BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a graph showing the transition between SI and HCCI combustion as internal EGR (a key engine parameter) is incrementally increased (as indicated on the horizontal axis) on a single-cylinder engine. On the left vertical axis is the coefficient of variance (COV) of the measured in-cylinder peak pressure. On the right vertical axis is the level of nitrogen oxide pollutants (NOx) in the engine exhaust. At low levels of EGR (left side of the plot), SI combustion occurs and the NOx emissions are high. At the highest levels of EGR, HCCI combustion occurs and NOx emissions are almost zero. At intermediate EGR levels between these limits, variable amounts of both HCCI and SI occur unstably over time, resulting in high COV and poor, erratic power output from the engine. NOx remains relatively low for the intermediate EGR zone.
FIG. 2 is another illustration of how combustion stability deteriorates at intermediate EGR levels (between full SI and HCCI). In this case, the vertical axis indicates the amount of heat released in combustion for individual cycles. For a given intermediate EGR level, the strength of successive combustion events can vary widely, producing a broad range of observed heat release values. These variations reflect varying degrees of both SI and HCCI in each cycle. Such behavior would of course be totally unacceptable performance for a realistic engine.
FIG. 3 includes return maps (from nonlinear theory) that reveal the deterministic patterns in the unstable combustion variations occurring at intermediate EGR levels. The presence of determinism demonstrates that it is possible to predict how the combustion will vary in the next cycle based on recent past history.
FIG. 4 includes graphs of symbol sequence histograms, which are another type of nonlinear tool for recognizing the deterministic combustion variations. Using such histograms, it is possible to precisely determine how far the SI-HCCI transition has progressed based on relative frequency of patterns in the combustion events.
FIG. 5 illustrates three different combustion sequences at an intermediate EGR level. The first two plots, (a) and (b), reflect undesirable behavior in which the combustion amplitude oscillates over very large values from one cycle to the next. In (c), there is a sequence of successive cycles in which the combustion stays close to its optimal value. With the proper control feedback, this optimal combustion magnitude indicates a potentially good control target.
FIG. 6 includes graphs of heat release rates corresponding to specific points on the integrated heat release return maps in the SI-HCCI transition. The sequence of highlighted combustion events represents one type of undesirable cycle-to-cycle combustion oscillation. Because heat release rate is resolved at each point along the combustion path of individual cycles, it reveals details about each combustion event not visible in the integrated values. By analyzing such information, it has been revealed that the combustion oscillations are due to time varying competition between SI and HCCI combustion mechanisms. This competition is responsible for the cycle-to-cycle deterministic patterns.
FIG. 7 is similar to FIG. 6 but reflects a different sequence of undesirable combustion events.
FIG. 8 illustrates the return maps and corresponding heat release rate profiles for a near-optimal sequence of combustion events. By maintaining the proper balance between SI and HCCI features through manipulation of combustion control variables such as spark timing, it should be possible to continue the near-optimal combustion over a much longer time.
FIG. 9 reveals the near-optimal combustion amplitude in the SI-HCCI transition region that could be stabilized by proper feedback control.
FIG. 10 illustrates how a simplified combustion rate function (referred to as Combustion Efficiency or CE) can be used to quantitatively describe the competition between SI and HCCI combustion mechanisms at intermediate EGR levels.
FIG. 11 illustrates how a simple deterministic model based on the CE function described above can produce behavior that closely mimics the experimental unstable combustion patterns in the SI-HCCI transition. FIG. 11a illustrates a return map predicted by such a model in the presence of stochastic noise (as would be the case in a real engine). FIG. 11b illustrates just the deterministic part of the model without the presence of noise.
FIG. 12
a and 12b illustrate how the deterministic pattern in combustion variations can be used to make predictions. The points indicated in blue are experimental integrated heat release measurements of combustion strength for several cycles at an intermediate EGR level in the SI-HCCI transition. The red points are predictions based on statistical patterns ‘learned’ and built into a simple model at a previous time for the same operating condition. Such models can be continuously updated and adapted as engine conditions change.
FIG. 13 illustrates how a return map produced by a simple statistical model (b) compares with experimental observations (a) for an intermediate EGR level.
FIG. 14 is a logic flow diagram of the software decision process.
Patent Application
20070181096
