The development of a control system able to detect the combustion phase within the engine cycle, would enable the engine efficiency (or the engine torque) to be maximized. Indeed, it is well known that the torque output of an engine is strongly correlated to the combustion phase. Many parameters can be defined to represent the combustion phase: the most commonly used are the Pressure Peak Position (PPP) and the Center of Combustion (CA50), which is defined as the crankshaft angle where the 50% of fuel mass is burned. The combustion phase and the previously mentioned efficiency-related parameters, are influenced by spark advance (for spark ignition engines), and by the start of main injection (for compressed ignition engines). Therefore, knowing these quantities would enable the implementation of a closed-loop combustion control system. Figure 1.a and 1.b show some experimental results, for the gasoline engine used in this work, in which the normalized Indicated Mean Effective Pressure (IMEP) is represented versus PPP and CASO. Figure 1. IMEP versus PPP (a) and IMEP versus CASO (b). Spark ignition engine. Table 1. Technical data of the gasoline engine Figure 2. Operating points tested on the gasoline engine. Figure 3. Sensors’ configuration for the gasoline and diesel engines. 3. Preliminary analysis of the vibration signal As the vibration measurement was not available, a Bruel&Kjaer 4393 accelerometer has been installed on the engine block, between cylinders 2 and 3. The in-cylinder pressure and vibration signals have been recorded at high frequency (simultaneously sampled at 100 kHz), whereas the most important ECU variables at 200 Hz. Figure 4. Coherence function between in-cylinder pressure and vibration trace for the gasoline engine, at 2000 rpm. Figure 5. Coherence function between in-cylinder pressure and vibration trace for the diesel engine, at 1500 rpm. Figure 6. (a) Comparison between in-cylinder pressure and accelerometer signal; (b) Comparison between in-cylinder pressure first derivative and accelerometer signal. Figure 6.a shows the comparison between cylinder pressure and vibration trace. The accelerometer signal has been low pass filtered, in order to isolate only the harmonic components useful for the identification of the combustion parameter. The correlation between the two waveforms is hard to tell, but if the pressure signal first derivative is used instead of the pressure trace itself (Figure 6.b), a clearer information can be extracted. The similarity between the acceleration signal and the pressure first derivative could be explained by assuming that a vibration that propagates through the engine block can be caused by the force impulses generated by the pressure gradient variation. Hence, it may be concluded that the accelerometer signal is particularly sensitive to pressure gradient variations. Figure 7. (a) Measured and estimated PPP; (b) Measured and estimated PPP during a spark advance sweep. Table 3. Correlation factor between PPP calculated via in-cylinder pressure signal and PPP estimated via acceleration signal, for cylinder 2. Table 4. Correlation factor between PPP calculated via in-cylinder pressure and PPP estimated via acceleration signal, for cylinder 3 The pressure peak position estimation methodology has been applied on each operating point listed in Figure 2 and the experimental results, in terms of linear correlation, for the cylinders 2 and 3, are shown in Table 3 and Table 4. The correlation coefficients have been calculated using Pearson's linear correlation, expressed by Equation (2): Figure 9 shows the comparison between measured and estimated pressure peak, during the same test (1500 rpn and 4 bar BMEP). The time delay between in-cylinder pressure and vibration signals may be easily compensated also in this case. The RMSE for PPP detection is 0.21 deg. Figure 8. Comparison between in-cylinder pressure and accelerometer signals. Figure 9. Measured PPP from in-cylinder pressure signal, and estimated from acceleration signal, for a test run at 1500 rpm and 4 bar BMEP. By deeply analyzing the in-cylinder pressure signal, accurate information about the combustion process can be extracted. For example: it is possible to identify the Start of Combustion of the Pre and Main injection, by analyzing the Rate of Heat Release (ROHR) calculated using the equation (3). compared to the Rate of Heat Release First Derivative (ROHRD), as shown in Figure 11, resulting in a much greate1 similarity between the two signals. Figure 10. Comparison between RoHR and structure-borne sound, after time delay compensation. Figure 11. Comparison between RoHR first derivative and acceleration signal, after time delay compensation.
