JP3758236B2 - Misfire detection device for internal combustion engine - Google Patents

Description

として算出される。ここで、「（Ｔ１２０ｉ−Ｔ１２０ｉ-6）／６」項は、過渡補正項であり、気筒間のクランク角偏差時間の算出にこうした過渡補正項を加味することにより、例えば急加速時や急減速時等、内燃機関の運転条件による過渡的な回転変動増減の影響は好適に排除されるようになる。
【００８５】
同様にして、第１気筒に対する第３〜第６気筒のクランク角偏差時間ΔＴ#3〜ΔＴ#6は、それぞれ
ΔＴ#3＝｛（Ｔ１２０ｉ＋２×Ｔ１２０ｉ-6）／３｝−Ｔ１２０ｉ-4…（４）
ΔＴ#4＝｛（Ｔ１２０ｉ＋Ｔ１２０ｉ-6）／２｝−Ｔ１２０ｉ-3…（５）
ΔＴ#5＝｛（２×Ｔ１２０ｉ＋Ｔ１２０ｉ-6）／３｝−Ｔ１２０ｉ-2…（６）
ΔＴ#6＝｛（５×Ｔ１２０ｉ＋Ｔ１２０ｉ-6）／６｝−Ｔ１２０ｉ-1…（７）
として算出される。
【００８６】
なお、上記第１気筒の前回のクランク軸１２０°ＣＡ回転時間Ｔ１２０ｉ-6を含め、第２〜第６気筒のクランク軸１２０°ＣＡ回転時間Ｔ１２０ｉ-5〜Ｔ１２０ｉ-1は、上記ステップＳ２０３を通じて算出され、後のステップＳ２１７を通じて更新されている値が用いられる。
【００８７】
こうして第１気筒に対する第２〜第６気筒のクランク角偏差時間ΔＴｎを算出した電子制御装置９は次に、ステップＳ２０６にて、次式に基づき、それらクランク角偏差時間ΔＴｎをクランク角偏差Δθｎ、すなわち回転角度の偏差に変換する。ただし、次式（８）式において、ｎは、＃２〜＃６の５気筒分である。
Δθｎ＝ΔＴｎ×（１２０°ＣＡ／Ｔ１２０ｉ） …（８）
この第１気筒に対する第２〜第６気筒のクランク角偏差Δθｎを求めると、同電子制御装置９では、次のステップＳ２０７にて、内燃機関１が現在、特定の運転条件下、例えば急加速や急減速等の過渡状態、シフトチェンジ状態、燃料カット時や復帰時、始動時や電気負荷投入時、アイドル状態、パージ制御状態、ＥＧＲ（排気還流制御）実行中、可変吸気実行中等々、クランク軸の大きな回転変動を招く特定の運転状態、或いは軽負荷運転域や高回転域等、いわゆる失火判定不能な運転域にないか否かをその都度の運転情報に基づき判断する。そして、同機関１がこうした特定の運転条件下にないことを条件に、ステップＳ２０８にて、各気筒別、且つ運転条件の別に上記求めた（変換した）クランク角偏差Δθｎを積算し、続くステップＳ２０９にて、前記積算カウンタ９０７をインクリメントする。
【００８８】
すなわち、内燃機関１が上記急加速や急減速等の過渡状態、シフトチェンジ状態、燃料カット時や復帰時、始動時や電気負荷投入時、等々の運転条件下にあった場合には、上記クランク角偏差Δθｎも、同機関１の正常な燃焼状態において求められた値ではない可能性が高い。そこで、内燃機関１のそのような運転条件下では、上記求めたクランク角偏差Δθｎについての積算処理を行わないようにしている。なお後述するように、同実施形態にかかる装置にあっては、この積算処理されるいわば正常なクランク角偏差Δθｎのみが、後に実施される公差学習処理に供されることとなる。
【００８９】
また、上記ステップＳ２０８におけるクランク角偏差Δθｎの各気筒別、且つ運転条件別の積算処理は前述のように、前記気筒間クランク角偏差（公差）積算値メモリ９０８に対して行われる。この積算値メモリ９０８のメモリ構造を図５に例示する。
【００９０】
この図５に示されるように、上記気筒間クランク角偏差（公差）積算値メモリ９０８は、第２〜第６気筒（＃２〜＃６）の別に、且つ機関１の運転条件である回転速度（ＮＥ）及び負荷（吸気管圧力ＰＭ）の別に、クランク角偏差Δθｎが積算登録される構造となっている。すなわち、本学習制御ルーチンの繰り返しの実行に基づき、同図５に示されるテーブルの各々には、それぞれ正常なクランク角偏差Δθｎが、「ΣΔθｎ(NE,PM) 」といったかたちで積算登録されるようになる。そして、前記積算カウンタ９０７は、こうして気筒間クランク角偏差（公差）積算値メモリ９０８に登録されたクランク角偏差ΣΔθｎ(NE,PM) の積算数をその計数値として示すこととなる。
【００９１】
こうしてクランク角偏差Δθｎの積算処理を行うと、電子制御装置９は次に、ステップＳ２１０にて、公差学習を行うべきか否か、その実行条件をチェックする。この実行条件のチェック処理については、後に図７及び図８を併せ参照して詳述する。
【００９２】
該公差学習実行条件についてのチェックを終えた電子制御装置９は、次のステップＳ２１１にて、前記点火数カウンタ９０９の計数値に基づき例えば「１００」点火等、所定の点火数が経過しているか否かを判断する。この結果、所定の点火数に達していない旨判断される場合には、ステップＳ２１６に移行して、前記点火数カウンタ９０９をインクリメントし、ステップＳ２１７にて、前記各気筒のクランク軸１２０°ＣＡ回転時間Ｔ１２０ｉの値を
Ｔ１２０ｉ-6 ＝Ｔ１２０ｉ-5 …（９）
Ｔ１２０ｉ-5 ＝Ｔ１２０ｉ-4 …（１０）
Ｔ１２０ｉ-4 ＝Ｔ１２０ｉ-3 …（１１）
Ｔ１２０ｉ-3 ＝Ｔ１２０ｉ-2 …（１２）
Ｔ１２０ｉ-2 ＝Ｔ１２０ｉ-1 …（１３）
Ｔ１２０ｉ-1 ＝Ｔ１２０ｉ …（１４）
といったかたちで更新した後、本ルーチンを一旦抜ける。
【００９３】
他方、所定の点火数を経過している旨判断される場合には、ステップＳ２１２にて、上記公差学習実行条件についてのチェック結果に基づき、同実行条件の成否判定を行う。この公差学習実行条件の成否判定処理については、後に図１１を併せ参照して詳述する。
【００９４】
電子制御装置９は次いで、ステップＳ２１３にて、該公差学習実行条件の成否判定が公差学習実行の「可」を示すものであるか「不可」を示すものであるかを判断する。そして、同成否判定が「公差学習実行不可」を示すものであった場合には、上記ステップＳ２１６及びステップＳ２１７の処理を実行して本ルーチンを一旦抜け、「公差学習実行可」を示すものであったときに、ステップＳ２１４にて公差学習を実行する。
【００９５】
この公差学習は、前記バックアップＲＡＭ９ｄ内の気筒間クランク角偏差（公差）学習値メモリ９１０に対して行われる。この学習値メモリ９１０のメモリ構造を図６に例示する。
【００９６】
この図６に示されるように、該学習値メモリ９１０も、上記気筒間クランク角偏差（公差）積算値メモリ９０８（図５）同様、第２〜第６気筒（＃２〜＃６）の別に、且つ機関１の運転条件である回転速度（ＮＥ）並びに負荷（吸気管圧力ＰＭ）の別に、前記クランク角偏差についての学習値ΔθｎＬが更新登録される構造となっている。
【００９７】
そしてここでは、上述した積算処理（ステップＳ２０８）において気筒間クランク角偏差（公差）積算値メモリ９０８に登録されている気筒別、運転条件別のクランク角偏差積算値ΣΔθｎ(NE,PM) を読み込んでその平均値Δθｎ(NE,PM)_AVを
Δθｎ(NE,PM)_AV ＝ΣΔθｎ(NE,PM) ／（積算カウンタ計数値）…（１５）
として求めるとともに、該求めたクランク角偏差平均値Δθｎ(NE,PM)_AVと上記学習値メモリ９１０内の当該気筒、並びに当該運転条件に対応する同クランク角偏差についての学習値ΔθｎＬ(NE,PM) とから、なまし（徐変）演算

を実行して、新たな学習値ΔθｎＬ(NE,PM) を求める。そして、この新たに求めた学習値ΔθｎＬ(NE,PM) を、上記学習値メモリ９１０の該当する欄に更新登録する。
【００９８】
なお、上記（１６）式において、値「８」は、なまし（徐変）係数であり、該値「８」以外にも処理系に応じた任意の値を採用することができることは云うまでもない。
【００９９】
また、上記学習値メモリ９１０において、その学習値ΔθｎＬ(NE,PM) の更新が行われるのは、上記積算値メモリ９０８にも対応するクランク角偏差積算値ΣΔθｎ(NE,PM) が存在している場合に限られる。すなわち、対応するクランク角偏差積算値ΣΔθｎ(NE,PM) が存在していなかった場合、その平均値Δθｎ(NE,PM)_AVも得られないことから、上記（１６）式のなまし（徐変）演算自体、その実行が不可能となる。
【０１００】
公差学習制御ルーチンにおいて、こうして公差学習を実行した電子制御装置９は、次のステップＳ２１５にて、前記積算値メモリ９０８、前記積算カウンタ９０７、及び前記点火数カウンタ９０９をそれぞれリセットする。そして次の学習に備えるべく、上述したステップＳ２１６並びにステップＳ２１７の処理を実行した後、本ルーチンを一旦抜ける。
【０１０１】
電子制御装置９（学習制御部９０６）を通じてこのような機関１の運転条件に応じた学習処理が行われることにより、前記メインルーチン（図３）において同学習値ΔθｎＬ(NE,PM) に基づき公差補正された値として算出されるクランク角速度ωｎの値も自ずとその信頼性が高められることとなる。そしてひいては、その後の失火判定に際しても、その判定精度は自ずと高いものとなる。
【０１０２】
次に、図７及び図８を参照して、上記公差学習制御ルーチンにおけるステップＳ２１０の処理として実行される公差学習実行条件のチェック処理について説明する。
【０１０３】
この図７及び図８に示す公差学習実行条件のチェックルーチンにおいて、電子制御装置９（学習制御部９０６）は、これまで同様、
（１）前回の割り込み時刻と今回の割り込み時刻との偏差から、クランク角が６０°ＣＡ回転するのに要した時間Ｔ６０ｉを算出する（ステップＳ３００）。
（２）今回の割り込みタイミングが上死点後（ＡＴＤＣ）６０°ＣＡであるか否かを前記基準信号ＣＹＬに基づき判断する（ステップＳ３０１）。
（３）この割り込みタイミングが上死点後６０°ＣＡではない旨判断される場合、上記求めた時間Ｔ６０ｉをＴ６０ｉ-1とした後、本ルーチンを一旦終了する（ステップＳ３１９）。
（４）同割り込みタイミングが上死点後６０°ＣＡである旨判断される場合には、前記基準信号ＣＹＬに基づき今回の気筒の気筒番号ｎを識別した後（ステップＳ３０２）、上記求めた時間Ｔ６０ｉについての過去２回分のデータを累積して、クランク軸が１２０°ＣＡ回転するのに要した時間Ｔ１２０ｉを算出する（ステップＳ３０３）。
といった処理を実行した後、次のステップＳ３０４にて、先の（１）’式に基づき気筒毎のクランク角速度ωｎを算出する。そして、更に次のステップＳ３０５にて、それら算出したクランク角速度ωｎに基づき、３６０°ＣＡ差分法、すなわち失火検出対象となる気筒及びその隣接気筒の回転角速度差分を３６０°ＣＡ離れた気筒の同差分から差し引いた２階差分
Δωｎ＝（ωｎ-1 − ωｎ）−（ωｎ-4 −ωｎ-3 ） …（１７）
を用いてクランク角速度変動量Δωｎを算出する。
【０１０４】
こうしてクランク角速度変動量Δωｎを算出した電子制御装置９は次に、ステップＳ３０６にて、このクランク角速度変動量Δωｎと同変動量Δωｎに対して予め設定されている失火判定値ＲＥＦ２とを比較する。そして、このクランク角速度変動量Δωｎが失火判定値ＲＥＦ２を超えている旨判断される場合には、ステップＳ３０７にて、前記仮失火カウンタ９１２のうちのＣＭＦカウンタをインクリメントして、ステップＳ３０８の処理に移行する。
【０１０５】
他方、ステップＳ３０６において、クランク角速度変動量Δωｎが失火判定値ＲＥＦ２以下である旨判断される場合には、そのままステップＳ３０８の処理に移行する。
【０１０６】
ステップＳ３０８においては、上記クランク角速度変動量Δωｎと同変動量Δωｎに対して予め設定されているラフロード（悪路走行）判定値ＲＥＦ３（＜ＲＥＦ２）とを更に比較する。
【０１０７】
ラフロードにあっては一般に、過渡的な回転変動が起こりやすい状況にあるため、こうした状況が継続される場合にも、公差学習は実行すべきではない。
そこで、電子制御装置９は、同ステップＳ３０８において、クランク角速度変動量Δωｎが該ラフロード判定値ＲＥＦ３を超えていて且つ上記失火判定値ＲＥＦ２以下である旨判断される場合には、現在ラフロードを走行中であるとして、ステップＳ３０９にて前記ラフロード（ＣＲＧ）カウンタ９１１をインクリメントする。
【０１０８】
他方、同ステップＳ３０８において、クランク角速度変動量Δωｎが上記ラフロード判定値ＲＥＦ３以下である旨判断される場合には、そのままステップＳ３１０の処理に移行する。
【０１０９】
この公差学習実行条件チェックルーチンにおいて、ステップＳ３１０（図８）以降の処理は、上記３６０°ＣＡ差分法では失火の判定が不可能である３６０°ＣＡ対向気筒連続失火を検出するための失火判定処理である。引き続き、それら処理の詳細について説明する。
【０１１０】
上記ラフロード判定を終えた電子制御装置９は、次のステップＳ３１０にて、前記酸素（Ｏ2 ）センサ１５の出力に基づく空燃比フィードバック（Ｆ／Ｂ）制御の実行中であるか否かをまず判断する。
【０１１１】
因みに、こうしたフィードバック制御が実行されている状態にあって機関１に失火が発生している場合には、その空燃比フィードバック補正係数ｃｆｂと同補正係数の平均値ｃｆｂＡＶとの和が、上記酸素センサ１５の特性や内燃機関個々の機差などによる初期公差よりも大きい側（空燃比のリーン（Ｌ）側）にずれることが発明者等によって確認されている。
【０１１２】
そこで、電子制御装置９は、上記ステップＳ３１０にて空燃比フィードバック制御中である旨判断される場合、ステップＳ３１１にて、空燃比フィードバック補正係数ｃｆｂ及び同補正係数の平均値ｃｆｂＡＶの和と上記初期公差とを比較し、それら和が上記初期公差以上であるときには、失火が発生しているとして、ステップＳ３１２にて前記仮失火カウンタ９１２のうちのＣＯＦカウンタをインクリメントする。
【０１１３】
ところで、上記ステップＳ３１１の判定処理は、空燃比のフィードバック制御が実行中であることが大前提となるが、例えば高負荷燃料増量中など、こうしたフィードバック制御が行われない場合であっても、当該機関１が正常点火されているか否かについての判断が行われることが望ましい。
【０１１４】
因みに、前記酸素（Ｏ2 ）センサ１５にあっては、その活性時、内燃機関１に失火が発生すると、
（Ａ）その出力周期が極端に短くなる。
或いは
（Ｂ）その出力がリーン（Ｌ）側にへばりつく。
といった何れかの状態を示すようになることが発明者等によって確認されている。これら（Ａ）及び（Ｂ）の状態についての測定結果をそれぞれ図９及び図１０に示す。
【０１１５】
例えば、内燃機関１の失火に伴い、酸素センサ１５の出力周期が短くなる場合には、図９において「ＦＯ2 センサ出力」として示されるように、正常点火時の振幅周期（およそ０．５〜２Ｈｚ）に対して明らかに区別できるような短い周期（同図９に「ｃｓｇｔ」として示される点火周期程度）となる。
【０１１６】
なお、この図９は、空燃比フィードバック制御が実行されている状態における上記空燃比フィードバック補正係数ｃｆｂの推移についても併せ示しており、機関１に失火が発生した場合にこの空燃比フィードバック補正係数ｃｆｂと同補正係数の平均値ｃｆｂＡＶとの和が大きな値をとるようになることは、この図９によっても明らかである。
【０１１７】
一方、内燃機関１の高負荷燃料増量時（ＷＯＴ）には、同機関１の失火に伴い、酸素センサ１５の出力がリーン（Ｌ）側にへばりつくようになる。そしてこの場合には、図１０においてこれも「ＦＯ2 センサ出力」として示されるように、上述した正常点火時の振幅周期よりも十分長い時間に亘って、その出力がリーン（Ｌ）側に固定されるようになる。
【０１１８】
このように、上記酸素センサ１５が活性状態にあれば、その出力（Ｒ／Ｌ）を監視することで、内燃機関１の失火発生の有無を判定することができるようになる。
【０１１９】
図８に示す同公差学習実行条件チェックルーチンにおいて、ステップＳ３１３以降の処理は、こうした原理に基づいて酸素センサ１５の出力から内燃機関１の失火発生の有無を判定するための処理である。
【０１２０】
すなわち、上記ステップＳ３１０にて空燃比フィードバック制御中ではない旨判断した、或いは空燃比フィードバック制御中であったとしても上記ステップＳ３１１にて空燃比フィードバック補正係数ｃｆｂ及び同補正係数の平均値ｃｆｂＡＶの和が上記初期公差未満である旨判断した電子制御装置９は、ステップＳ３１３にて、上記酸素（Ｏ2 ）センサ１５が活性状態にあるか否かをまず判断する。そして、同酸素センサ１５が活性状態にあることを条件に、それぞれ
・その出力周期（振幅周期）が正常点火時の振幅周期Ｆｓよりも短いか否か（ステップＳ３１４）。
・そのリーン（Ｌ）側の出力時間が正常点火時のリーン側出力時間ＴLOW よりも長いか否か（ステップＳ３１６）。
といった比較を行い、同出力周期が正常点火時の振幅周期Ｆｓよりも短い旨判断される場合には、ステップＳ３１５にて前記仮失火カウンタ９１２のうちのＣＦカウンタをインクリメントし、同出力のリーン側出力時間が正常点火時のリーン側出力時間ＴLOW よりも長い旨判断される場合には、ステップＳ３１７にて前記仮失火カウンタ９１２のうちのＣＴカウンタをインクリメントする。
【０１２１】
こうして全ての項目についてのチェックを終えた電子制御装置９は、最後に、ステップＳ３１８にて、上記ステップＳ３０４において算出したクランク角速度ωｎをはじめとするそれら気筒毎のクランク角速度の値に対し、前述のようにω(n-5) →廃棄、ω(n-4) →ω(n-5) 、ω(n-3) →ω(n-4) 、ω(n-2) →ω(n-3) 、ω(n-1) →ω(n-2) 、ωｎ→ω(n-1) といった更新処理を施して、同公差学習実行条件チェックルーチンを抜ける。
【０１２２】
なお、こうした公差学習実行条件のチェックルーチンが、前記点火数カウンタ９０９の計数値に基づき、例えば「１００」点火等を経過するまで繰り返し実行されるようになることは公差学習制御ルーチン（図４）の説明において既述した通りである。
【０１２３】
次に、図１１を更に参照して、上記公差学習制御ルーチンにおけるステップＳ２１２の処理として実行される公差学習実行条件の成否判定処理について説明する。
【０１２４】
この図１１に示す公差学習実行条件の成否判定ルーチンは前述のように、公差学習制御ルーチン（図４）のステップＳ２１１において上記所定の点火数を経過している旨判断される場合に起動される。
【０１２５】
こうして公差学習実行条件の成否判定ルーチンが起動されると、電子制御装置９（学習制御部９０６）はまず、ステップＳ４００にて、前記仮失火カウンタ９１２を構成する各カウンタ（ＣＭＦカウンタ、ＣＯＦカウンタ、ＣＦカウンタ、及びＣＴカウンタ）の計数値が何れか１つでも「１」以上となっているか否か、或いは前記ラフロードカウンタ（ＣＲＧカウンタ）９１１の計数値が同計数値に対する所定のラフロード判定値ＫＲＧ以上となっているか否かを判断する。
【０１２６】
その結果、前記仮失火カウンタ９１２の計数値が何れか１つでも「１」以上となっている場合、或いは前記ラフロードカウンタ９１１の計数値が上記判定値ＫＲＧ以上となっている場合には、ステップＳ４０１にて、前記ＲＡＭ９ｃ内の適宜の領域に「公差学習実行不可」を示すフラグをセットする。
【０１２７】
他方、前記仮失火カウンタ９１２の計数値が何れも「０」であり、且つ前記ラフロードカウンタ９１１の計数値が上記判定値ＫＲＧ未満である場合には、ステップＳ４０２にて、同ＲＡＭ９ｃ内の適宜の領域に「公差学習実行可」を示すフラグをセットする。
【０１２８】
こうしてフラグ処理を終えると、同電子制御装置９は、前記仮失火カウンタ９１２並びにラフロードカウンタ９１１をリセットして、同公差学習実行条件の成否判定ルーチンを抜ける。
【０１２９】
公差学習制御ルーチン（図４）のステップＳ２１３において、電子制御装置９は、こうして処理した「公差学習実行可」を示すフラグ、或いは「公差学習実行不可」を示すフラグに基づいて前述した公差学習実行の「可」若しくは「不可」を判断することとなる。換言すれば、図７及び図８に示した公差学習実行条件のチェックルーチンにおいて、その全てのチェック項目が正常である場合にのみ、前記態様での公差学習、すなわちその学習値ΔθｎＬ(NE,PM) の更新が行われるようになる。そしてこのため、同学習値ΔθｎＬ(NE,PM) の信頼性も自ずと高く維持されるようになる。
【０１３０】
以上説明したように、同実施形態にかかる失火検出装置によれば、
（１）内燃機関１の気筒別、且つ、運転条件の別にクランク角偏差（気筒間角度公差）についての学習を行うようにしたことで、同機関１のその都度の気筒、並びに運転条件に応じた極めて正確なクランク角速度ωｎを算出することができるようになる。そしてこのため、それらクランク角速度ωｎの推移に基づき算出されるクランク角速度変動量Δω(n-α-1)も自ずと正確な値となり、該クランク角速度変動量Δω(n-α-1)と失火判定値ＲＥＦ１との比較のもとに行われる失火の判定精度も極めて高く維持されるようになる。
（２）３６０°ＣＡ差分法に基づき判定される失火はもとより、該３６０°ＣＡ差分法では失火の判定が不可能である３６０°ＣＡ対向気筒連続失火等についてもその発生の有無を判定することができるとともに、それら判定において正常な点火が確認された場合にのみ上記学習を実行するようにしたことから、学習値の信頼性も高く維持されるようになる。
（３）機関１が例えば急加速や急減速等の過渡状態、シフトチェンジ状態、燃料カット時や復帰時、始動時や電気負荷投入時、アイドル状態、パージ制御状態、ＥＧＲ（排気還流制御）実行中、可変吸気実行中等々、クランク軸の大きな回転変動を招く特定の運転状態にあるとき、或いは軽負荷運転域や高回転域等、いわゆる失火判定不能な運転域にあるときにも学習の実行を禁止するようにしているため、これによっても学習値の信頼性は高く維持される。
（４）前記（３）式〜（７）式によるように、気筒間のクランク角偏差時間ΔＴｎの算出に過渡補正項を加味したことで、例えば急加速時や急減速時等、機関１の運転条件による過渡的な回転変動増減の影響も上記学習値から好適に排除されるようになる。
等々、多くの優れた効果が得られるようになる。
【０１３１】
なお、同実施形態の装置にあっては、公差学習制御ルーチン（図４）のステップＳ２０４において第１気筒（＃１）を判別した際、この第１気筒と他の第２〜第６気筒（＃２〜＃６）との間の全てのクランク角偏差を求めるようにした。
【０１３２】
しかし、偶数気筒からなる内燃機関であれば、３６０°ＣＡ離れた気筒同士は同じロータ被検出部を通じてそのクランク角度が検出されるため、それら気筒間のクランク角偏差はそもそも小さい。
【０１３３】
したがって、これら３６０°ＣＡ離れた気筒同士を１組とし（６気筒の場合であれば３組となる）、それら組毎に気筒間の（組間の）公差学習を行う構成とすることもできる。このような構成によれば、電子制御装置９において必要とされる演算量やＲＯＭ、ＲＡＭ等のメモリ容量を大幅に削減することができるようになる。
【０１３４】
また、公差学習制御ルーチンの同ステップＳ２０４において判別する気筒は、第１気筒（＃１）に限らず、他の任意の気筒であってもよい。要は、特定の気筒（若しくは組）に対する他の気筒（若しくは組）のクランク角偏差が算出される構成でありさえすればよい。
【０１３５】
また、同実施形態の装置にあっては、同じく公差学習制御ルーチン（図４）のステップＳ２０８及びステップＳ２１４において、それぞれ図５及び図６に示される態様で、機関１の運転条件（回転速度ＮＥ，機関負荷ＰＭ）の別にクランク角偏差Δθｎを積算し、或いは学習するようにした。
【０１３６】
しかし、内燃機関の上記運転条件に鑑みた場合、図１２（ａ）及び（ｂ）に、第１、第４気筒グループに対するそれぞれ第２、第５気筒グループ、及び第３、第６気筒グループの機関負荷に対する回転変動公差を例示するように、機関負荷が変化しても、それら回転変動公差の傾向はほぼ一定となっている。
【０１３７】
したがって、図５及び図６に例示したメモリ構造においても、その運転条件として機関負荷の欄を削除し、気筒並びに回転速度の別に、上述したクランク角偏差Δθｎの積算、或いは学習が行われる構成とすることもできる。こうした構成によっても、電子制御装置９において必要とされる演算量やＲＯＭ、ＲＡＭ等のメモリ容量は大幅に削減されるようになる。
【０１３８】
また、上述したクランク角偏差Δθｎの積算、或いは学習を機関１の運転条件の別に行うにしろ、同機関１の高回転域ではそれら積算、或いは学習が行われる機会は少ない。そしてこのため、学習値が求まらず、失火が発生してもその旨を検出することができないこともある。
【０１３９】
しかし、先の図１３に例示したように、各気筒間のクランク角偏差（公差）には、回転速度が増加するとそれら公差もほぼ直線的に増加する傾向がある。
すなわち、機関１の例えば低回転域における頻度の高い２運転条件でそれら公差が学習されたときには、いわゆる線形補間を行うことによって、同機関１の高回転域での公差を割り出りだすことが可能となる。こうした原理に基づいて機関１の高回転域での公差を割り出し、該割り出した公差を学習するようにすれば、上記不都合も好適に解消されるようになる。
【０１４０】
また同原理によれば、機関１の回転変動が大きくなることを予想して公差学習実行条件から外した領域についても、上記線形補間によってそれら領域の公差を学習することができるようになる。
【０１４１】
また、内燃機関１において失火が発生した場合、その未燃ガスが排気管１４内で後燃えし、公差学習実行条件のチェックルーチン（図７、図８）においてその酸素（Ｏ2 ）センサ１５の出力に基づく正確な失火判定（ステップＳ３１４及びステップＳ３１６）が不能となることがある。
【０１４２】
しかし、上記実施形態の装置において、
・排気温センサを追加し、同センサを通じて検出される排気温度が所定温度以上となるときには学習の実行を禁止する。
或いは、
・高負荷状態での運転時等、後燃えが発生しやすい運転条件では学習の実行を禁止する。
といった構成を併せ具えるようにすれば、こうした不都合も好適に回避されるようになる。
【０１４３】
また、同実施形態の装置にあっては、上記酸素センサ１５の出力に基づいて空燃比のフィードバック制御を行うシステムを想定した。しかし、機関の燃焼ガスに基づき空燃比をリニアに検出するリニア空燃比センサを用い、該リニア空燃比センサの出力に基づいて同空燃比のフィードバック制御を行うシステムにあっては、このリニア空燃比センサの出力を利用して、公差学習実行条件チェックルーチン（図７、図８）における前記ステップＳ３１４及びステップＳ３１６の処理に相当する失火判定を行うこともできる。
【０１４４】
因みに、リニア空燃比センサの場合、当該機関に失火が発生すると、
（ａ）その出力がリーン側に変化する。
或いは
（ｂ）その出力が全体的にリーン側へのオフセットを持つようになる。
といった何れかの状態を示すようになる。したがってこの場合、前記学習制御部９０６としては、
・該リニア空燃比センサの出力が所定期間以上リーン側にあるとき前記ＣＦカウンタをインクリメントする。
・該リニア空燃比センサの出力の平均値が所定値以上リーン側にあるとき前記ＣＴカウンタをインクリメントする。
といった構成を採ることとなる。
【０１４５】
なお、こうした失火判定に寄与し得るセンサとしては、ＨＣ濃度センサなどもある。
また、同公差学習実行条件チェックルーチン（図７、図８）におけるチェック項目の選択、或いは組み合わせ等は任意であり、対象となるシステムの規模に応じて自由にそれら項目の選択、或いは組み合わせを行うことができる。もっとも、前述した項目の全てが選択されるとき、前記学習値の信頼性が最大となことは云うまでもない。
【０１４６】
また、同実施形態の装置にあって、上記公差学習実行条件のチェックルーチン（図７、図８）におけるステップＳ３０５のクランク角速度変動量Δωｎの算出には、前記（１７）式による３６０°ＣＡ差分法を用いるとした。
【０１４７】
しかし、ある気筒において失火が発生した場合、クランク角速度ωｎは通常、その後徐々に正常な角速度に戻るようになる。このため、上記３６０°ＣＡ差分法においても次式（１７）’式として示すように、
Δωｎ＝（ωｎ-1 − ωｎ）−（ωｎ+2 −ωｎ+3 ） …（１７）’
といったかたちで、その後のクランク角速度「ωｎ+2」及び「ωｎ+3」を導入することが望ましい。これにより、失火発生の際にはクランク角速度変動量Δωｎとしてより大きな値が得られ、Ｓ／Ｎ（信号／雑音）比の向上が図られるようになる。
【０１４８】
また、同実施形態の装置にあっては、公差学習実行条件の成否判定（図１１）において「公差学習実行可」を示すフラグがセットされることを条件に前記学習値の更新が行われるとしたが、他に例えば、
・更新しようとする値とそれまでの学習値との差が所定以上に大きいときには、その更新しようとする値が複数回連続してほぼ同じ値となるときに限り、その値による学習値の更新を許可する。
といった学習アルゴリズムを採用するようにしてもよい。このようなアルゴリズムによれば、偶然に求まった値によって誤った学習が行われることもなく、それら学習値の信頼性が更に向上されるようになる。
【０１４９】
なお、この公差学習値の算出に際し、同実施形態の装置では上述のように、所定のサンプル数となるまで運転条件別のクランク角偏差を積算し、その平均値（Δθｎ(NE,PM)_AV）に基づき（正確には（１６）式のなまし演算によって）学習値（ΔθｎＬ(NE,PM) ）を求めたが、この平均値に代えて、所定のサンプル数となるまで同運転条件別のクランク角偏差をなまし処理した値なども適宜採用することができる。
【０１５０】
また、同実施形態の装置にあっては、メインルーチン（図３）での失火判定の際、失火判定値ＲＥＦ１と比較されるクランク角速度変動量Δω(n-α-1)を前記学習値に基づき補正することとしたが、失火判定値ＲＥＦ１の側を前記学習値に基づき補正するようにしても勿論よい。
【０１５１】
また、同学習値としても、前記クランク角偏差（気筒間角度公差）に限らず、それに相当する値、例えばクランク角速度、或いはその変動量、等々を採用することもできる。
【０１５２】
ところで、同実施形態では触れなかったが、内燃機関の特性として、機関本体の振動等により、ある特定の回転速度においてクランク角偏差が著しく不均一となることがある。
【０１５３】
一例として、機関本体の振動により、クランク角度を検出するための前述した電磁ピックアップの取り付け腕（ステー）が共振し、同ピックアップとロータ被検出部との位置関係が変化することなどが挙げられる。
【０１５４】
このような場合、電磁ピックアップによるロータ被検出部の検出間隔（検出時間）が変則的になってしまうことから、同検出時間に基づき前記（８）式を通じて算出されるクランク角偏差（角度公差）Δθにも図１４に示されるような特異点ＳＰ、ＳＰ’が生じることとなる。因みに図１４は、排気量１８００ｃｃ直列４気筒エンジンについて測定した回転速度と同角度公差Δθとの関係についての実測データである。
【０１５５】
そして、このような特異点ＳＰ、ＳＰ’が生じる場合、同クランク角偏差Δθについての前記運転条件（回転速度）別の公差学習を行い、それ以外の回転速度領域ではそれら公差学習値からの直線補間によってその角度公差を求めたとしても当該角度公差特性を正確に角速度変動量に反映させることはできず、ひいては前記メインルーチン（図３）を通じて実行される失火判定についての誤判定をも招きかねなくなる。
【０１５６】
なお、このような特異点ＳＰ、ＳＰ’の生じ方は、機関の種類や形状、更にはそのおかれる環境等によって区々であり、機関のどのような運転条件で該特異点ＳＰ、ＳＰ’生じるかを特定することは困難である。
【０１５７】
また、そうかといって、対象となる内燃機関の全運転領域に亘ってそれら角度公差を全て学習するにはメモリ容量等の制限を受けることとなり、やはり現実的ではない。
【０１５８】
そこで以下に、この発明にかかる失火検出装置の他の実施形態として、クランク角偏差（角度公差）Δθに上記特異点が生じる場合であっても、少ないメモリ容量で、しかも好適に該特異点による影響を回避することのできる装置についてその一例を示す。
【０１５９】
ここでは、先の実施形態の装置による前記公差学習に併せて、その公差学習値と上記特異点を含む実公差との偏差についての図１５及び図１６に示されるような偏差学習制御を実行し、その偏差学習値に基づき前記メインルーチン（図３）で用いられる失火判定値ＲＥＦ１を補正して上記特異点の存在に起因する誤った失火判定が行われることを回避する。
【０１６０】
因みにこの場合、前記公差学習を実行した各々特定の回転速度（運転条件）の合間の補間領域で上記公差学習値と実公差との偏差を求める必要があるため、同図１５及び図１６に示す偏差学習制御ルーチンでは、公差学習を実行した回転速度区間をそれら学習域に対応した所定の回転数毎の（例えば５００ｒｐｍ毎の）ゾーンに区分けし、それら区分けしたゾーンの単位で上記公差学習値（補間値）と実公差との偏差による影響を抑制するようにしている。なおこの偏差学習制御ルーチンは、前記内燃機関１の一点火毎（６気筒の場合には１２０゜ＣＡ毎、４気筒の場合には１８０゜ＣＡ毎）に、前記電子制御装置９を通じて起動、実行される。
【０１６１】
以下、この図１５及び図１６に示す偏差学習制御ルーチンについてその詳細を順次説明する。
すなわちいま、内燃機関１の任意気筒の点火に伴って同偏差学習制御ルーチンが起動されると、電子制御装置９はまず、ステップＳ５００にて、同機関の現在の回転速度（運転条件）に対応したゾーンにおいて前記公差学習が完了しているか否かを判断する。公差学習が完了していなければ、その対応する実公差との比較もできないため、本ルーチンを一旦終了する。
【０１６２】
一方、当該ゾーンにおいて公差学習が完了していれば、電子制御装置９は次のステップＳ５０１にて、偏差学習実行中フラグがセットされているか否かを判断する。この偏差学習実行中フラグとは、通常はセット状態におかれ、次に述べる条件によってはリセットされて、不十分な偏差学習の完了を未然に防止するためのフラグである。
【０１６３】
すなわち、上記区分けした各々のゾーンに対応した偏差学習を行う上で、あるゾーンでの公差学習値（補間値）と実公差との偏差測定が部分的に行われただけでは、最も影響の大きい上記特異点が測定されていない可能性がある。そこでここでは、例えば
・先の図４に例示した公差学習制御ルーチンのステップＳ２１０にかかる「公差学習実行条件のチェック処理（図７、図８）」やステップＳ２１２にかかる「公差学習実行条件の成否判定処理（図１１）」において学習を実行してはいけない状態であることが認識されている場合（ステップＳ５０２）。
或いは、
・回転速度が急激に変動するなどして、特定ゾーンの偏差測定を入念に行うことができない場合（ステップＳ５０３）。
等々、当該ゾーンで偏差学習が完了したとするには不十分な状態では、ステップＳ５０４にて上記偏差学習実行中フラグをリセットして、同ゾーンでの少なくとも今回の偏差学習を完了させないようにしている。
【０１６４】
上記ステップＳ５０１にて偏差学習実行中フラグがセットされている旨判断され、且つこれら学習をキャンセルすべき要因が生じていない旨判断される場合、電子制御装置９は、ステップＳ５０５にて、該当する公差学習値ΔθＬと実公差Δθとの偏差Δ（Δθ）を求める。実公差Δθが電磁ピックアップによるロータ被検出部の検出間隔（検出時間）に基づき前記（８）式を通じて算出されることは上述した通りである。
【０１６５】
こうして偏差Δ（Δθ）を求めた電子制御装置９は、次のステップＳ５０６にて、同偏差Δ（Δθ）についての最も大きな値を求めるべく、当該ゾーンにおいて保持している偏差Δ（Δθ）の値と今回求めた偏差Δ（Δθ）の値との大きい方の値を前記ＲＡＭ９ｃ（図１）内の所定の領域に保持していく。これは、上記特異点等、最も影響の大きい偏差を学習値とするための配慮である。
【０１６６】
こうしてより大きな偏差Δ（Δθ）を保持した、若しくはステップＳ５０４にて偏差学習実行中フラグをリセットした、若しくはステップＳ５０１にて同偏差学習実行中フラグがセットされていない旨判断した電子制御装置９は、ステップＳ５０７にて、回転速度がそれまでのゾーンを抜け、新しいゾーンに移行したか否かを判断する。新しいゾーンに移行していない場合には、当該ゾーンでの上記偏差Δ（Δθ）の算出、並びにそのより大きな値による更新と、後述するステップＳ５１６（図１６）以降の処理のみが繰り返し実行される。
【０１６７】
同ステップＳ５０７において、新しいゾーンに移行している旨判断される場合電子制御装置９は更に、次のステップＳ５０８にて、その移行がもといたゾーンへの逆戻りではなく、移行前のゾーンを全て通った次のゾーン（１段階だけ高速側のゾーン）への移行であるか否かを判断する。同移行がもといたゾーンへの逆戻りであった場合、移行前のゾーンの全域に亘って上記偏差Δ（Δθ）の測定を行ったことにはならないため、以下に説明する学習処理は行われずに、後述するステップＳ５１４（図１６）以降の処理が行われる。
【０１６８】
一方、ステップＳ５０８において、新しいゾーンへの移行が上記次のゾーンへの移行である旨判断される場合、電子制御装置９は、ステップＳ５０９にて上記偏差学習実行中フラグがセットされていることを確認した上で、次のステップＳ５１０〜ステップＳ５１３にかかる偏差学習を実行する。偏差学習実行中フラグがセットされていない場合、すなわち移行前のゾーンにおいて一度、ステップＳ５０２〜ステップＳ５０４を通じて学習の実行が不適当である旨判断されている場合にも移行前のゾーンの全域に亘って上記偏差Δ（Δθ）の測定を行ったことにはならないため、以下に説明する学習処理は行われずに、後述するステップＳ５１４以降の処理が行われる。
【０１６９】
偏差学習の実行に際してはまず、ステップＳ５１０（以下、図１６）にて、該学習対象となるゾーン、すなわち移行前のゾーンでの偏差学習実行条件の成立が初回であるか否かが判断される。
【０１７０】
この結果、同条件の成立が初回である旨判断される場合には、ステップＳ５１２にて、上記移行前のゾーンに関して上記ＲＡＭ９ｃ内の所定領域に保持されている最大の偏差Δ（Δθ）を同ゾーンの偏差学習値として偏差学習値メモリに登録し、次のステップＳ５１３にて、同ゾーンについての偏差学習が完了したことを示す偏差学習完了フラグをセットする。この偏差学習値メモリも、先の図６に例示した公差学習値メモリ９１０と同様、前記バックアップＲＡＭ９ｄ（図１）内の所定領域に予め用意されていて、上述した各ゾーン（回転速度範囲）の別にそれら偏差学習値Δ（Δθ）が登録される構造となっている。
【０１７１】
他方、ステップＳ５１０において、上記移行前のゾーンでの偏差学習実行条件の成立が初回ではなく、２回目以降である旨判断される場合には、ステップＳ５１１にて、上記偏差学習値メモリに登録されている同ゾーンについての偏差学習値Δ（Δθ）を同ゾーンに関して上記ＲＡＭ９ｃ内の所定領域に保持されている最大の偏差Δ（Δθ）によって更新する。なお、この更新に際しては、前述したなまし（徐変）処理を併用するようにしてもよい。
【０１７２】
こうして偏差学習の実行を終えると電子制御装置９は次に、該移行した新しいゾーンについての偏差学習を行うため、ステップＳ５１４にて、上記ＲＡＭ９ｃ内の所定領域に保持されている偏差Δ（Δθ）の値を「０」にクリアするとともに、ステップＳ５１５にて、上記偏差学習実行中フラグを標準（デフォルト）の状態であるセット状態とする。
【０１７３】
その後、電子制御装置９は、ステップＳ５１６にてその対象となっているゾーン（移行前のゾーン）に関する上記偏差学習完了フラグがセットされていることを確認した上で、次のステップＳ５１７〜ステップＳ５１８にかかる失火判定値補正処理を実行する。同ゾーンに関する偏差学習完了フラグがセットされていない場合には、このステップＳ５１７〜ステップＳ５１８にかかる失火判定値補正処理を行わずに、本ルーチンを一旦終了する。
【０１７４】
失火判定値補正処理の実行に際してはまず、ステップＳ５１７にて、当該ゾーンの偏差学習値Δ（Δθ）から前記失火判定値ＲＥＦ１に加えるべきオフセット量ＲＥＦｏｆｓを算出する。このオフセット量ＲＥＦｏｆｓの算出は、
ＲＥＦｏｆｓ＝Ｋｏｆｓ×Δ（Δθ）×回転速度 …（１８）
といったように、偏差学習値Δ（Δθ）の角度（ｒａｄ）情報を角速度（ｒａｄ／ｓｅｃ）の変動量に換算するかたちで行われる。ここで係数Ｋｏｆｓは、偏差学習値Δ（Δθ）をこうした失火判定値ＲＥＦ１と同じ次元の値に換算するための換算係数である。
【０１７５】
こうしてオフセット量ＲＥＦｏｆｓを算出した電子制御装置９は最後に、ステップＳ５１８にて同算出したオフセット量ＲＥＦｏｆｓを前記失火判定値ＲＥＦ１に加えて、本ルーチンを終了する。
【０１７６】
このような偏差学習制御が内燃機関１の一点火毎に行われることにより、上記各ゾーン毎に測定された公差学習値補間値と実公差との偏差Δ（Δθ）の最大値（偏差学習値）に応じた角速度変動量が別途求められるとともに、この求められた角速度変動量がオフセット量ＲＥＦｏｆｓとして、その都度、前記失火判定値ＲＥＦ１に加えられるようになる。
【０１７７】
したがって、たとえクランク角偏差（実公差）Δθに上述した特異点が生じる場合であっても、すなわちメインルーチン（図３）において失火判定値ＲＥＦ１と比較されるクランク角速度変動量Δω(n-α-1)に該特異点に基づく増加が生じる場合であっても、その角速度変動量増加分に応じたオフセット量ＲＥＦｏｆｓが失火判定値ＲＥＦ１に加わることで、同特異点に起因する誤った失火判定が行われることも好適に回避されるようになる。
【０１７８】
しかも、同偏差学習制御ルーチンによれば、上記ゾーンを単位として偏差学習を行うようにしたことで、その必要とされるメモリ容量の増加を最小限に抑えることができるようにもなる。
【０１７９】
なお、同実施形態の装置にあっては、失火判定値ＲＥＦ１に上記オフセット量ＲＥＦｏｆｓを加えて上記特異点に対処することとしたが、同メインルーチンにおいて比較対象となるクランク角速度変動量Δω(n-α-1)から上記オフセット量ＲＥＦｏｆｓを引いてその対処とする構成であっても勿論よい。
【０１８０】
また、上記偏差学習値としても、クランク角偏差（気筒間角度公差）との偏差量に限らず、それに相当する値、すなわち公差学習値に応じて、例えばクランク角速度との偏差量、或いはその変動量、等々を採用することができる。
【０１８１】
また、以上の実施形態では何れも、学習の実行条件を適正に判断するための要素として上記空燃比センサ（酸素センサ、リニア空燃比センサ）の出力や空燃比フィードバック制御にかかる空燃比補正係数を参照するようにした。
【０１８２】
しかし、図８に例示したそれら判断内容からも明らかなように、こうした空燃比センサの出力や空燃比フィードバック制御にかかる空燃比補正係数から直接、当該機関の失火発生の有無を検出する構成とすることもできる。
【０１８３】
すなわち、同図８に例示した正常点火を判定するルーチンのみを同実施形態にかかる装置のクランク角速度変動量を用いた失火検出に代えて用いる構成とすることもできる。
【０１８４】
そしてこのときであれ、同図８におけるチェック項目の選択、或いは組み合わせ等は任意であり、対象となるシステムの規模に応じて自由にそれら項目の選択、或いは組み合わせを行うことができる。
【０１８５】
また更には、それら自由に選択、若しくは組み合わせた失火検出方法を、上記実施形態にかかる装置の失火検出方法以外の方法と組み合わせて、それら方法による失火検出精度の更なる向上を図るようにすることもできる。
【０１８６】
また、図８に例示したチェックルーチンでは、３６０゜ＣＡ対向気筒の連続失火を検出する３つの方法が示されているが、この中のステップＳ３１１の処埋を図１７のステップＳ３１１’の処理として示すように変更してもよい。
【０１８７】
すなわち、図８のチェックルーチンでは空燃比補正係数ｃｆｂとその平均値との和を初期公差と比較して失火検出していたが、他に図１７に示すように、空燃比補正係数と同補正係数の学習値との和を初期公差と比較して失火検出するようにしてもよい。
【０１８８】
また更に、図８のチェックルーチンでは、そのステップＳ３１４の処理において、Ｏ2 センサ振幅周期が予め設定されている正常点火時の振幅周期Ｆｓよりも短いとき失火が発生している旨判断しているが、図１７のチェックルーチンにおけるステップＳ３１４’の処理として示すように、下限の判定値（Ｆｓ）だけでなく、所定範囲を設定して、この範囲外のときには失火が発生している旨判断するようにしてもよい。このように所定範囲を設定することにより、図９に示すような失火だけでなく、図１０に示すような失火も検出することができるようになる。
【０１８９】
また、以上の各実施形態では、回転角速度変動量としてクランク角偏差Δθｎを運転条件の別に学習しているが、これに限られることはなく、これに相当する値として、例えばクランク角偏差Δθｎを求めるために用いるクランク角偏差時間ΔＴｎを学習するようにしてもよい。
【図面の簡単な説明】
【図１】この発明にかかる失火検出装置の一実施形態を示すブロック図。
【図２】同実施形態にかかる電子制御装置の機能的構成を示すブロック図。
【図３】同実施形態の失火判定のメインルーチンを示すフローチャート。
【図４】同実施形態の公差学習制御ルーチンを示すフローチャート。
【図５】気筒間クランク角偏差（公差）の積算値メモリ構造例を示す略図。
【図６】気筒間クランク角偏差（公差）の学習値メモリ構造例を示す略図。
【図７】公差学習実行条件のチェックルーチンを示すフローチャート。
【図８】公差学習実行条件のチェックルーチンを示すフローチャート。
【図９】失火時における酸素センサの出力例を示すタイムチャート。
【図１０】失火時における酸素センサの出力例を示すタイムチャート。
【図１１】公差学習実行条件の成否判定ルーチンを示すフローチャート。
【図１２】負荷−気筒間クランク角偏差（公差）特性を示すグラフ。
【図１３】回転速度−気筒間クランク角偏差（公差）特性を示すグラフ。
【図１４】クランク角偏差（公差）に生じる特異点の様子を示すグラフ。
【図１５】特異点対策である偏差学習制御ルーチンを示すフローチャート。
【図１６】特異点対策である偏差学習制御ルーチンを示すフローチャート。
【図１７】公差学習実行条件の他のチェックルーチンを示すフローチャート。
【符号の説明】
１…内燃機関、２…吸気管、３…吸気管圧力センサ、５…回転角センサ、６…基準位置センサ、７…ディストリビュータ、８…水温センサ、９…電子制御装置、１０…インジェクタ、１１…イグナイタ、１２…警告ランプ、１３…ピストン、１４…排気管、１５…酸素（Ｏ2 ）センサ、９ａ…ＣＰＵ、９ｂ…ＲＯＭ、９ｃ…ＲＡＭ、９ｄ…バックアップＲＡＭ、９ｅ…Ｉ／Ｏポート、９０１…角速度変動量演算部、９０２…失火判定部、９０３…点火数カウンタ、９０４…仮失火カウンタ、９０５…角速度記憶部、９０６…学習制御部、９０７…積算カウンタ、９０８…気筒間クランク角偏差（公差）積算値メモリ、９０９…点火数カウンタ、９１０…気筒間クランク角偏差（公差）学習値メモリ、９１１…ラフロードカウンタ（ＣＲＧ）、９１２…仮失火カウンタ。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a misfire detection device for an internal combustion engine that detects misfire that has occurred in the internal combustion engine, and more particularly to a device that has a deviation in crank angular velocity and that improves the misfire detection accuracy.
[0002]
[Prior art]
In general, in an internal combustion engine, when a misfire occurs in an explosion stroke of a certain cylinder, the crank angular speed at that time, that is, the rotational angular speed of the crankshaft that is the engine output shaft becomes small. For this reason, by monitoring such a change in the crank angular velocity, it is possible to detect whether or not misfire has occurred for each cylinder.
[0003]
However, normally, even if the cylinders are normally ignited due to the influence of the difference in the combustion amount for each cylinder or the tolerance between the cylinders of the sensor for detecting the crank angle, the measured crank Angular velocity changes. In other words, there is a risk of misjudgment that the measured cylinder with a low crank angular velocity has misfired despite being normally ignited.
[0004]
So conventionally,
(A) When the fuel is not affected by the presence or absence of misfire, the crank angular speed deviation between the cylinders is obtained, and the misfire judgment value for the crank angular speed at the time of fuel injection is corrected based on the obtained deviation (for example, Japanese Patent Laid-Open No. Hei 4). -265265 reference).
(B) Always obtain the crank angular speed deviation between the cylinders and assume that the ignition is normal when the obtained deviation is the same under different operating conditions, and correct the crank angular speed based on the obtained deviation. (See, for example, JP-A-4-110632).
Etc. By learning the crank angular velocity deviation (rotational angle deviation) between cylinders other than at the time of misfire and correcting the judgment value or crank angular speed based on the learned deviation at the time of misfire judgment, the accuracy for detecting the misfire is improved. I try to let them.
[0005]
[Problems to be solved by the invention]
In this way, by learning the crank angular speed deviation (rotational angle deviation) between cylinders other than at the time of misfire, the influence due to the difference in the combustion amount that differs for each cylinder or the tolerance between the cylinders of the sensor that detects the crank angle can be confirmed. Will be absorbed.
[0006]
However, such crank angular velocity deviations (rotational angle deviations) between the cylinders are shown in FIG. 13, for example, in a six-cylinder internal combustion engine as an example, the second, fifth cylinder groups, and third cylinders for the first and fourth cylinder groups. As exemplified by the angular tolerance of the sixth cylinder group, it also varies depending on the operating conditions of the engine such as the rotational speed of the internal combustion engine.
[0007]
Therefore, just by learning the crank angular velocity deviation (rotational angle deviation) between the cylinders other than at the time of misfire, the influence and the crank angle due to the difference in the combustion amount differing for each cylinder are detected under all operating conditions of the internal combustion engine. Sensor-to-cylinder tolerance and the like are not completely removed.
[0008]
In addition, in the method exemplified as the conventional (A) above, it is assumed that normal misfire detection is impossible and learning is performed until the fuel is cut, because the deviation learning is not performed. However, since the learning value itself does not take into account the engine operating condition of such an engine or the original crank angular velocity deviation between the cylinders due to combustion, the judgment result itself regarding the presence or absence of misfire remains doubtful.
[0009]
On the other hand, even in the method exemplified as the conventional (b) above, a sensor for detecting the crank angular speed deviation between cylinders that changes depending on the operating conditions of such an engine, that is, the effect of the difference in combustion amount for each cylinder and the crank angle It is not possible to completely eliminate the tolerance between cylinders.
[0010]
The sensor for detecting the crank angle usually includes a rotor that has a projection or the like to be detected and is mounted on the crankshaft of the engine, and an electromagnetic pickup that is disposed close to the rotor. Although it is configured, it is usually considered that the sensor causes the crank angular velocity deviation between cylinders.
・ "Manufacturing tolerance of rotor detected part"
other than,
・ "Variation of air gap between rotor detection part and electromagnetic pickup"
There are also.
[0011]
That is, in this sensor, an AC signal induced in the electromagnetic pickup as it passes through the rotor detection part is taken into an appropriate waveform shaping circuit and converted into a binary signal, and the logic of the converted binary signal is converted into a binary signal. The rotation time or rotation angular velocity for each cylinder is obtained according to the timing at which the level is reversed. At this time, the level of the induced AC signal changes depending on the air gap between the rotor detected portion and the electromagnetic pickup, and the timing at which the logic level of the binary signal is inverted is the level of the AC signal. It depends on. For this reason, if the air gap is different for each cylinder, a deviation also occurs in the timing at which the logic level of the binary signal is inverted. This deviation further changes according to the engine operating conditions (rotational speed) described above.
[0012]
The present invention has been made in view of such circumstances, and has an object of improving the accuracy of detecting misfires generated in an internal combustion engine, and in particular, always with high accuracy regardless of the operating conditions of the internal combustion engine. An object of the present invention is to provide a misfire detection device for an internal combustion engine that can detect misfire occurring in the engine.
[0013]
Another object of the present invention is to provide a misfire detection device for an internal combustion engine that can detect misfire occurring in the internal combustion engine with high accuracy using the output of the air-fuel ratio sensor.
[0014]
[Means for Solving the Problems]
In order to achieve such an object, according to the present invention, the crank angular speed deviation (rotational angular speed fluctuation amount of the engine output shaft) for each cylinder is learned for each engine operating condition.
[0015]
That is, in the first aspect of the invention, the learning means learns the amount of fluctuation in the rotational angular velocity for each cylinder of the engine output shaft according to the operating condition of the engine. Here, the rotational angular velocity fluctuation amount does not necessarily have to be itself, but may be a value corresponding thereto, for example, a rotational angle deviation as described in claim 2. Furthermore, a value corresponding to the rotation angle deviation, or a required rotation time deviation can also be used as a value corresponding to the rotation angular velocity fluctuation amount.
[0016]
  In any case, as in the first aspect of the invention, the amount of variation in the rotational angular velocity for each cylinder (or a value corresponding to the amount of variation in angular velocity) is compared with a predetermined misfire determination value to detect the occurrence of misfire in the engine. If the misfire determination value or the cylinder-specific rotational angular velocity fluctuation amount (or a value corresponding to the angular velocity fluctuation amount) is corrected through the correction means based on the learning value under the corresponding operating condition in each case, Regardless of engine operating conditions, high misfire detection accuracy is maintained.
  By the way, in order to maintain the learned value at a more reliable value in the learning means, it is determined whether or not the engine is normally ignited, and the learning is executed only when the engine is normally ignited. It is desirable.
  By adopting such a configuration, when it is determined that the engine is not normally ignited, the execution of the learning is prohibited, and the reliability of the learning value corresponding to the operating condition of the engine is suitably maintained. It becomes like this.
In addition, according to experiments by the inventors, it has been confirmed that a so-called singular point in which the crank angle deviation becomes extremely nonuniform at a specific rotation speed is generated as a characteristic of the internal combustion engine due to vibrations of the engine body. . The cause is considered to be that the mounting arm (stay) of the electromagnetic pickup for detecting the crank angle resonates due to the vibration of the engine body, and the positional relationship between the pickup and the rotor detected part changes.
And, when such a singular point occurs, even if the tolerance learning according to the driving condition for the crank angle deviation is performed, and the angular tolerance is obtained by linear interpolation from those tolerance learning values in other operating regions, The angle tolerance characteristic including the singular point cannot be accurately reflected in the angular velocity fluctuation amount or its equivalent value, and thus misjudgment regarding misfire judgment performed based on comparison between those values and the misfire judgment value. May be invited.
In such a case, according to the invention of claim 1,
The correction means measures a deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft, and is compared with the misfire determination value or the determination value according to the measured deviation. Deviation correction means for correcting the cylinder-specific rotational angular velocity fluctuation amount is included.
Such a configuration is effective in avoiding the influence of the singular point.
That is, even when the singular point occurs, the deviation from the tolerance learning value caused by the singular point is corrected in the above manner so that the influence of the singular point on the misfire determination is preferably offset. Become.
[0017]
Further, as in the invention described in claim 2, when the rotation angle deviation is used as a value corresponding to the rotation angular velocity fluctuation amount, the correcting means is
・ Cylinder rotation angle deviation of the engine output shaft is obtained by correcting the rotation angle deviation between cylinders based on the corresponding value of each internal cylinder engine and the learning value in the operating condition, and the change of the rotation angle velocity for each cylinder is calculated. Based on the cylinder, the amount of fluctuation in rotational angular velocity for each cylinder to be compared with the misfire determination value is calculated.
It is preferable to constitute as follows.
[0018]
With such a configuration, in the construction of a recent misfire detection device assisted by a microcomputer, the learning structure and the correction (calculation of the rotational angular velocity for each cylinder and the rotational angular velocity fluctuation amount for each cylinder) structure are extremely high in calculation efficiency. It will be possible to.
[0021]
As a configuration for determining whether or not the engine is normally ignited, there is the following configuration. For example, misfire determination using a 360 ° CA (crank angle) difference method, that is, a second-order difference method in which the difference in rotational angular velocity between a cylinder subject to misfire detection and its adjacent cylinder is subtracted from the same difference between cylinders separated by 360 ° CA is possible. If so, the claim3According to the described invention,
-It is determined that the engine is normally ignited based on the obtained amount of fluctuation of the rotational angular velocity for each cylinder of the engine output shaft being equal to or less than the temporary misfire determination value. Such a configuration is effective.
[0022]
Incidentally, according to the 360 ° CA difference method, the variation between the cylinders separated by 360 ° CA of the sensor for detecting the output shaft rotation angle (crank angle) of the engine is preferably canceled out. For this reason, in the case where misfire occurs in a state that does not include the case where cylinders separated by 360 ° CA are misfired, the claims3This can be determined easily and reliably through the configuration of the described invention. When it is determined that the engine is not normally ignited, the execution of the learning is prohibited, and the reliability of the learning value corresponding to the operating condition of the engine is suitably maintained.
[0023]
In particular, when such a 360 ° CA difference method is adopted,4According to the described invention,
The engine output shaft obtained on the basis of the rotation angle signal by the second-order difference obtained by subtracting the rotation angle difference between the cylinder after 360 ° CA and the adjacent cylinder from the rotation angle difference between the cylinder subject to misfire detection and the adjacent cylinder. It is determined whether the cylinder-by-cylinder rotational angular velocity fluctuation amount is equal to or less than a temporary misfire determination value. Such a configuration is desirable for improving the S / N (signal / noise) ratio.
[0024]
That is, when a misfire occurs in a certain cylinder, the angular velocity (crank angular velocity) of the engine output shaft usually returns gradually to a normal angular velocity thereafter. For this reason, by taking the second-order difference between the rotation angle difference of the (future) cylinder after 360 ° CA and its adjacent cylinder in this way, in the event of a misfire, the rotation angular velocity for each cylinder of the engine output shaft A larger value can be obtained as the fluctuation amount.
[0025]
On the other hand, when it is impossible to determine misfire by the 360 ° CA difference method, that is, when cylinders separated by 360 ° CA are misfiring together (hereinafter referred to as 360 ° CA facing cylinder continuous misfire), Claim5According to the described invention,
During execution of the air-fuel ratio feedback control, the sum of the air-fuel ratio correction coefficient and the average value of the correction coefficient, or the learning ratio of the air-fuel ratio correction coefficient, if the learning control of the air-fuel ratio correction coefficient is being executed, Based on the fact that the sum with the learning value is not on the lean side of the predetermined value, it is determined that the engine is normally ignited. Such a configuration is effective. Here, the predetermined value to be compared with the sum of the air-fuel ratio correction coefficient and the average value thereof, or the sum of the air-fuel ratio correction coefficient and the learning value of the correction coefficient is the characteristic of the air-fuel ratio sensor or the individual internal combustion engine. It may be an initial tolerance due to machine differences.
[0026]
With such a configuration, even when misfire is not determined by the 360 ° CA differential method due to continuous misfire of the 360 ° CA facing cylinder, it is possible to accurately determine whether or not the engine has misfired. become. In this case as well, when it is determined that the engine is not normally ignited, the execution of the learning is prohibited, and the reliability of the learning value corresponding to the operating condition of the engine is suitably maintained.
[0027]
By the way, claims5In the above-described configuration of the described invention, it is a major premise that the air-fuel ratio feedback control is being executed. However, it is desirable to determine whether or not the engine is normally ignited even when such feedback control is not performed, for example, during high load fuel increase.
[0028]
And in such cases, the claims6According to the described invention,
When the air-fuel ratio sensor is active, it is determined that the engine is normally ignited based on the fact that its output is not on the lean side for a predetermined period or longer. In particular, when an oxygen sensor is used as an air-fuel ratio sensor,7As described in the invention, when the oxygen sensor is activated, the engine is normally ignited based on the fact that its output cycle is within a predetermined range (including the meaning that it is not lower than a predetermined value). Judging. In the case where a linear air-fuel ratio sensor is used as the air-fuel ratio sensor,8According to the described invention,
It is determined that the engine is normally ignited based on the fact that the output of the linear air-fuel ratio sensor (including the average value of the output and the smoothed value) is not on the lean side for a predetermined value or more. Such a configuration is effective.
[0029]
In other words, when an air-fuel ratio sensor such as an oxygen sensor or a linear air-fuel ratio sensor is activated, if a misfire occurs in the engine,
・ The output is fixed to the lean side.
In particular, when the air-fuel ratio sensor is an oxygen sensor,
・ The output cycle becomes extremely short.
And when the air-fuel ratio sensor is a linear air-fuel ratio sensor,
-The output will have an overall lean offset.
It has been confirmed by the inventors that such a state is indicated.
[0030]
Therefore, the above claims6-8By monitoring each of these states according to the configuration of the described invention, it is possible to determine whether or not the engine has misfired even when the air-fuel ratio feedback control is not performed. Also in this case, when it is determined that the engine is not normally ignited, the execution of the learning is prohibited, and the reliability of the learning value corresponding to the operating condition of the engine is suitably maintained. become.
[0031]
On the other hand, in order to maintain the reliability of such learning values, the learning means is9As described in the invention, when the difference between the value to be updated and the learned value so far is larger than a predetermined value, only when the value to be updated becomes the same value continuously several times, Allows the learning value to be updated with that value. It is effective to adopt a learning algorithm such as According to such a learning algorithm, the erroneous learning is not performed by the value obtained by chance, and the reliability of the learning value is further improved.
[0032]
On the other hand, the learning means is further claimed.10According to the described invention,
-When the operating state of the engine is in a state that causes a large rotational fluctuation of the engine output shaft, the execution of learning is stopped. Or a claim11According to the described invention,
・ When the relevant engine is in an operating range where misfire determination is impossible, the execution of learning is stopped. Such a configuration is also effective. By stopping the execution of learning in these driving states or driving ranges, the reliability of the learning value is naturally maintained high.
[0033]
Note that the “operating state that causes a large rotational fluctuation of the engine output shaft” includes, for example, a transient state such as sudden acceleration or sudden deceleration, a shift change state, a fuel cut or return, a start-up or electric load application, an idle state , Purge control state, EGR (exhaust gas recirculation control) execution, variable intake air execution, and so on.
[0034]
  In addition, examples of the “operating range where misfire determination is impossible” include a light load operating range and a high rotation range..
[0040]
  AlsoClaims 1-11In the case of2According to the described invention,
The deviation correction means learns the maximum deviation between the learning value from the learning means and the cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft according to the operating conditions of the engine when measuring the deviation.
Such a configuration is effective in minimizing the memory capacity required for the correction. Here again, the rotational angular velocity fluctuation amount does not necessarily have to be itself.With lightIt is the same.
[0041]
  Further, at this time, claim 1.3As in the described invention, the deviation correction means may update the maximum deviation as the learning value of the driving condition when it is determined that the driving conditions for one learning value have been operated.
[0042]
  Claim 14As described above, misfire detection may not be performed until deviation learning by the deviation correction unit is completed, and by adopting such a configuration, misjudgment of misfire due to crank angular speed deviation is suppressed. You will be able to
[0048]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows an embodiment of a misfire detection apparatus for an internal combustion engine according to the present invention.
[0049]
In this embodiment, a 6-cylinder internal combustion engine is used as an internal combustion engine, and an apparatus for detecting misfire occurring in the 6-cylinder internal combustion engine is shown.
That is, in the apparatus of this embodiment shown in FIG. 1, the internal combustion engine 1 is an internal combustion engine composed of the six cylinders.
[0050]
The internal combustion engine 1 is provided with an intake pipe 2, and intake air introduced from an air cleaner (not shown) is taken into the engine 1 through the intake pipe 2. The intake pipe 2 is provided with an intake pipe pressure sensor 3 through which the pressure PM in the intake pipe 2 is sequentially detected. The detected pressure PM in the intake pipe 2 is taken into the electronic control unit 9 described later as one parameter indicating the operating state of the internal combustion engine 1.
[0051]
On the other hand, the crankshaft (not shown) of the internal combustion engine 1 is provided with a rotation angle sensor 5 that outputs a rotation signal NE for each predetermined crank angle of the crankshaft. The rotational speed of the engine 1 is calculated based on the rotation signal NE output from the rotation angle sensor 5. This rotation signal NE is also taken into the electronic control unit 9 described later as one parameter indicating the operating state of the internal combustion engine 1.
[0052]
Further, the internal combustion engine 1 is provided with a distributor 7 for controlling the ignition timing and ignition sequence for each cylinder, and the distributor 7 further outputs a reference position signal CYL for discriminating each cylinder. A reference position sensor 6 is built in. In the reference position sensor 6, the reference position signal CYL is also sent to the electronic control unit 9 every time the piston 13 of the first cylinder of the engine 1 reaches the top, that is, the compression top dead center (# 1 TDC). Output. The distributor 7 usually obtains rotational power from the internal combustion engine 1 and rotates at a rotational speed of (1/2).
[0053]
The cooling water passage of the internal combustion engine 1 is provided with a water temperature sensor 8 for detecting the temperature of the cooling water circulating in the water passage, and the exhaust pipe 14 has a rich air-fuel ratio (based on the oxygen concentration of the combustion gas). An oxygen (O2) sensor 15 for detecting R) / lean (L) is provided. The temperature of the cooling water detected through the water temperature sensor 8 and the signal indicating the rich (R) / lean (L) of the air-fuel ratio detected through the oxygen sensor are also used as parameters indicating the operating state of the engine 1 as an electronic control unit 9. Is taken in.
[0054]
The electronic controller 9 in which the detection signals from the intake pipe pressure sensor 3, the rotation angle sensor 5, and the reference position sensor 6 described above, including the water temperature sensor 8 and the oxygen sensor 15, are also shown in FIG. 1. In addition to a CPU (Central Processing Unit) 9a, a ROM 9b that is a read-only memory for storing control programs and control constants necessary for arithmetic processing, a so-called data memory for temporarily storing arithmetic data, etc. As a RAM 9c, a backup RAM 9d that backs up its storage contents through a battery (not shown), and an I / 0 port 9e for inputting / outputting signals to / from an external device.
[0055]
In this electronic control unit 9, it is largely
(A) Based on various detection signals from the above sensors, the optimal control amounts of the fuel system and the ignition system of the internal combustion engine 1 are calculated to accurately determine the injector 10 that is the fuel injection means, the igniter 11 that is the ignition means, and the like. A control signal for controlling is output.
(B) Whether or not misfire has occurred in each cylinder of the internal combustion engine 1 is detected based on various detection signals from the sensor.
The process is executed. In the electronic control unit 9, when the injector 10 is driven, the well-known air-fuel ratio feedback control based on the output of the oxygen sensor 15 is also executed. Further, in the detection of whether or not a misfire has occurred (b) above, when it is determined that a misfire has occurred, for example, the warning lamp 12 is controlled to be turned on to notify the driver or the like of the occurrence of misfire, Appropriate fail-safe processing is executed.
[0056]
FIG. 2 functionally shows the configuration of the electronic control device 9 mainly as a misfire detection device. Next, referring to FIG. 2 together, the configuration of the misfire detection device according to the embodiment is shown. Further, the function will be described in detail.
[0057]
In the electronic control unit 9 shown in FIG. 2, the angular velocity fluctuation amount calculation unit 901 has a crankshaft angular velocity (crank angular velocity) ωn for each cylinder based on the intake pipe pressure PM, the rotation signal NE, and the reference position signal CYL. (N = 1 to 6) is obtained, and the angular velocity fluctuation amount Δω (n−α−1) between the cylinders is calculated from the angular velocity ωn.
[0058]
Here, when a six-cylinder internal combustion engine is targeted as in the apparatus according to the embodiment, the time T120i (i) required for the crankshaft to rotate by 120 ° CA in calculating the crank angular velocity ωn is calculated. Indicates the number of times of processing by the electronic control unit 9),
ωn = (KDSOMG−ΔθnL) / T120i (1)
In this manner, the crank angular velocity ωn is calculated.
[0059]
In this equation (1), the coefficient KDSOMG is a conversion coefficient for obtaining the rotational angular velocity (rad: radians) of the crankshaft, and the value ΔθnL is an inter-cylinder crank angle deviation (tolerance) described later in the backup RAM 9d. This is a learned value for the crank angle deviation between the cylinders stored in the learned value memory 910.
[0060]
As described above, the angular velocity fluctuation amount calculation unit 901 corrects the inter-cylinder crank angle deviation with the learned value ΔθnL to obtain the crank angular velocity ωn.
Incidentally, when targeting a four-cylinder internal combustion engine, the time T180i required for the crankshaft to rotate 180 ° CA is used to calculate the crank angular velocity ωn.
[0061]
Further, in the calculation of the angular velocity fluctuation amount Δω (n−α−1) in the angular velocity fluctuation amount calculation unit 901, based on the present and past values for the obtained crank angular velocity ωn,
Δω (n-α-1) = (ω (n-α-2) −ω (n-α-1)) − (ω (n−1) −ωn) (2)
The second-order difference calculation is executed.
[0062]
In the equation (2), the value ωn is the crank angular velocity obtained this time, and the value ω (n−1) is the crank angular velocity obtained last time. The difference between these values (ω (n−1) −ωn) is the angular velocity fluctuation amount between the cylinders in which the explosion stroke continues.
[0063]
Further, in the equation (2), the value α can take a value of “0 to 5” in the case of the device of the same embodiment intended for a 6-cylinder internal combustion engine, and usually the same angular velocity fluctuation amount Δω. As (n-α-1), a value that tends to cause a change in angular velocity due to misfire is used. In the apparatus of the embodiment, “1” is adopted as the value α. Incidentally, when targeting a four-cylinder internal combustion engine, a value of “0 to 3” is used as this value α.
[0064]
The past values before the value ω (n−1) are sequentially updated and registered in the angular velocity storage unit 905 including the RAM 9c or the backup RAM 9d, for example. In the case of the apparatus of the same embodiment for a six-cylinder internal combustion engine, it is sufficient that the past values have five values ω (n−1) to ω (n−5) at the maximum.
[0065]
Also, in the electronic control unit 9 shown in FIG. 2, the misfire determination unit 902 has a predetermined misfire determination value corresponding to the calculated angular velocity fluctuation amount Δω (n−α−1) between the cylinders and the angular velocity fluctuation amount. This is a part for comparing with REF1 and determining whether or not misfire has occurred in the internal combustion engine 1. Here, the angular velocity fluctuation amount Δω (n−α−1) between the cylinders is compared with the misfire determination value REF1 for each cylinder, and when the angular velocity fluctuation amount Δω (n−α−1) exceeds the misfire determination value REF1. The counter CMIS (n−α−1) corresponding to the cylinder of the temporary misfire counter 904 in the RAM 9c is incremented.
[0066]
The increment of the counter CMIS (n-α-1) corresponding to each cylinder is continuously executed until the ignition number counted through the ignition number counter 903 reaches a predetermined ignition number such as “100” or “500”. Is done. For example, when the count value of the counter CMIS (n-α-1) corresponding to a certain cylinder among the ignition numbers “100” is equal to or greater than “30”, a catalytic converter (not shown) due to misfire is generated. Since there is a concern about damage or the like, the electronic control device 9 warns the driver through lighting control of the warning lamp 12 or the like.
[0067]
On the other hand, in the electronic control unit 9, the learning control unit 906 is a part that learns and controls the crank angle deviation (tolerance) between the cylinders based on the intake pipe pressure PM, the rotation signal NE, and the reference position signal CYL that are taken in. is there.
[0068]
Here, among the above six cylinders, the crank angle deviation of the second to sixth cylinders (# 2 to # 6) with respect to the first cylinder (# 1) is learned.
(1) Crank angle deviation Δθn (n) of the second to sixth cylinders (# 2 to # 6) with respect to the first cylinder (# 1) based on the time T120i required for the crankshaft to perform 120 ° CA rotation = 2 to 6) is integrated by a predetermined number for each cylinder and for each operating condition of the engine 1.
(2) On the condition that the engine 1 is normally ignited, the integrated values of the crank angle deviation Δθn for each cylinder and for each operating condition are averaged, and the average value is further smoothed (gradual change process). This is used as a learning value ΔθnL for the crank angle deviation.
The process is executed.
[0069]
The cumulative counter 907 is used for counting the cumulative number in the process (1), and the crank angle deviation between the cylinders in the RAM 9c is registered for registering the cumulative value for each cylinder and operating condition of the crank angle deviation Δθn. A (tolerance) integrated value memory 908 is used. The learning value ΔθnL is also obtained for each cylinder of the engine 1 and for each operating condition, and the obtained learning value ΔθnL is an inter-cylinder crank angle deviation (tolerance) learning value memory in the backup RAM 9d. Each update registration is performed for 910.
[0070]
In addition, the rough load (CRG) counter 911 and the temporary misfire counter 912 indicate that the check result on whether or not the engine 1 is normally ignited in the process (2) of the learning control unit 906 is a predetermined ignition number (for example, “ The ignition number counter 909 is a counter for repeatedly counting the number of ignitions.
[0071]
FIG. 3 shows a main routine for misfire determination executed through the angular velocity fluctuation amount calculation unit 901 and misfire determination unit 902 of the electronic control unit 9, and FIG. 4 shows the learning control unit 906 of the electronic control unit 9. Each of the tolerance learning control routines to be executed through is shown. Hereinafter, the misfire determination operation of the apparatus according to the embodiment will be described more specifically with reference to FIGS. 3 and 4 together.
[0072]
First, the main routine shown in FIG. 3 will be described.
This main routine is started as an angle interruption process every time the crank angle of the internal combustion engine 1 recognized based on the rotation signal NE reaches 60 ° CA.
[0073]
That is, when the crankshaft is rotated by 60 ° CA and such an interrupt condition is satisfied, the electronic control unit 9 first determines in step S100 from the deviation between the previous interrupt time of this routine and the current interrupt time, A time T60i required for the shaft to rotate 60 ° CA is calculated.
[0074]
The electronic control unit 9 which has thus calculated the time T60i then determines in step S101 whether or not the current interrupt timing is after top dead center (ATDC) 60 ° CA based on the reference signal CYL. If it is determined that the interrupt timing is not 60 ° CA after top dead center, the electronic control unit 9 sets the obtained time T60i to T60i-1 in step S110, and then ends this routine once. . As described above, the subscript i of these times indicates the number of times of processing by the electronic control unit 9.
[0075]
On the other hand, when it is determined that the interrupt timing is 60 ° CA after top dead center, the electronic control unit 9 executes misfire determination processing from step S102.
That is, in this case, after identifying the cylinder number n of the current cylinder based on the reference signal CYL in step S102, the electronic control unit 9 data for the past two times for the obtained time T60i in step S103. Is accumulated to calculate a time T120i required for the crankshaft to rotate 120 ° CA.
[0076]
Then, the electronic control unit 9 that has calculated the time T120i in this way corresponds to the learning value corresponding to the crank angle deviation (tolerance) learning value memory 910, that is, the operation condition (rotational speed NE / engine load PM) of the cylinder. On the condition that the learning value ΔθnL (NE, PM) exists (step S104), in the next step S105, the tolerance is corrected based on the previous equation (1), that is, the cylinder corrected based on the learning value ΔθnL. Is calculated. Here, when it is determined that there is no learning value corresponding to the operating condition of the cylinder, for example,
ωn = KDSOMG / T120i (1) ′
In this manner, only the crank angular velocity ωn is calculated, and this is updated and registered in the angular velocity storage unit 905 (step S109).
[0077]
When the crank angle speed ωn corrected for the tolerance is obtained, the electronic control unit 9 further calculates the amount of variation in crank angular speed Δω (n−α-1) between the cylinders based on the previous equation (2) in the next step S106. To do. In step S107, the calculated crank angular speed fluctuation amount Δω (n−α−1) is compared with the misfire determination value REF1, and the crank angular speed fluctuation amount Δω (n−α-1) is compared with the misfire determination value. If it is determined that REF1 is exceeded, the CMIS counter corresponding to the cylinder number “n−α−1” of the temporary misfire counter 904 is incremented in step S108.
[0078]
On the other hand, if the crank angular speed fluctuation amount Δω (n−α−1) does not exceed the misfire determination value REF1, the process proceeds to the process of step S109, which is the final step of this routine, without incrementing the temporary misfire counter 904. To do. In this step S109, ω (n-5) → discard, ω (n-4) → ω (n-5), ω (n-3) with respect to the crank angular velocity data already stored in the angular velocity storage unit 905. ) → ω (n-4), ω (n-2) → ω (n-3), ω (n-1) → ω (n-2), ωn → ω (n-1) Is given.
[0079]
The increment of the temporary misfire counter 904 is continuously executed until a predetermined ignition number counted through the ignition number counter 903 is reached, and the warning lamp 12 according to the misfire count value by the temporary misfire counter 904 each time. As described above, the lighting control and the like are performed.
[0080]
Next, the tolerance learning control routine shown in FIG. 4 will be described.
This tolerance learning control routine is also started as an angle interruption process every time the crank angle of the internal combustion engine 1 recognized based on the rotation signal NE reaches 60 ° CA, as in the main routine.
[0081]
In this tolerance learning control routine, every time such an interrupt condition is satisfied when the crankshaft rotates 60 ° CA,
(1) The time T60i required for the crank angle to rotate by 60 ° CA is calculated from the deviation between the previous interrupt time and the current interrupt time of this routine (tolerance learning control routine) (step S200).
(2) It is determined based on the reference signal CYL whether or not the current interrupt timing is 60 ° CA after top dead center (ATDC) (step S201).
(3) If it is determined that the interrupt timing is not 60 ° CA after top dead center, the routine is terminated once after the obtained time T60i is set to T60i−1 (step S218).
(4) If it is determined that the interrupt timing is 60 ° CA after top dead center, the cylinder number n of the current cylinder is identified based on the reference signal CYL (step S202), and then the obtained time Data for the past two times for T60i are accumulated to calculate a time T120i required for the crankshaft to rotate by 120 ° CA (step S203).
Such processing is executed through the electronic control device 9 (learning control unit 906).
[0082]
In step S204, the electronic control unit 9 that has thus calculated the time T120i next determines whether or not the identified cylinder number n is the first cylinder (# 1). If it is determined that the identified cylinder number n is not the first cylinder, the electronic control unit 9 directly proceeds to the tolerance learning execution condition check process in step S210.
[0083]
On the other hand, if it is determined that the identified cylinder number n is the first cylinder, the crank angle of the second to sixth cylinders (# 2 to # 6) with respect to the first cylinder is determined in the next step S205. A deviation (cylinder tolerance) time ΔTn is calculated.
[0084]
For example, the crank angle deviation time ΔT # 2 of the second cylinder relative to the first cylinder is

Is calculated as Here, the term “(T120i−T120i-6) / 6” is a transient correction term. By adding such a transient correction term to the calculation of the crank angle deviation time between the cylinders, for example, when sudden acceleration or sudden deceleration occurs. The influence of a transient increase or decrease in rotational fluctuation due to the operating conditions of the internal combustion engine, for example, is suitably eliminated.
[0085]
Similarly, the crank angle deviation times ΔT # 3 to ΔT # 6 of the third to sixth cylinders with respect to the first cylinder are respectively
ΔT # 3 = {(T120i + 2 × T120i-6) / 3} -T120i-4 (4)
ΔT # 4 = {(T120i + T120i-6) / 2} -T120i-3 (5)
ΔT # 5 = {(2 × T120i + T120i−6) / 3} −T120i−2 (6)
ΔT # 6 = {(5 × T120i + T120i−6) / 6} −T120i−1 (7)
Is calculated as
[0086]
The crankshaft 120 ° CA rotation times T120i-5 to T120i-1 of the second to sixth cylinders, including the previous crankshaft 120 ° CA rotation time T120i-6 of the first cylinder, are calculated through step S203. The value updated through the subsequent step S217 is used.
[0087]
In step S206, the electronic control unit 9 that has thus calculated the crank angle deviation time ΔTn of the second to sixth cylinders with respect to the first cylinder next converts the crank angle deviation time ΔTn to the crank angle deviation Δθn, based on the following equation. That is, it is converted into a rotation angle deviation. However, in the following equation (8), n is for five cylinders # 2 to # 6.
Δθn = ΔTn × (120 ° CA / T120i) (8)
When the crank angle deviation Δθn of the second to sixth cylinders with respect to the first cylinder is obtained, in the electronic control unit 9, the internal combustion engine 1 is currently in a specific operating condition, for example, sudden acceleration or Crankshaft such as sudden deceleration, transition change state, fuel cut or return, start-up or electric load application, idle state, purge control state, EGR (exhaust gas recirculation control) execution, variable intake air execution, etc. It is determined on the basis of each operation information whether or not there is a specific operation state that causes a large rotation fluctuation, or an operation region in which a so-called misfire determination cannot be made, such as a light load operation region or a high rotation region. Then, on the condition that the engine 1 is not under such specific operating conditions, in step S208, the obtained (converted) crank angle deviation Δθn is integrated for each cylinder and for each operating condition, and the subsequent step In S209, the integration counter 907 is incremented.
[0088]
That is, when the internal combustion engine 1 is in an operating condition such as a transient state such as sudden acceleration or sudden deceleration, a shift change state, a fuel cut or return, a start-up or an electric load application, etc. The angular deviation Δθn is also likely not to be a value obtained in the normal combustion state of the engine 1. Therefore, under such operating conditions of the internal combustion engine 1, the integration process for the obtained crank angle deviation Δθn is not performed. As will be described later, in the apparatus according to the embodiment, only the normal crank angle deviation Δθn that is subjected to the integration process is subjected to a tolerance learning process that is performed later.
[0089]
Further, the integration processing of the crank angle deviation Δθn in each cylinder and operation condition in step S208 is performed on the inter-cylinder crank angle deviation (tolerance) integration value memory 908 as described above. The memory structure of the integrated value memory 908 is illustrated in FIG.
[0090]
As shown in FIG. 5, the inter-cylinder crank angle deviation (tolerance) integrated value memory 908 is a rotational speed that is an operating condition of the engine 1 separately from the second to sixth cylinders (# 2 to # 6). The crank angle deviation Δθn is accumulated and registered separately from (NE) and load (intake pipe pressure PM). That is, based on the repeated execution of this learning control routine, the normal crank angle deviation Δθn is accumulated and registered in each of the tables shown in FIG. 5 in the form of “ΣΔθn (NE, PM)”. become. Then, the integration counter 907 indicates the integration number of the crank angle deviation ΣΔθn (NE, PM) registered in the inter-cylinder crank angle deviation (tolerance) integration value memory 908 as the count value.
[0091]
When the crank angle deviation Δθn is integrated in this way, the electronic control unit 9 next checks whether or not the tolerance learning should be performed in step S210. The execution condition check process will be described in detail later with reference to FIGS.
[0092]
In the next step S211, the electronic control unit 9 that has finished checking the tolerance learning execution condition determines whether a predetermined ignition number such as “100” ignition has elapsed based on the count value of the ignition number counter 909. Judge whether or not. As a result, when it is determined that the predetermined ignition number has not been reached, the routine proceeds to step S216, where the ignition number counter 909 is incremented, and in step S217, the crankshaft 120 ° CA rotation of each cylinder is performed. The value of time T120i
T120i-6 = T120i-5 (9)
T120i-5 = T120i-4 (10)
T120i-4 = T120i-3 (11)
T120i-3 = T120i-2 (12)
T120i-2 = T120i-1 (13)
T120i-1 = T120i (14)
After updating in such a way, this routine is temporarily exited.
[0093]
On the other hand, when it is determined that the predetermined number of ignitions has elapsed, in step S212, whether or not the execution condition is successful is determined based on the check result for the tolerance learning execution condition. The success / failure determination process for the tolerance learning execution condition will be described in detail later with reference to FIG.
[0094]
Next, in step S213, the electronic control unit 9 determines whether the success / failure determination of the tolerance learning execution condition indicates “permitted” or “impossible” of the tolerance learning execution. If the success / failure determination indicates “tolerance learning cannot be performed”, the processing of step S216 and step S217 is executed to temporarily exit this routine and indicate “tolerance learning can be performed”. If so, tolerance learning is executed in step S214.
[0095]
This tolerance learning is performed on the inter-cylinder crank angle deviation (tolerance) learning value memory 910 in the backup RAM 9d. The memory structure of the learning value memory 910 is illustrated in FIG.
[0096]
As shown in FIG. 6, the learning value memory 910 is also provided separately from the second to sixth cylinders (# 2 to # 6), similarly to the inter-cylinder crank angle deviation (tolerance) integrated value memory 908 (FIG. 5). In addition, the learning value ΔθnL for the crank angle deviation is updated and registered separately for the rotational speed (NE) and the load (intake pipe pressure PM) which are the operating conditions of the engine 1.
[0097]
Here, the crank angle deviation integrated value ΣΔθn (NE, PM) for each cylinder and operation condition registered in the inter-cylinder crank angle deviation (tolerance) integrated value memory 908 in the above-described integration process (step S208) is read. The average value Δθn (NE, PM) _AV
Δθn (NE, PM) _AV = ΣΔθn (NE, PM) / (integrated counter count value) (15)
And the learned crank angle deviation average value Δθn (NE, PM) _AV, the cylinder in the learned value memory 910, and the learned value ΔθnL (NE, PM) for the same crank angle deviation corresponding to the operating condition. ) And annealing (gradual change)

To obtain a new learning value ΔθnL (NE, PM). Then, the newly obtained learning value ΔθnL (NE, PM) is updated and registered in the corresponding column of the learning value memory 910.
[0098]
In the above equation (16), the value “8” is a smoothing (gradual change) coefficient, and it goes without saying that any value other than the value “8” can be adopted according to the processing system. Nor.
[0099]
In addition, the learning value ΔθnL (NE, PM) is updated in the learning value memory 910 because there is a crank angle deviation integrated value ΣΔθn (NE, PM) corresponding to the integrated value memory 908 as well. Limited to That is, if the corresponding crank angle deviation integrated value ΣΔθn (NE, PM) does not exist, the average value Δθn (NE, PM) _AV cannot be obtained. Weird) The operation itself is impossible to execute.
[0100]
In the tolerance learning control routine, the electronic control unit 9 that has executed the tolerance learning in this way resets the integrated value memory 908, the integrated counter 907, and the ignition number counter 909 in the next step S215. Then, in order to prepare for the next learning, after executing the processing of step S216 and step S217 described above, this routine is temporarily exited.
[0101]
By performing learning processing according to the operating conditions of the engine 1 through the electronic control unit 9 (learning control unit 906), tolerance is based on the learning value ΔθnL (NE, PM) in the main routine (FIG. 3). The reliability of the crank angular velocity ωn calculated as the corrected value is naturally enhanced. As a result, even in the subsequent misfire determination, the determination accuracy is naturally high.
[0102]
Next, with reference to FIG. 7 and FIG. 8, the tolerance learning execution condition check process executed as the process of step S210 in the tolerance learning control routine will be described.
[0103]
In the tolerance learning execution condition check routine shown in FIGS. 7 and 8, the electronic control unit 9 (learning control unit 906)
(1) The time T60i required for the crank angle to rotate by 60 ° CA is calculated from the deviation between the previous interrupt time and the current interrupt time (step S300).
(2) It is determined based on the reference signal CYL whether the current interrupt timing is 60 ° CA after top dead center (ATDC) (step S301).
(3) If it is determined that the interrupt timing is not 60 ° CA after top dead center, the routine is terminated once after the obtained time T60i is set to T60i−1 (step S319).
(4) If it is determined that the interrupt timing is 60 ° CA after top dead center, the cylinder number n of the current cylinder is identified based on the reference signal CYL (step S302), and the obtained time Data for the past two times for T60i are accumulated to calculate a time T120i required for the crankshaft to rotate 120 ° CA (step S303).
After the above processing is executed, in the next step S304, the crank angular speed ωn for each cylinder is calculated based on the previous equation (1) ′. Further, in the next step S305, based on the calculated crank angular velocity ωn, the 360 ° CA difference method, that is, the same difference between the cylinders subject to misfire detection and the rotation angular velocity difference between the adjacent cylinders 360 ° CA apart. 2nd floor difference subtracted from
Δωn = (ωn−1−ωn) − (ωn−4−ωn−3) (17)
Is used to calculate the crank angular velocity fluctuation amount Δωn.
[0104]
In step S306, the electronic control unit 9 that has thus calculated the crank angular speed fluctuation amount Δωn compares the crank angular speed fluctuation amount Δωn with the misfire determination value REF2 preset for the fluctuation amount Δωn. If it is determined that the crank angular speed fluctuation amount Δωn exceeds the misfire determination value REF2, the CMF counter of the temporary misfire counter 912 is incremented in step S307, and the process of step S308 is performed. Transition.
[0105]
On the other hand, if it is determined in step S306 that the crank angular speed fluctuation amount Δωn is equal to or less than the misfire determination value REF2, the process proceeds to step S308 as it is.
[0106]
In step S308, the crank angular speed fluctuation amount Δωn and the rough road (rough road running) determination value REF3 (<REF2) set in advance for the fluctuation amount Δωn are further compared.
[0107]
In rough roads, in general, transient rotational fluctuations are likely to occur, and therefore tolerance learning should not be performed even if such conditions continue.
Therefore, if it is determined in step S308 that the crank angular speed fluctuation amount Δωn exceeds the rough road determination value REF3 and is equal to or less than the misfire determination value REF2, the electronic control device 9 is currently traveling on rough road. In step S309, the rough load (CRG) counter 911 is incremented.
[0108]
On the other hand, if it is determined in step S308 that the crank angular speed fluctuation amount Δωn is equal to or less than the rough road determination value REF3, the process directly proceeds to step S310.
[0109]
In this tolerance learning execution condition check routine, the processing after step S310 (FIG. 8) is misfire determination processing for detecting 360 ° CA on-cylinder continuous misfire that cannot be determined by the 360 ° CA difference method. It is. Next, details of these processes will be described.
[0110]
After completing the rough road determination, the electronic control unit 9 first determines in step S310 whether air-fuel ratio feedback (F / B) control based on the output of the oxygen (O2) sensor 15 is being executed. To do.
[0111]
Incidentally, when such a feedback control is being performed and misfire has occurred in the engine 1, the sum of the air-fuel ratio feedback correction coefficient cfb and the average value cfbAV of the correction coefficient is the oxygen sensor. It has been confirmed by the inventors that there is a shift to a side larger than the initial tolerance due to the characteristics of 15 or the individual machine differences of the internal combustion engine (air-fuel ratio lean (L) side).
[0112]
Accordingly, when it is determined in step S310 that the air-fuel ratio feedback control is being performed, the electronic control unit 9 determines in step S311 the sum of the air-fuel ratio feedback correction coefficient cfb and the average value cfbAV of the correction coefficient and the initial value. When the sum is equal to or greater than the initial tolerance, it is determined that misfire has occurred, and the COF counter of the temporary misfire counter 912 is incremented in step S312.
[0113]
By the way, the determination process in step S311 is based on the premise that the feedback control of the air-fuel ratio is being executed. However, even if such feedback control is not performed, for example, during high load fuel increase, It is desirable to determine whether or not the engine 1 is normally ignited.
[0114]
Incidentally, in the oxygen (O2) sensor 15, when misfire occurs in the internal combustion engine 1 when activated,
(A) The output cycle becomes extremely short.
Or
(B) The output sticks to the lean (L) side.
It has been confirmed by the inventors that one of these states is indicated. The measurement results for the states (A) and (B) are shown in FIGS. 9 and 10, respectively.
[0115]
For example, when the output cycle of the oxygen sensor 15 is shortened due to the misfire of the internal combustion engine 1, as shown as “FO2 sensor output” in FIG. 9, the amplitude cycle at the time of normal ignition (approximately 0.5 to 2 Hz). ) With a short cycle (about the ignition cycle shown as “csgt” in FIG. 9).
[0116]
FIG. 9 also shows the transition of the air-fuel ratio feedback correction coefficient cfb in a state where the air-fuel ratio feedback control is being executed, and this air-fuel ratio feedback correction coefficient cfb when a misfire occurs in the engine 1. It is also clear from FIG. 9 that the sum of the correction coefficient and the average value cfbAV of the correction coefficient takes a large value.
[0117]
On the other hand, when the internal combustion engine 1 is increased in high-load fuel (WOT), the output of the oxygen sensor 15 spreads to the lean (L) side as the engine 1 misfires. In this case, the output is fixed to the lean (L) side for a time sufficiently longer than the amplitude cycle at the time of normal ignition, as also indicated as “FO2 sensor output” in FIG. Become so.
[0118]
Thus, if the oxygen sensor 15 is in an active state, it is possible to determine whether or not misfire has occurred in the internal combustion engine 1 by monitoring its output (R / L).
[0119]
In the tolerance learning execution condition check routine shown in FIG. 8, the processing after step S313 is processing for determining whether or not misfire has occurred in the internal combustion engine 1 from the output of the oxygen sensor 15 based on such a principle.
[0120]
That is, it is determined in step S310 that the air-fuel ratio feedback control is not being performed, or even if the air-fuel ratio feedback control is being performed, the sum of the air-fuel ratio feedback correction coefficient cfb and the average value cfbAV of the correction coefficient is determined in step S311. In step S313, the electronic control unit 9 having determined that is less than the initial tolerance first determines whether or not the oxygen (O2) sensor 15 is in an active state. And on condition that the oxygen sensor 15 is in an active state,
Whether the output cycle (amplitude cycle) is shorter than the amplitude cycle Fs during normal ignition (step S314).
Whether the lean (L) output time is longer than the lean output time TLOW during normal ignition (step S316).
If it is determined that the output cycle is shorter than the amplitude cycle Fs during normal ignition, the CF counter in the temporary misfire counter 912 is incremented in step S315, and the lean side of the output is If it is determined that the output time is longer than the lean side output time TLOW during normal ignition, the CT counter of the temporary misfire counter 912 is incremented in step S317.
[0121]
The electronic control unit 9 having finished checking all the items in this way, finally, in step S318, the above-described crank angular velocity values for each cylinder including the crank angular velocity ωn calculated in step S304 are described above. Ω (n-5) → discard, ω (n-4) → ω (n-5), ω (n-3) → ω (n-4), ω (n-2) → ω (n- 3) Update processing such as ω (n-1) → ω (n-2), ωn → ω (n-1) is performed, and the tolerance learning execution condition check routine is exited.
[0122]
It should be noted that the tolerance learning execution condition check routine is repeatedly executed based on the count value of the ignition number counter 909 until, for example, “100” ignition or the like has elapsed, a tolerance learning control routine (FIG. 4). As already described in the explanation.
[0123]
Next, the success / failure determination process of the tolerance learning execution condition executed as the process of step S212 in the tolerance learning control routine will be described further with reference to FIG.
[0124]
The success / failure determination routine for the tolerance learning execution condition shown in FIG. 11 is started when it is determined in step S211 of the tolerance learning control routine (FIG. 4) that the predetermined ignition number has elapsed. .
[0125]
When the tolerance learning execution condition success / failure determination routine is started in this way, the electronic control unit 9 (learning control unit 906) first, in step S400, each counter (CMF counter, COF counter, Whether any one of the counter values of the CF counter and the CT counter is “1” or more, or the count value of the rough load counter (CRG counter) 911 is a predetermined rough load determination value KRG for the same count value. It is determined whether or not this is the case.
[0126]
As a result, when any one of the temporary misfire counters 912 is “1” or more, or when the rough load counter 911 is more than the determination value KRG, In S401, a flag indicating "tolerance learning cannot be executed" is set in an appropriate area in the RAM 9c.
[0127]
On the other hand, when the count values of the temporary misfire counter 912 are all “0” and the count value of the rough load counter 911 is less than the determination value KRG, in step S402, an appropriate value in the RAM 9c is selected. A flag indicating that “tolerance learning can be performed” is set in the area.
[0128]
When the flag processing is thus completed, the electronic control unit 9 resets the temporary misfire counter 912 and the rough load counter 911, and exits the routine for determining success or failure of the same tolerance learning execution condition.
[0129]
In step S213 of the tolerance learning control routine (FIG. 4), the electronic control unit 9 executes the above-described tolerance learning based on the flag indicating “tolerance learning can be performed” or the flag indicating “tolerance learning cannot be performed” thus processed. It will be judged whether it is “possible” or “impossible”. In other words, in the tolerance learning execution condition check routine shown in FIGS. 7 and 8, only when all the check items are normal, the tolerance learning in the above-described manner, that is, the learning value ΔθnL (NE, PM ) Will be updated. For this reason, the reliability of the learning value ΔθnL (NE, PM) is naturally maintained high.
[0130]
As described above, according to the misfire detection device according to the embodiment,
(1) By learning about the crank angle deviation (inter-cylinder angle tolerance) for each cylinder of the internal combustion engine 1 and depending on the operating conditions, the engine 1 according to each cylinder and the operating conditions In addition, an extremely accurate crank angular speed ωn can be calculated. For this reason, the crank angular speed fluctuation amount Δω (n−α−1) calculated based on the transition of the crank angular speed ωn is naturally an accurate value, and the crank angular speed fluctuation amount Δω (n−α-1) and misfire determination are determined. The determination accuracy of misfire performed based on the comparison with the value REF1 is also maintained extremely high.
(2) In addition to misfires determined based on the 360 ° CA differential method, it is also determined whether or not misfire has occurred in a 360 ° CA opposed cylinder continuous misfire that cannot be determined by the 360 ° CA differential method. In addition, since the learning is performed only when normal ignition is confirmed in these determinations, the reliability of the learning value is also maintained high.
(3) The engine 1 is in a transient state such as sudden acceleration or sudden deceleration, shift change state, fuel cut or return, start-up or electric load application, idle state, purge control state, EGR (exhaust gas recirculation control) execution Execute learning even when the engine is in a specific operating state that causes large crankshaft fluctuations, such as during middle or variable intake, or when it is in a so-called misfiring detection range such as a light-load operating range or high-speed operating range. Therefore, the reliability of the learning value is maintained at a high level.
(4) By adding a transient correction term to the calculation of the crank angle deviation time ΔTn between the cylinders as in the equations (3) to (7), for example, during sudden acceleration or sudden deceleration, the engine 1 The influence of the transient increase / decrease in rotational fluctuation due to operating conditions is also suitably excluded from the learned value.
Many excellent effects can be obtained.
[0131]
In the apparatus of the same embodiment, when the first cylinder (# 1) is determined in step S204 of the tolerance learning control routine (FIG. 4), this first cylinder and the other second to sixth cylinders ( All crank angle deviations between # 2 and # 6) are obtained.
[0132]
However, in the case of an internal combustion engine composed of an even number of cylinders, the crank angles of the cylinders separated by 360 ° CA are detected through the same rotor detected part, so that the crank angle deviation between the cylinders is small in the first place.
[0133]
Therefore, the cylinders separated by 360 ° CA may be set as one set (or 3 sets in the case of 6 cylinders), and tolerance learning between cylinders (between sets) may be performed for each set. . According to such a configuration, the amount of calculation required in the electronic control unit 9 and the memory capacity of ROM, RAM, etc. can be greatly reduced.
[0134]
Further, the cylinder determined in step S204 of the tolerance learning control routine is not limited to the first cylinder (# 1), and may be any other cylinder. The point is that the crank angle deviation of another cylinder (or set) with respect to a specific cylinder (or set) may be calculated.
[0135]
Further, in the apparatus of the same embodiment, the operating conditions (rotational speed NE) of the engine 1 in the manner shown in FIGS. 5 and 6 respectively in step S208 and step S214 of the tolerance learning control routine (FIG. 4). The crank angle deviation Δθn is integrated or learned separately from the engine load PM).
[0136]
However, in view of the above operating conditions of the internal combustion engine, FIGS. 12 (a) and 12 (b) show the second, fifth cylinder group, and the third, sixth cylinder group, respectively, for the first and fourth cylinder groups. As exemplified by the rotational fluctuation tolerance with respect to the engine load, even if the engine load changes, the tendency of the rotational fluctuation tolerance is substantially constant.
[0137]
Therefore, in the memory structure illustrated in FIGS. 5 and 6 as well, the engine load column is deleted as the operation condition, and the above-described integration or learning of the crank angle deviation Δθn is performed for each cylinder and rotation speed. You can also Even with such a configuration, the amount of calculation required in the electronic control unit 9 and the memory capacity of ROM, RAM, and the like are greatly reduced.
[0138]
Even if the above-described crank angle deviation Δθn is accumulated or learned according to the operating conditions of the engine 1, there is little opportunity for the accumulation or learning to be performed in the high rotation range of the engine 1. For this reason, the learning value is not obtained, and even if a misfire occurs, that fact may not be detected.
[0139]
However, as illustrated in FIG. 13, the crank angle deviation (tolerance) between the cylinders tends to increase almost linearly as the rotational speed increases.
That is, when these tolerances are learned under two frequent operating conditions of the engine 1, for example, in a low engine speed range, the tolerance in the engine 1 at a high engine speed range can be determined by performing so-called linear interpolation. It becomes possible. If the tolerance in the high rotation range of the engine 1 is determined based on such a principle and the calculated tolerance is learned, the above-described inconvenience is preferably eliminated.
[0140]
Further, according to the same principle, the tolerance of the regions can be learned by the above-described linear interpolation even for regions excluded from the tolerance learning execution condition in anticipation that the rotational fluctuation of the engine 1 becomes large.
[0141]
Further, when a misfire occurs in the internal combustion engine 1, the unburned gas burns afterward in the exhaust pipe 14, and the output of the oxygen (O2) sensor 15 is output in a check routine for tolerance learning execution conditions (FIGS. 7 and 8). An accurate misfire determination (step S314 and step S316) may not be possible.
[0142]
However, in the apparatus of the above embodiment,
-An exhaust temperature sensor is added, and execution of learning is prohibited when the exhaust temperature detected through the sensor exceeds a predetermined temperature.
Or
・ Prohibit learning in operating conditions where afterburning is likely to occur, such as when driving under high load conditions.
If such a configuration is also provided, such inconvenience can be suitably avoided.
[0143]
In the apparatus of the embodiment, a system that performs feedback control of the air-fuel ratio based on the output of the oxygen sensor 15 is assumed. However, in a system that uses a linear air-fuel ratio sensor that linearly detects the air-fuel ratio based on the combustion gas of the engine and performs feedback control of the air-fuel ratio based on the output of the linear air-fuel ratio sensor, this linear air-fuel ratio By using the output of the sensor, it is also possible to perform misfire determination corresponding to the processing of step S314 and step S316 in the tolerance learning execution condition check routine (FIGS. 7 and 8).
[0144]
Incidentally, in the case of a linear air-fuel ratio sensor, if a misfire occurs in the engine,
(A) The output changes to the lean side.
Or
(B) The output has an offset toward the lean side as a whole.
Any one of the states is displayed. Therefore, in this case, as the learning control unit 906,
When the output of the linear air-fuel ratio sensor is on the lean side for a predetermined period or longer, the CF counter is incremented.
The CT counter is incremented when the average value of the output of the linear air-fuel ratio sensor is on the lean side over a predetermined value.
The following structure will be adopted.
[0145]
An example of a sensor that can contribute to such misfire determination is an HC concentration sensor.
In addition, the selection or combination of check items in the tolerance learning execution condition check routine (FIGS. 7 and 8) is arbitrary, and these items can be freely selected or combined according to the scale of the target system. be able to. However, it goes without saying that the reliability of the learning value is maximized when all of the above-described items are selected.
[0146]
Further, in the apparatus of the same embodiment, the calculation of the crank angular velocity fluctuation amount Δωn in step S305 in the tolerance learning execution condition check routine (FIGS. 7 and 8) is performed by the 360 ° CA difference according to the above equation (17) The method was used.
[0147]
However, when a misfire occurs in a certain cylinder, the crank angular velocity ωn usually returns gradually to a normal angular velocity thereafter. Therefore, also in the 360 ° CA difference method, as shown in the following equation (17) ′,
Δωn = (ωn−1−ωn) − (ωn + 2−ωn + 3) (17) ′
Thus, it is desirable to introduce the subsequent crank angular speeds “ωn + 2” and “ωn + 3”. As a result, when a misfire occurs, a larger value is obtained as the crank angular speed fluctuation amount Δωn, and the S / N (signal / noise) ratio is improved.
[0148]
In the apparatus according to the embodiment, when the learning value is updated on condition that a flag indicating that “tolerance learning can be performed” is set in the success / failure determination of the tolerance learning execution condition (FIG. 11). However, for example,
・ When the difference between the value to be updated and the learning value so far is larger than a predetermined value, the learning value is updated by that value only when the value to be updated becomes the same value continuously several times. Allow.
Such a learning algorithm may be adopted. According to such an algorithm, the erroneous learning is not performed by the value obtained by chance, and the reliability of the learning value is further improved.
[0149]
When calculating the tolerance learning value, as described above, the apparatus of the same embodiment integrates the crank angle deviation for each operating condition until the predetermined number of samples is obtained, and the average value (Δθn (NE, PM) _AV The learning value (ΔθnL (NE, PM)) was calculated based on (Accurate calculation of equation (16)), but instead of this average value, the same values were obtained for the same operating conditions. A value obtained by smoothing the crank angle deviation can be appropriately employed.
[0150]
In the apparatus of the same embodiment, when the misfire determination is performed in the main routine (FIG. 3), the crank angular speed fluctuation amount Δω (n−α−1) compared with the misfire determination value REF1 is set as the learning value. The correction based on the misfire determination value REF1 may be corrected based on the learning value.
[0151]
Further, the learning value is not limited to the crank angle deviation (inter-cylinder angle tolerance), and a value corresponding thereto, for example, a crank angular velocity or a variation amount thereof can be employed.
[0152]
By the way, although not mentioned in the embodiment, as a characteristic of the internal combustion engine, the crank angle deviation may become extremely nonuniform at a specific rotation speed due to vibration of the engine body or the like.
[0153]
As an example, the above-described electromagnetic pickup mounting arm (stay) for detecting the crank angle resonates due to vibration of the engine body, and the positional relationship between the pickup and the rotor detected portion changes.
[0154]
In such a case, since the detection interval (detection time) of the rotor detected part by the electromagnetic pickup becomes irregular, the crank angle deviation (angle tolerance) calculated through the equation (8) based on the detection time. Singular points SP and SP ′ as shown in FIG. 14 are also generated in Δθ. FIG. 14 shows actual measurement data regarding the relationship between the rotational speed measured for the in-line four-cylinder engine with a displacement of 1800 cc and the same angle tolerance Δθ.
[0155]
When such singular points SP and SP ′ are generated, tolerance learning is performed for each operating condition (rotational speed) with respect to the crank angle deviation Δθ, and in other rotational speed regions, a straight line from these tolerance learned values is obtained. Even if the angle tolerance is obtained by interpolation, the angle tolerance characteristic cannot be accurately reflected in the amount of change in angular velocity, which may result in misjudgment regarding misfire judgment executed through the main routine (FIG. 3). Disappear.
[0156]
Note that such singular points SP and SP ′ are generated depending on the type and shape of the engine and the environment in which the engine is placed. The singular points SP and SP ′ are generated under any operating conditions of the engine. It is difficult to identify what happens.
[0157]
On the other hand, learning all these angular tolerances over the entire operating range of the target internal combustion engine is limited by memory capacity and the like, which is not practical.
[0158]
Accordingly, in the following, as another embodiment of the misfire detection device according to the present invention, even when the singular point occurs in the crank angle deviation (angle tolerance) Δθ, it is preferable to use the singular point with a small memory capacity. An example of an apparatus that can avoid the influence will be described.
[0159]
Here, in addition to the tolerance learning by the apparatus of the previous embodiment, deviation learning control as shown in FIG. 15 and FIG. 16 is executed for the deviation between the tolerance learning value and the actual tolerance including the singular point. The misfire determination value REF1 used in the main routine (FIG. 3) is corrected based on the deviation learning value to avoid erroneous misfire determination due to the presence of the singular point.
[0160]
Incidentally, in this case, since it is necessary to obtain the deviation between the tolerance learning value and the actual tolerance in the interpolation region between the specific rotational speeds (operating conditions) at which the tolerance learning is performed, it is shown in FIG. 15 and FIG. In the deviation learning control routine, the rotation speed sections in which the tolerance learning is performed are divided into zones for each predetermined number of rotations (for example, every 500 rpm) corresponding to the learning areas, and the tolerance learning value ( The influence of the deviation between the interpolation value) and the actual tolerance is suppressed. This deviation learning control routine is started and executed through the electronic control unit 9 every ignition of the internal combustion engine 1 (every 120 ° CA in the case of 6 cylinders and every 180 ° CA in the case of 4 cylinders). Is done.
[0161]
Hereinafter, the details of the deviation learning control routine shown in FIGS. 15 and 16 will be sequentially described.
That is, when the deviation learning control routine is started in association with ignition of an arbitrary cylinder of the internal combustion engine 1, the electronic control unit 9 first responds to the current rotational speed (operating condition) of the engine in step S500. It is determined whether or not the tolerance learning is completed in the zone. If the tolerance learning has not been completed, comparison with the corresponding actual tolerance cannot be performed, and thus this routine is terminated once.
[0162]
On the other hand, if the tolerance learning is completed in the zone, the electronic control unit 9 determines whether or not the deviation learning execution flag is set in the next step S501. The deviation learning execution flag is a flag for preventing the completion of insufficient deviation learning from being set in a normal state and being reset depending on the following conditions.
[0163]
In other words, when performing deviation learning corresponding to each of the above-mentioned zones, it is most influential if a deviation measurement between a tolerance learning value (interpolation value) and an actual tolerance in a certain zone is only partially performed. The singular point may not be measured. So here, for example,
The “tolerance learning execution condition check process (FIGS. 7 and 8)” according to step S210 of the tolerance learning control routine illustrated in FIG. 4 and the “tolerance learning execution condition success / failure determination process” (FIG. 11) according to step S212. ) ”Is recognized as a state in which learning should not be executed (step S502).
Or
A case where the deviation measurement of the specific zone cannot be performed carefully due to a sudden change in the rotation speed (step S503).
If the deviation learning is not sufficient for completion in the zone, the deviation learning execution flag is reset in step S504 so that at least the current deviation learning in the zone is not completed. Yes.
[0164]
If it is determined in step S501 that the deviation learning execution flag is set, and it is determined that there is no cause for canceling these learnings, the electronic control unit 9 corresponds in step S505. A deviation Δ (Δθ) between the tolerance learning value ΔθL and the actual tolerance Δθ is obtained. As described above, the actual tolerance Δθ is calculated through the equation (8) based on the detection interval (detection time) of the rotor detected portion by the electromagnetic pickup.
[0165]
The electronic control unit 9 that has obtained the deviation Δ (Δθ) in this way determines the deviation Δ (Δθ) held in the zone so as to obtain the largest value for the deviation Δ (Δθ) in the next step S506. The larger value of the value and the value of the deviation Δ (Δθ) obtained this time is held in a predetermined area in the RAM 9c (FIG. 1). This is a consideration for using a deviation having the largest influence such as the singular point as a learning value.
[0166]
In this way, the electronic control unit 9 that holds a larger deviation Δ (Δθ), resets the deviation learning execution flag in step S504, or determines that the deviation learning execution flag is not set in step S501. In step S507, it is determined whether or not the rotational speed has passed through the previous zone and shifted to a new zone. If the zone has not been shifted to a new zone, only the calculation of the deviation Δ (Δθ) in the zone, the update by the larger value, and the processing after step S516 (FIG. 16) to be described later are repeatedly executed. .
[0167]
If it is determined in step S507 that the zone has been shifted to a new zone, the electronic control unit 9 does not go back to the zone in which the transition originated in step S508, but passes through all the zones before the transition. It is determined whether or not it is a transition to the next zone (zone on the high speed side for only one stage). If the transition is a return to the original zone, the deviation Δ (Δθ) has not been measured over the entire zone before the transition, so the learning process described below is not performed. Then, the process after step S514 (FIG. 16) described later is performed.
[0168]
On the other hand, if it is determined in step S508 that the transition to the new zone is the transition to the next zone, the electronic control unit 9 confirms that the deviation learning execution flag is set in step S509. After confirmation, deviation learning according to the next steps S510 to S513 is executed. Even when the deviation learning execution flag is not set, that is, when it is determined that the execution of learning is inappropriate through steps S502 to S504 in the zone before the transition, the entire zone before the transition is covered. Thus, the measurement of the deviation Δ (Δθ) is not performed, and thus the learning process described below is not performed, and the processes after step S514 described later are performed.
[0169]
When executing the deviation learning, first, in step S510 (hereinafter, FIG. 16), it is determined whether or not the deviation learning execution condition is established for the first time in the zone to be learned, that is, the zone before the transition. .
[0170]
As a result, if it is determined that the same condition is established for the first time, in step S512, the maximum deviation Δ (Δθ) held in the predetermined area in the RAM 9c with respect to the zone before the transition is set to the same. The deviation learning value is registered in the deviation learning value memory as a zone deviation learning value, and in step S513, a deviation learning completion flag indicating that deviation learning has been completed for the zone is set. Similar to the tolerance learning value memory 910 illustrated in FIG. 6, the deviation learning value memory is also prepared in advance in a predetermined area in the backup RAM 9d (FIG. 1), and each of the above-described zones (rotational speed ranges) is provided. Separately, the deviation learning value Δ (Δθ) is registered.
[0171]
On the other hand, if it is determined in step S510 that the deviation learning execution condition is not established in the zone before the transition but the first time, it is registered in the deviation learning value memory in step S511. The deviation learning value Δ (Δθ) for the same zone is updated with the maximum deviation Δ (Δθ) held in the predetermined area in the RAM 9c for the zone. In this update, the above-described annealing (gradual change) process may be used in combination.
[0172]
When the execution of the deviation learning is finished in this way, the electronic control unit 9 next performs deviation learning for the new zone that has been transferred, so that the deviation Δ (Δθ) held in the predetermined area in the RAM 9c in step S514. In step S515, the deviation learning execution flag is set to a standard (default) state.
[0173]
Thereafter, the electronic control unit 9 confirms that the deviation learning completion flag relating to the target zone (the zone before the transition) is set in step S516, and then performs the next steps S517 to S518. The misfire determination value correction process concerning is performed. If the deviation learning completion flag relating to the same zone is not set, the routine is temporarily terminated without performing the misfire determination value correction processing in steps S517 to S518.
[0174]
When executing the misfire determination value correction process, first, in step S517, an offset amount REFofs to be added to the misfire determination value REF1 is calculated from the deviation learning value Δ (Δθ) of the zone. The calculation of the offset amount REFofs is as follows:
REFofs = Kofs × Δ (Δθ) × rotational speed (18)
In this way, the angle (rad) information of the deviation learning value Δ (Δθ) is converted into the amount of change in the angular velocity (rad / sec). Here, the coefficient Kofs is a conversion coefficient for converting the deviation learning value Δ (Δθ) into a value of the same dimension as the misfire determination value REF1.
[0175]
The electronic control unit 9 which has thus calculated the offset amount REFofs finally adds the offset amount REFofs calculated in step S518 to the misfire determination value REF1 and ends this routine.
[0176]
By performing such deviation learning control for each ignition of the internal combustion engine 1, the maximum value (deviation learning value) of the deviation Δ (Δθ) between the tolerance learning value interpolated value measured for each zone and the actual tolerance. ) Is separately obtained, and the obtained angular velocity fluctuation amount is added to the misfire determination value REF1 as an offset amount REFofs each time.
[0177]
Therefore, even if the singular point described above occurs in the crank angle deviation (actual tolerance) Δθ, that is, the crank angular speed fluctuation amount Δω (n−α−) compared with the misfire determination value REF1 in the main routine (FIG. 3). Even when an increase based on the singular point occurs in 1), an erroneous misfire determination due to the singular point is caused by adding the offset amount REFofs corresponding to the increase in the angular velocity fluctuation amount to the misfire determination value REF1. This is also preferably avoided.
[0178]
In addition, according to the deviation learning control routine, since the deviation learning is performed in units of the zones, an increase in the required memory capacity can be minimized.
[0179]
In the apparatus of the embodiment, the offset amount REFofs is added to the misfire determination value REF1 to deal with the singular point. However, the crank angular speed fluctuation amount Δω (n to be compared in the main routine is compared. Of course, the configuration may be such that the offset amount REFofs is subtracted from -α-1) to cope with it.
[0180]
Further, the deviation learning value is not limited to the deviation amount from the crank angle deviation (inter-cylinder angle tolerance), but the corresponding value, that is, the deviation amount from the crank angular speed, for example, or the variation thereof, according to the tolerance learning value. Amount, etc. can be employed.
[0181]
In any of the above embodiments, the output of the air-fuel ratio sensor (oxygen sensor, linear air-fuel ratio sensor) or the air-fuel ratio correction coefficient for air-fuel ratio feedback control is used as an element for properly determining the learning execution condition. I tried to refer to it.
[0182]
However, as is clear from the determination contents illustrated in FIG. 8, it is configured to detect whether or not misfire has occurred in the engine directly from the output of the air-fuel ratio sensor and the air-fuel ratio correction coefficient for air-fuel ratio feedback control. You can also.
[0183]
That is, only the routine for determining normal ignition illustrated in FIG. 8 may be used in place of misfire detection using the crank angular speed fluctuation amount of the apparatus according to the embodiment.
[0184]
Even at this time, the selection or combination of check items in FIG. 8 is arbitrary, and these items can be freely selected or combined according to the scale of the target system.
[0185]
Furthermore, a misfire detection method that is freely selected or combined can be combined with a method other than the misfire detection method of the apparatus according to the above-described embodiment to further improve the misfire detection accuracy by these methods. You can also.
[0186]
Further, in the check routine illustrated in FIG. 8, three methods for detecting continuous misfiring of the 360 ° CA on-cylinder cylinder are shown, but the processing in step S311 is the processing in step S311 ′ in FIG. Changes may be made as shown.
[0187]
That is, in the check routine of FIG. 8, misfire detection is performed by comparing the sum of the air-fuel ratio correction coefficient cfb and its average value with the initial tolerance, but as shown in FIG. 17, the same correction as the air-fuel ratio correction coefficient is performed. Misfire detection may be performed by comparing the sum of the coefficient and the learning value with the initial tolerance.
[0188]
Furthermore, in the check routine of FIG. 8, it is determined in the process of step S314 that a misfire has occurred when the O2 sensor amplitude cycle is shorter than the preset amplitude cycle Fs during normal ignition. As shown in the process of step S314 ′ in the check routine of FIG. 17, not only the lower limit determination value (Fs) but also a predetermined range is set, and when it is outside this range, it is determined that misfire has occurred. It may be. By setting the predetermined range in this way, not only a misfire as shown in FIG. 9 but also a misfire as shown in FIG. 10 can be detected.
[0189]
Further, in each of the above embodiments, the crank angle deviation Δθn is learned as the rotational angular velocity fluctuation amount according to the operating conditions. However, the present invention is not limited to this, and for example, the crank angle deviation Δθn is calculated as a corresponding value. The crank angle deviation time ΔTn used for obtaining may be learned.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of a misfire detection apparatus according to the present invention.
FIG. 2 is an exemplary block diagram showing a functional configuration of the electronic control apparatus according to the embodiment;
FIG. 3 is a flowchart showing a main routine for misfire determination according to the embodiment;
FIG. 4 is a flowchart showing a tolerance learning control routine of the embodiment.
FIG. 5 is a schematic diagram showing an example of an integrated value memory structure of crank angle deviation (tolerance) between cylinders.
FIG. 6 is a schematic diagram showing an example of a learned value memory structure of crank angle deviation (tolerance) between cylinders.
FIG. 7 is a flowchart showing a check routine for a tolerance learning execution condition.
FIG. 8 is a flowchart showing a check routine for a tolerance learning execution condition.
FIG. 9 is a time chart showing an output example of the oxygen sensor at the time of misfire.
FIG. 10 is a time chart showing an output example of the oxygen sensor at the time of misfire.
FIG. 11 is a flowchart showing a success / failure determination routine for a tolerance learning execution condition.
FIG. 12 is a graph showing a load-cylinder crank angle deviation (tolerance) characteristic;
FIG. 13 is a graph showing a rotational speed-cylinder crank angle deviation (tolerance) characteristic.
FIG. 14 is a graph showing a state of singular points generated in crank angle deviation (tolerance).
FIG. 15 is a flowchart showing a deviation learning control routine which is a countermeasure against singular points.
FIG. 16 is a flowchart showing a deviation learning control routine which is a countermeasure against singular points.
FIG. 17 is a flowchart showing another check routine for tolerance learning execution conditions;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Internal combustion engine, 2 ... Intake pipe, 3 ... Intake pipe pressure sensor, 5 ... Rotation angle sensor, 6 ... Reference position sensor, 7 ... Distributor, 8 ... Water temperature sensor, 9 ... Electronic control unit, 10 ... Injector, 11 ... Igniter, 12 ... warning lamp, 13 ... piston, 14 ... exhaust pipe, 15 ... oxygen (O2) sensor, 9a ... CPU, 9b ... ROM, 9c ... RAM, 9d ... backup RAM, 9e ... I / O port, 901 ... Angular velocity variation calculation unit, 902 ... misfire determination unit, 903 ... ignition number counter, 904 ... temporary misfire counter, 905 ... angular velocity storage unit, 906 ... learning control unit, 907 ... integration counter, 908 ... inter-cylinder crank angle deviation (tolerance) ) Integrated value memory, 909 ... Ignition number counter, 910 ... Inter-cylinder crank angle deviation (tolerance) learning value memory, 911 ... Rough road counter (CRG), 12 ... provisional misfire counter.

Claims (18)

Based on the rotation angle signal corresponding to the rotation of the output shaft of the internal combustion engine, the amount of fluctuation in the rotational angular velocity for each cylinder of the engine output shaft is obtained, and the obtained amount of fluctuation in the rotational angular velocity for each cylinder is compared with a predetermined misfire determination value. In a misfire detection device for an internal combustion engine that detects the occurrence of misfire,
Learning means for learning the amount of variation in rotational angular velocity for each cylinder of the engine output shaft according to the operating conditions of the engine;
A correction means for correcting the misfire determination value or the variation amount of the rotational angular velocity for each cylinder to be compared with the determination value based on a learning value in a corresponding operation condition in each case,
The learning means executes the learning on the condition that the engine is normally ignited ,
The correction means measures a deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft, and is compared with the misfire judgment value or the same judgment value according to the measured deviation. A misfire detection apparatus for an internal combustion engine, comprising deviation correction means for correcting a cylinder-specific rotational angular velocity fluctuation amount . The misfire detection device for an internal combustion engine according to claim 1,
The learning means obtains an inter-cylinder rotation angle deviation of the engine output shaft, and learns the obtained inter-cylinder rotation angle deviation for each cylinder of the same engine and operating conditions.
The correction means corrects the inter-cylinder rotation angle deviation based on the corresponding cylinder in each case of the same engine and the learned value under the operating condition to determine the rotation angular velocity for each cylinder of the same engine output shaft, A misfire detection apparatus for an internal combustion engine, which calculates a cylinder-by-cylinder rotation angular velocity fluctuation amount to be compared with the misfire determination value based on a change in rotation angular velocity. The learning means determines that the engine is normally ignited based on the fact that the cylinder-by-cylinder rotational angular velocity fluctuation amount of the engine output shaft obtained based on the rotational angle signal is equal to or less than a temporary misfire determination value. 3. A misfire detection apparatus for an internal combustion engine according to 1 or 2. The learning means obtains based on the rotation angle signal by a second-order difference obtained by subtracting the rotation angle difference between the cylinder and the adjacent cylinder after 360 ° crank angle from the rotation angle difference between the cylinder subject to misfire detection and the adjacent cylinder. The misfire detection device for an internal combustion engine according to claim 1 or 2, wherein it is determined whether or not the cylinder-by-cylinder rotation angular velocity fluctuation amount of the engine output shaft is equal to or less than a temporary misfire determination value. The misfire detection device for an internal combustion engine according to claim 1 or 2,
An air-fuel ratio sensor for detecting the air-fuel ratio based on the oxygen concentration of the combustion gas of the internal combustion engine;
Air-fuel ratio feedback control means for feedback-controlling the fuel injection amount to the internal combustion engine using an air-fuel ratio correction coefficient so as to obtain a desired air-fuel ratio based on the output of the air-fuel ratio sensor,
During the execution of the feedback control by the air-fuel ratio feedback control means, the learning means calculates the sum of the air-fuel ratio correction coefficient and the average value of the correction coefficient, or the sum of the air-fuel ratio correction coefficient and the learning value of the correction coefficient. A misfire detection apparatus for an internal combustion engine, comprising: means for determining that the engine is normally ignited based on being not leaner than a predetermined value. The misfire detection device for an internal combustion engine according to claim 1 or 2,
An air-fuel ratio sensor for detecting the air-fuel ratio based on the oxygen concentration of the combustion gas of the internal combustion engine,
The learning means includes means for determining that the engine is normally ignited based on the fact that the output is not on the lean side for a predetermined period or more when the air-fuel ratio sensor is active. Misfire detection device. The misfire detection device for an internal combustion engine according to claim 1 or 2,
An oxygen sensor for detecting the rich / lean air-fuel ratio based on the oxygen concentration of the combustion gas of the internal combustion engine,
The learning means includes a means for judging that the engine is normally ignited based on the output cycle being within a predetermined range when the oxygen sensor is active. Detection device. The misfire detection device for an internal combustion engine according to claim 1 or 2,
A linear air-fuel ratio sensor for linearly detecting the air-fuel ratio based on the combustion gas of the internal combustion engine,
The learning means includes a means for determining that the engine is normally ignited based on the fact that the output of the linear air-fuel ratio sensor is not on the lean side beyond a predetermined value. . The misfire detection device for an internal combustion engine according to any one of claims 1 to 8,
When the difference between the value to be updated and the learned value so far is larger than a predetermined value, the learning means depends on the value only when the value to be updated becomes the same value continuously several times. A misfire detection apparatus for an internal combustion engine, wherein update of a learning value is permitted. The misfire detection device for an internal combustion engine according to any one of claims 1 to 9,
The learning means monitors the operating state of the engine, and stops the execution of the learning when the operating state causes a large rotational fluctuation of the engine output shaft. apparatus. The misfire detection device for an internal combustion engine according to any one of claims 1 to 10,
The learning means monitors the operating state of the engine, and stops the execution of the learning when the engine is in an operating range where the misfire determination is impossible. The misfire detection device for an internal combustion engine according to any one of claims 1 to 11,
The deviation correction means learns a maximum deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft according to an operating condition of the engine when measuring the deviation. Misfire detection device. In misfire detecting device according to claim 1 wherein the internal combustion engine,
The deviation correcting means includes means for updating the maximum deviation as a learned value of the operating condition when it is determined that the operating condition for one learned value has been operated as a whole. . The misfire detection device for an internal combustion engine according to claim 12 or 13,
A misfire detection apparatus for an internal combustion engine, wherein misfire detection is not performed until deviation learning by the deviation correction means is completed . Based on the rotation angle signal according to the rotation of the output shaft of the internal combustion engine, the amount of fluctuation in the rotational angular velocity for each cylinder of the engine output shaft is obtained, and the obtained amount of fluctuation in the rotational angular velocity for each cylinder is compared with a predetermined misfire determination value. In a misfire detection device for an internal combustion engine that detects the occurrence of misfire,
Learning means for learning the amount of variation in rotational angular velocity for each cylinder of the engine output shaft according to the operating conditions of the engine;
A correction means for correcting the misfire determination value or the variation amount of the rotational angular velocity for each cylinder to be compared with the determination value based on a learning value in a corresponding operation condition in each case,
When the difference between the value to be updated and the learning value so far is larger than a predetermined value, the learning means depends on the value only when the value to be updated becomes the same value continuously several times. Allow learning values to be updated,
The correction means measures a deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft, and is compared with the misfire judgment value or the same judgment value according to the measured deviation. A misfire detection apparatus for an internal combustion engine, comprising deviation correction means for correcting a cylinder-specific rotational angular velocity fluctuation amount . Based on the rotation angle signal according to the rotation of the output shaft of the internal combustion engine, the amount of fluctuation in the rotational angular velocity for each cylinder of the engine output shaft is obtained, and the obtained amount of fluctuation in the rotational angular velocity for each cylinder is compared with a predetermined misfire determination value. In a misfire detection device for an internal combustion engine that detects the occurrence of misfire,
Learning means for learning the amount of variation in rotational angular velocity for each cylinder of the engine output shaft according to the operating conditions of the engine;
A correction means for correcting the misfire determination value or the variation amount of the rotational angular velocity for each cylinder to be compared with the determination value based on a learning value in a corresponding operation condition in each case,
The learning means monitors the operating state of the engine, and when the engine is in an operating range where the misfire determination is impossible, stops the execution of the learning,
The correction means measures a deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft, and is compared with the misfire judgment value or the same judgment value according to the measured deviation. A misfire detection apparatus for an internal combustion engine, comprising deviation correction means for correcting a cylinder-specific rotational angular velocity fluctuation amount . Based on the rotation angle signal according to the rotation of the output shaft of the internal combustion engine, the amount of fluctuation in the rotational angular velocity for each cylinder of the engine output shaft is obtained, and the obtained amount of fluctuation in the rotational angular velocity for each cylinder is compared with a predetermined misfire determination value. In a misfire detection device for an internal combustion engine that detects the occurrence of misfire,
Learning means for learning the amount of variation in rotational angular velocity for each cylinder of the engine output shaft according to the operating conditions of the engine;
A correction means for correcting the misfire determination value or the variation amount of the rotational angular velocity for each cylinder to be compared with the determination value based on a learning value in a corresponding operation condition in each case,
The learning means monitors the operating state of the engine, and when the engine is in an operating range where misfire determination is not possible, learns the amount of variation in rotational angular velocity for each cylinder in the operating range by linear interpolation,
The correction means measures a deviation between a learning value obtained by the learning means and a cylinder-specific rotational angular velocity fluctuation amount of the engine output shaft, and is compared with the misfire judgment value or the same judgment value according to the measured deviation. A misfire detection apparatus for an internal combustion engine, comprising deviation correction means for correcting a cylinder-specific rotational angular velocity fluctuation amount . The misfire detection device for an internal combustion engine according to any one of claims 15 to 17,
The learning means obtains an inter-cylinder rotation angle deviation of the engine output shaft, and learns the obtained inter-cylinder rotation angle deviation for each cylinder of the same engine and operating conditions.
The correction means corrects the inter-cylinder rotation angle deviation based on the corresponding cylinder of the same engine each time and the learned value under the operating condition to determine the rotation angular velocity for each cylinder of the same engine output shaft, A misfire detection apparatus for an internal combustion engine, which calculates a cylinder-by-cylinder rotation angular velocity fluctuation amount to be compared with the misfire determination value based on a change in rotation angular velocity .

Images