To reduce emissions, CNG engines are commonly equipped with three-way catalytic converters. However when the engines run at transient conditions, the air fuel ratio can not be precisely controlled at theoretical value by traditional means, so the catalytic converters can not achieve their desired effect. This paper presents a new method for CNG engines to control air fuel ratio at transient conditions. The moments which intake and exhaust valves open are used as the trigger signals for ECU to collect the test data simultaneously. The dynamic information of CNG engine is detected by multiple sensors; the nonlinear coupling relationship between air fuel ratio of CNG engine and the operating conditions are established through information fusion and neural network control. The requirement of real time control for air fuel ratio is achieved, so the emissions of CNG engines are reduced further.