BO：輝度が大津いき値以下の領域の平均輝度
μ_WO：輝度が大津いき値より大きい領域の平均輝度
μ_T：画像全体の平均輝度
である。
【００７６】
ｃ．領域間コントラスト差
白領域の輝度平均と黒領域の輝度平均との差を求め、この差を全体の階調数２５６で除したものである。
領域間コントラスト差Ｗ_A-Bは次式で表される。
Ｗ_A-B＝（μ_W−μ_B）／２５６
【００７７】
ｄ．全体の輝度標準偏差／クラス内の輝度標準偏差
黒白両領域を含む全体の輝度標準偏差σ_W+Bをクラス内の輝度標準偏差
σ_classで除したものである。
輝度標準偏差σ_W+Bは次式で表される。
【数８】

【００７８】
クラス内の輝度標準偏差σ_classは、クラス内のばらつきを数量化するために、各画素についてその画素の所属するクラスの平均輝度（白領域なら白領域の平均輝度）との差を計算し、この差よりクラス内の輝度の標準偏差を求めたものである。このクラス内の輝度標準偏差σ_classは次式で表される。
【数９】

【００７９】
ただし、Ｐ_i：画素ｉの輝度値
ｎ_W：白領域の画素数
ｎ_B：黒領域の画素数
（ｎ_W＋ｎ_B≠Ｎ：画像全体の画素数）
μ_W：白領域の平均輝度
μ_B：黒領域の平均輝度
μ_W+B：白黒両領域合わせた領域の平均輝度
ｉ∈Ｗ：白領域の画素
ｉ∈Ｂ：黒領域の画素
【００８０】
ｅ．クラス間輝度差／クラス内の輝度標準偏差
白のクラスの平均輝度と黒のクラスの平均輝度の差μ_W−μ_Bをクラス内の輝度標準偏差σ_classで除したものである。
ｆ．全体の輝度標準偏差
白領域と黒領域の両領域の輝度の標準偏差σ_W+Bであり、既に説明したものである。
【００８１】
（２）「白領域」の析出炭化物に基づくもの
ｇ．白領域輝度分散
白領域の輝度の分散状態を示す指標であり、炭化物の析出状況による輝度分散の有為差の有無を表したものである。
白領域輝度分散σ_W ²は次式で表される。
【数１０】

ただし、Ｎ_W：白領域の画素数
Ｐ_i（ｉ∈ｗ）：白領域に属する各画素の輝度
Ｐ_W：白領域の平均輝度
【００８２】
ｈ．白領域内炭化物密度
炭化物画像（白領域を再２値化した画像）の黒点面積の総和を白領域の総面積で除したものである。
【００８３】
（３）「黒領域」のラス構造に基づくもの
ｉ．黒領域輝度分散
黒領域の輝度の分散状態を示す指標であり、ラス構造による輝度分散の有為差の有無を表したものである。白領域輝度分散と同様に計算する。
【００８４】
（４）その他
ｊ．黒面積／白面積
「黒領域」の面積を「白領域」の面積で除したものであり、実施された画像処理が妥当なものか否かを判定する指標となる。
【００８５】
これらの特徴量は、２つの相間の濃淡の度合い、すなわち母材の劣化の度合いを数値化したものであるから、これらの特徴量を用いることにより母材の劣化の度合い（経時変化の度合い）が定量化される。これにより、組織像の評価に際し個人の主観が入る余地が無くなり、母材の劣化の度合いを客観的に評価することが可能になる。また、この評価方法はシステム化できるので、評価の過程を自動化することが可能である。
【００８６】
図２〜図５は、上記で求められた特徴量の処理結果を示す図であり、Ｎｏ．１の試料（供試材：Ｎｏ．１）を例にとり、それぞれの特徴量に対して、ＬＭパラメータと画像処理結果との相関を整理したものである。各図中の括弧（）内にＬＭパラメータとの相関係数を表示している。なお、未使用材のＬＭパラメータは計算できないので、便宜上１７．０の値を用いた。
【００８７】
ここで、図２（ａ）は領域間クラス分離度を示す図、図２（ｂ）は正準判別クラス分離度を示す図、図３（ｃ）は領域間コントラスト差を示す図、図３（ｄ）は白領域輝度分散を示す図、図４（ｅ）は白領域内炭化物密度を示す図、図４（ｆ）は黒領域輝度分散を示す図、図５（ｇ）は黒面積／白面積を示す図である。
【００８８】
図６及び図７は、特徴量の処理結果の他の例を示す図であり、Ｎｏ．２の試料（供試材：Ｎｏ．２）を例にとり、それぞれの特徴量に対して、ＬＭパラメータと画像処理結果との相関を整理したものである。各図中の括弧（）内にＬＭパラメータとの相関係数を表示し、未使用材のＬＭパラメータを１７．０とした点も上記Ｎｏ．１の試料と同様である。
【００８９】
ここで、図６（ａ）は全体の輝度標準偏差／クラス内の輝度標準偏差を示す図、図６（ｂ）はクラス間輝度差／クラス内の輝度標準偏差を示す図、図７（ｃ）は全体の輝度標準偏差を示す図である。
【００９０】
図８及び図９は、特徴量の処理結果の他の例を示す図であり、Ｎｏ．３の試料（供試材：Ｎｏ．３）を例にとり、それぞれの特徴量に対して、ＬＭパラメータと画像処理結果との相関を整理したものである。各図中の括弧（）内にＬＭパラメータとの相関係数を表示し、未使用材のＬＭパラメータを１７．０とした点も上記Ｎｏ．１の試料と同様である。
【００９１】
ここで、図８（ａ）は全体の輝度標準偏差／クラス内の輝度標準偏差を示す図、図８（ｂ）はクラス間輝度差／クラス内の輝度標準偏差を示す図、図９（ｃ）は全体の輝度標準偏差を示す図である。
【００９２】
本実施形態において行った画像処理の結果の一例を図１０〜図１７に示す。
図１０〜図１３は、ＬＭパラメータ値が低い（劣化度が小さい）Ｎｏ．１の試料（供試材：Ｎｏ．１）の画像処理結果を示す図であり、図１０（ａ）は原画像、図１０（ｂ）は黒領域画像、図１１（ｃ）は白領域画像、図１１（ｄ）は原画像ヒストグラム、図１２（ｅ）は領域分割画像、図１２（ｆ）は炭化物画像、図１３（ｇ）はラス画像である。
【００９３】
図１４〜図１７は、ＬＭパラメータ値が高い（劣化度が大きい）Ｎｏ．１の試料（供試材：Ｎｏ．１）の画像処理結果を示す図であり、図１４（ａ）は原画像、図１４（ｂ）は黒領域画像、図１５（ｃ）は白領域画像、図１５（ｄ）は原画像ヒストグラム、図１６（ｅ）は領域分割画像、図１６（ｆ）は炭化物画像、図１７（ｇ）はラス画像である。
【００９４】
この画像処理の結果によれば、領域間のコントラスト比は、ＬＭパラメータと高い相関があることがわかった。コントラスト比に基づく特徴量としては、（ａ）領域間クラス分離度〜（ｃ）領域間コントラスト差があるが、（ｃ）領域間コントラスト差が最も優れていることがわかった。また、（ｅ）白領域内炭化物密度は、ＬＭパラメータと相関がとれているので、析出炭化物の抽出が可能であることがわかった。
【００９５】
以上説明したように、本実施形態の材料の寿命評価方法によれば、次のような効果を奏することができる。
（１）２値化像の２つの相の濃淡の度合いを特徴量化することができ、この２つの相の濃淡の度合いに基づき材料の経時変化の度合いを推定することができる。したがって、組織像の判定に際し、個人の主観による影響を除外することができ、材料の経時変化の度合いを客観的に評価することができる。その結果、評価結果の定量化及び自動化を図ることができる。
【００９６】
（２）２値化像とした後、フェライト領域（白領域）の微小構造部分である炭化物を所定の面積いき値により除去するので、フェライト領域（白領域）に微小構造部分が発生した場合においても、２値化像の２つの相の濃淡の度合いを特徴量化することができる。
（３）２値化像を原画像に重ね合わせ、再度２値化するので、炭化物画像やラス構造の画像等のような２値化画像を得ることができる。
【００９７】
［参考例］
本発明の参考例による材料の寿命評価方法について、１相の材料として火力発電設備の伝熱管の溶接熱影響部（以下、ＨＡＺとも称する）の寿命を推定する場合を例に採り説明する。
ここでは、第１の実施形態で用いた供試材（Ｎｏ．１〜３）各々に様々な条件で熱処理を施し、溶接熱影響部（ＨＡＺ）の寿命評価のための試料とした。
熱処理における加熱温度及び加熱時間は、第１の実施形態と全く同様としたので、試料点数の合計は、第１の実施形態と全く同様である。
【００９８】
次いで、熱処理を施した試料（Ｎｏ．１〜３）各々の溶接熱影響部（ＨＡＺ）の表面からレプリカを採取し、このレプリカを光学顕微鏡を用いて観察し、その光学顕微鏡組織を撮影するとともに、これらの光学顕微鏡組織像を従来技術で既に説明したラーソン・ミラー・パラメータ法を用いて分類した。なお、パラメータ定数Ｃの値は、各試料（Ｎｏ．１〜３）ともに２０とした。
【００９９】
次いで、溶接熱影響部（ＨＡＺ）の光学顕微鏡組織の画像処理を図１８に基づき実施した。
Ａ．原画像（光学顕微鏡組織像）の変換
この画像処理を実施するに先だって、原画像（光学顕微鏡組織像）として、グレースケール２５６階調で、８００ピクセル（画素）×５８０ピクセル（画素）の大きさの画像を用いた。
【０１００】
このＨＡＺの原画像に画像処理を施す際の着目点は、次の通りである。
（ａ）材料は高温での使用を受けることにより劣化（経時変化）するが、劣化が進行するにしたがって、金属組織中のマルテンサイトラス構造（線状構造）が崩れ、線状の構造から点状の構造に変化していく。
【０１０１】
（ｂ）上記の変化は、マルテンサイト領域のみで発生し、フェライト領域では発生しない。したがって、複数視野を処理してその処理結果を平均することにより、処理対象となる画像中のフェライト領域の面積比率を一定値に近づけ、影響を低減することで対処する。
【０１０２】
上記の着目点を考慮し、図１８の画像処理を示す流れ図にしたがって、ラス構造部分（線状構造部分）の面積比率の特徴量化を行った。
Ｂ．原画像の２値化
（１）いき値処理
前記変換を施した原画像に対して、正準判別法を用いていき値処理を施し、ラス構造及び粒界線の領域と、それ以外の領域に分割された２値化像とした。
低合金ＣｒＭｏ鋼では、マルテンサイト領域及びフェライト領域が白領域、マルテンサイト領域内のラス構造部が黒領域となる。
このいき値の選び方は、第１の実施形態のいき値の選び方と同様である。
【０１０３】
（２）２値化像の反転
いき値処理で得られた２値化像を反転して黒領域を抽出する。
この２値化像を、黒領域（ラス構造部）に注目して処理を進めるためには、得られた２値化像を反転して黒領域を抽出する必要がある。
【０１０４】
（３）微小ラベルの除去
白領域（マルテンサイト領域）内のラス構造部以外の点状の領域を除去するために、ある面積いき値以下の領域を消去した。
ラス構造部は、崩れると、２値化したときに点状構造へと変化するので、この点状構造を面積でいき値を設けて除外する。
【０１０５】
ここでは、面積いき値を０画素（除外処理無し）、５０画素、２００画素に設定した。面積いき値を５０画素とすると、白領域内から５０画素以下の点状の領域が消去され、２００画素とすると、白領域内から２００画素以下の点状の領域が消去される。また、面積いき値を０画素とすると、白領域内の点状の領域は消去されないまま残る。
【０１０６】
（２）膨張・収縮処理
白領域（マルテンサイト領域）内の黒領域（ラス構造部）を領域として把握するために、黒領域に対して膨張・収縮処理（クロージング：closing）を施した。
膨張処理は、第１の実施形態で実施された膨張処理と同様であり、ここでは、５×５のフィルタを用いて膨張処理（最大値フィルタ処理）を２回実施した。
収縮処理も、第１の実施形態で実施された収縮処理と同様であり、ここでは、５×５のフィルタを用いて収縮処理（最小値フィルタ処理）を２回実施した。
【０１０７】
この膨張・収縮処理では、最大値フィルタ及び最小値フィルタを用いて白領域（マルテンサイト領域）内の黒領域（ラス構造部）を１つの領域として抽出した。
黒領域に対して膨張・収縮処理を施した結果、黒領域内の小さな白い部分が黒で埋められ、断線等の無い完全な線状の黒領域になった。
【０１０８】
Ｃ．特徴量の計算
上記により得られた２値化画像を用いて、次の６つの特徴量を計算した。
ａ．２値化後の黒領域面積
２値化した画像の黒領域の面積を全画像面積で除したものである。
ｂ．２値化後、微小ラベルの除去（５０画素以下）を施した黒領域面積
２値化して、微小黒ラベル（５０画素以下）を除去した後の黒領域の面積を全画像面積で除したものである。
【０１０９】
ｃ．２値化後、微小ラベルの除去（２００画素以下）を施した黒領域面積
２値化して、微小黒ラベル（２００画素以下）を除去した後の黒領域の面積を全画像面積で除したものである。
ｄ．２値化後、膨張・収縮処理を施した後の黒領域面積
２値化した画像に膨張・収縮処理を施した２値画像の黒領域の面積を全画像面積で除したものである。
【０１１０】
ｅ．２値化後、微小ラベルの除去（５０画素以下）を施し、さらに膨張・収縮処理を施した後の黒領域面積
これは、２値化して、微小黒ラベル（５０画素以下）を除去し、この画像にさらに膨張・収縮処理を施した２値画像の黒領域の面積を全画像面積で除したものである。
【０１１１】
ｆ．２値化後、微小ラベルの除去（２００画素以下）を施し、さらに膨張・収縮処理を施した後の黒領域面積
これは、２値化して、微小黒ラベル（２００画素以下）を除去し、この画像にさらに膨張・収縮処理を施した２値画像の黒領域の面積を全画像面積で除したものである。
【０１１２】
これらの特徴量は、白領域（マルテンサイト領域）内の黒領域（ラス構造部）の消失の度合いを表すものであるから、これらの特徴量を用いることにより溶接熱影響部（ＨＡＺ）の劣化の度合い（経時変化の度合い）が定量化される。これにより、組織像の評価に際し個人の主観が入る余地が無くなり、溶接熱影響部（ＨＡＺ）の劣化の度合いを客観的に評価することが可能になる。また、この評価方法はシステム化できるので、評価の過程を自動化することが可能である。
【０１１３】
図１９〜図２１は、上記で求められた特徴量の処理結果を示す図であり、Ｎｏ．１の溶接熱影響部（ＨＡＺ）（供試材：Ｎｏ．１）を例にとり、それぞれの特徴量に対して、ＬＭパラメータとの相関、及び温度別の熱付与時間との相関を整理したものである。各図中の括弧（）内にＬＭパラメータとの相関係数を表示している。なお、未使用材のＬＭパラメータは計算できないので、便宜上１７．０の値を用いた。
【０１１４】
図１９（ａ）は２値化後の黒領域面積を示す図、図１９（ｂ）は２値化後微小ラベルの除去（５０画素以下）を施した黒領域面積を示す図、図２０（ｃ）は２値化後微小ラベルの除去（２００画素以下）を施した黒領域面積を示す図、図２０（ｄ）は２値化後膨張・収縮処理を施した後の黒領域面積を示す図、図２１（ｅ）は２値化後微小ラベルの除去（５０画素以下）を施しさらに膨張・収縮処理を施した後の黒領域面積を示す図、図２１（ｆ）は２値化後微小ラベルの除去（２００画素以下）を施しさらに膨張・収縮処理を施した後の黒領域面積を示す図である。
【０１１５】
本参考例において行った画像処理の結果の一例を図２２〜図２７に示す。
図２２〜図２４は、ＬＭパラメータ値が低い（劣化度が小さい）Ｎｏ．１の溶接熱影響部（ＨＡＺ）（供試材：Ｎｏ．１）の画像処理結果を示す図であり、図２２（ａ）は原画像、図２２（ｂ）は原画像ヒストグラム、図２２（ｃ）は２値化画像、図２３（ｄ）は２値化及び微小ラベル除去（５０画素以下）の画像、図２３（ｅ）は２値化及び微小ラベル除去（２００画素以下）の画像、図２３（ｆ）は２値化及び膨張収縮処理の画像、図２４（ｇ）は２値化微小ラベル除去（５０画素以下）及び膨張収縮処理の画像、図２４（ｈ）は２値化微小ラベル除去
（２００画素以下）及び膨張収縮処理の画像である。
【０１１６】
図２５〜図２７は、ＬＭパラメータ値が高い（劣化度が大きい）Ｎｏ．１の溶接熱影響部（ＨＡＺ）（供試材：Ｎｏ．１）の画像処理結果を示す図であり、図２５（ａ）は原画像、図２５（ｂ）は原画像ヒストグラム、図２５（ｃ）は２値化画像、図２６（ｄ）は２値化及び微小ラベル除去（５０画素以下）の画像、図２６（ｅ）は２値化及び微小ラベル除去（２００画素以下）の画像、図２６（ｆ）は２値化及び膨張収縮処理の画像、図２７（ｇ）は２値化微小ラベル除去（５０画素以下）及び膨張収縮処理の画像、図２７（ｈ）は２値化微小ラベル除去（２００画素以下）及び膨張収縮処理の画像である。
【０１１７】
この画像処理の結果によれば、６つの特徴量共にＬＭパラメータと高い相関があり、点状構造を除去した面積の比を特徴量化することで、溶接熱影響部（ＨＡＺ）の劣化の度合いを定量化できることが明らかになった。
【０１１８】
以上説明したように、本参考例の材料の寿命評価方法によれば、次のような効果を奏することができる。
（１）２値化像のマルテンサイト領域内のラス構造部の面積の比率を特徴量化し、この面積の比率に基づき材料の劣化の度合いを推定することができる。したがって、組織像の判定に際し、個人の主観による影響を除外することができ、材料の劣化の度合いを客観的に評価することができる。その結果、評価の定量化及び自動化を図ることができる。
【０１１９】
（２）２値化像とした後、マルテンサイト領域内の点状構造部分を所定の面積いき値により除去するので、健全なラス構造のみが残ると考えられ、マルテンサイト領域内に点状構造部分が発生した場合においても、２値化像のマルテンサイト領域内のラス構造部分の比率を特徴量化することができる。
（３）点状構造部分を除去した後、得られた像に膨張・収縮処理を施すので、２値化像のマルテンサイト領域内のラス構造部分の面積の比率の特徴量化を、より単純なものとすることができる。
【０１２０】
以上、本発明の材料の寿命評価方法の各実施形態について図面に基づき説明してきたが、具体的な構成は上記各実施形態に限定されるものではなく、適宜変更可能である。例えば、レプリカ像の替わりに、材料の表面を直接撮影した光学顕微鏡及び電子顕微鏡像あるいは金属顕微鏡像を用いても、同様の効果が得られる。
また、微小ラベル除去の際のいき値となる画素数は、除去の対象となる点状構造を十分除去できる値であればよく、５０画素、２００画素に限定されることはない。
また、評価の対象となる材料は、高温での使用により組織に経時変化が生じる材料であればよく、上記の材料に限定されない。
【０１２１】
【発明の効果】
以上説明した様に、本発明の材料の寿命評価方法によれば、２相からなる材料の組織像を所定のいき値により２値化して２つの相に分割された２値化像とし、次いで、該２値化像に基づき前記材料を特徴付ける１つ以上の特徴量を求め、該特徴量に基づき前記材料の経時変化の度合いを推定するので、組織像の判定に際し、個人の主観による影響が除外される結果、２相からなる材料の経時変化の度合いを客観的に評価することができる。
また、前記２値化像を前記材料の組織像に重ね合わせ、再度２値化するので、前記２値化像の各領域内の細かい凹凸等を消去し特徴量化することで、目視により２値化像の２つの相各々の抽出の成否を確認することができる。
【０１２２】
また、前記２値化像に基づき前記材料を特徴付ける１つ以上の特徴量を求めるので、評価結果を定量化し、この寿命評価の過程を自動化することができる。その結果、寿命評価に要する時間を短縮し、検査コストを削減することができる。
【０１２３】
本発明の材料の寿命評価方法において、前記２値化像に膨張・収縮処理を施すことにより、２値化像の２つの相各々の特徴量化を容易に行うことができる。
【０１２４】
本発明の寿命評価方法において、前記２値化像の少なくとも一方の相内の微小構造部分を所定の面積いき値により除去することにより、材料の経時変化に伴い少なくとも一方の相内に発生した微小構造部分を除去することができる。したがって、前記相内に微小構造部分が発生した場合の２値化像の２つの相の面積の比率を特徴量化することができる。
【０１２５】
本発明の寿命評価方法において、前記微小構造部分を除去した後、前記２値化像に膨張・収縮処理を施すことにより、特徴量化をより容易に行うことができ、評価結果の定量化の精度を向上させることができる。
【０１２６】
本発明の寿命評価方法において、前記２値化像を前記材料の組織像に重ね合わせ、再度２値化することにより、前記２値化像の各領域内の細かい凹凸等を消去し特徴量化することで、目視により２値化像の２つの相各々の抽出の成否を確認することができる。
【０１２７】
本発明の参考例による材料の寿命評価方法によれば、１相からなる材料の組織像を所定のいき値により２値化して２つの領域に分割された２値化像とし、該２値化像に基づき前記材料を特徴付ける１つ以上の特徴量を求め、該特徴量に基づき前記材料の経時変化の度合いを推定するので、組織像の判定に際し、個人の主観による影響が除外される結果、１相からなる材料の経時変化の度合いを客観的に評価することができる。また、２値化像に基づき前記材料を特徴付ける１つ以上の特徴量を求めるので、評価結果を定量化し、この寿命評価の過程を自動化することができる。その結果、寿命評価に要する時間を短縮し、検査コストを削減することができ、製品の製造コストの削減を図ることができる。
【０１２８】
前記本発明の参考例による材料の寿命評価方法において、前記２値化像の２つの領域各々に膨張・収縮処理を施した場合、目視では２つの領域各々の判別が困難なものであっても、特徴量化を容易に行うことができ、評価結果の定量化の精度を向上させることができる。
【０１２９】
前記本発明の参考例による材料の寿命評価方法において、前記２値化像の少なくとも一方の領域内の微小構造部分を所定の面積いき値により除去する場合、前記領域内に微小構造部分が発生した場合においても、２値化像の２つの領域を特徴量化することができる。
【０１３０】
前記本発明の参考例による材料の寿命評価方法において、前記微小構造部分を除去した後、前記２つの領域各々に膨張・収縮処理を施した場合、前記２つの領域各々の特徴量化をより容易に行うことができ、評価結果の定量化の精度を向上させることができる。
【０１３１】
前記本発明または本発明の参考例による材料の寿命評価方法において、前記組織像を、前記材料のレプリカ像、または前記材料の光学顕微鏡像あるいは電子顕微鏡像とすることにより、前記材料の表面の組織像を容易に２値化することができ、評価結果の定量化、自動化を容易に実施することができる。
【０１３２】
前記本発明または本発明の参考例による材料の寿命評価方法において、前記特徴量を、前記２つの相の原画像におけるコントラスト比とすることにより、２相からなる材料の経時変化の度合いを客観的に評価することができる。
【０１３３】
前記本発明または本発明の参考例による材料の寿命評価方法において、前記特徴量を、前記２つの相のうち一方の像と、前記材料の経時変化に伴って該一方の像の領域内に観測される炭化物とに基づいたものとすることにより、前記領域内に炭化物が発生した場合においても、２値化像の２つの領域を特徴量化することができる。
【０１３４】
前記本発明または本発明の参考例による材料の寿命評価方法において、前記特徴量を、前記２つの相のうち一方の像と、前記材料の経時変化に伴って該一方の像の領域内で観測されるラス構造とに基づいたものとすることにより、前記領域内にラス構造が発生した場合においても、２値化像の２つの領域を特徴量化することができる。
【０１３５】
前記本発明または本発明の参考例による材料の寿命評価方法において、前記特徴量を、前記２値化像のうち一方の像と他方の像との面積比とすることにより、２値化像の２つの領域を特徴量化することができる。
【０１３６】
上記いずれかの材料の寿命評価方法において、前記２値化像を、組織の地部分と、該地部分の内部に形成されたマルテンサイトラス構造の面積比とした場合、前記地部分の内部にマルテンサイトラス構造が発生した場合においても、２値化像の２つの領域を特徴量化することができる。
【図面の簡単な説明】
【図１】 本発明の第１の実施形態の材料の寿命評価方法を示す流れ図である。
【図２】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図３】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図４】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図５】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図６】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図７】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図８】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図９】 本発明の第１の実施形態の特徴量の処理結果を示す図である。
【図１０】 本発明の第１の実施形態のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図１１】 本発明の第１の実施形態のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図１２】 本発明の第１の実施形態のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図１３】 本発明の第１の実施形態のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図１４】 本発明の第１の実施形態のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図１５】 本発明の第１の実施形態のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図１６】 本発明の第１の実施形態のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図１７】 本発明の第１の実施形態のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図１８】 本発明の参考例による材料の寿命評価方法を示す流れ図である。
【図１９】 本発明の参考例の特徴量の処理結果を示す図である。
【図２０】 本発明の参考例の特徴量の処理結果を示す図である。
【図２１】 本発明の参考例の特徴量の処理結果を示す図である。
【図２２】 本発明の参考例のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図２３】 本発明の参考例のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図２４】 本発明の参考例のＬＭパラメータ値が低い（劣化度が小さい）母材の画像処理の結果を示す図である。
【図２５】 本発明の参考例のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図２６】 本発明の参考例のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図２７】 本発明の参考例のＬＭパラメータ値が高い（劣化度が大きい）母材の画像処理の結果を示す図である。
【図２８】 低合金ＣｒＭｏ鋼の各劣化段階に対応した標準組織を示す模式図である。
【図２９】 低合金ＣｒＭｏ鋼の各劣化段階に対応した標準組織を示す模式図である。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a material life evaluation method, and in particular, a life when a high-temperature pressure-resistant metal member such as a low alloy CrMo steel used as piping material for a boiler for a thermal power plant is changed over time due to use at a high temperature. The present invention relates to a method for evaluating the life of a material applied for evaluation.
[0002]
[Prior art]
In recent years, in thermal power plants, maintenance management that fully considers thermal fatigue due to equipment deterioration due to long-term use, frequent start / stop, rapid load fluctuations, etc. is becoming more and more important as the operation time becomes longer. It has become to.
For example, in large-diameter thick-walled pipes that use high-temperature pressure-resistant metal members such as low-alloy CrMo steel, the base material that is the pipe body and its heat affected zone (Heat Affected Zone: HAZ) change over time due to use at high temperatures. This change over time becomes a factor that deteriorates the pipe, and therefore it is necessary to estimate the life of the pipe from the change over time in order to ensure the reliability of the pipe.
Therefore, as a conventional material life evaluation method, a magnetic particle inspection, an ultrasonic inspection, a radiation transmission inspection, and the like are known in addition to a tissue inspection using an optical microscope and an electron microscope.
[0003]
Here, the structure inspection (determination by an expert) using an optical microscope and an electron microscope will be described.
This structure inspection is called a tissue contrast method. For example, when estimating the life of a boiler (base material) of a boiler for a thermal power plant, the degree of deterioration of the structure of the pipe material is classified according to the degree of deterioration in advance. A standard optical microscope image and an electron microscope structure image (referred to as a standard structure) are prepared for each stage. Next, a replica is taken from the surface of the pipe whose life is to be estimated, this replica is observed using an optical microscope and an electron microscope, and the optical microscope and the electron microscope structure are compared with the standard structure described above. The stage to which the standard structure having the closest structure to the electron microscope structure belongs is defined as the degree of deterioration of the pipe.
Instead of observing the replica, a method of directly observing the surface of the pipe or the extracted material using an optical microscope may be used.
[0004]
FIG. 28 and FIG. 29 are schematic views showing standard structures corresponding to respective deterioration stages of a low-phase CrMo steel having a two-phase structure used as a high-temperature pressure-resistant piping material, and are classified into four stages here.
FIG. 28A shows an initial stage deterioration state, that is, a first stage deterioration state (I_M), The difference in contrast between ferrite (white part in the figure) and bainite (or pearlite) (black part in the figure) is clear, which is substantially the same as that of the unused material.
FIG. 28B shows the second stage deterioration state (II_M), The contrast difference between ferrite and bainite (or pearlite) is reduced, and carbides (black spots in the white region) are precipitated in the ferrite grains. Turn into.
[0005]
FIG. 29 (c) shows the third stage deterioration state (III_M), The carbides in the ferrite grains are further coarsened, and the linear structure called lath is reduced in the bainite region. However, the variation is large.
FIG. 29 (d) shows the final stage degradation state, that is, the fourth stage degradation state (IV_M) In the pearlite region, the pearlite is decomposed and finally disappears. At this stage, there is almost no contrast difference between ferrite and bainite.
[0006]
The deterioration degree of this low alloy CrMo steel is determined by applying the Larson Miller parameter method (LM method) to the Larson Miller parameter (LM) which is a deterioration degree index calculated from the heating time and temperature. ). This Larson-Miller parameter (LM) can be obtained by the following equation.
L · M = T (C + logt) / 1000
Where T: Absolute temperature
C: Constant
t: heating time
It is.
[0007]
Also, when estimating the life of the weld heat affected zone (HAZ) of a pipe made of a low alloy CrMo steel having a single-phase structure, the life should be estimated using the structure contrast method, as with the pipe (base material) described above. Can do.
In this case, the degree of deterioration of the structure of the weld heat affected zone (HAZ) is classified in advance according to the degree of deterioration, and a standard optical microscope and electron microscope structure image (standard structure) are prepared for each stage. Next, a replica is taken from the surface of the weld heat affected zone (HAZ) whose life is to be estimated, and this replica is observed using an optical microscope and an electron microscope to obtain an optical microscope and an electron microscope structure and the above-described standard structure. In comparison, the stage to which the standard structure closest in structure to the optical microscope and electron microscope structure belongs is defined as the deterioration degree of the pipe.
[0008]
The degree of deterioration of the weld heat affected zone (HAZ) is classified into three stages as follows.
First stage: The area ratio of the martensite region and the linear structure (martensite lath) in the martensite region is almost equal to that of the unused material.
Second stage: The area ratio between the martensite region and the martensite lath decreases, and carbides precipitate in the martensite region.
Third stage: The precipitated carbide is dispersed in the martensite region and martensite lath disappears.
The degree of deterioration of the weld heat affected zone (HAZ) can also be expressed by using the Larson mirror parameter (L · M) as in the case of the pipe described above.
Instead of observing the replica, there may be a method of directly observing the structure of the surface of the pipe using a metal microscope or the like.
[0009]
[Problems to be solved by the invention]
By the way, in the conventional structure | tissue examination (judgment by an expert) mentioned above, the optical microscope and electron microscope structure of piping or welding heat-affected zone (HAZ) used as the object of life estimation are compared with a standard structure, and piping is compared. Since the degree of deterioration is estimated, there is a problem in that the subjectivity of the individual inspector greatly affects and there is a possibility that the determination of the degree of deterioration may vary among the inspectors.
In addition, when the optical microscope and the electron microscope tissue are located in the middle of two standard tissues, it is difficult to determine which standard tissue is close, and therefore it is difficult to quantify the determination result. there were.
[0010]
The present invention has been made in view of the above circumstances, and it is possible to exclude the influence of individual subjectivity during the evaluation, and as a result, the evaluation result can be quantified, and the evaluation work It is an object of the present invention to provide a material life evaluation method capable of shortening the evaluation work time and reducing the cost by improving the efficiency.
[0011]
[Means for Solving the Problems]
  In order to solve the above problems, the present invention provides the following material lifetime evaluation method.
  That is,The present inventionThe method for evaluating the lifetime of a material is a method for evaluating a lifetime when a material having two phases undergoes a change with time in at least one phase due to use at a high temperature, and the tissue image of the material is based on the gradation of the image. The binarized image is binarized by a threshold value to obtain a binarized image divided into two phases. Then, the binarized image is superimposed on the tissue image of the material, and the material is changed based on the binarized image again. One or more characteristic quantities to be characterized are obtained, and the degree of change with time of the material is estimated based on the characteristic quantities.
[0012]
In this method, a tissue image of a material composed of two phases, which is a gradation image, is subjected to threshold processing to obtain a binary image divided into two phases. For example, the density information of the tissue image is divided by a certain level of intensity (threshold), and “1” is assigned to an area darker than this level (black area), and “0” is assigned to an area lighter than this level. Then, only the area darker than the level (black area) is displayed as a binary image.
[0013]
  As a result, the degree of shading between the two phases of the binarized image is converted into a feature amount, and the degree of change with time of the material composed of the two phases is estimated based on the degree of shading.
  As described above, in the determination of the tissue image, the influence of individual subjectivity is excluded. As a result, the degree of change with time of the material composed of two phases is objectively evaluated. Further, by converting one or more feature quantities characterizing the material on the basis of the binarized image, for example, the degree of shading, the evaluation result is quantified and automation is possible.
  In the above method, the binary image is superimposed on the structure image of the material and binarized again to clarify the fine structure of one phase extracted by the binarization process. It becomes possible to make the structure a feature amount.
[0014]
  In addition, the present inventionThe life evaluation method of the material ofSaidIn the material life evaluation method, the binarized image is subjected to expansion / contraction processing.
[0015]
In this method, by performing expansion / contraction processing on the binarized image, fine irregularities and the like in each region of the binarized image are erased, and the feature value is converted into a simpler figure. As a result, it is possible to facilitate the feature quantity of each of the two phases of the binarized image. For example, even if it is difficult to visually discriminate the degree of shading between the two phases, the degree of shading can be converted into a feature amount.
[0016]
  The present inventionIn the method for evaluating the lifetime of a material, a microstructure portion in at least one phase of the binarized image is removed with a predetermined area threshold value.May.
[0017]
In this method, for example, even in the case where precipitated carbides of fine particles are generated inside the ferrite particles, which is one of the two phases, the precipitated carbides are removed by a predetermined area threshold value. It becomes possible to distinguish phases.
That is, as the material changes with time, at least one microstructure portion generated in one phase is removed by a predetermined area threshold value, so that 2 of the binarized image when the microstructure portion is generated in the phase is obtained. It becomes possible to characterize the ratio of the areas of two phases.
[0018]
The material life evaluation method according to claim 4 is characterized in that, in the material life evaluation method according to claim 3, after the fine structure portion is removed, the binarized image is subjected to expansion / contraction processing. .
[0019]
In this method, by performing expansion / contraction processing on the binarized image from which the minute structure portion has been removed, it becomes easier to obtain a feature amount, and the accuracy of quantifying the evaluation result is improved.
[0022]
  According to a reference example of the present inventionThe material life evaluation method is a method for evaluating the life when a material having one phase changes with time due to use at a high temperature. The material image of the material is binarized by a predetermined threshold value, A binarized image divided into regions, one or more feature quantities characterizing the material are obtained based on the binarized image, and a degree of temporal change of the material is estimated based on the feature quantities. Yes.
[0023]
In this method, a threshold value process is performed on a tissue image of a material that is a gradation image to obtain a binary image. For example, the density information of the tissue image is divided by a certain level of intensity (threshold), and “1” is assigned to an area darker than this level (black area), and “0” is assigned to an area lighter than this level. Then, only the area darker than the level (black area) is displayed as a binary image.
[0024]
Thereby, the ratio of the area of the two regions of the binarized image is converted into a feature amount, and the degree of change with time of the material composed of one phase is estimated based on the ratio of the areas.
As described above, in the determination of the tissue image, the influence of individual subjectivity is excluded, and the degree of change with time of the material composed of one phase is objectively evaluated. Further, by converting one or more feature amounts characterizing the material based on the binarized image into feature amounts, the evaluation result is quantified, and automation is possible.
[0025]
  According to the reference example of the present inventionIn the material life evaluation method, each of the two regions of the binarized image is subjected to expansion / contraction processing.May be.
[0026]
In this method, by performing expansion / contraction processing on each of the two regions of the binarized image, for example, even if it is difficult to discriminate each of the two regions by visual observation, the feature value can be easily obtained. And the accuracy of quantification of evaluation results is improved.
[0027]
  According to the reference example of the present inventionIn the material life evaluation method, a microstructure portion in at least one region of the binarized image is removed with a predetermined area threshold value.May be.
[0028]
In this method, for example, by removing a minute structure portion generated in at least one region with a change in material with a predetermined area threshold value, it is possible to make a feature amount excluding the minute structure portion in the region. Become.
[0029]
  Any of the aboveIn the material life evaluation method, after removing the microstructure portion, each of the two regions is subjected to expansion / contraction treatment.May be.
[0030]
In this method, by performing expansion / contraction processing on each of the two regions of the binarized image from which the minute structure portion has been removed, it becomes easier to quantify the characteristics and improve the accuracy of quantification of the evaluation results.
[0031]
  Any of the aboveIn the material lifetime evaluation method, the structure image is a replica image of the material or an optical microscope image of the material.May be.
[0032]
In this method, since the tissue image is a replica image of the material or an optical microscope image of the material, it becomes easy to binarize the structure image of the surface of the pipe or the extubated material. Quantification of evaluation results and automation of evaluation become easy.
[0033]
  Any of the aboveIn the material lifetime evaluation method, the feature amount is a contrast ratio in the original image of the two phases.May be.
  In this method, it is possible to objectively evaluate the degree of change with time of a material composed of two phases.
[0034]
  Any of the aboveIn the material lifetime evaluation method, the feature amount is based on one image of the two phases and carbides observed in a region of the one image as the material changes with time.May be.
[0035]
  Any of the aboveIn the material lifetime evaluation method, the feature amount is based on an image of one of the two phases and a lath structure observed in a region of the one image as the material changes with time.May be.
[0036]
  Any of the aboveIn the material lifetime evaluation method, the feature amount is an area ratio between one image and the other image of the binarized image.May be.
  In this method, it is possible to convert the two regions of the binarized image into features.
[0037]
  Any of the aboveIn the material life evaluation method, the binarized image is an area ratio of a matrix structure and a martensitic structure formed in crystal grains.May be.
[0038]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the material life evaluation method of the present invention will be described with reference to the drawings.
[First Embodiment]
The material lifetime evaluation method according to the first embodiment of the present invention will be described by taking as an example the case of estimating the lifetime of the base material of the heat transfer tube of the thermal power generation facility.
Here, three types of materials (STBA22, STBA23 and STBA24) are selected from metal materials having a two-phase structure, that is, alloy steel pipes for boiler heat exchangers (JIS G 3462), and the base materials (No. 1-3). The shapes of these base materials were cylindrical with an outer diameter of 50.8 mm, an inner diameter of 10.0 mm, and a length of 100 mm.
[0039]
[Table 1]

[0040]
Next, each base material (No. 1 to 3) was abutted against two base materials, and the butt portion was welded by a covered arc welding method to obtain a welded joint (test material: No. 1 to 3).
The post-weld heat treatment conditions after the covering arc welding are as follows.
Base material (No. 1 to 3): 715 ° C. × 15 minutes / AC
[0041]
Subsequently, each of these test materials (Nos. 1 to 3) was subjected to heat treatment under various conditions to obtain samples for life evaluation.
The heating temperature (maximum holding temperature) and heating time (holding time at the heating temperature) in the heat treatment were as follows for each of the test materials (Nos. 1 to 3).
Heating temperature: 550 ° C, 575 ° C, 600 ° C, 625 ° C, 650 ° C, 5 points
Heating time: 30 h (hours), 100 h, 300 h in each of the above heating times
6 points of 1000h, 3000h, 10000h
As a result, the total number of sample points was 93 [= {untreated material (1 point) + heat treated material (temperature 5 points × time 6 points)} × test material (3 types)].
[0042]
Next, a replica is collected from the surface of each of the heat-treated samples (No. 1 to 3), this replica is observed using an optical microscope, the optical microscope structure is photographed, and these optical microscope structure images are displayed. Classification was made using the Larson-Miller parameter method already described in the prior art. The value of parameter constant C was 20 for each sample (No. 1 to 3).
[0043]
Subsequently, the image processing of this optical microscope structure | tissue was implemented based on FIG.
A. Conversion of original image (optical microscope tissue image)
Prior to this image processing, an image having a gray scale of 256 gradations and a size of 800 pixels (pixels) × 580 pixels (pixels) was used as an original image (optical microscope tissue image).
[0044]
The points of interest when image processing is performed on the original image are as follows.
(A) It has a two-phase metal structure represented by a white region and a black region, and the contrast ratio between the two phases decreases due to deterioration (time-dependent change) due to use at a high temperature such as external heating.
(B) In the white region, granular carbides grow and become visible, and the density increases as a whole.
(C) In the black region, the linear structure called lath is reduced. However, the variation is large.
[0045]
With the above points of interest in mind, a “black region image” and a “white region image” were extracted according to the flowchart showing the image processing of FIG.
B. Binary conversion of original image
(1) threshold processing
The converted original image was subjected to threshold processing using a canonical discriminant method most well known as a binarization threshold value determination method to obtain a binarized image.
[0046]
For example, the shading information of this original image is divided by a certain level of shading (threshold), and “1” is assigned to a region darker than this level (black region), and “0” is assigned to a region lighter than this level. Then, only a region (black region) darker than the level is displayed as a binary image.
In the low alloy CrMo steel, the ferrite region is a white region and the bainite (or pearlite) is a black region.
[0047]
For example, when the image is divided into two classes (0 class and 1 class) according to a certain threshold value, the threshold value between the classes σ_B ²Select the threshold value k that maximizes (k). This interclass variance σ_B ²(k) is expressed by the following equation.
σ_B ²(k) = ω₀(Μ₀−μ_T)²+ Ω₁(Μ₁−μ_T)²
[0048]
here,
[Expression 1]

[Expression 2]

[Equation 3]

[Expression 4]

[Equation 5]

P_i= N_i/ N
ω₀+ Ω₁= 1
It is.
[0049]
However,
i: Light and shade level (gradation) (i = integer between 1 and L)
n_i: Number of pixels at level i
N: Total number of pixels
P_i: Ratio of area occupied by points of luminance i when luminance 1 to k is class 0 (black region) and luminance k + 1 to L is class 1 (white region)
ω₀: 0 class area ratio
ω₁: 1 class area ratio
μ₀: Class 0 average brightness
μ₁: 1 class average brightness
μ_T: Average brightness of the entire image
It is.
[0050]
(2) Inversion of binarized image
A black image is extracted by inverting the binarized image obtained by threshold processing.
In this embodiment, the reason for inverting the binarized image is that black (0) is not processed and white (255) is processed when converting to a monochrome image of 256 gradations. The binarized image obtained by the value processing may be used as it is when the processing proceeds while focusing on the white region (ferrite region), but the processing is performed focusing on the black region (bainite (or pearlite) region). This is because in order to proceed, it is necessary to invert the obtained binarized image and extract a black region.
[0051]
C. White area extraction
(1) Removal of minute labels
In order to remove black spot-like carbides scattered in the white region (ferrite region), a region having a certain area threshold value or less was deleted. Here, since the area threshold is set to 200 pixels, black spots having an area of 200 pixels or less in the white region are removed.
[0052]
(2) Black area expansion / contraction processing
In order to fill the gap in the black area, the black area was subjected to expansion / contraction processing (closing).
For the expansion process, an n × n (n is an odd number) filter is prepared, and for example, a 3 × 3 filter is applied to all the pixels of the binarized image, and a white pixel adjacent to the black pixel. Is converted to black pixels, and is also referred to as maximum value filter processing.
[0053]
For example, if there is a 3 × 3 pixel having the highest luminance (black pixel) among the eight pixels surrounding the central pixel, the luminance of this pixel is set as the luminance of the central pixel. Here, since the maximum value filter is applied after binarization, the luminance can be any one of 0 and 255. Therefore, if at least one of the eight pixels surrounding the central pixel is a black pixel, the central pixel is also a black pixel.
Here, expansion processing (maximum value filter processing) was performed twice using a 5 × 5 filter.
[0054]
The contraction process is an inverse operation of the expansion process, has a function of thinly removing the black pixel portion, and is also referred to as a minimum value filter process.
Here, contraction processing (minimum value filter processing) was performed twice using a 5 × 5 filter.
[0055]
In this expansion / contraction process, a maximum value filter and a minimum value filter are used to fill a gap corresponding to a white point / black point in each region, and the region is extracted as one region. The reason is that, in simple binarization, extracted areas such as white spots (lass structure) in the black area (bainite) and black spots (carbide particles) in the white area (ferrite) are intermingled. However, for example, since the black spot of the white area is important information corresponding to the above-described attention point (b), this white area needs to be reliably extracted as “one white area”.
[0056]
As a result of performing expansion / contraction processing on the black area, a small white portion in the black area was filled with black, and the unevenness of the boundary area between the black area and the white area was leveled and flattened.
In the low alloy CrMo steel, since the bainite region is a black region, the lath structure which is a linear structure in the bainite region can be eliminated.
[0057]
However, if the density of the carbide particles that are precipitates in the “white area” is high, there is a possibility that the expansion process may result in erroneous extraction as a “black area”. Removal of the microlabel was performed.
[0058]
(3) Maximum value filter processing
In order to narrow down the white area image (to ensure the white area), the black area was expanded using a maximum value filter.
[0059]
(4) Brightness inversion
The black area including the boundary line was inverted to form a white area.
Before the luminance reversal, for example, in the low alloy CrMo steel, the bainite region including the boundary line is a black region (white on the screen (luminance 255)), and the ferrite region is a white region (black on the screen (luminance 0)). )It has become.
In the next masking, the white (brightness 255) portion (ON portion) remains, so that the ferrite region of black (brightness 0) is inverted to white (brightness 255). Was set to ON.
[0060]
In summary, the extraction of the “white area” is performed by expanding the “black area” to the maximum by the expansion / contraction process and the maximum value filter process for the “black area” obtained by simple binarization. This is done by setting the white area as the “white area”.
[0061]
(5) Masking
By masking the binary image with inverted luminance and the original image, only the ferrite region in which the luminance was inverted to become white (luminance 255) was obtained.
This completes the creation of the white area image.
[0062]
(6) Median filter processing
Similar to the maximum value filter processing described above, a 5 × 5 filter was used to replace the median (median value) in the filter as a representative value (as the new brightness of the pixel at the center of the filter).
As a result, even if there is a protruding point of brightness in the filter, the influence can be mitigated. In other words, a steep change in luminance in the image is suppressed, and a gentle luminance change (low frequency component of the image) can be extracted.
[0063]
(7) Difference processing
The white area image obtained by masking is subtracted from the image after the median filter processing (the difference in brightness of each pixel is taken), and the part where the change in brightness is steep, that is, the part that seems to be a precipitate, is extracted. Ingredients.
[0064]
(8) Re-binarization
Furthermore, in order to extract the precipitated carbide from the “white region”, binarization was performed again.
In order to suppress the adverse effects due to the low frequency components, binarization was performed by the canonical discrimination method using an image in which only the high frequency components were left by differential processing.
As described above, a binarized carbide image could be created.
[0065]
D. Black area extraction
(1) Black area expansion → contraction processing
In order to fill a lath linear structure (white region) that is a gap in the bainite region (black region), the black region was subjected to expansion / contraction processing (closing: closing).
Here, expansion processing (maximum value filter processing) was performed twice using a 5 × 5 filter.
Thereafter, contraction processing (minimum value filter processing) was performed twice using a 5 × 5 filter.
In this expansion / contraction process, a maximum value filter and a minimum value filter are used to fill a gap corresponding to a white point / black point in each region, that is, a lath linear structure (white region), and extract as one region.
[0066]
(2) Black area shrinkage / expansion processing
In the case of a material composed of two phases, it is generally considered that an image is composed of three regions of “black region”, “white region”, and “boundary region” having different properties. Among these areas, the “boundary area” is extracted in accordance with the “black area” in simple binarization, and therefore it is difficult to extract only the “black area” alone. Therefore, by applying contraction / expansion processing to the “black area”, the “boundary area” is removed from the “black area” and only the black area is extracted.
[0067]
Here, in order to remove the thin linear structure that is the boundary line of the bainite region (black region), the black region was subjected to contraction / expansion processing (opening).
For example, contraction processing (minimum value filter processing) was performed twice using a 5 × 5 filter, and then expansion processing (maximum value filter processing) was performed twice using a 5 × 5 filter.
As a result, in this contraction / expansion processing, the thin line portion that is the boundary line of each region is removed using the minimum value filter and the maximum value filter, and only the black region can be extracted.
[0068]
(3) Masking
In the binary image obtained by the contraction → expansion processing, the bainite region from which the boundary line is removed is a black region (white on the screen (luminance 255)), and the ferrite region is a white region (black on the screen). Therefore, by masking the binary image and the original image, an image in which only the bainite region, which is a black region, is extracted is obtained.
Thereby, the creation of the black area image is completed.
[0069]
(4) Rebinarization
Furthermore, binarization was performed by the canonical discrimination method using an image obtained by extracting only the bainite region which is a black region by masking.
As described above, an image having a binarized lath structure could be created.
[0070]
E. Feature calculation
Using the white area image, black area image, carbide image, and lath image obtained as described above, the following seven feature amounts were calculated.
(1) Based on the contrast ratio of “black area” and “white area”
a. Inter-class class separation
The variance between classes between the black region and the white region is divided by the overall variance, and takes a value between 0 and 1. This separation degree indicates that the closer to 1, the better the separation between classes, that is, the higher the contrast ratio.
[0071]
Inter-class class separation η_AIs expressed by the following equation.
η_A= Σ_BA ²/ Σ_BW ²
However, σ_BA ²Is the variance between classes when the black area and the white area are two classes, σ_BW ²Is the overall luminance dispersion of both black and white areas.
[0072]
Interclass variance σ_BA ²Is expressed by the following equation.
[Formula 6]

Where n_B: Number of pixels in the black area
n_W: Number of pixels in white area
μ_B: Average brightness of black area
μ_W: Average brightness of white area
μ_BW: Average brightness of the combined black and white area
It is.
[0073]
b. Canonical discriminant class separation
The variance between classes based on the canonical discriminant method of the histogram is divided by the overall variance. The threshold value is obtained on the luminance histogram by the so-called Otsu's method, and the pixels belonging to the upper and lower sides of the luminance histogram with the threshold as a boundary. A group (two pixel groups) is regarded as a class, and takes a value between 0 and 1. This separation degree indicates that the closer to 1, the better the separation between classes, that is, the higher the contrast ratio.
[0074]
Canonical classification class separation η₀Is expressed by the following equation.
η₀= Σ_BO ²/ Σ_T ²
However, σ_BO ²Is the variance between classes when the image is divided into two classes with luminance threshold k, σ_T ²Is the luminance variance of the entire image.
[0075]
Interclass variance σ_BO ²Is expressed by the following equation.
[Expression 7]

Where n_BO: Number of pixels in the area where the luminance is below the Otsu threshold
n_WO: Number of pixels in the area where the brightness is greater than the Otsu threshold
N (= n_BO+ N_WO): Number of pixels in the entire image
μ_BO: Average brightness in areas where the brightness is below the Otsu threshold
μ_WO: Average brightness in areas where brightness is greater than Otsu threshold
μ_T: Average brightness of the entire image
It is.
[0076]
c. Contrast difference between regions
The difference between the brightness average of the white area and the brightness average of the black area is obtained, and this difference is divided by the total number of gradations 256.
Contrast difference between regions W_ABIs expressed by the following equation.
W_AB= (Μ_W−μ_B) / 256
[0077]
d. Overall brightness standard deviation / In-class brightness standard deviation
Overall brightness standard deviation σ including both black and white areas_{W + B}The brightness standard deviation within the class
σ_classDivided by.
Luminance standard deviation σ_{W + B}Is expressed by the following equation.
[Equation 8]

[0078]
In-class brightness standard deviation σ_classCalculates the difference between the average brightness of the class to which each pixel belongs (or the average brightness of the white area if it is a white area) for each pixel. The standard deviation is obtained. Luminance standard deviation σ within this class_classIs expressed by the following equation.
[Equation 9]

[0079]
However, P_i: Luminance value of pixel i
n_W: Number of pixels in white area
n_B: Number of pixels in the black area
(N_W+ N_B≠ N: number of pixels in the entire image)
μ_W: Average brightness of white area
μ_B: Average brightness of black area
μ_{W + B}: Average luminance of both black and white areas
i∈W: pixel in white area
i∈B: pixel in black region
[0080]
e. Luminance difference between classes / Luminance standard deviation within class
Difference between the average brightness of the white class and the average brightness of the black class μ_W−μ_BThe brightness standard deviation within the class σ_classDivided by.
f. Overall brightness standard deviation
Standard deviation σ of brightness in both white and black areas_{W + B}It has already been explained.
[0081]
(2) Based on precipitated carbides in the “white region”
g. White area luminance dispersion
This is an index indicating the luminance dispersion state of the white region, and represents the presence or absence of a significant difference in luminance dispersion depending on the state of carbide precipitation.
White area luminance dispersion σ_W ²Is expressed by the following equation.
[Expression 10]

However, N_W: Number of pixels in white area
P_i(I∈w): luminance of each pixel belonging to the white area
P_W: Average brightness of white area
[0082]
h. Carbide density in white region
The sum of the black spot areas of the carbide image (image obtained by binarizing the white area) is divided by the total area of the white area.
[0083]
(3) Based on the lath structure of the “black area”
i. Black area luminance dispersion
This is an index indicating the luminance dispersion state of the black region, and indicates the presence or absence of a significant difference in luminance dispersion due to the lath structure. Calculation is performed in the same manner as the white area luminance dispersion.
[0084]
(4) Other
j. Black area / White area
The area of the “black area” is divided by the area of the “white area”, and serves as an index for determining whether or not the image processing performed is appropriate.
[0085]
Since these feature values are numerical values of the degree of shading between the two phases, that is, the degree of deterioration of the base material, the degree of deterioration of the base material (degree of change over time) by using these feature values. Is quantified. As a result, there is no room for individual subjectivity when evaluating the tissue image, and the degree of deterioration of the base material can be objectively evaluated. Further, since this evaluation method can be systematized, the evaluation process can be automated.
[0086]
2 to 5 are diagrams showing the processing results of the feature values obtained above. No. 1 sample (test material: No. 1) is taken as an example, and the correlation between the LM parameter and the image processing result is arranged for each feature amount. The correlation coefficient with the LM parameter is displayed in parentheses () in each figure. Since the LM parameter of the unused material cannot be calculated, a value of 17.0 is used for convenience.
[0087]
2A is a diagram showing the class separation between regions, FIG. 2B is a diagram showing the canonical discrimination class separation, FIG. 3C is a diagram showing the contrast difference between regions, and FIG. (D) is a graph showing the luminance dispersion in the white region, FIG. 4 (e) is a diagram showing the carbide density in the white region, FIG. 4 (f) is a graph showing the luminance dispersion in the black region, and FIG. It is a figure which shows a white area.
[0088]
6 and 7 are diagrams illustrating other examples of the processing result of the feature amount. Taking the sample 2 (test material: No. 2) as an example, the correlation between the LM parameter and the image processing result is arranged for each feature amount. The correlation coefficient with the LM parameter is displayed in parentheses () in each figure, and the LM parameter of the unused material is set to 17.0. This is the same as the sample No. 1.
[0089]
Here, FIG. 6A is a diagram showing the overall luminance standard deviation / luminance standard deviation within the class, FIG. 6B is a diagram showing the inter-class luminance difference / class luminance standard deviation, and FIG. ) Is a diagram showing the overall luminance standard deviation.
[0090]
8 and 9 are diagrams illustrating other examples of the processing result of the feature amount. Taking the sample 3 (test material: No. 3) as an example, the correlation between the LM parameter and the image processing result is arranged for each feature amount. The correlation coefficient with the LM parameter is displayed in parentheses () in each figure, and the LM parameter of the unused material is set to 17.0. This is the same as the sample No. 1.
[0091]
Here, FIG. 8A is a diagram showing the overall luminance standard deviation / luminance standard deviation within the class, FIG. 8B is a diagram showing the inter-class luminance difference / intra-class luminance standard deviation, and FIG. 9C. ) Is a diagram showing the overall luminance standard deviation.
[0092]
An example of the result of the image processing performed in this embodiment is shown in FIGS.
10 to 13 show Nos. With low LM parameter values (low degradation degree). FIG. 10 is a diagram showing image processing results of a sample 1 (test material: No. 1), FIG. 10A is an original image, FIG. 10B is a black region image, and FIG. 11C is a white region image. 11 (d) is an original image histogram, FIG. 12 (e) is a segmented image, FIG. 12 (f) is a carbide image, and FIG. 13 (g) is a lath image.
[0093]
FIGS. 14 to 17 show the case where the LM parameter value is high (the deterioration degree is large). 15A and 15B are diagrams showing image processing results of a sample No. 1 (test material: No. 1), in which FIG. 14A is an original image, FIG. 14B is a black region image, and FIG. 15C is a white region image. 15 (d) is an original image histogram, FIG. 16 (e) is a segmented image, FIG. 16 (f) is a carbide image, and FIG. 17 (g) is a lath image.
[0094]
According to the result of this image processing, it was found that the contrast ratio between the regions has a high correlation with the LM parameter. The feature amount based on the contrast ratio includes (a) class separation between regions to (c) contrast difference between regions, and (c) it was found that the contrast difference between regions was the most excellent. Further, (e) the carbide density in the white region correlates with the LM parameter, and thus it was found that the precipitated carbide can be extracted.
[0095]
As described above, according to the material lifetime evaluation method of the present embodiment, the following effects can be obtained.
(1) The degree of shading of the two phases of the binarized image can be characterized, and the degree of change with time of the material can be estimated based on the degree of shading of the two phases. Therefore, in determining the tissue image, it is possible to exclude the influence due to the subjectivity of the individual, and it is possible to objectively evaluate the degree of change with time of the material. As a result, the evaluation result can be quantified and automated.
[0096]
(2) After the binarized image is formed, the carbide, which is the fine structure portion of the ferrite region (white region), is removed by a predetermined area threshold value, so that when the fine structure portion is generated in the ferrite region (white region) In addition, the degree of shading of the two phases of the binarized image can be featured.
(3) Since the binarized image is superimposed on the original image and binarized again, a binarized image such as a carbide image or a lath structure image can be obtained.
[0097]
  [Reference example]
  Of the present inventionAccording to reference examplesThe material life evaluation method will be described by taking as an example the case of estimating the life of a heat affected zone (hereinafter also referred to as HAZ) of a heat transfer tube of a thermal power generation facility as a one-phase material.
  Here, the specimens (Nos. 1 to 3) used in the first embodiment were each subjected to heat treatment under various conditions, and used as samples for evaluating the life of the weld heat affected zone (HAZ).
  Since the heating temperature and heating time in the heat treatment are exactly the same as in the first embodiment, the total number of sample points is exactly the same as in the first embodiment.
[0098]
Next, a replica is taken from the surface of each heat-affected zone (HAZ) of each heat-treated sample (No. 1 to 3), this replica is observed using an optical microscope, and the optical microscope structure is photographed. These optical microscope tissue images were classified using the Larson-Miller parameter method already described in the prior art. The value of parameter constant C was 20 for each sample (No. 1 to 3).
[0099]
Subsequently, the image processing of the optical microscope structure of a welding heat affected zone (HAZ) was implemented based on FIG.
A. Conversion of original image (optical microscope tissue image)
Prior to this image processing, an image having a gray scale of 256 gradations and a size of 800 pixels (pixels) × 580 pixels (pixels) was used as an original image (optical microscope tissue image).
[0100]
The points of interest when image processing is performed on the HAZ original image are as follows.
(A) The material deteriorates (changes with time) due to the use at high temperature, but as the deterioration progresses, the martensitic structure (linear structure) in the metal structure collapses, and the linear structure changes to a dotted shape. The structure will change.
[0101]
(B) The above change occurs only in the martensite region and does not occur in the ferrite region. Therefore, by dealing with a plurality of fields of view and averaging the processing results, the area ratio of the ferrite region in the image to be processed is brought close to a constant value, and the influence is reduced.
[0102]
In consideration of the above-mentioned point of interest, the area ratio of the lath structure portion (linear structure portion) is characterized according to the flowchart showing the image processing of FIG.
B. Binary conversion of original image
(1) threshold processing
The original image subjected to the conversion was subjected to value processing using a canonical discrimination method to obtain a binarized image divided into a lath structure and grain boundary line region and other regions.
In the low alloy CrMo steel, the martensite region and the ferrite region are white regions, and the lath structure portion in the martensite region is black region.
The method of selecting the threshold value is the same as the method of selecting the threshold value in the first embodiment.
[0103]
(2) Inversion of binarized image
A black image is extracted by inverting the binarized image obtained by threshold processing.
In order to process the binarized image while paying attention to the black region (lass structure portion), it is necessary to invert the obtained binarized image and extract the black region.
[0104]
(3) Removal of minute labels
In order to remove point-like regions other than the lath structure portion in the white region (martensite region), a region having a certain area threshold value or less was deleted.
When the lath structure portion collapses, it changes to a point-like structure when binarized, and therefore this point-like structure is excluded by providing a threshold value in terms of area.
[0105]
Here, the area threshold is set to 0 pixel (no exclusion process), 50 pixels, and 200 pixels. If the area threshold is 50 pixels, a dotted area of 50 pixels or less is erased from the white area, and if it is 200 pixels, a dotted area of 200 pixels or less is erased from the white area. If the area threshold is 0 pixel, the dotted area in the white area remains unerased.
[0106]
(2) Expansion / contraction treatment
In order to grasp the black region (lass structure portion) in the white region (martensite region) as the region, the black region was subjected to expansion / contraction processing (closing).
The expansion process is the same as the expansion process performed in the first embodiment. Here, the expansion process (maximum value filter process) was performed twice using a 5 × 5 filter.
The contraction process is the same as the contraction process performed in the first embodiment. Here, the contraction process (minimum value filter process) was performed twice using a 5 × 5 filter.
[0107]
In this expansion / contraction process, the black region (lass structure portion) in the white region (martensite region) is extracted as one region using the maximum value filter and the minimum value filter.
As a result of performing expansion / contraction processing on the black area, a small white portion in the black area is filled with black, and a complete black area without disconnection or the like is obtained.
[0108]
C. Feature calculation
The following six feature amounts were calculated using the binarized image obtained as described above.
a. Black area after binarization
The black area of the binarized image is divided by the total image area.
b. After binarization, the area of the black area where the minute label is removed (50 pixels or less)
The area of the black region after binarization and removal of the minute black label (50 pixels or less) is divided by the total image area.
[0109]
c. After binarization, the area of the black area after removing the minute label (200 pixels or less)
The area of the black region after binarization and removal of the minute black label (200 pixels or less) is divided by the total image area.
d. Black area after binarization and expansion / contraction treatment
This is obtained by dividing the area of the black region of the binary image obtained by performing the expansion / contraction process on the binarized image by the total image area.
[0110]
e. After binarization, the black area after removal of minute labels (50 pixels or less) and further expansion / contraction processing
This is obtained by binarizing, removing a minute black label (50 pixels or less), and dividing the area of the black area of the binary image obtained by further expanding and contracting the image by the total image area.
[0111]
f. After binarization, the black area after removal of minute labels (200 pixels or less) and further expansion / contraction processing
This is binarized to remove a minute black label (200 pixels or less), and the area of the black area of the binary image obtained by further expanding and contracting the image is divided by the total image area.
[0112]
These feature amounts represent the degree of disappearance of the black region (lass structure portion) in the white region (martensite region). The degree of change (degree of change with time) is quantified. As a result, there is no room for an individual's subjectivity when evaluating the tissue image, and it is possible to objectively evaluate the degree of deterioration of the weld heat affected zone (HAZ). Further, since this evaluation method can be systematized, the evaluation process can be automated.
[0113]
19 to 21 are diagrams showing the processing results of the feature values obtained above. No. 1 welding heat affected zone (HAZ) (test material: No. 1) as an example, for each feature amount, the correlation with the LM parameter and the correlation with the heat application time by temperature are arranged. It is. The correlation coefficient with the LM parameter is displayed in parentheses () in each figure. Since the LM parameter of the unused material cannot be calculated, a value of 17.0 is used for convenience.
[0114]
19A is a diagram showing the black area after binarization, FIG. 19B is a diagram showing the black area after the binarization of minute labels (50 pixels or less), and FIG. FIG. 20C is a diagram showing the area of the black region after the binarized minute label is removed (200 pixels or less), and FIG. 20D is the area of the black region after the binarization is expanded / contracted. FIG. 21 (e) is a diagram showing the black area after binarization after removing minute labels (50 pixels or less) and further performing expansion / contraction processing, and FIG. 21 (f) is after binarization. It is a figure which shows the black area | region area after performing a removal (200 pixels or less) of a micro label, and also performing an expansion / contraction process.
[0115]
  BookReference exampleAn example of the result of the image processing performed in Fig. 22 to Fig. 27 is shown.
  22 to 24 show the case where the LM parameter value is low (the degree of deterioration is small). FIG. 22 is a diagram showing an image processing result of welding heat affected zone (HAZ) 1 (test material: No. 1), FIG. 22A is an original image, FIG. 22B is an original image histogram, and FIG. c) is a binarized image, FIG. 23 (d) is an image of binarization and minute label removal (50 pixels or less), FIG. 23 (e) is an image of binarization and minute label removal (200 pixels or less), FIG. 23 (f) is an image of binarization and expansion / contraction processing, FIG. 24 (g) is an image of binarization micro label removal (50 pixels or less) and expansion / contraction processing, and FIG. 24 (h) is a binarization micro image. Label removal
It is an image of (200 pixels or less) and expansion / contraction processing.
[0116]
FIGS. 25 to 27 show Nos. With high LM parameter values (high degradation degree). FIG. 25 is a diagram showing an image processing result of welding heat affected zone (HAZ) 1 (test material: No. 1), FIG. 25A is an original image, FIG. 25B is an original image histogram, and FIG. c) is a binarized image, FIG. 26 (d) is an image of binarization and minute label removal (50 pixels or less), FIG. 26 (e) is an image of binarization and minute label removal (200 pixels or less), Fig. 26 (f) is an image of binarization and expansion / contraction processing, Fig. 27 (g) is an image of binarization minute label removal (50 pixels or less) and expansion / contraction processing, and Fig. 27 (h) is a binarization minute image. It is an image of label removal (200 pixels or less) and expansion / contraction processing.
[0117]
According to the result of this image processing, all of the six feature quantities have a high correlation with the LM parameter, and the degree of deterioration of the welding heat affected zone (HAZ) can be determined by converting the ratio of the area from which the point-like structure is removed to the feature quantity. It became clear that it could be quantified.
[0118]
  As explained above, the bookReference exampleAccording to this material lifetime evaluation method, the following effects can be obtained.
(1) The ratio of the area of the lath structure portion in the martensite region of the binarized image can be characterized and the degree of deterioration of the material can be estimated based on the area ratio. Therefore, in determining the tissue image, it is possible to exclude the influence due to the subjectivity of the individual and to objectively evaluate the degree of material deterioration. As a result, the evaluation can be quantified and automated.
[0119]
(2) After forming the binarized image, the point-like structure portion in the martensite region is removed with a predetermined area threshold value, so that it is considered that only a healthy lath structure remains, and the point-like structure in the martensite region. Even when a portion is generated, the ratio of the lath structure portion in the martensite region of the binarized image can be converted into a feature amount.
(3) Since the obtained image is subjected to expansion / contraction processing after the point-like structure portion is removed, the characteristic amount of the ratio of the area of the lath structure portion in the martensite region of the binarized image can be simplified. Can be.
[0120]
As mentioned above, although each embodiment of the lifetime evaluation method of the material of this invention has been described based on drawing, a specific structure is not limited to each said embodiment, It can change suitably. For example, the same effect can be obtained by using an optical microscope and an electron microscope image or a metal microscope image obtained by directly photographing the surface of the material instead of the replica image.
Further, the threshold value for removing the minute label is not limited to 50 pixels and 200 pixels as long as it is a value that can sufficiently remove the point-like structure to be removed.
The material to be evaluated is not limited to the above materials as long as it is a material that changes with time in the tissue when used at a high temperature.
[0121]
【The invention's effect】
  As explained above, the present inventionMaterial ofAccording to the material life evaluation method, a structure image of a material composed of two phases is binarized by a predetermined threshold value to obtain a binarized image divided into two phases, and then based on the binarized image, Since one or more feature quantities characterizing the material are obtained, and the degree of change with time of the material is estimated based on the feature quantities, the influence of individual subjectivity is excluded when determining the tissue image. The degree of change with time of the material can be objectively evaluated.
  In addition, since the binarized image is superimposed on the texture image of the material and binarized again, the binarized image can be visualized by erasing fine irregularities and the like in each area of the binarized image and making it a feature amount. The success or failure of extraction of each of the two phases of the chemical image can be confirmed.
[0122]
Further, since one or more feature quantities characterizing the material are obtained based on the binarized image, the evaluation result can be quantified and the process of life evaluation can be automated. As a result, the time required for the life evaluation can be shortened and the inspection cost can be reduced.
[0123]
  The present inventionIn the material life evaluation method, expansion / contraction processing is performed on the binarized image.ByIt is possible to easily convert the feature amounts of the two phases of the binarized image.
[0124]
  The present inventionFor life evaluation methodLeave, And removing a microstructure portion in at least one phase of the binarized image with a predetermined area threshold valueByIn addition, it is possible to remove a microstructure portion generated in at least one of the phases as the material changes with time. Therefore, the ratio of the areas of the two phases of the binarized image when a microstructure portion is generated in the phase can be converted into a feature amount.
[0125]
  The present inventionFor life evaluation methodLeaveThen, after removing the microstructure portion, the binarized image is subjected to expansion / contraction processingByTherefore, it is possible to easily perform feature quantification and improve the accuracy of quantification of evaluation results.
[0126]
  The present inventionFor life evaluation methodLeaveThe binarized image is superimposed on the structure image of the material and binarized again.ByThe success or failure of the extraction of each of the two phases of the binarized image can be confirmed by visual observation by erasing fine irregularities or the like in each region of the binarized image and making it a feature amount.
[0127]
  According to a reference example of the present inventionAccording to the material life evaluation method, a tissue image of a material composed of one phase is binarized by a predetermined threshold value to obtain a binarized image divided into two regions, and the material is obtained based on the binarized image. Since one or more feature quantities to be characterized are obtained and the degree of temporal change of the material is estimated based on the feature quantities, the influence of individual subjectivity is excluded when determining the tissue image. The degree of change with time can be objectively evaluated. Further, since one or more feature quantities characterizing the material are obtained based on the binarized image, the evaluation result can be quantified and the process of life evaluation can be automated. As a result, the time required for the life evaluation can be shortened, the inspection cost can be reduced, and the manufacturing cost of the product can be reduced.
[0128]
  According to the reference example of the present inventionFor material life evaluation methodLeave, Expansion / contraction processing is performed on each of the two areas of the binary image.When appliedEven if it is difficult to discriminate each of the two regions by visual observation, it is possible to easily perform the feature amount and improve the accuracy of quantifying the evaluation result.
[0129]
  According to the reference example of the present inventionFor material life evaluation methodLeave, And removes the microstructure portion in at least one region of the binarized image with a predetermined area threshold value.CaseEven when a microstructure portion is generated in the region, the two regions of the binarized image can be converted into feature amounts.
[0130]
  According to the reference example of the present inventionFor material life evaluation methodLeave, After removing the microstructure portion, each of the two regions is subjected to expansion / contraction treatmentWhen appliedThe feature amounts of each of the two regions can be easily obtained, and the accuracy of quantifying the evaluation result can be improved.
[0131]
  According to the present invention or the reference example of the present inventionFor material life evaluation methodLeaveThe tissue image is a replica image of the material, or an optical microscope image or an electron microscope image of the material.ByThe tissue image on the surface of the material can be easily binarized, and the evaluation result can be easily quantified and automated.
[0132]
  According to the present invention or the reference example of the present inventionFor material life evaluation methodLeave, The feature amount and the contrast ratio in the original image of the two phasesByThe degree of change with time of a material composed of two phases can be objectively evaluated.
[0133]
  According to the present invention or the reference example of the present inventionFor material life evaluation methodLeaveThe feature amount is based on an image of one of the two phases and a carbide observed in a region of the one image as the material changes over time;ByEven when carbide is generated in the region, the two regions of the binarized image can be characterized.
[0134]
  According to the present invention or the reference example of the present inventionFor material life evaluation methodLeaveThe feature amount is based on an image of one of the two phases and a lath structure observed in a region of the one image as the material changes over time.ByEven when a lath structure is generated in the region, the two regions of the binarized image can be featured.
[0135]
  According to the present invention or the reference example of the present inventionFor material life evaluation methodLeave, And the area ratio between one image and the other image of the binarized image.ByTwo regions of the binarized image can be featured.
[0136]
  the aboveAnyFor material life evaluation methodLeaveThe binarized image was defined as the area ratio of the ground portion of the tissue and the martensitic structure formed inside the ground portion.CaseEven when a martensitic structure is generated inside the ground portion, the two regions of the binarized image can be featured.
[Brief description of the drawings]
FIG. 1 is a flowchart showing a material lifetime evaluation method according to a first embodiment of the present invention.
FIG. 2 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 3 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 4 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 5 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 6 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 7 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 8 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 9 is a diagram illustrating a processing result of a feature amount according to the first embodiment of this invention.
FIG. 10 is a diagram illustrating a result of image processing of a base material having a low LM parameter value (low degradation degree) according to the first embodiment of the present invention.
FIG. 11 is a diagram illustrating a result of image processing of a base material having a low LM parameter value (low degradation degree) according to the first embodiment of the present invention.
FIG. 12 is a diagram illustrating a result of image processing of a base material having a low LM parameter value (low degradation degree) according to the first embodiment of the present invention.
FIG. 13 is a diagram illustrating a result of image processing of a base material having a low LM parameter value (low degradation degree) according to the first embodiment of the present invention.
FIG. 14 is a diagram illustrating a result of image processing of a base material having a high LM parameter value (a high degree of deterioration) according to the first embodiment of the present invention.
FIG. 15 is a diagram illustrating a result of image processing of a base material having a high LM parameter value (a high degree of deterioration) according to the first embodiment of the present invention.
FIG. 16 is a diagram illustrating a result of image processing of a base material having a high LM parameter value (a high degree of deterioration) according to the first embodiment of the present invention.
FIG. 17 is a diagram illustrating a result of image processing of a base material having a high LM parameter value (a high degree of deterioration) according to the first embodiment of the present invention.
FIG. 18 shows the present invention.According to reference examplesIt is a flowchart which shows the lifetime evaluation method of material.
FIG. 19 shows the present invention.Reference exampleIt is a figure which shows the processing result of the feature-value.
FIG. 20 shows the present invention.Reference exampleIt is a figure which shows the processing result of the feature-value.
FIG. 21 shows the present invention.Reference exampleIt is a figure which shows the processing result of the feature-value.
FIG. 22 shows the present invention.Reference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is low (deterioration degree is small).
FIG. 23 shows the present invention.Reference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is low (deterioration degree is small).
FIG. 24 shows the present invention.Reference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is low (deterioration degree is small).
FIG. 25 shows the present invention.Reference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is high (deterioration degree is large).
FIG. 26 of the present inventionReference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is high (deterioration degree is large).
Fig. 27 of the present inventionReference exampleIt is a figure which shows the result of the image processing of the base material whose LM parameter value of is high (deterioration degree is large).
FIG. 28 is a schematic diagram showing a standard structure corresponding to each stage of deterioration of low alloy CrMo steel.
FIG. 29 is a schematic diagram showing a standard structure corresponding to each stage of deterioration of low alloy CrMo steel.

Claims (4)

A method for evaluating a lifetime when a change with time occurs in at least one phase of a metal material composed of two phases due to use at a high temperature,
The binarized image obtained by binarizing the metal structure image into a white area and a black area by binarizing using a canonical discrimination method according to a predetermined gray level based on the gray level information of the image,
Extracts the histology of the white area by masking the histology of the metal material using the binarized image after performing the expansion and contraction processing a black region of the 2 Nekazo was extracted The white region tissue image is binarized again using a canonical discrimination method, and the white region tissue image is a re-binarized image from which carbide is extracted,
Further, after performing expansion / contraction processing and contraction / expansion processing on the black region of the binarized image, the tissue image of the black region can be obtained by masking the tissue image of the metal material using the binarized image. Extracting and binarizing the extracted black region tissue image using a canonical discriminant method, and re-binarized image from which the lath structure is extracted for the black region tissue image,
A method for evaluating the lifetime of a metal material, wherein the degree of change with time of the metal material is estimated based on a feature quantity based on the carbide and lath structure. A method for evaluating a lifetime when a change with time occurs in at least one phase of a metal material composed of two phases due to use at a high temperature,
The binarized image obtained by binarizing the metal structure image into a white area and a black area by binarizing using a canonical discrimination method according to a predetermined gray level based on the gray level information of the image,
Extracting the tissue image of the white region by masking the tissue image of the metal material using the binary image after performing expansion / contraction processing on the black region of the binary image ,
Further, after performing expansion / contraction processing and contraction / expansion processing on the black region of the binarized image, the tissue image of the black region can be obtained by masking the tissue image of the metal material using the binarized image. Extract and
Based on the contrast ratio between the tissue image of the black region and the tissue image of the white region, which is calculated from the luminance of the tissue image of the black region and the luminance of the tissue image of the white region , A method for evaluating the life of a metal material, characterized by estimating a degree. A method for evaluating a lifetime when a change with time occurs in at least one phase of a metal material composed of two phases due to use at a high temperature,
The binarized image obtained by binarizing the tissue image of the metal material into a white region and a black region by binarizing using a canonical discrimination method according to a predetermined shading degree based on shading information of the image,
Extracting the tissue image of the white region by masking the tissue image of the metal material using the binary image after performing expansion / contraction processing on the black region of the binary image ,
Further, after performing expansion / contraction processing and contraction / expansion processing on the black region of the binarized image, the tissue image of the black region can be obtained by masking the tissue image of the metal material using the binarized image. Extract and
A method for evaluating the lifetime of a metal material, wherein the degree of change with time of the metal material is estimated based on an area ratio between the white region and the black region. Wherein prior to application of the expansion and contraction processing, claims 1 and removing a predetermined area living values at least one of the infinitesimal structure portion of the white and black areas of the 2 Nekazo 3 The lifetime evaluation method of the metal material of any one of these.

Images