Knock recognition based on vibration signal and Wiebe function in a heavy-duty spark ignited engine fueled with Methane Pierpaolo Napolitano b , Irina Jimenez a , Benjam´ ın Pla a , Carlo Beatrice b a CMT-Motores Termicos, Universitat Politecnica de Valencia, Camino de Vera s/n, E-46022 Valencia, Spain. b Consiglio Nazionale delle Ricerche, Piazzale Aldo Moro,7, Rome RM, Italy. Abstract Increasing demands on higher performance and lower fuel consumption and emissions have lead the path for internal combustion engine development; this race is nowadays directly related of CO2 emissions reduction. In spark-ignited (SI) engines, knock is one of the major barriers to achieve high thermal efficiency at high loads. The knocking risk is even higher in heavy-duty (HD) engines due to the size of the cylinders and to the low rotation speed. This paper proposes a knock detection strategy based on the combination of knock sensors and combustion modeling applied to a HD natural gas (NG) engine. The aim is to have a reliable, economic and computationally efficient algorithm to be implemented directly on the engine ECU. The method proposed has been applied to an extensive set of experimental data acquired on a SI NG heavy-duty engine. The results of the proposed knock estimation method are benchmarkt with those based on in-cylinder pressure analysis using piezoelectric transducers. The extension of the method based on in-cylinder pressure to a high displace- ment heavy-duty NG engine not only represents an innovation, but improves the knock recognition based on in-cylinder pressure compared with conventional methods as MAPO or IMAP. Besides, the development of an alternative method based on knock sensor signal, allows to obtain a higher or equal sensitivity compared to the traditional MAPO method based on in-cylinder pressure, with the advantage of only using knock sensors. Keywords: knock, heavy-duty engine, resonance, wiebe function 1. Introduction In recent years, diverse legislative measures have been implemented by international regulatory agencies in order to increase the development of alternative fuels in trans- port systems [1]. One of the most widely investigated al- ternative fuels found in literature is Natural Gas (NG) [2], which is a gas mixture consisting primary of methane, with smaller percentages of other gases such as ethane, propane, and butane. The recent sensibility in the reduction of CO2 emis- sions from thermal system has pushed the research to the use of alternative low carbon content fuels, and they dif- fusion to any possible combustion system [3]. The sector of heavy duty (HD) engine for road and off-road applica- tion was not excluded from this phenomena [4]. Since NG has a higher octane number than gasoline, it is possible to work with higher compression ratio in spark ignition (SI) engines [5, 6]. Nevertheless, using a gas instead of a liquid fuel involves the displacement of some air by NG, then leading to a reduction in the engine power output in port fuel injection cases [7]. In order to overcome this problem, two solutions can be found in literature: on the one hand, increasing the compression ratio, or in the other hand using lean combustion [8, 9]. However, the compres- sion ratio increase is limited by knock phenomena in SI engines due to higher combustion pressures and tempera- tures [9, 6], and as regards lean combustion, this has an operation limit, i.e over lean the mixture may lead to in- stability and misfire [10, 11]. Knock is an abnormal combustion phenomena in SI en- gines, related with the uncontrolled combustion of the end gas [12]. When knock occurs, a rapid combustion is ob- served due to the high local pressure, which produces shock waves that heavily excite the in-cylinder resonant modes [13, 14]. The engine exposition to knock during several cycles may lead to piston rings braking, piston melting, engine efficiency decrease and engine damage in general [15]. On this way, knock recognition techniques are impor- tant in order to achieve high thermal efficiency. These methods can be mainly classified in two principal groups: direct and indirect knock recognition methods [15]. The first group is based on the in-cylinder pressure measure- ment, which is directly influenced by the phenomena [13, Preprint submitted to Fuel October 18, 2021