Cycle-to-cycle variation analysis of early flame propagation in engine cylinder using proper orthogonal decomposition

https://doi.org/10.1016/j.expthermflusci.2014.06.017Get rights and content

Highlights

  • Cross-correlation of flame pattern is developed to compute the velocity fields.

  • Weak spiral flow pattern within the flame was only found in higher POD modes.

  • Flame speed and POD coefficients are well correlated with pressure-derived results.

Abstract

Experimental investigations on the cycle-to-cycle variations of early flame propagation were conducted at two intake air swirl ratio of 0.55 and 5.68 in a single-cylinder spark-ignition direct-injection (SIDI) optical engine. Both the crank-angle resolved flame images through the quartz insert in the piston and the in-cylinder pressure were simultaneously recorded for 250 consecutive cycles. An algorithm based on the cross-correlation of flame pattern was implemented to compute the two-dimensional velocity fields representing an early flame propagation. Afterwards, the proper orthogonal decomposition (POD) was performed to access the cycle-to-cycle variations of flame propagation. During early flame formation, the uncertainty of velocity was evaluated to be less than 5%. The velocity fields of early flame propagation were significantly affected by in-cylinder air flow. Increasing the swirl ratio from 0.55 to 5.68 augmented the average early flame propagation speed from 2.1 m/s to 3.2 m/s, and the COV of flame speed was reduced from 49.7% to 25.1%. Good correlations were identified between early flame speed and the pressure-derived results. Similar magnitude of reduction in COV (coefficient of variation) of IMEP (indicated mean efficient pressure) & PP was also found. The coefficients of POD mode 1 were found to represent the flame speed, and the remaining higher POD modes were related to the fluctuation in kinetic energy. The POD coefficients showed good correlations with the pressure-derived results such as CA05 (5% of total heat release) & PP (peak pressure). The weak spiral flow pattern, observed within the flame, was only found in higher POD modes. In summary, using the coefficients and modes of POD, the average and fluctuating parts of early flame propagation can be resolved, thereby providing more quantitative information of the cycle-to-cycle variation during the early combustion process in an engine cylinder.

Introduction

Strong cycle-to-cycle variations of engine flows often prohibit SIDI engines from reaching their full potential of efficient and clean combustion. It is because the variations of air flow, fuel and temperature distributions in the vicinity of spark plug prior to ignition all affect the early flame formation, propagation and the subsequent combustion processes [1], [2], [3], [4]. Therefore, it is essential to investigate the flame propagation at early stage under realistic engine operating conditions. While in-cylinder pressure sensor is widely adopted to collect combustion data, pressure-based data can only provide very limited information when the flame begins at early stage. With the rapid development of high speed imaging hardware [5], [6], taking combustion images inside an engine has evolved as a powerful technique to visualize the early flame formation.

The flame propagation speed, which is defined as the propagation rate of the flame through the air–fuel mixture, is a key combustion characteristic of early flame behavior. Previous studies [7], [8] quantify the flame growth speed by taking the temporal derivative of flame radius. However, in their studies, only a single speed value was extracted from two sequential flame images, and no information about the velocity field within the flame structure was obtained. In addition, strong cycle-to-cycle variations must also be quantified to provide a complete spatial and temporal description of flame propagation. Since the calculations of flame speed based on flame sizes are inadequate to reveal the spatial information, a novel technique capable of resolving the velocity fields of flame propagating and providing quantitative cycle-to-cycle information is needed.

The current objective is to investigate the cycle-to-cycle variations of velocity fields within the flame pattern propagation at early stage using the proper orthogonal decomposition (POD) technique. POD technique has been widely applied by researchers to study the variations in engine turbulent flows [9], [10], [11], quantitative comparisons between the velocity fields of simulations and experiments [12], [13], [14], misfire analysis [15] and more recently on the variations of scalar fields in engine research [16], [17], [18].

In this study, the simultaneous high speed combustion imaging and in-cylinder pressure recording under different intake air swirl conditions are performed in an optical SIDI engine. An algorithm based on the cross-correlation of flame patterns is implemented to compute the velocity fields of flame propagation. Then, POD is conducted on the velocity fields, and the coefficients and modes of POD are used to resolve the average and fluctuating parts of early flame velocity field. It is believed that more quantitative information of the cycle-to-cycle variation at early combustion process will be revealed by this approach. Other POD-based studies on fuel spray development [19] and in-cylinder air flow evolution and variations [20] have been carried out using the same engine. Furthermore, it has been concluded from a previous study [2] that the flame speed increased nearly twice for high swirl compared with that of low swirl, and the flame location was much more repeatable under high swirl condition.

Section snippets

Imaging setup and operating parameters

An optical SIDI engine was utilized in this investigation. Fig. 1 depicts the experimental apparatus. Optical access into combustion chamber was achieved through two pent-roof windows and a quartz-insert piston. A spark plug and eight-hole fuel injector were centrally installed in the cylinder head. A pressure transducer (Kistler 6125A) was mounted to acquire in-cylinder pressure measurements. Heat release analysis of in-cylinder pressure was processed to obtain the engine combustion metrics

Procedure for proper orthogonal decomposition analysis

The POD analysis procedure is briefly described in this section. A practical guide for using POD can be found in previous papers [22], [23]. POD conducts an optimal linear decomposition of a set of velocity fields, V(k), k = 1, 2, …, K, creates an orthonormal spatial basis functions (POD modes% φm, m = 1, 2, …, M) and coefficients cm(k).V(k)=m=1Mcm(k)φm

The POD code in previous paper [22] is employed here. The code mathematically minimizes the following:k=1kV(k)-m=1Mcm(k)φm2min

The minimization

Flow structure of early flame propagation

Fig. 4 depicts the velocity fields of early flame propagation at low swirl condition which is overlapped on greyscale images of the flame. It can be seen that the velocity within the flame is smaller than that on the flame front. This is expected since the motion within the flame is only due to convection, while at the front it is the combination of convection and burning velocity. While cycle #52 shows a clockwise flame propagation, cycle #162 depicts a totally different pattern with the flame

Conclusions

High-speed images and in-cylinder pressure were concurrently recorded during early flame propagation in a single-cylinder SIDI optical engine under two extreme intake air swirl conditions. For each condition, test data of 250 engine cycles was obtained in order to achieve a reasonable cycle-to-cycle variation analysis. A robust cross-correlation technique was implemented to compute the projected two-dimensional velocity fields of early flame propagation. The accuracy of velocity measurements

Acknowledgements

This research is sponsored by GM R&D and NSFC under Grants Nos. 51076093/E060702 and 51176115/E060404, and carried out at the National Engineering Laboratory for Automotive Electronic Control Technology of the SJTU.

References (23)

  • W. Zeng, C.A. Idicheria, T.D. Fansler, M.C. Drake, Conditional analysis of enhanced combustion luminosity imaging in a...
  • Cited by (100)

    View all citing articles on Scopus
    View full text