JP2015222059A - Internal combustion engine control unit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an internal combustion engine control unit capable of learning an abnormal combustion occurrence probability per operating region.SOLUTION: An internal combustion engine control unit detects knocking using a cylinder internal pressure sensor. The internal combustion engine control unit includes a learning map that has ignition timing and an intake air quantity as map axes, that includes a plurality of grid points, and in which learned values of a knocking occurrence probability correspond to the respective grid points in an updatable manner. The control unit learns the knocking occurrence probability using the learning map during operation of an internal combustion engine. This learning is weighting learning for updating the learned values of learning target grid points using a weight win response to a distance between the learning target grid points and learning data on the learning map. Using a variable qindicating whether knocking occurs by 1 or 0 as the learning data, weighted averaging is performed by adding the weight wto the variable q, thereby calculating the learned values of the knocking occurrence probability.

Description

  The present invention relates to a control device for an internal combustion engine.

  Conventionally, for example, Patent Document 1 discloses a control device for an internal combustion engine. This conventional control device determines whether or not pre-ignition is likely to occur based on a combustion speed parameter correlated with the combustion speed.

JP 2013-104323 A Japanese Patent Laying-Open No. 2005-084834 JP 2004-308450 A

  If the above-mentioned pre-ignition, and further, the probability of occurrence of abnormal combustion such as knocking and misfire can be learned by mapping for each operating region of the internal combustion engine (for example, the region defined by the ignition timing and the intake air amount), Engine control (for example, ignition timing control) using the acquired occurrence probability can be suitably performed. However, the occurrence probability of abnormal combustion is a statistic and is not obtained as an instantaneous value. For this reason, if the driving region changes during acquisition of the occurrence probability, there is a problem that correct statistics cannot be obtained and an appropriate map cannot be obtained for each driving region.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a control device for an internal combustion engine in which the occurrence probability of abnormal combustion can be learned for each operation region.

A first invention is a control device for an internal combustion engine,
Abnormal combustion detecting means for detecting abnormal combustion;
A learning map that includes a region defining parameter that defines the operating region of the internal combustion engine as a map axis, has a plurality of lattice points, and the learning value of the occurrence probability of abnormal combustion is associated with each of the plurality of lattice points so as to be updatable. When,
Abnormal combustion learning means for learning the occurrence probability of abnormal combustion using the learning map during operation of the internal combustion engine;
With
In learning of the learning map, when learning data of the occurrence probability of abnormal combustion is acquired, the weight of the learning data is set to be larger as the distance between the learning target lattice point on the learning map and the learning data is closer In addition, every time learning data of the probability of occurrence of abnormal combustion is acquired, the learning value of the learning target lattice point is set so that the learning data is more greatly reflected in the learning value as the weight is increased at the learning target lattice point. Weighted learning to update,
The abnormal combustion learning means calculates a learning value of the occurrence probability of abnormal combustion by using as a learning data a variable indicating the presence or absence of occurrence of abnormal combustion as 1 and 0 and assigning the weight to the variable and performing a weighted average. It is characterized by doing.

  According to the first invention, even when the operation region changes during acquisition of the occurrence probability of abnormal combustion, the occurrence probability of abnormal combustion that is essentially a statistic can be learned for each operation region. . Then, a learning map that enables learning of the occurrence probability of abnormal combustion for each operation region can be acquired.

It is a figure showing a knock probability map, a knock intensity map, and a map of a knock limit line obtained by combining these maps.

Embodiment 1 FIG.
[System configuration of internal combustion engine]
The internal combustion engine of this embodiment is a spark ignition type internal combustion engine. The internal combustion engine includes an intake passage for taking intake air into each cylinder. In the intake passage, an electronically controlled throttle valve is provided to adjust the intake air amount. Each cylinder of the internal combustion engine is provided with a fuel injection valve for injecting fuel into the intake port and an ignition plug for igniting the air-fuel mixture.

  Next, a control system for the internal combustion engine will be described. The system according to the present embodiment includes a sensor system including various sensors (partially exemplified below) necessary for operation of the internal combustion engine, and an ECU (Electronic Control Unit) that controls the operation state of the internal combustion engine. Yes. First, the sensor system will be described. The crank angle sensor outputs a signal synchronized with the rotation of the crankshaft, and the air flow meter measures the intake air amount. The in-cylinder pressure sensor detects the in-cylinder pressure.

  The ECU is configured by an arithmetic processing unit that includes a storage circuit including a ROM, a RAM, a nonvolatile memory, and the like, and an input / output port. A learning map (to be described later) is stored in the nonvolatile memory of the ECU. Each sensor of the sensor system is connected to the input side of the ECU. An actuator such as an ignition device having a throttle valve, a fuel injection valve, and an ignition plug is connected to the output side of the ECU. And ECU drives each actuator based on the operation information of the internal combustion engine detected by the sensor system, and performs operation control.

[Map learning performed in Embodiment 1]
According to the in-cylinder pressure sensor described above, the in-cylinder pressure waveform during one cycle of the internal combustion engine can be acquired. When knocking occurs, a high-frequency vibration waveform is detected in the in-cylinder pressure waveform. By using this vibration waveform, it is possible to calculate the presence / absence of knocking and the knock intensity. That is, it is possible to determine whether or not knocking has occurred based on whether or not the vibration waveform is generated. Further, the knock intensity can be calculated based on the amplitude of the vibration waveform. It should be noted that the determination of the presence or absence of occurrence of knock and the calculation of the knock intensity are performed using, for example, a method using a knock sensor as an abnormal combustion detection means in addition to a method using an in-cylinder pressure sensor. Also good.

  FIG. 1 is a diagram showing a knock probability map, a knock intensity map, and a knock limit line map obtained by combining these maps. In the present embodiment, the knock occurrence probability map shown in FIG. 1A and the knock intensity map shown in FIG. 1B are created with a learning process during operation of the internal combustion engine 10 (that is, onboard). The Based on the knock occurrence probability and the knock intensity obtained from these maps, a map of the knock limit line (thick broken line) where the knock occurrence probability and the knock intensity fall within the predetermined limits is shown in FIG. C). Each of these maps uses region defining parameters (in this example, ignition timing and intake air amount) that define the operating region of the internal combustion engine 10 as map axes.

  In the knock occurrence probability map shown in FIG. 1 (A), the driving region is divided into four according to the level of the knock occurrence probability. That is, the “no knock” region in FIG. 1A indicates an operation region in which the knock occurrence probability is zero, and the “knock probability low” region has a knock occurrence probability greater than zero and is greater than the first predetermined value. Is a region where the knock occurrence probability is greater than the first predetermined value and smaller than the second predetermined value, and the “high knock probability” region is the first knock occurrence probability. 2 shows an area that is a predetermined value or more. Similarly, in the knock intensity map shown in FIG. 1B, the operation region is divided into four according to the level of the knock intensity.

  According to the map of the knock limit line shown in FIG. 1C, it is possible to obtain the knock limit ignition timing that does not exceed the knock limit line under the current intake air amount. Therefore, by using this map and controlling the ignition timing under the current intake air amount so as not to advance beyond the knock limit ignition timing, the knock occurrence probability and the knock intensity become a predetermined level or more. Knock generation can be suppressed.

ただし、上記（１）式において、ｎ_ｐはノックの発生回数であり、ｎはノック発生確率Ｐの算出のためのデータの取得回数（言い換えれば、ノック発生確率の算出対象となるエンジンサイクル数）である。 (Learning of knock occurrence probability map)
The system of the present embodiment has a main feature in the learning method of the knock occurrence probability map. Here, generally, the knock occurrence probability P can be determined according to the following equation (1).

However, in the above equation (1), n _p is the number of occurrences of knock, and n is the number of data acquisitions for calculating the knock occurrence probability P (in other words, the number of engine cycles for which the knock occurrence probability is to be calculated). It is.

  The knock occurrence probability P generally used as described above is a statistic, and is not obtained as an instantaneous value. For this reason, if the operating region changes during acquisition of the knock occurrence probability, correct statistics cannot be obtained, and there is a problem that an appropriate map cannot be obtained for each operating region. Therefore, in the present embodiment, the knock occurrence probability P (n) is calculated with weight for learning the knock occurrence probability map. Unlike the knock occurrence probability, the knock strength is a value that can be calculated from the in-cylinder pressure data during one cycle. Therefore, the other knock intensity map can be learned by using a known weighting learning method (for example, the method described in International Patent Application No. 2014/002189).

The knock occurrence probability map includes the ignition timing and intake air amount (region defining parameter) as map axes, has a plurality of lattice points, and the learning value of the occurrence probability of abnormal combustion is associated with each lattice point so that it can be updated. Learning map. Then, learning of this learning map is performed when the learning data of the occurrence probability of abnormal combustion is acquired, the closer the distance between the learning target lattice point and the learning data on the learning map (on the operation region), the more the learning data Each time the weight w _k is set to be large and the learning data of the occurrence probability of abnormal combustion is acquired, the learning data is more greatly reflected in the learning value as the weight w _k is larger at the learning target lattice point. This is weighted learning for updating the learning value of the learning target lattice point. In addition, in this embodiment, a variable q _k that indicates the presence or absence of occurrence of knock as 1 and 0 is used as learning data. Then, the weighting average is performed by assigning the weight w _k to the variable q _k , whereby the occurrence probability P (n) of abnormal combustion is calculated as the learning value of the learning target lattice point.

Specifically, the occurrence probability P (n) of abnormal combustion can be expressed as the following equation (2) using the variable q _k and the weight w _k . The numerator of equation (2) is the sum of products of n variables q _k and weights w _k , and the denominator of equation (2) is the sum of weights w _k for n times of data acquisition. When the numerator and denominator of the formula (2) are expressed by a recurrence formula, the numerator N (n) and the denominator M (n) when the number of data acquisition times is the kth are respectively the following (3) and (4). It becomes like the formula. The expressions N (1) and M (1) are for defining an initial value (value when k = 1). The setting of the weight w _k according to the distance between the lattice point and the learning data is preferably performed using a Gaussian function as disclosed in the above-mentioned document (International Publication No. 2014/002189), for example. Further, not only a Gaussian function but also a linear function or a trigonometric function can be used.

For learning of the knock occurrence probability map, every time learning data (variable q _k ) is acquired (that is, every engine cycle for which the knock occurrence probability P is calculated), abnormal combustion of each lattice point of the map is performed. The learning value of the occurrence probability P (n) may be calculated using the above equations (2) to (4).

As described above, by calculating the knock occurrence probability P with the weight using the weight w _k , the distance of the learning data (variable q _k ) to the lattice point on the operation region of the knock occurrence probability map It is possible to reflect the difference in the calculation of the knock occurrence probability P. In other words, according to this method, when calculating a knock generation probability P in some grid points, rather than simply increasing by one the knock occurrence count n _P when knocking occurs, the grid point and knock The degree to which the variable q _k is reflected in the knock occurrence probability P is changed according to the distance from the variable q _k indicating the presence or absence of the variable (that is, according to the magnitude of the weight w _k ).

  As described above, according to the present method, the calculation of the knock occurrence probability P with the weight in consideration of the position of the generated knock on the operation region enables learning of the grid points in each operation region. It becomes like this. For this reason, even when the driving region changes during acquisition of the knock occurrence probability, the knock occurrence probability, which is essentially a statistic, can be learned for each driving region. Then, as shown in FIG. 1A, a knock occurrence probability map that enables learning of the knock occurrence probability P for each operation region can be acquired.

  By the way, in the above-described first embodiment, the knock generation probability for learning of the knock occurrence probability map is described with weighting as an example of knock that is one aspect of abnormal combustion. However, the above method may be used for a learning map that learns the occurrence probability of other forms of abnormal combustion such as pre-ignition other than knock or misfire. Specifically, the presence / absence of pre-ignition can also be detected using, for example, an in-cylinder pressure sensor. For this reason, the said method is applicable also in the case of pre-ignition. In addition, when misfire occurs, the amount of heat generated by combustion in the cylinder becomes zero, and therefore whether or not misfire occurs for each cycle can be detected using, for example, an in-cylinder pressure sensor. Further, in the case of misfire, it is preferable to apply the above method to a misfire occurrence probability map having the ignition timing, the intake air amount, and the fuel injection amount as map axes. Then, it is preferable to control the ignition timing and the fuel injection amount using the misfire occurrence probability map so that the misfire occurrence probability is equal to or less than a predetermined threshold.

Claims (1)

Abnormal combustion detecting means for detecting abnormal combustion;
A learning map that includes a region defining parameter that defines the operating region of the internal combustion engine as a map axis, has a plurality of lattice points, and the learning value of the occurrence probability of abnormal combustion is associated with each of the plurality of lattice points so as to be updatable. When,
Abnormal combustion learning means for learning the occurrence probability of abnormal combustion using the learning map during operation of the internal combustion engine;
With
In learning of the learning map, when learning data of the occurrence probability of abnormal combustion is acquired, the weight of the learning data is set to be larger as the distance between the learning target lattice point on the learning map and the learning data is closer In addition, every time learning data of the probability of occurrence of abnormal combustion is acquired, the learning value of the learning target lattice point is set so that the learning data is more greatly reflected in the learning value as the weight is increased at the learning target lattice point. Weighted learning to update,
The abnormal combustion learning means calculates a learning value of the occurrence probability of abnormal combustion by using as a learning data a variable indicating the presence or absence of occurrence of abnormal combustion as 1 and 0 and assigning the weight to the variable and performing a weighted average. A control device for an internal combustion engine.

Images