A Non-Linear Regression Technique to Estimate from Vibrational Engine Data the Instantaneous In-Cylinder Pressure Peak and Related Angular Position

In this paper, a downsized twin-cylinder turbocharged spark-ignition engine is experimentally investigated at test-bench in order to verify the potential to estimate the peak pressure value and the related crank angle position, based on vibrational data acquired by an accelerometer sensor. Purpose of the activity is to provide the ECU of additional information to establish a closed-loop control of the spark timing, on a cycle-by-cycle basis. In this way, an optimal combustion phasing can be more properly accomplished in each engine operating condition. Engine behavior is firstly characterized in terms of average thermodynamic and performance parameters and cycle-by-cycle variations (CCVs) at high-load operation. In particular, both a spark advance and an A/F ratio sweep are actuated. In-cylinder pressure data are acquired by pressure sensors flush-mounted within the combustion chamber of both cylinders. The Coefficient of Variation of the net Indicated Mean Effective Pressure (CoVIMEP) and of in-cylinder peak pressure (CoVp,max) are utilized to quantify the cyclic dispersion and identify its dependency on combustion phasing and duration. Vibrational data are provided by a non-intrusive accelerometer sensor located on the head of cylinder #1. In particular, a proper processing of the accelerometer signal is applied to build correlations able to estimate with a relevant accuracy the cycle-by-cycle scattering of the crank angle position and amplitude of the in-cylinder pressure peak, as well as the related CoV. A maximum in-cylinder pressure error below 2 bar and a maximum crank angle error below 1 degree was obtained in most of data points