A Strategy Based on the Architecture ANFIS (Adaptive Neuro-Fuzzy Inference System) for Calibration of Internal Combustion Engine

Nowadays the necessity of diminish the processing time is searched incessantly in the industry in general, what it is not different for the automotive industry. The test of internal combustion engines (ICE) adds in general, a significant cost for the automobiles manufacturers due to difficulties found during the calibration of the engines[1][2]. Regarding the higher costs that involves all the process, we consider the use of Neuro-Fuzzy systems to facilitate the calibration, to diminish the running time and to avoid the using of expensive equipments by means of the introduction of a new strategy for controlling of air-fuel mixture. This strategy can describe the non linear characteristics of the ICE and adds a level of adaptability to the system, through the adjustment of its parameters by functional data of the engine[1]. The controller's system is adjusted to supply the adequate amount of fuel in each operational condition of the engine, providing better dynamic performance and increasing efficiency. In this article a strategy based on architecture ANFIS (Adaptative Neuro-Fuzzy Inference System) will be described, as well as the definition of rules for input variables of the system by the application of a simple and efficient methodology of training of the polynomial output, allowing the ICE to operate with the ideal mixture for each point of operation, with fast convergence and without a high computational cost.