Application Research On Knock Detection And Control Of Natural Gas Engine

Natural gas is an attractive alternative fuel because of its clean and efficient combustion.And some worldwide problems about oil shortage and environmental pollution could be alleviated by using it. Thereby natural gas has a good application prospect. Natural gas engine has been widely used as the propulsion of marine power plant. In order to pursue a better fuel economy, the compression ratio is raised continually and lean burn technology is adopted by the marine engine manufacturing companies, as a result, the possibility of engine knock increases. However, intense engine knock will worsen the in-cylinder combustion condition and damage the important components of engine,causing the engine life reduced. However,studies showed that slight knock improves the dynamic performance and fuel economy of engine. Hence, many researchers focus on finding knock threshold accurately and controlling the engine to operate under slight knock condition.In this thesis, the research objective is an electronically controlled multi-point injection natural gas engine converted from a marine diesel engine. The corresponding knock signal detection system is established via detecting the engine block vibration. In order to find the knock threshold and its feature frequency band,the engine vibration signals are analyzed and processed with the method of short time Fourier transform and discrete wavelet transform.This feature frequency band is used as the band pass filter frequency input for the knock hardware processing chip HIP9011. Reasonable knock criterion index and knock intensity criterion are selected based on the integral of HIP9011 output voltage. Besides, the software and hardware for knock closed-loop control system are developed successfully based on the real-time feedback of chip integral, the real-time feedback of ignition advance angle and the throttle angle.As the knock detection experiments show, both the short-time Fourier transform and discrete wavelet transform can realize the generation time of knock. Short-time Fourier transform can obtain the knock feature frequency band efficiently whose frequency is about 2.5kHz. And discrete wavelet transform can efficiently monitor the knock intensity. In the knock control experiment, by calculating the average integral voltage amplitude of 100 cycles nearby the knock threshold, the knock threshold Vp is identified as 0.69, and the intervals of voltage amplitudes in different knock intensity are determined. In the control experiment for knock real-time closed-loop, when knock occurs in the current cycle, the engine will operate in non-knock or slight knock condition in the next cycle and in slight condition in the later cycle by the proposed controller .Therefore, effective knock control can be realized.