Study On Dynamic Characteristics Of The Sensors For Engines Air/Fuel Control System

Dynamic characteristics of the sensors are one of main technical matters required to be studied in automobile engine electronic control field. Thermal mass air flow sensor (MAF) and exhaust gas oxygen (EGO) sensor are key components in engine air-fuel ratio (AFR) feed-forward control and feedback control loop, and their nonlinear dynamic characteristics have direct influences on the measurement and control accuracy of instantaneous AFR.In order to design efficiently the anti-aliasing filter of signal conditioning electro-circuit in electronic control unit (ECU) and to choose adaptable signal processing method, and to study further the method to improve sensors dynamic performance and to reduce or eliminate dynamic measurement errors, the dynamic performance of engine MAF and EGO sensors under different operation condition were studied by establishing sensors' nonlinear dynamic models.Hammerstein models of hot wire type,hot film type MAF sensor and EGO sensor were established by a two-step identification method using static state and dynamic state calibration experiment data of the sensors. The time domain and frequency domain performance indices of above-mentioned sensors under different step excitation condition were calculated by means of the established models, and main characters of the two types of sensors were analyzed, respectively.During modeling, the delay behaviors of excitation signal source were considered sufficiently, the actual inputs acting on the sensor were estimated using different kinds of methods, and modeling accuracy was improved. According to control-oriented demands, the influences of model order on modeling accuracy were analyzed, and the appropriate model order of the sensors was determined.Asymptotic Method (ASYM) was used to build dynamic linear model of the MAF and the EGO sensors Hammerstein model, respectively. The model frequency errors were described exactly. Firstly, high order model of the sensors and measurement errors model were established. Then model order was reduced. Finally, the direct modeling method was compared with ASYM, and the modeling effects of different identification method were verified.