Control devices for internal combustion engines

Abstract

An object of the present invention is to update the learning values ​​of a large number of grid points by one learning operation, and to easily adjust the speed and efficiency of learning in a wide learning area. The engine has an ECU that performs engine control using various control parameters. The ECU has a learning map in which learned values ​​of control parameters are stored, and executes weighted learning control of the learned values. In the weighted learning control, each time a control parameter is acquired, a weight wkij that decreases as the distance from the position of the acquired value zk of the control parameter to the grid point is set for each grid point of the learning map. Then, the learned value Zij(k) of all grid points is updated based on the acquired value zk of the control parameter and the weight wkij. Thereby, all learned values ​​can be efficiently updated by one learning operation.

Description

technical field [0001] The present invention relates to a control device for an internal combustion engine provided with a learning map of control parameters. Background technique [0002] As prior art, there is known a control device for an internal combustion engine including a learning map of control parameters as disclosed in, for example, Patent Document 1 (JP-A-2009-046988). Learning values ​​for correcting control parameters are stored in each grid point of the learning map. In the prior art, when a control parameter to be learned is acquired, four grid points located around the acquired value are selected on the learning map, and the values ​​of these four grid points are updated. learning value. In this learning control, after weighting the obtained value of the control parameter, it is reflected on the learning value of the surrounding grid points. , the greater the weight. [0003] In addition, as documents related to the present invention, the applicants are ...

AI Technical Summary

Problems solved by technology

However, in the prior art, the number of learning values ​​updated by one learning action is limited to four, and learning values ​​are not updated at grid points far from the obtained values ​​of the control parameters, so there is a problem of low learning efficiency.

Images