为满足飞行器的大范围推力需求，超声速燃烧室需工作在不同的燃烧模态。而已有研究表明壁面压力受当量比历史调节方向的影响存在突变和迟滞现象，且发生在亚燃/超燃模态转换阶段，但还欠缺从真实燃烧/流动特征的角度深刻认识这些现象。此外，上述现象可能导致推力的突变与迟滞，进而不利于飞行器的准确控制，但模态转换是否导致推力突变尚有争议。

  本文利用直连燃烧试验平台和数值仿真方法，拟对乙烯燃料在凹腔上游喷注的超声速燃烧室中的出现上述现象和推力变化特性开展系统研究。从三维燃烧/流动结构与演变的角度，掌握了不同稳焰模式的稳焰和振荡机制，深入分析了两种燃烧模态转换和相应迟滞现象的物理机制。并澄清了传统认识的偏差，推力的迟滞和突变现象可发生在超燃模态，且该现象与不同稳焰模式的转换关联起来更恰当。此外，针对迟滞环内点，分别通过试验和仿真揭示了当量比初值耦合当量比降低速率，以及点火器功率对稳焰模式的影响规律。

  首先通过定工况的试验和数值研究，从三维燃烧/流动结构的角度，探寻火焰稳定和燃烧振荡的物理机制。发现凹腔剪切层稳焰模式基本无振荡，且可由任意展向的二维分布表征其燃烧/流动特点。射流尾迹稳焰模式则显现出更多三维特征，由于拐角分离效应，侧壁附近有更大的低速区和更多释热，并衍生出两种稳焰机制：第一种紧靠喷孔上游存在分离，可卷吸燃料到上游掺混燃烧，因此侧壁附近喷孔上游的低速区常驻高温燃气，其稳焰效果与凹腔类似；第二种喷孔上游的分离和喷孔较远，无法向上游卷吸燃料，整体稳焰更依靠凹腔，但侧壁附近仍有低速区，因此火焰偶尔可传播到凹腔上游。上述两种稳焰机制均存在燃烧振荡，第一种的振荡来源于侧壁附近低速区的热-声耦合不稳定性。第二种除了上述振荡外，部分工况还伴随着火焰展向传播、火焰与激波相互作用、激波/边界层干扰以及拐角分离效应多因素耦合导致的非声振荡。

  试验和仿真通过台阶式当量比调节方式均捕获到一类燃烧迟滞现象，归因于凹腔剪切层稳焰模式与射流尾迹稳焰模式的转换，并伴随着激波串消失和重建所引起的突变和迟滞。期间流动分离和释热分布的突变导致沿程静压的突变，体现为推力性能的明显突变。并且相比于历史当量比降低路径，在当量比增加路径上产生的推力突变程度更大。这是由于从凹腔剪切层稳焰模式转换为射流尾迹稳焰模式，当量比需增加到足够大，才能形成激波并诱导流动分离。但反过来，由于现有的激波和流动分离会提升燃烧效率当量比需降到足够低才能触发转换。模态转换发生前燃烧/流动结构和推力连续变化，因而当量比增加路径下模态转换的临界当量比更高，也就意味着其推力突变更大。

  进一步研究发现，来流条件的不同不仅会影响上述燃烧迟滞的当量比区间，以及推力的突变程度，甚至在来流马赫数3.0条件下还同时捕捉到第二类燃烧迟滞现象。与第一类迟滞不同，第二类迟滞区间完全处于射流尾迹稳焰模式，发现其与激波-火焰的强/弱切换相关，并伴随着激波串强度的突变和迟滞现象。第二类模态转换期间的流动分离、释热分布和沿程静压变化幅度明显减小，因而推力并未出现明显突变。

  针对处于第一类迟滞环的目标当量比，试验发现若从很高的初始当量比快速降低到该目标当量比过程中，则由于流动分离减弱过快，燃料来不及掺混燃烧，到一定程度，射流尾迹稳焰模式会转换为凹腔剪切层稳焰模式。如果当量比缓慢下降或者初始当量比较低，则流动分离减弱较慢，燃料仍来得及掺混和燃烧，不存在燃烧效率的突降，可避免出现上述模态转换。

  依旧针对第一类迟滞环内工况，以高温燃气发生器（HGG）作为点火器，借助仿真手段发现不同于小功率点火下的凹腔稳焰模式，大功率点火可使剪切层火焰产生足够高压升，进而转换为射流尾迹稳焰模式，该结论可推广到其他类型点火装置。在凹腔内安装HGG对两类稳焰模式影响都很小。而在凹腔上游安装HGG，与凹腔协同能够更好地促进燃烧和稳焰，在射流尾迹稳焰模式，甚至可能引起类似第二类迟滞环中的激波-火焰强/弱切换，进而改变燃烧振荡特性。

   试验和仿真均发现亚燃模态与有分离的超燃模态之间切换不一定导致燃烧迟滞，且沿程静压和马赫数随时间连续变化。此外，从燃烧/流动特征角度，这两种模态均处于射流尾迹稳焰模式，且等价于燃烧区高压、高推力的强燃烧模态。因此，上述两种模态没有明显界限，可归为有分离的燃烧模态。相对地，纯超燃模态可称为无分离的燃烧模态，等价于燃烧区低压、低推力的弱燃烧模态，也等价于凹腔剪切层稳焰模式。

        Supersonic combustor should operate in different combustion modes for wide-range aircraft thrust requirements. Former researchers demonstrated that combustor wall pressure showed catastrophe and hysteresis phenomena depending on historical equivalence ratio (ER) variation directions. They argued these occurred in ramjet/scramjet mode transition region, but lacked in-depth data depicting combustion/flow characteristics of these phenomena. Meanwhile, the above phenomena might cause thrust catastrophe and hysteresis, which is not conducive to precise control of aircrafts. But whether mode transition would cause thrust catastrophe is still controversial.

        The current work utilized direct-connect combustion test facility and numerical simulation method, and investigated the above phenomena and thrust variation characteristics systematically in a supersonic combustor with transverse ethylene fuel injection upstream cavity flameholders. From the viewpoint of 3-D combustion/flow structures and evolutions, flame stabilization and oscillation mechanisms in different flame stabilization modes were elucidated, physical mechanisms of two different mode transitions and relevant combustion hystereses were further analyzed. Meanwhile, different from traditional knowledge, thrust hysteresis and catastrophe phenomena could occur in scramjet mode, and they were more appropriately related to transition between different flame stabilization modes. Additionally, for certain objective condition in hysteresis loop, the effect of ER initial value coupled with ER decreasing rate, and the effect of ignition power on the flame stabilization mode were revealed by experiments and simulations, respectively.

        At first, direct-connect combustion tests and numerical simulations were performed under typical inflow conditions and certain equivalence ratio (ER) conditions, to explore physical mechanism of flame stabilization and combustion oscillation from the viewpoint of 3-D combustion/flow structures. It was found that the cavity shear-layer stabilized mode had negligible oscillation, and it could be shown by any spanwise 2-D combustion/flow distribution. Relatively, the jet-wake stabilized mode showed obvious 3-D distribution. Due to corner separation effect, there was larger low-speed region and more heat release near sidewall, and two slightly different flame stabilization mechanisms existed. For the first, there were flow separation upstream nearby the fuel injectors, and could bring fuel upstream for mix and combustion, and thus the upstream high-temperature low-speed region could promote flame stabilization as the cavity. For the second, the upstream flow separation was far away from the fuel injectors, which could not bring fuel upstream, and thus flame stabilization mainly depended on the cavity. But there was still low-speed region near sidewall, and thus flame could occasionally propagate upstream the cavity. The above two flame stabilization mechanisms both had combustion oscillations. In the first, oscillations were mainly driven by thermo-acoustic instabilities in the low-speed region nearby the sidewall, which were also usual in the second. Besides, there was non-acoustic oscillation in certain case of the second mechanism, which was driven by the coupling of spanwise flame propagation, flame/shock interaction, shock/boundary-layer interaction and corner separation effect.

        By step-by-step ER regulation method, a kind of combustion hysteresis was captured both experimentally and numerically, and attributed to transitions between the cavity shear-layer stabilized and the jet-wake stabilized mode, along with catastrophe and hysteresis of the shock train’s establishment and vanishment. During either transition, wall pressure distribution would change suddenly because of catastrophe of flow separation and heat release distribution, and result in obvious thrust catastrophe. Meanwhile, compared to the transition in historical ER decreasing path, historical ER increasing would result in larger thrust catastrophe, and the reason was as follows. For the transition from the cavity shear-layer stabilized to the jet-wake stabilized mode, ER should increase high enough to generate shock and flow separation. But for the inverse transition, the initial flow separation could enhance combustion, thus ER should decrease lower enough to trigger the transition. Combustion/flow structures and thrust varied continuously before either of the above transitions, and thus the critical mode transition ER was higher in the historical ER increasing path, signifying larger thrust catastrophe.

        Further study showed that inflow conditions would influence ER ranges and thrust catastrophe degrees of the above hysteresis. Furthermore, under Mach 3.0 inflow condition, a second type of combustion hysteresis was simultaneously captured. Different from the first hysteresis, the second was totally in the jet-wake stabilized mode, and it was related to the flame-shock intensive/weak interaction mode transitions, along with catastrophe and hysteresis of the shock train intensity. During either mode transition, the separation, heat release and wall pressure all changed slightly, and thus thrust showed negligible catastrophe.

For an objective ER condition in the first hysteresis, it was experimentally observed that if the ER decrease very quickly from high enough inital level to the objective level, then due to fast decrease of flow separation, the fuel would not have enough time for mix and combustion. To some extent, flame would transit from the jet-wake stabilized to the cavity shear-layer stabilized mode. Otherwise, if the initial ER was not so high or the ER decreased more slowly, then flow separation decreased more slowly, and fuel would have enough time for mix and combustion, then the combustion efficiency would not change suddenly with no mode transition.

        Still for an objective ER condition in the first hysteresis, a high-temperature gas generator (HGG) was used as an ignitor, it was found that low-power ignition resulted in the cavity shear-layer stabilized mode, but high enough ignition power would produce high enough pressure in the cavity shear-layer stabilized mode, and trigger mode transition to the jet-wake stabilized mode. This conclusion can be generalized to other ignition devices. Additionally, HGG inside the cavity showed slight influence in both flame stabilization modes. But if the HGG was configured upstream the cavity, then it could better cooperate with cavity for promoting combustion and flame stabilization. In the jet-wake stabilized mode, the upstream HGG even could induce the flame-shock weak/intensive interaction mode transition which similarly occurred in the second hysteresis, and thus changed the combustion oscillation characteristics.

        It was observed experimentally and numerically that ramjet mode/separated scramjet mode transition would not necessarily cause combustion hysteresis, and streamwise pressure and Mach number distributions showed continuous change. Additionally, from the viewpoint of combustion/flow structures, the above two modes were both in the jet-wake stabilized mode, which was equivalent to the intensive combustion mode characterized by combustion-zone high-pressure and high thrust. Therefore, these two modes had no clear distinction, and could be classified as the separated combustion mode. Relatively, the shock-free scramjet mode could be called as the non-separated combustion mode, which was also equivalent to the weak combustion mode characterized by combustion-zone low-pressure and low thrust, and also equivalent to the cavity shear-layer stabilized mode.