Neuro-Fuzzy Modeling of Diesel Fuel Consumption Under Dynamic Loads

Abstract

The article discusses the construction of a neuro-fuzzy model of diesel fuel consumption in various modes of its operation. The structure of a neuro-fuzzy network trained according to the results of field tests is presented. To train the neural network, a hybrid method was used in the form of an error back propagation algorithm and a least squares method. In the process of training, the parameters of the neural-fuzzy network were selected that provide the maximum error of an individual measurement of 3.5375% with an average error of the model of 0.4429%. The neuro-fuzzy diesel model makes it possible to simulate various modes of its operation in accordance with the test cycle ETC EK UN No. 49. Using well-known optimization methods from the obtained model of fuel consumption under unsteady loads, it is possible to determine diesel parameters close to optimal when calibrating the electronic control unit.