JP3602575B2 - Air-fuel ratio estimator for each cylinder of internal combustion engine - Google Patents

Description

とするとき、値（Ｙ（ｋ）−ＣＸハット（ｋ））がオブザーバの収束状況を示していることになるので、その値からオブザーバの収束状況を簡易かつ正確に判定することができる。
【００１４】
また、推定された状態変数Ｘハット（ｋ）から集合部空燃比の推定値ＣＸハット（ｋ）を算出し、前記出力Ｙ（ｋ）と比較してオブザーバの収束状況を判定する判定手段、を備え、フィードバック制御手段は、オブザーバの収束状況が不良と判定されたとき、フィードバック補正項（ゲイン）の変更を停止する如く構成したので、オブザーバの収束状況を簡易かつ正確に判定することができると共に、それに応じてフィードバック補正項の変更を停止することができることから、制御値を誤ることがない。
【００１５】
請求項２項にあっては、前記出力Ｙ（ｋ）として前記空燃比センサの出力を用いる如く構成したので、一層容易かつ正確にオブザーバの収束状況を判定することができる。
【００１６】
請求項３項にあっては、前記判定手段は、集合部空燃比の推定値ＣＸハット（ｋ）と出力Ｙ（ｋ）の差Ｅ．ＨＡＴ（ｋ）およびその平均値のいずれか求め、所定値と比較してオブザーバの収束状況を判定する如く構成したので、簡易かつ正確にオブザーバの収束状況を判定することができる。特に平均値を用いるときは、一過性の原因によって収束状況が低下したときは収束状況の低下と判定しないので、実際の運転状態に良くマッチした空燃比推定、ないしはそれに基づく空燃比フィードバック制御を行うことができる。
【００１７】
請求項４項にあっては、前記判定手段は、集合部空燃比の推定値ＣＸハット（ｋ）と出力Ｙ（ｋ）の差Ｅ．ＨＡＴ（ｋ）およびその平均値のいずれかを求め、第１の所定値と比較してそれを超えた回数をカウントし、カウント値を第２の所定値と比較してオブザーバの収束状況を判定する如く構成したので、一過性の原因によって収束状況が低下したときは収束状況の低下と判定しないので、実際の運転状態に良くマッチした空燃比推定、ないしはそれに基づく空燃比フィードバック制御を行うことができる。
【００１８】
【実施例】
以下、添付図面に即してこの発明の実施例を説明する。
【００１９】
図１はこの発明に係る内燃機関の気筒別空燃比推定装置を全体的に示す概略図である。図において符号１０は４気筒の内燃機関を示しており、吸気路１２の先端に配置されたエアクリーナ１４から導入された吸気は、スロットル弁１６でその流量を調節されつつインテークマニホルド１８を経て第１ないし第４気筒に流入される。各気筒の吸気弁（図示せず）の付近にはインジェクタ２０が設けられて燃料を噴射する。
【００２０】
噴射されて吸気と一体となった混合気は、各気筒内で図示しない点火プラグで点火されて燃焼してピストン（図示せず）を駆動する。燃焼後の排気ガスは排気弁（図示せず）を介してエキゾーストマニホルド２２に排出され、エキゾーストパイプ２４を経て三元触媒コンバータ２６で浄化されつつ機関外に排出される。また、吸気路１２には、スロットル弁配置位置付近に、それをバイパスするバイパス路２８が設けられる。
【００２１】
内燃機関１０のディストリビュータ（図示せず）内にはピストン（図示せず）のクランク角度位置を検出するクランク角センサ３４が設けられると共に、スロットル弁１６の開度を検出するスロットル開度センサ３６、スロットル弁１６下流の吸気圧力を絶対圧力で検出する絶対圧センサ３８も設けられる。更に、排気系においてエキゾーストマニホルド２２と三元触媒コンバータ２６の間には酸素濃度検出素子からなる広域空燃比センサ４０が設けられ、排気ガス中の酸素濃度に比例した値を出力する。これらセンサ３４などの出力は、制御ユニット４２に送られる。
【００２２】
図２は制御ユニット４２の詳細を示すブロック図である。広域空燃比センサ４０の出力は検出回路４６に入力され、そこで適当な線形化処理が行われ、理論空燃比を中心としてリーンからリッチにわたる広い範囲において排気ガス中の酸素濃度に比例したリニアな特性からなる空燃比（Ａ／Ｆ）が検出される。その詳細は先に本出願人が提案した別の出願（特開平４−３６９４７１号）に述べられているので、これ以上の説明は省略する。尚、以下の説明において、このセンサを「ＬＡＦセンサ」（リニア・エーバイエフ・センサ）と称する。検出回路４６の出力はＡ／Ｄ変換回路４８を介してＣＰＵ５０，ＲＯＭ５２，ＲＡＭ５４などからなるマイクロコンピュータに取り込まれ、ＲＡＭ５４に格納される。
【００２３】
同様に、スロットル開度センサ３６などのアナログ出力は、レベル変換回路５６、マルチプレクサ５８および第２のＡ／Ｄ変換回路６０を介してマイクロコンピュータに入力される。またクランク角センサ３４の出力は波形整形回路６２で波形整形された後、カウンタ６４で出力値がカウントされ、カウント値はマイクロ・コンピュータ内に入力される。マイクロコンピュータにおいてＣＰＵ５０は、ＲＯＭ５２に格納された命令に従って検出値から制御値を演算し、駆動回路６６を介して各気筒のインジェクタ２０を駆動すると共に、第２の駆動回路６８を介して電磁弁７０を駆動し、図１に示したバイパス路２８を通る２次空気量を制御する。
【００２４】
図３はこの発明に係る内燃機関の気筒別空燃比推定装置の動作を説明するフロー・チャートであるが、これは先に提案した排気系の挙動を状態方程式で同定すると共に、オブザーバを介して各気筒空燃比を推定する技術を前提としているので、理解の便宜上、それについて簡単に説明する。
【００２５】
先ず、１個のＬＡＦセンサの出力から各気筒の空燃比を精度良く分離抽出するためには、ＬＡＦセンサの検出応答遅れを正確に解明する必要がある。そこで、とりあえずこの遅れを１次遅れ系と擬似的にモデル化し、図４に示す如きモデルを作成した。ここでＬＡＦ：ＬＡＦセンサ出力、Ａ／Ｆ：入力Ａ／Ｆ、とすると、その状態方程式は下記の数１で示すことができる。
【００２６】
【数１】

【００２７】
これを周期ΔＴで離散化すると、数２で示すようになる。図５は数２をブロック線図で表したものである。
【００２８】
【数２】

【００２９】
従って、数２を用いることによってセンサ出力より真の空燃比を求めることができる。即ち、数２を変形すれば数３に示すようになるので、時刻ｋのときの値から時刻ｋ−１のときの値を数４のように逆算することができる。
【００３０】
【数３】

【００３１】
【数４】

【００３２】
具体的には数２をＺ変換を用いて伝達関数で示せば数５の如くになるので、その逆伝達関数を今回のセンサ出力ＬＡＦに乗じることによって前回の入力空燃比をリアルタイムに推定することができる。図６にそのリアルタイムのＡ／Ｆ推定器のブロック線図を示す。
【００３３】
【数５】

【００３４】
続いて、上記の如く求めた真の空燃比に基づいて各気筒の空燃比を分離抽出する手法について説明すると、先願でも述べたように、排気系の集合部の空燃比を各気筒の空燃比の時間的な寄与度を考慮した加重平均であると考え、時刻ｋのときの値を、数６のように表した。尚、Ｆ（燃料量）を制御量としたため、ここでは『燃空比Ｆ／Ａ』を用いているが、後の説明においては理解の便宜のため、支障ない限り「空燃比」を用いる。尚、空燃比（ないしは燃空比）は、先に数５で求めた応答遅れを補正した真の値を意味する。
【００３５】
【数６】

【００３６】
即ち、集合部の空燃比は、気筒ごとの過去の燃焼履歴に重みＣ（例えば直近に燃焼した気筒は４０％、その前が３０％．．．など）を乗じたものの合算で表した。ここでＣはｎ×ｍの行列である。このモデルをブロック線図であらわすと、図７のようになる。
【００３７】
また、その状態方程式は数７のようになる。
【００３８】
【数７】

【００３９】
また集合部の空燃比をｙ（ｋ）とおくと、出力方程式は数８のように表すことができる。
【００４０】
【数８】

【００４１】
上記において、ｕ（ｋ）は観測不可能のため、この状態方程式からオブザーバを設計してもｘ（ｋ）は観測することができない。そこで４ＴＤＣ前（即ち、同一気筒）の空燃比は急激に変化しない定常運転状態にあると仮定してｘ（ｋ＋１）＝ｘ（ｋ−３）とすると、数９のようになる。これは、ｕ（ｋ）を目標空燃比としても同様である。
【００４２】
【数９】

【００４３】
ここで、上記の如く求めたモデルについてシミュレーション結果を示す。図８は４気筒内燃機関について３気筒の空燃比を１４．７にし、１気筒だけ１２．０にして燃料を供給した場合を示す。図９はそのときの集合部の空燃比を上記モデルで求めたものを示す。同図においてはステップ状の出力が得られているが、ここで更にＬＡＦセンサの応答遅れを考慮すると、センサ出力は図１０に「モデル出力値」と示すようになまされた波形となる。図中「実測値」は同じ場合のＬＡＦセンサ出力の実測値であるが、これと比較し、上記モデルが多気筒内燃機関の排気系を良くモデル化していることを検証している。
【００４４】
よって、数１０で示される状態方程式と出力方程式にてｘ（ｋ）を観察する通常のカルマンフィルタの問題に帰着する。その荷重行列Ｑ，Ｒを数１１のようにおいてリカッチの方程式を解くと、ゲイン行列Ｋは数１２のようになる。
【００４５】
【数１０】

【００４６】
【数１１】

【００４７】
【数１２】

【００４８】
これより行列Ａ−ＫＣを求めると、数１３のようになる。
【００４９】
【数１３】

【００５０】
一般的なオブザーバの構成は図１１に示されるようになるが、今回のモデルでは入力ｕ（ｋ）がないので、図１２に示すようにｙ（ｋ）のみを入力とする構成となり、これを数式で表すと数１４のようになる。
【００５１】
【数１４】

【００５２】
ここでｙ（ｋ）を入力とするオブザーバ、即ちカルマンフィルタのシステム行列は数１５のように表される。
【００５３】
【数１５】

【００５４】
今回のモデルで、リカッチ方程式の荷重配分Ｒの要素：Ｑの要素＝１：１のとき、カルマンフィルタのシステム行列Ｓは、数１６で与えられる。
【００５５】
【数１６】

【００５６】
図１３に上記したモデルとオブザーバを組み合わせたものを示す。シミュレーション結果は先の出願に示されているので省略するが、これにより集合部空燃比より各気筒の空燃比を的確に抽出することができる。
【００５７】
上記を前提として図３フロー・チャートに従って説明する。
【００５８】
尚、実施例の場合、各気筒空燃比を推定した後、推定値に基づいて各気筒空燃比を目標値にフィードバック制御する例を示す。即ち、オブザーバによって集合部空燃比より各気筒空燃比を推定することができたことから、ＰＩＤなどの制御則を用いて空燃比を気筒別に制御することが可能となるからである。
【００５９】
具体的には図１４に示すように、集合部空燃比を示すＬＡＦセンサ出力ＫＡＣＴと目標空燃比とからＰＩＤ制御則を用いて集合部フィードバック補正項（ゲイン）ＫＬＡＦを求めると共に、オブザーバ推定値＃ｎＡ／Ｆから気筒毎のフィードバック補正項（ゲイン）＃ｎＫＬＡＦ（ｎ：気筒）を求める。
【００６０】
気筒毎のフィードバック補正項＃ｎＫＬＡＦはより具体的には、集合部Ａ／Ｆを気筒毎のフィードバック補正項＃ｎＫＬＡＦの平均値の前回演算値で除算して求めた目標値とオブザーバ推定値＃ｎＡ／Ｆとの偏差を解消するようにＰＩＤ則を用いて求める。これにより、各気筒の空燃比は集合部空燃比に収束し、集合部空燃比は目標空燃比に収束することとなって、結果的に全ての気筒の空燃比が目標空燃比に収束する。
【００６１】
ここで、各気筒の燃料噴射量＃ｎＴｏｕｔ （インジェクタの開弁時間で規定される）は、
＃ｎＴｏｕｔ ＝Ｔｉｍ×ＫＣＭＤ×ＫＴＯＴＡＬ×＃ｎＫＬＡＦ×ＫＬＡＦ
で求められる。上記で、Ｔｉｍ：基本値、ＫＣＭＤ：目標空燃比、ＫＴＯＴＡＬ：その他の補正項、である。これ以外にバッテリ補正項などの加算項もあるが、省略する。尚、かかる制御の詳細は本出願人が先に提案した特願平６−３３２０１号に述べられているので、これ以上の説明は省略する。
【００６２】
以上を前提として実施例を図３フロー・チャートを参照して説明する。尚、このプログラムはＴＤＣからの所定のクランク角度において、即ち、噴射順位（第１、第３、第４、第２気筒の順）毎に各気筒の燃料噴射量を決定する。
【００６３】
先ず、Ｓ１００において機関回転数Ｎｅ、吸気圧力Ｐｂ、ＬＡＦセンサ出力ＫＡＣＴ（ｋ）（（ｋ）は前記の如く離散時間系での時刻を示す）などを読み込み、Ｓ１０２に進んでクランキングか否か判断し、否定されるときはＳ１０４に進んでフューエルカットか否か判断する。Ｓ１０４でも否定されるときはＳ１０６に進んで前記した基本値Ｔｉｍを検索する。これは機関回転数と吸気圧力とから設定されたマップ（特性図示せず）を検索して行う。
【００６４】
続いてＳ１０８に進んでＬＡＦセンサ４０の活性化が完了したか否か判断し、肯定されるときはＳ１１０に進んで前記したオブザーバ行列演算を行って各気筒空燃比（気筒毎Ａ／Ｆ）を推定する。続いてＳ１１２に進んで値Ｅ．ＨＡＴ（ｋ）を算出し、Ｓ１１４に進んで算出値Ｅ．ＨＡＴ（ｋ）を所定値と比較し、オブザーバの収束状況を判定する。
【００６５】
これについて説明すると、オブザーバの推定機能を表現する数１４を書き直すと、数１７に示すようになる。
【００６６】
【数１７】

【００６７】
数１７において、ｙ（ｋ）−ＣＸハット（ｋ）に着目すると、数１８のように示すことができる。
【００６８】
【数１８】

【００６９】
即ち、ｙ（ｋ）は前記した出力方程式で表される、排気系集合部空燃比、具体的にはＬＡＦセンサが検出する実際の集合部空燃比に相当し、ＣＸハット（ｋ）は推定された各気筒空燃比（内部状態変数Ｘハット）に各気筒排気ガスの混ざり度合いを示す行列Ｃを乗じたことにより、集合部空燃比推定値に相当する。従って、ｙ（ｋ）−ＣＸハット（ｋ）は、集合部空燃比の実際値（ＬＡＦセンサ出力ＫＡＣＴ）と推定値の差を表しており、換言すれば、差Ｅ．ＨＡＴ（ｋ）は、オブザーバの推定精度、即ち、オブザーバの収束状況を表す指標とみなすことができる。この発明は上記した知見に基づいてなされた。
【００７０】
図３フロー・チャートに戻ると、Ｓ１１２でこの差Ｅ．ＨＡＴ（ｋ）を求め、Ｓ１１４に進んで所定値と比較する。所定値は、オブザーバの収束状況を判定するに適当な値、例えば０．１などとする。
【００７１】
Ｓ１１４で差Ｅ．ＨＡＴ（ｋ）が所定値以下と判断されるときはオブザーバの収束状況が良好、即ち、所期の推定精度が得られていると判断し、Ｓ１１６に進んで気筒毎フィードバック補正項＃ｎＫＬＡＦを図１４に関して述べたように演算する。
【００７２】
続いてＳ１１８に進み、気筒毎フィードバック補正項＃ｎＫＬＡＦの学習値を演算する。具体的には、移動平均（ないしは加重移動平均）を求めて行う。尚、かく求めた学習値は、運転状態、具体的には燃料噴射量に応じて格納する。即ち、気筒毎フィードバック補正項＃ｎＫＬＡＦはインジェクタ２０の特性を示しているので、次に述べる出力燃料噴射量＃ｎＴｏｕｔと対にして格納しておく。
【００７３】
続いてＳ１２０に進み、図１４に関して述べたように集合部フィードバック補正項ＫＬＡＦを演算し、Ｓ１２２に進んで気筒毎の燃料噴射量＃ｎＴｏｕｔ を決定し、Ｓ１２４に進んで当該気筒のインジェクタ２０の駆動回路に出力する。
【００７４】
またＳ１１４で差Ｅ．ＨＡＴ（ｋ）が所定値を超えてオブザーバの収束状況が不良と判断されるときはＳ１２６に進んでオブザーバ（行列演算）の推定値（図１４に＃ｎＡ／Ｆで示す）を所定の値に固定する。所定の値は例えば、前回値＃ｎＡ／Ｆ（ｋ−１） （図示の簡略化のため、今回値にサブスクリプト（ｋ）を付すのは省略する）とする。更に、この推定値をパラメータとして格納している場合、差Ｅ．ＨＡＴ（ｋ）に応じてパラメータの更新を禁止する。
【００７５】
続いてＳ１２８に進んで気筒毎フィードバック補正項＃ｎＫＬＡＦを前回値＃ｎＫＬＡＦ（ｋ−１） とする。即ち、気筒との整合性からオブザーバ行列演算を間引くことなく、完全に中止し、オブザーバが停止される間は、停止直前の値に基づいて燃料噴射量を決定するようにした。これは、気筒間の空燃比のバラツキの度合いは本来的にそう変化するものではなく、気筒毎フィードバック補正項の値は集合部フィードバック補正項の値に比較すれば小さい値となり、１付近の値となるからである。そこで、オブザーバ推定値の信頼性が低い場合は、むしろ前回値を使用して誤った推定値を用いて却って制御性を悪化させることを回避した。
【００７６】
尚、Ｓ１０８でＬＡＦセンサ４０の活性化が完了していないと判断されるときはＳ１３０に進んで学習で求めた値＃ｎＫＬＡＦｓｔｙ を読み出し、Ｓ１３２に進んでその値を気筒毎フィードバック補正項＃ｎＫＬＡＦとした。即ち、気筒毎フィードバック補正項＃ｎＫＬＡＦの値をアイドル運転状態において、前回値（ｋ−１） との加重平均を
＃ｎＫＬＡＦｓｔｙ ＝Ｗ×＃ｎＫＬＡＦ＋（１−Ｗ）×＃ｎＫＬＡＦｓｔｙ（ｋ−１）
と求めて学習しておき、センサ活性化が完了していないときは最新の学習値を使用するようにした。ここで、＃ｎＫＬＡＦｓｔｙ ：最新学習値、Ｗ：重み係数、＃ｎＫＬＡＦｓｔｙ（ｋ−１）： 前回学習値、である。これによって気筒毎フィードバック補正項＃ｎＫＬＡＦに過去の燃焼履歴を一層良く反映させることができる。
【００７７】
続いてＳ１３４に進んで集合部フィードバック補正項ＫＬＡＦの値を１とし（集合部フィードバック制御の停止を意味する）、Ｓ１２２に進んで燃料噴射量＃ｎＴｏｕｔ を算出する。
【００７８】
また、Ｓ１０２でクランキングと判断されるときはＳ１３６に進んで水温センサ（図１で図示省略）を通じて検出した水温から所定の特性に従ってクランキング時の燃料噴射量Ｔｉｃｒ を算出し、Ｓ１３８に進んで始動モードの式（説明省略）に基づいて燃料噴射量Ｔｏｕｔ を決定する。
【００７９】
更に、Ｓ１０４でフューエル・カットと判断されるときはＳ１４０に進んで燃料噴射量Ｔｏｕｔ を零とし、Ｓ１４２に進んでＳ１２６と同様にオブザーバ（行列演算）の推定値を所定の値に固定し、Ｓ１４４に進んでＳ１２８と同様に気筒毎フィードバック補正項＃ｎＫＬＡＦを停止直前の値とする（これは学習値であっても良い）。尚、Ｓ１４２でオブザーバ（行列演算）推定値を所定の値に固定するのは、燃焼が行われず、正しい空燃比が計測できないためである。
【００８０】
この実施例は上記の如く構成したので、オブザーバの収束状況（推定精度）を簡易に判定することができ、推定値の信頼性が低い際に、誤った推定値による制御を行って却って制御性を悪化させることがない。
【００８１】
そしてオブザーバの収束状況が低下した場合には、燃料供給系の気筒間のバラツキは領域によって極端に変化することはないと思われることから、気筒毎フィードバック補正項＃ｎＫＬＡＦは停止直前に演算された値を用いるようにしたため、気筒間の空燃比のバラツキがかなりの程度まで吸収されて各気筒の空燃比を目標値に精度良く収束させることが可能となる。それによって、目標空燃比を理論空燃比とするときは、三元触媒２６の浄化率を向上させることができる。また、目標空燃比をリーン側に設定すれば、リーンバーン制御を精度良く実現することができる。
【００８２】
図１５はこの発明の第２実施例を示す、フロー・チャートの部分図である。第１実施例と相違する点に焦点をおいて説明すると、図３のＳ１１４に相当するＳ１１４ａにおいて差Ｅ．ＨＡＴ（ｋ）の移動平均値Ｅ．ＨＡＴｍｏｖ （ｋ）を求めて所定値と比較するようにした。
【００８３】
これにより、例えば一過性の原因から差Ｅ．ＨＡＴ（ｋ）が増加したときなども収束状況低下と判定することがなく、その結果、実際の運転状態によりマッチしてオブザーバの収束状況を判定することができる。残余の構成および効果は、第１実施例と異ならない。尚、移動平均を用いたが、移動加重平均、加重平均などを用いても良い。
【００８４】
図１６はこの発明の第３実施例を示す、フロー・チャートの部分図である。第２実施例と相違する点に焦点をおいて説明すると、図３のＳ１１４に相当するステップの後にＳ１１４ｂ，Ｓ１１４ｃ，Ｓ１１４ｄを追加し、差Ｅ．ＨＡＴ（ｋ）が所定値を超える回数をカウントし、カウント値が第２の所定値ＣＮＴＲＥＦ、例えば３回を超えるときオブザーバ収束状況低下と判断し、カウント値を０にリセットした後、Ｓ１２６に進むようにした。
【００８５】
これにより、第２実施例と同様に、例えば一過性の原因から差Ｅ．ＨＡＴ（ｋ）が増加したときなど収束状況低下と判定することがなく、その結果より実際の運転状態によりマッチしてオブザーバの収束状況を判定することができる。残余の構成および効果は、第１実施例と異ならない。
【００８６】
尚、上記において、気筒毎の空燃比の推定動作とそれに基づく空燃比フィードバック制御動作とを併せて示したが、この発明の要旨は第１の目的に限っては気筒毎の空燃比の推定動作、より正確には気筒毎の空燃比を推定するオブザーバの収束状況の判定にあり、空燃比フィードバック制御動作を必須とするものではない。
【００８７】
更には、空燃比センサとして広域空燃比センサを使用する場合を例にとって説明したが、いわゆるＯ_２ センサを用いる場合にも妥当する。
【００８８】
【発明の効果】
請求項１項にあっては、オブザーバの収束状況を簡易かつ正確に判定することができる。
【００８９】
請求項２項にあっては、オブザーバの収束状況を簡易かつ正確に判定することができると共に、それに応じてフィードバック補正項の変更を停止することができて制御値を誤ることがない。
【００９０】
請求項３項にあっては、一層容易かつ正確にオブザーバの収束状況を判定することができる。
【００９１】
請求項４項にあっては、簡易かつ正確にオブザーバの収束状況を判定することができる。特に平均値を用いるときは、一過性の原因によって収束状況が低下したときは収束状況の低下と判定しないので、実際の運転状態に良くマッチした空燃比推定、ないしはそれに基づく空燃比フィードバック制御を行うことができる。
【００９２】
請求項５項にあっては、一過性の原因によって収束状況が低下したときは収束状況の低下と判定しないので、実際の運転状態に良くマッチした空燃比推定、ないしはそれに基づく空燃比フィードバック制御を行うことができる。
【図面の簡単な説明】
【図１】この発明に係る内燃機関の気筒別空燃比推定装置を全体的に示すブロック図である。
【図２】図１中の制御ユニットの詳細を示すブロック図である。
【図３】この発明に係る内燃機関の気筒別空燃比推定装置の動作を示すフロー・チャートである。
【図４】この発明で前提とする空燃比センサの検出動作をモデル化した例を示すブロック図である。
【図５】図４に示すモデルを周期ΔＴで離散化したモデルである。
【図６】空燃比センサの検出挙動をモデル化した真の空燃比推定器を示すブロック線図である。
【図７】内燃機関の排気系の挙動を示すモデルを表すブロック線図である。
【図８】図７に示すモデルを用いて４気筒内燃機関について３気筒の空燃比を１４．７に、１気筒の空燃比を１２．０にして燃料を供給する場合を示すデータ図である。
【図９】図８に示す入力を与えたときの図７モデルの集合部の空燃比を表すデータ図である。
【図１０】図８に示す入力を与えたときの図７モデルの集合部の空燃比をＬＡＦセンサの応答遅れを考慮して表したデータと、同じ場合のＬＡＦセンサ出力の実測値を比較するグラフ図である。
【図１１】一般的なオブザーバの構成を示すブロック線図である。
【図１２】この発明で用いるオブザーバの構成を示すブロック線図である。
【図１３】図７に示すモデルと図１２に示すオブザーバを組み合わせた構成を示す説明ブロック図である。
【図１４】この発明で併用する空燃比の気筒別フィードバック制御を示すブロック図である。
【図１５】この発明の第２実施例を示す図３フロー・チャートの部分図である。
【図１６】この発明の第３実施例を示す図３フロー・チャートの部分図である。
【符号の説明】
１０ 内燃機関
１８ インテークマニホルド
２０ インジェクタ
２２ エキゾーストマニホルド
４０ 空燃比センサ（ＬＡＦセンサ）
４２ 制御ユニット[0001]
[Industrial applications]
The present invention relates to an exhaust system collecting part of a multi-cylinder internal combustion engine. To The convergence state of the observer in the cylinder-by-cylinder air-fuel ratio estimation device of the internal combustion engine, which is provided with an air-fuel ratio sensor, identifies the behavior of the exhaust system by a state equation, and estimates each cylinder air-fuel ratio (air-fuel ratio per cylinder) via the observer That is, the present invention relates to a method for simply and accurately determining the estimation accuracy of each cylinder air-fuel ratio.
[0002]
[Prior art]
It is common practice to provide an air-fuel ratio sensor in an exhaust system of an internal combustion engine to detect an air-fuel ratio, and an example thereof is a technology described in Japanese Patent Application Laid-Open No. S59-101562. Also, the present applicant has previously proposed in Japanese Patent Application No. 3-359339 (JP-A-5-180059) a technique for identifying the behavior of an exhaust system by a state equation and estimating the air-fuel ratio of each cylinder via an observer. are doing. In this case, the air-fuel ratio sensor is a wide-range air-fuel ratio sensor, that is, an O-type sensor whose output is inverted at the stoichiometric air-fuel ratio ₂ A sensor having an output characteristic proportional to the oxygen concentration in the exhaust gas before and after the stoichiometric air-fuel ratio is used instead of a sensor.
[0003]
[Problems to be solved by the invention]
Although the air-fuel ratio of each cylinder can be estimated by the above-described configuration, the estimation accuracy of the observer, in other words, the convergence state of the observer may not always be good depending on the operation state. When the reliability of such an observer estimated value is low, if control is performed using the estimated value, a situation may occur in which controllability deteriorates.
[0004]
Accordingly, an object of the present invention is to provide an exhaust system collecting section of a multi-cylinder internal combustion engine. To In the cylinder-by-cylinder air-fuel ratio estimating device of the internal combustion engine that provides an air-fuel ratio sensor, identifies the behavior of the exhaust system by a state equation, and estimates the air-fuel ratio of each cylinder via an observer, the observer converges, that is, It is an object of the present invention to provide a cylinder-by-cylinder air-fuel ratio estimating apparatus for easily and accurately determining the estimation accuracy of a fuel ratio.
[0005]
A second object is an exhaust system collecting section of a multi-cylinder internal combustion engine. To An air-fuel ratio sensor is provided, the behavior of the exhaust system is identified by a state equation, the air-fuel ratio of each cylinder is estimated via an observer, and the air-fuel ratio is determined based on the estimated cylinder air-fuel ratio using the fuel injection amount as an operation amount. In the cylinder-by-cylinder air-fuel ratio estimating apparatus of the internal combustion engine that performs feedback control, the convergence state of the observer, that is, the estimation accuracy of each cylinder air-fuel ratio is determined easily and accurately, and the possibility of changing the feedback correction term (gain) is determined accordingly. It is an object of the present invention to provide a cylinder-by-cylinder air-fuel ratio estimating apparatus for preventing a control value from being erroneously determined.
[0006]
[Means for Solving the Problems]
In order to solve the above-mentioned object, according to claim 1, an air-fuel ratio sensor is disposed in an exhaust system collecting part of a multi-cylinder internal combustion engine, and an air-fuel ratio of an input air-fuel mixture of each cylinder is estimated from an output thereof. The system is characterized in that the behavior of the exhaust system is identified by a state equation in which each cylinder air-fuel ratio is an input U (k), and a junction air-fuel ratio is an output Y (k), and the input U (k) is a predetermined value. An observer is constructed as a value, the output Y (k) is input, and each cylinder air-fuel ratio is estimated as a state variable X from a state equation by X hat (k). Estimation means Feedback control means for performing feedback control of the air-fuel ratio using the fuel injection amount as an operation amount based on the estimated cylinder air-fuel ratio Determining means for calculating an estimated value CX hat (k) of the collecting portion air-fuel ratio from the estimated state variable X hat (k) and comparing the calculated value with the output Y (k) to determine the convergence state of the observer Equipped The feedback control means stops changing the feedback correction term used to feedback-control the air-fuel ratio when the convergence state of the observer is determined to be poor. It was configured as follows.
[0007]
still, In the above, “the input U (k) is a predetermined value” means that, for example, the air-fuel ratio of the same cylinder is assumed to be the same as the air-fuel ratio before the previous combustion, or the air-fuel ratio of each cylinder is set as the target air-fuel ratio. It is.
[0010]
the above Claims to achieve the purpose of 2 In the paragraph, the output of the air-fuel ratio sensor is used as the output Y (k).
[0011]
the above Claims to achieve the purpose of 3 In the above item, the determination means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). Any one of HAT (k) and its average value is determined and compared with a predetermined value to determine the convergence state of the observer.
[0012]
the above Claims to achieve the purpose of 4 In the above item, the determination means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). HAT (k) and its average value are obtained, compared with a first predetermined value, counted the number of times exceeding the first predetermined value, and compared with the second predetermined value to determine the convergence state of the observer. It was configured so that
[0013]
[Action]
According to claim 1, from the estimated state variable X hat (k) Estimation of air-fuel ratio at the junction Since the apparatus is provided with a determination means for calculating the value CX hat (k) and comparing it with the output Y (k) to determine the convergence state of the observer, it is possible to easily and accurately determine the convergence state of the observer. . That is,

Then, since the value (Y (k) -CX hat (k)) indicates the convergence state of the observer, the convergence state of the observer can be easily and accurately determined from the value.
[0014]
Also Determining means for calculating an estimated value CX hat (k) of the collecting portion air-fuel ratio from the estimated state variable X hat (k), and comparing the calculated value with the output Y (k) to determine the convergence state of the observer. , The feedback control means , When it is determined that the convergence state of the observer is bad, Since the change of the feedback correction term (gain) is configured to be stopped, the convergence state of the observer can be easily and accurately determined, and the change of the feedback correction term can be stopped accordingly. There is no mistake.
[0015]
Claim 2 In the item, since the output of the air-fuel ratio sensor is used as the output Y (k), the convergence state of the observer can be more easily and accurately determined.
[0016]
Claim 3 In the above item, the determination means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). Since HAT (k) and its average value are obtained and compared with a predetermined value to determine the convergence state of the observer, it is possible to easily and accurately determine the convergence state of the observer. In particular, when the average value is used, when the convergence state is reduced due to a transient cause, it is not determined that the convergence state is reduced.Therefore, the air-fuel ratio estimation that is well matched to the actual operation state, or the air-fuel ratio feedback control based thereon is performed. It can be carried out.
[0017]
Claim 4 In the above item, the determination means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). HAT (k) and its average value are obtained, compared with a first predetermined value, counted the number of times exceeding the first predetermined value, and compared with the second predetermined value to determine the convergence state of the observer. When the convergence situation is reduced due to a transient cause, it is not determined that the convergence situation is reduced.Therefore, it is necessary to perform an air-fuel ratio estimation that is well matched to the actual operating state, or perform air-fuel ratio feedback control based on the estimation. Can be.
[0018]
【Example】
Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.
[0019]
FIG. 1 is a schematic diagram showing an overall cylinder-by-cylinder air-fuel ratio estimating apparatus according to the present invention. In the figure, reference numeral 10 denotes a four-cylinder internal combustion engine, and intake air introduced from an air cleaner 14 disposed at the tip of an intake passage 12 passes through an intake manifold 18 while its flow rate is adjusted by a throttle valve 16, and the first intake air passes through a first intake manifold 18. Or into the fourth cylinder. An injector 20 is provided near an intake valve (not shown) of each cylinder to inject fuel.
[0020]
The air-fuel mixture injected and integrated with the intake air is ignited by an ignition plug (not shown) in each cylinder, burns, and drives a piston (not shown). The exhaust gas after the combustion is discharged to an exhaust manifold 22 via an exhaust valve (not shown), and is discharged outside the engine through an exhaust pipe 24 while being purified by a three-way catalytic converter 26. The intake passage 12 is provided with a bypass passage 28 near the position where the throttle valve is disposed.
[0021]
A crank angle sensor 34 for detecting a crank angle position of a piston (not shown) is provided in a distributor (not shown) of the internal combustion engine 10, a throttle opening sensor 36 for detecting an opening of the throttle valve 16, An absolute pressure sensor 38 for detecting the intake pressure downstream of the throttle valve 16 as an absolute pressure is also provided. Further, in the exhaust system, a wide-range air-fuel ratio sensor 40 including an oxygen concentration detecting element is provided between the exhaust manifold 22 and the three-way catalytic converter 26, and outputs a value proportional to the oxygen concentration in the exhaust gas. Outputs of these sensors 34 and the like are sent to the control unit 42.
[0022]
FIG. 2 is a block diagram showing details of the control unit 42. The output of the wide-range air-fuel ratio sensor 40 is input to a detection circuit 46, where appropriate linearization processing is performed, and a linear characteristic proportional to the oxygen concentration in the exhaust gas in a wide range from lean to rich around the stoichiometric air-fuel ratio. The air-fuel ratio (A / F) is detected. The details are described in another application (Japanese Patent Application Laid-Open No. 4-369471) proposed by the present applicant, and further description is omitted. In the following description, this sensor will be referred to as a “LAF sensor” (linear EV sensor). The output of the detection circuit 46 is taken into a microcomputer including a CPU 50, a ROM 52, a RAM 54 and the like via an A / D conversion circuit 48, and stored in the RAM 54.
[0023]
Similarly, an analog output from the throttle opening sensor 36 and the like is input to the microcomputer via the level conversion circuit 56, the multiplexer 58, and the second A / D conversion circuit 60. After the output of the crank angle sensor 34 is shaped by a waveform shaping circuit 62, the output value is counted by a counter 64, and the count value is input into the microcomputer. In the microcomputer, the CPU 50 calculates a control value from the detected value according to a command stored in the ROM 52, drives the injector 20 of each cylinder via the drive circuit 66, and controls the solenoid valve 70 via the second drive circuit 68. To control the amount of secondary air passing through the bypass 28 shown in FIG.
[0024]
FIG. 3 is a flow chart for explaining the operation of the cylinder-by-cylinder air-fuel ratio estimating device according to the present invention. Since the technique for estimating the air-fuel ratio of each cylinder is premised, it will be briefly described for convenience of understanding.
[0025]
First, in order to accurately extract and extract the air-fuel ratio of each cylinder from the output of one LAF sensor, it is necessary to accurately clarify the detection response delay of the LAF sensor. Therefore, this delay is tentatively modeled as a first-order delay system, and a model as shown in FIG. 4 is created. Here, assuming that LAF: LAF sensor output and A / F: input A / F, the state equation can be expressed by the following Equation 1.
[0026]
(Equation 1)

[0027]
When this is discretized by the period ΔT, it becomes as shown in Expression 2. FIG. 5 is a block diagram of Equation 2.
[0028]
(Equation 2)

[0029]
Therefore, by using Equation 2, the true air-fuel ratio can be obtained from the sensor output. That is, if Equation 2 is modified, Equation 3 is obtained, so that the value at Time k-1 can be inversely calculated from the value at Time k as in Equation 4.
[0030]
(Equation 3)

[0031]
(Equation 4)

[0032]
Specifically, if Equation 2 is expressed as a transfer function using Z-transformation, Equation 5 is obtained. Therefore, the previous input air-fuel ratio is estimated in real time by multiplying the inverse transfer function by the current sensor output LAF. Can be. FIG. 6 shows a block diagram of the real-time A / F estimator.
[0033]
(Equation 5)

[0034]
Next, a method of separating and extracting the air-fuel ratio of each cylinder based on the true air-fuel ratio obtained as described above will be described. Considering the weighted average in consideration of the temporal contribution of the fuel ratio, the value at the time k was represented as in Expression 6. Note that, since F (fuel amount) is a control amount, “fuel / air ratio F / A” is used here, but in the following description, for convenience of understanding, “air / fuel ratio” will be used unless troublesome. Note that the air-fuel ratio (or the fuel-air ratio) means a true value obtained by correcting the response delay previously obtained by Expression 5.
[0035]
(Equation 6)

[0036]
That is, the air-fuel ratio of the collecting portion is represented by the sum of the past combustion history of each cylinder multiplied by a weight C (for example, 40% for the most recently burned cylinder, 30% before that, etc.). Here, C is an n × m matrix. This model is represented by a block diagram as shown in FIG.
[0037]
The equation of state is as shown in Equation 7.
[0038]
(Equation 7)

[0039]
When the air-fuel ratio of the collecting portion is set to y (k), the output equation can be expressed as shown in Expression 8.
[0040]
(Equation 8)

[0041]
In the above, since u (k) cannot be observed, x (k) cannot be observed even if an observer is designed from this equation of state. Therefore, if x (k + 1) = x (k−3) assuming that the air-fuel ratio before 4TDC (that is, the same cylinder) is in a steady operation state in which the air-fuel ratio does not suddenly change, Equation 9 is obtained. This is the same even when u (k) is set as the target air-fuel ratio.
[0042]
(Equation 9)

[0043]
Here, simulation results are shown for the model obtained as described above. FIG. 8 shows a case where the fuel is supplied to the four-cylinder internal combustion engine by setting the air-fuel ratio of three cylinders to 14.7 and setting only one cylinder to 12.0. FIG. 9 shows the air-fuel ratio of the collecting portion at that time obtained by the above model. Although a step-like output is obtained in the same figure, if the response delay of the LAF sensor is further taken into consideration, the sensor output has a waveform smoothed as shown in FIG. 10 as “model output value”. In the figure, “actual measurement value” is an actual measurement value of the output of the LAF sensor in the same case, and it is verified by comparison with this that the above model models the exhaust system of the multi-cylinder internal combustion engine well.
[0044]
Therefore, this results in a problem of a normal Kalman filter for observing x (k) using the state equation and the output equation shown in Expression 10. Solving the Riccati equation with the weight matrices Q and R as shown in Equation 11, the gain matrix K becomes as shown in Equation 12.
[0045]
(Equation 10)

[0046]
(Equation 11)

[0047]
(Equation 12)

[0048]
From this, the matrix A-KC is obtained as shown in Expression 13.
[0049]
(Equation 13)

[0050]
Although a general observer configuration is as shown in FIG. 11, since there is no input u (k) in this model, the configuration is such that only y (k) is input as shown in FIG. When expressed by a mathematical expression, it becomes as shown in Expression 14.
[0051]
[Equation 14]

[0052]
Here, an observer that receives y (k) as an input, that is, a system matrix of a Kalman filter is represented by Expression 15.
[0053]
[Equation 15]

[0054]
In this model, when the elements of the weight distribution R of the Riccati equation: the elements of Q = 1: 1, the system matrix S of the Kalman filter is given by Expression 16.
[0055]
(Equation 16)

[0056]
FIG. 13 shows a combination of the above-described model and observer. The simulation result is omitted since it is shown in the previous application, but the air-fuel ratio of each cylinder can be accurately extracted from the air-fuel ratio of the collecting section.
[0057]
The description will be made with reference to the flowchart of FIG.
[0058]
In the case of the embodiment, an example is shown in which, after estimating each cylinder air-fuel ratio, each cylinder air-fuel ratio is feedback-controlled to a target value based on the estimated value. That is, because the observer can estimate the air-fuel ratio of each cylinder from the air-fuel ratio of the collecting section, it is possible to control the air-fuel ratio for each cylinder using a control law such as PID.
[0059]
More specifically, as shown in FIG. 14, an aggregate feedback correction term (gain) KLAF is obtained from the LAF sensor output KACT indicating the aggregate air-fuel ratio and the target air-fuel ratio using the PID control law, and the observer estimated value # A feedback correction term (gain) #nKLAF (n: cylinder) for each cylinder is obtained from nA / F.
[0060]
More specifically, the feedback correction term #nKLAF for each cylinder is a target value and an observer estimated value #nA obtained by dividing the collective portion A / F by the last calculated value of the average value of the feedback correction term #nKLAF for each cylinder. / F is determined using the PID rule so as to eliminate the deviation from / F. As a result, the air-fuel ratio of each cylinder converges to the air-fuel ratio of the converging portion, and the air-fuel ratio of the converging portion converges to the target air-fuel ratio.
[0061]
Here, the fuel injection amount #nTout of each cylinder (defined by the valve opening time of the injector) is
# NTout = Tim × KCMD × KTOTAL × # nKLAF × KLAF
Is required. In the above, Tim: basic value, KCMD: target air-fuel ratio, and KTOTAL: other correction terms. There are other addition terms such as a battery correction term, but they are omitted. The details of such control are described in Japanese Patent Application No. Hei 6-33201 previously proposed by the present applicant, so that further description will be omitted.
[0062]
An embodiment will be described with reference to the flowchart of FIG. Note that this program determines the fuel injection amount of each cylinder at a predetermined crank angle from the TDC, that is, for each injection order (first, third, fourth, and second cylinders).
[0063]
First, in S100, the engine speed Ne, the intake pressure Pb, the LAF sensor output KACT (k) ((k) indicates a time in a discrete time system as described above) and the like are read. If the determination is negative, the routine proceeds to S104, in which it is determined whether or not fuel cut has been performed. If the result in S104 is negative, the program proceeds to S106, in which the aforementioned basic value Tim is searched. This is performed by searching a set map (characteristics not shown) based on the engine speed and the intake pressure.
[0064]
Then, the program proceeds to S108, in which it is determined whether the activation of the LAF sensor 40 has been completed. presume. Then, the process proceeds to S112, where the value E. HAT (k) is calculated, and the routine proceeds to S114, where the calculated value E. HAT (k) is compared with a predetermined value to determine the convergence state of the observer.
[0065]
Describing this, Expression 14 expressing the observer estimation function is rewritten as Expression 17.
[0066]
[Equation 17]

[0067]
In Expression 17, when attention is paid to y (k) -CX hat (k), it can be expressed as Expression 18.
[0068]
(Equation 18)

[0069]
That is, y (k) corresponds to the exhaust system air-fuel ratio, specifically the actual air-fuel ratio detected by the LAF sensor, represented by the output equation described above, and the CX hat (k) is estimated. By multiplying each of the cylinder air-fuel ratios (internal state variable X hat) by a matrix C indicating the degree of mixing of each cylinder exhaust gas, it corresponds to the estimated air-fuel ratio of the collecting portion. Therefore, y (k) -CX hat (k) represents the difference between the actual value (LAF sensor output KACT) and the estimated value of the air-fuel ratio at the collecting portion. HAT (k) can be regarded as an index representing the estimation accuracy of the observer, that is, the convergence state of the observer. The present invention has been made based on the above findings.
[0070]
Returning to the flowchart of FIG. 3, the difference E.S. HAT (k) is obtained, and the process proceeds to S114 and is compared with a predetermined value. The predetermined value is a value appropriate for determining the convergence state of the observer, for example, 0.1.
[0071]
The difference E.S. When it is determined that HAT (k) is equal to or less than the predetermined value, it is determined that the observer has a good convergence state, that is, it is determined that the intended estimation accuracy has been obtained. Compute as described for 14.
[0072]
Then, the process proceeds to S118, in which a learning value of the feedback correction term #nKLAF for each cylinder is calculated. Specifically, a moving average (or a weighted moving average) is obtained. The learning value thus obtained is stored according to the operating state, specifically, the fuel injection amount. That is, since the cylinder-by-cylinder feedback correction term #nKLAF indicates the characteristics of the injector 20, it is stored in pairs with the output fuel injection amount #nTout described below.
[0073]
Subsequently, the process proceeds to S120, where the collective feedback correction term KLAF is calculated as described with reference to FIG. 14, the process proceeds to S122 to determine the fuel injection amount #nTout for each cylinder, and the process proceeds to S124 to drive the injector 20 of the cylinder. Output to the circuit.
[0074]
The difference E.S. When HAT (k) exceeds a predetermined value and the convergence state of the observer is determined to be poor, the process proceeds to S126, where the estimated value of the observer (matrix operation) (indicated by # nA / F in FIG. 14) is set to a predetermined value. Fix it. The predetermined value is, for example, a previous value # nA / F (k-1) (for the sake of simplicity of illustration, the addition of the subscript (k) to the current value is omitted). Further, when this estimated value is stored as a parameter, the difference E.D. Updating of parameters according to HAT (k) is prohibited.
[0075]
Then, the process proceeds to S128, in which the cylinder-by-cylinder feedback correction term #nKLAF is set to the previous value #nKLAF (k-1). In other words, the observer matrix calculation is completely stopped without thinning out the observer matrix calculation from the consistency with the cylinder, and while the observer is stopped, the fuel injection amount is determined based on the value immediately before the stop. This is because the degree of variation in the air-fuel ratio between the cylinders does not inherently change so, and the value of the feedback correction term for each cylinder is smaller than the value of the collective feedback correction term, and is close to 1. This is because Therefore, when the reliability of the observer estimated value is low, it is rather avoided that the controllability is deteriorated by using an incorrect estimated value using the previous value.
[0076]
If it is determined in step S108 that the activation of the LAF sensor 40 has not been completed, the process proceeds to step S130 to read the value #nKLAFsty obtained by learning, and proceeds to step S132 to determine the value as the feedback correction term #nKLAF for each cylinder. did. That is, the weighted average of the value of the cylinder-by-cylinder feedback correction term #nKLAF and the previous value (k-1) in the idling operation state is calculated.
# NKLAFsty = W × # nKLAF + (1−W) × # nKLAFsty (k−1)
, And the latest learning value is used when the sensor activation is not completed. Here, #nKLAFsty: latest learning value, W: weighting factor, #nKLAFsty (k-1): last learning value. As a result, the past combustion history can be better reflected in the cylinder-by-cylinder feedback correction term #nKLAF.
[0077]
Subsequently, the process proceeds to S134, where the value of the collective feedback correction term KLAF is set to 1 (meaning that the collective feedback control is stopped), and the process proceeds to S122 to calculate the fuel injection amount #nTout.
[0078]
If it is determined in step S102 that cranking is to be performed, the process proceeds to step S136, where the fuel injection amount Ticr during cranking is calculated according to predetermined characteristics from the water temperature detected through a water temperature sensor (not shown in FIG. 1), and the process proceeds to step S138. The fuel injection amount Tout is determined on the basis of the start mode equation (the description is omitted).
[0079]
Further, when it is determined in S104 that the fuel is cut, the flow proceeds to S140 to set the fuel injection amount Tout to zero, and proceeds to S142 to fix the estimated value of the observer (matrix operation) to a predetermined value as in S126. Then, as in S128, the cylinder-by-cylinder feedback correction term #nKLAF is set to the value immediately before the stop (this may be a learning value). The reason that the estimated value of the observer (matrix operation) is fixed to a predetermined value in S142 is that combustion is not performed and a correct air-fuel ratio cannot be measured.
[0080]
Since this embodiment is configured as described above, it is possible to easily determine the convergence state (estimation accuracy) of the observer, and when the reliability of the estimated value is low, control by an erroneous estimated value is performed and controllability is rather reduced. Does not worsen.
[0081]
When the convergence state of the observer is reduced, the variation between the cylinders of the fuel supply system does not seem to change drastically depending on the region. Therefore, the feedback correction term #nKLAF for each cylinder is calculated immediately before the stop. Since the value is used, the variation of the air-fuel ratio between the cylinders is absorbed to a considerable extent, and the air-fuel ratio of each cylinder can be accurately converged to the target value. Thus, when the target air-fuel ratio is set to the stoichiometric air-fuel ratio, the purification rate of the three-way catalyst 26 can be improved. If the target air-fuel ratio is set to the lean side, lean burn control can be realized with high accuracy.
[0082]
FIG. 15 is a partial flowchart showing the second embodiment of the present invention. Focusing on the differences from the first embodiment, the difference E.S. in S114a corresponding to S114 in FIG. Moving average value of HAT (k) HATmov (k) was determined and compared with a predetermined value.
[0083]
Thus, for example, the difference E.P. Even when HAT (k) increases, it is not determined that the convergence state is reduced. As a result, the convergence state of the observer can be determined by matching the actual operation state. The remaining configuration and effects are not different from the first embodiment. Although the moving average is used, a moving weighted average, a weighted average, or the like may be used.
[0084]
FIG. 16 is a partial flow chart showing the third embodiment of the present invention. Explaining focusing on the differences from the second embodiment, S114b, S114c, and S114d are added after the step corresponding to S114 in FIG. The number of times that HAT (k) exceeds a predetermined value is counted, and when the count value exceeds a second predetermined value CNTREF, for example, three times, it is determined that the observer convergence state is low. I did it.
[0085]
Thus, similarly to the second embodiment, the difference E.D. The convergence state of the observer can be determined by matching the actual operation state based on the result without determining that the convergence state has decreased, such as when the HAT (k) increases. The remaining configuration and effects are not different from the first embodiment.
[0086]
In the above description, the air-fuel ratio estimation operation for each cylinder and the air-fuel ratio feedback control operation based thereon are shown together. However, the gist of the present invention is to limit the air-fuel ratio estimation operation for each cylinder only to the first object. More precisely, the determination of the convergence state of the observer for estimating the air-fuel ratio for each cylinder is performed, and the air-fuel ratio feedback control operation is not essential.
[0087]
Further, the case where the wide-range air-fuel ratio sensor is used as the air-fuel ratio sensor has been described as an example. ₂ This is also valid when using a sensor.
[0088]
【The invention's effect】
According to the first aspect, the convergence state of the observer can be easily and accurately determined.
[0089]
According to the second aspect, the convergence state of the observer can be easily and accurately determined, and the change of the feedback correction term can be stopped accordingly, so that the control value is not mistaken.
[0090]
According to the third aspect, the convergence state of the observer can be more easily and accurately determined.
[0091]
According to the fourth aspect, the convergence state of the observer can be determined easily and accurately. In particular, when the average value is used, when the convergence state is reduced due to a transient cause, it is not determined that the convergence state is reduced.Therefore, the air-fuel ratio estimation that is well matched to the actual operation state, or the air-fuel ratio feedback control based thereon is performed. It can be carried out.
[0092]
According to the fifth aspect, when the convergence state is reduced due to a transient cause, it is not determined that the convergence state is reduced. Therefore, the air-fuel ratio estimation that well matches the actual operation state, or the air-fuel ratio feedback control based on the estimation is performed. It can be performed.
[Brief description of the drawings]
FIG. 1 is a block diagram generally showing a cylinder-by-cylinder air-fuel ratio estimating apparatus according to the present invention.
FIG. 2 is a block diagram showing details of a control unit in FIG. 1;
FIG. 3 is a flowchart showing the operation of the cylinder-by-cylinder air-fuel ratio estimating apparatus according to the present invention.
FIG. 4 is a block diagram showing an example in which a detection operation of an air-fuel ratio sensor premised in the present invention is modeled.
FIG. 5 is a model obtained by discretizing the model shown in FIG. 4 with a period ΔT.
FIG. 6 is a block diagram showing a true air-fuel ratio estimator that models the detection behavior of an air-fuel ratio sensor.
FIG. 7 is a block diagram illustrating a model showing a behavior of an exhaust system of the internal combustion engine.
8 is a data diagram showing a case in which fuel is supplied with the air-fuel ratio of three cylinders set to 14.7 and the air-fuel ratio of one cylinder set to 12.0 for a four-cylinder internal combustion engine using the model shown in FIG. 7; .
9 is a data diagram showing the air-fuel ratio of the aggregate of the model of FIG. 7 when the input shown in FIG. 8 is given.
10 compares the air-fuel ratio of the collective part of the model of FIG. 7 when the input shown in FIG. 8 is given in consideration of the response delay of the LAF sensor, and the measured value of the LAF sensor output in the same case. FIG.
FIG. 11 is a block diagram showing a configuration of a general observer.
FIG. 12 is a block diagram showing a configuration of an observer used in the present invention.
13 is an explanatory block diagram showing a configuration in which the model shown in FIG. 7 and the observer shown in FIG. 12 are combined.
FIG. 14 is a block diagram showing cylinder-by-cylinder feedback control of the air-fuel ratio used in the present invention.
FIG. 15 is a partial view of the flow chart of FIG. 3 showing a second embodiment of the present invention.
FIG. 16 is a partial view of the flow chart of FIG. 3 showing a third embodiment of the present invention.
[Explanation of symbols]
10 Internal combustion engine
18 Intake manifold
20 Injector
22 Exhaust manifold
40 air-fuel ratio sensor (LAF sensor)
42 control unit

Claims (4)

An apparatus for arranging an air-fuel ratio sensor in an exhaust system collecting part of a multi-cylinder internal combustion engine and estimating an air-fuel ratio of an input air-fuel mixture of each cylinder from an output thereof,
a. The behavior of the exhaust system is identified by a state equation in which each cylinder air-fuel ratio is set as an input U (k), and an air-fuel ratio of a collecting section is set as an output Y (k), and an observer is constructed using the input U (k) as a predetermined value. Estimating means for receiving the output Y (k) as input, estimating each cylinder air-fuel ratio as a state variable X from a state equation by X hat (k),
And b. Feedback control means for performing feedback control of the air-fuel ratio using the fuel injection amount as an operation amount based on the estimated cylinder air-fuel ratio,
With,
c. Determining means for calculating an estimated value CX hat (k) of the air-fuel ratio at the collecting portion from the estimated state variable X hat (k) and comparing the calculated value with the output Y (k) to determine the convergence state of the observer;
Wherein the feedback control means, when the convergence status of observer is is determined to be defective, the internal combustion engine, characterized by stopping the change of the feedback correction term is used to feedback control the air-fuel ratio Air-fuel ratio estimation device for each cylinder. The output Y (k) is characterized by using the output of the air-fuel ratio sensor as claimed in claim 1 Koko cylinder air-fuel ratio estimation apparatus for an internal combustion engine according. The determining means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). The cylinder-by-cylinder air-fuel ratio estimating device for an internal combustion engine according to claim 1 or 2 , wherein any one of HAT (k) and its average value is obtained and compared with a predetermined value to determine the convergence state of the observer. The determining means determines the difference E.E. between the estimated value CX hat (k) of the collecting portion air-fuel ratio and the output Y (k). HAT (k) and its average value are obtained, compared with a first predetermined value, counted the number of times exceeding the first predetermined value, and compared with the second predetermined value to determine the convergence state of the observer. The cylinder-by-cylinder air-fuel ratio estimating device according to claim 1 or 2, wherein:

Images