吸气式高超声速动力是低成本跨大气层飞行和全球快速运输的关键技术，高超声速进气道不起动/再起动对动力系统的设计和控制提出了挑战。不起动/再起动本质上是多变量影响的激波-边界层相互作用现象，经过数十年研究仍未发现统一规律、难以准确预测不起动/再起动边界。本文突破传统起动理论的一维、无黏假设，对二元进气道不起动/再起动现象展开系统化的二维、黏性数值模拟，利用量纲分析方法发现了进气道不起动/再起动相似律，并结合定量规律深入揭示了相关机理，形成了二元进气道不起动/再起动新理论。

  本文对简化内收缩通道（简化构型）的不起动/再起动特性展开了数值研究，通过绕唇罩前缘顺时针/逆时针转动唇罩实现不起动/再起动，得到了各种条件下的不起动/再起动边界。提出了无量纲唇罩长度和等效内收缩比概念，建立了不起动/再起动相似律。对于给定的来流马赫数，进气道不起动/再起动特性可由唇罩角度、无量纲唇罩长度和等效收缩比三者中的两个独立变量来描述。在此基础上，得到了不起动/再起动边界的自相似经验公式，可同时描述唇罩角度、唇罩长度、唇口高度、入口边界层厚度（雷诺数）和来流马赫数的影响。本文发现，Kantrowitz极限接近于极端短唇罩或大唇罩角度条件下的不起动边界，而等熵极限接近于极端长唇罩或小唇罩角度条件下的不起动边界。Kantrowitz极限预测的再起动边界不具有普遍性。

  根据不起动边界分布和流动结构，将简化构型的不起动分为三类。短唇罩、过渡型和长唇罩不起动的临界状态分别显示单个分离泡、扁平分离区和靠近喉道的小分离区。短唇罩不起动伴随分离泡持续增大及其导致的内流道壅塞。过渡型不起动由分离区尺寸在流向压差驱动下不断增大导致，在唇口溢流发生前无壅塞。长唇罩不起动表现为喉道处小分离区向上游发展的过程，不起动过程中出现类激波串结构和流动壅塞。短唇罩和过渡型不起动导致稳定的终态流场，而长唇罩不起动导致振荡的流场。

  简化构型的再起动也可分为三类。短唇罩再起动临界状态下，单个大分离泡停留在唇口，这是唇口激波系与唇罩尾缘膨胀波竞争的结果。过渡型再起动对应再起动唇罩角度的平台，临界状态下无唇口溢流，分离泡由唇口激波系单独维持，与喉道高度或内收缩比无关。长唇罩再起动临界流场显示微弱振荡的长分离区，这是唇口激波系与几何喉道流量限制互补作用的结果，且后者占主导地位。将简化构型的不起动/再起动迟滞现象分为四类，指出起动/不起动流场的分离区维持机制不同是导致双解现象的根本原因。

  在简化构型基础上，考虑实际二元进气道构型，对具有膨胀转角构型的内收缩通道（实用构型）进一步展开数值研究。通过平移唇罩减小/增大几何喉道，以实现不起动/再起动。发现了实用构型的不起动/再起动相似律，并提出了适用于大收缩角不起动的替代性收缩比概念。不起动数值结果表明，在小收缩角下，肩部膨胀转角相比简化构型对于不起动有较弱的抑制作用。中等收缩角下，膨胀转角反而促进了不起动。大收缩角下，膨胀转角对于不起动产生显著的抑制作用，使得不起动边界偏离Kantrowitz极限。实用构型的不起动分类体系与简化构型大致相同，区别在于实用构型的短唇罩不起动过程中无壅塞。再起动数值结果表明，在6°~11°范围内，膨胀转角对于再起动的促进作用强于简化构型唇罩尾缘的膨胀波，但是膨胀转角的作用随着收缩角增大而减弱。在12°以上，实用构型的再起动性能较差。实用构型的再起动均属于短唇罩再起动，具体可细分为两种亚型。

  对于简化构型和实用构型，考察了不起动/再起动临界状态的壁面压力分布，提出了基于临界状态压力峰值的无量纲不起动/再起动压力界限。并将其与无量纲唇罩长度、临界状态分离区特征尺度关联。不起动压力界限反映内收缩通道的最大压缩能力。考虑到压缩能力和起动性能之间的矛盾，提出了基于等效收缩比和替代性收缩比的自相似性压缩效率概念，以衡量单位无黏流道面积的收缩带来的压力升高。

  为了评估实用流道构型对不起动/再起动性能的影响，并扩展相似律的应用范围，对具有圆弧/折线肩部转折和折线唇罩的五类修正构型开展了系统化的不起动/再起动数值研究。发现当肩部/唇罩构型导致相对于基准构型的不起动模式转变时，能够有效抑制不起动。圆弧/折线肩部仅在大收缩角时具有不起动抑制作用，在各种压缩角下对再起动特性影响较小。具有较长唇口弯折段的折线唇罩及其组合构型能够显著改善不起动/再起动特性。

   Hypersonic airbreathing propulsion is a key technology for fast global transport and low-cost trans-atmospheric flight. The unstart/restart of hypersonic inlets brings challenges to the design and control of propulsion system. The unstart/restart is essentially a shockwave/boundary-layer interaction phenomenon affected by multiple variables. The universal laws of unstart/restart remain unclear after decades of investigation. To correct the traditional one-dimensional inviscid starting theory, the inlet unstart/restart was studied by two-dimensional viscous numerical simulation. The similarity laws of unstart/restart were observed, and the mechanisms were revealed. This dissertation established a relatively holonomic theory framework for two-dimensional inlets unstart/restart.

   The unstart/restart characteristics of simplified contraction ducts were studied firstly. The cowl plate was rotated clockwise/anti-clockwise around the cowl lip to induce unstart/restart, and the unstart/restart boundaries under different conditions were obtained. The concepts of dimensionless cowl length and effective contraction ratio were proposed, and the unstart/restart similarity laws were developed. For a given Mach number, the unstart/restart characteristics can be described with any two variables in cowl angle, dimensionless cowl length, and effective area ratio. The empirical equations of unstart/restart boundary were obtained based on the similarity laws. The influences of cowl angle, cowl length, cowl height, thickness of entrance boundary layer (Reynolds number), and cowl lip Mach number were described simultaneously in a self-similar formulation. It is found that Kantrowitz limit is close to the effective unstart area ratio for short cowl or large cowl angle, and the isentropic limit is the other extreme for long cowl or small cowl angle, while Kantrowitz limit is not generalizable for restart boundaries.

   The unstarts of the simplified ducts are classified into three types according to the unstart boundary distribution and flow structure. The unstart critical state of the short-cowl, transitional and long-cowl unstart shows a single separation bubble, a flat separation region, and a near-throat small separation respectively. The short-cowl unstart occurs as the separation bubble enlarges continuously and thereby chokes the internal flow duct. The transitional unstart is caused by the separation region growth driven by the increasing streamwise pressure difference, and there is no flow choke before the cowl spillage occurs. The long-cowl unstart happens as the small separation region at the geometric throat develops upstream. The separation region shows “shock train” structure, and flow choke occurs. The short-cowl and transitional unstart result in stable final flow fields, while the long-cowl unstart results in an oscillatory flow field.

   The rsatarts of the simplified ducts can also be classified into three types. The short-cowl restart critical state shows a massive separation bubble near the cowl entrance, which is a result of the competition between the cowl shock system and the expansion waves from the cowl tail. The transitional restart (corresponding to the plateau of restart cowl angle) shows a massive separation bubble without cowl spillage at the critical state. The separation bubble is sustained by the cowl shock system alone and independent of the throat height or internal contraction ratio. The long-cowl restart critical state shows a long separation region with slight oscillation, which is a result of the complementation between the cowl shock system and the mass flow constraint of the geometric throat, and the former is less significant. The unstart/restart hysteresis is classified into four types. The different sustaining mechanism of the separation region in the started/unstarted flow fields is the fundamental cause of the dual-solution phenomena.

   Based on the similarity laws of the simplified ducts, the unstart/restart of the practical contraction ducts with an expansion corner at the shoulder were numerically studied. The unstart/restart was induced by translating the cowl horizontally and decreasing/increasing the height of the geometric throat. The similarity laws for practical configurations were observed and the alternative contraction ratio suitable for practical configurations was proposed. According to the numerical results of unstarts, the expansion corner has weak suppression effect under small contraction angles. For medium contraction angles, the expansion corner facilitates the unstart instead. For large contraction angles, the unstart is suppressed significantly by the expansion corner, which makes the unstart boundary deviate from Kantrowitz limit. The classification scheme of the practical ducts is roughly the same as the simplified ducts, the only difference is that there is no flow choke during the short-cowl unstart process of the practical ducts. The numerical results of the restarts indicates that the expansion corner facilitates the restart more significantly than the expansion waves from the cowl tail of the simplified configurations, but the expansion corner effect becomes weaker as the contraction angle increases. For the contraction angles over 12°, the restart ability of the practical ducts is worse than the simplified ducts. The restarts of the practical ducts can be classified into two subtypes, both of which belong to the short-cowl restart.

   The wall pressure distributions at the unstart/restart critical state were examined for both simplified ducts and practical ducts. The dimensionless unstart/restart pressure criteria based on the wall pressure peak were proposed and were correlated with the dimensionless cowl length and critical separation scale. The unstart pressure criteria reflects the maximum compression ability of the contraction ducts. Considering the contradiction between the compression and starting ability, the self-similar compression efficiency based on the effective/alternative contraction ratio was proposed, which can weigh the pressure rise caused by unit reduction of the flow area.

   Systematic unstart/restart numerical simulations on five modified configurations were conducted to assese the influence of the geometric factors in practical inlet design and extend the similarity law. The unstart is suppressed effectively as the flow path design causes the conversion of unstart mode relative to the baseline configuration. The arc/secant shoulder hinders the unstart only at large contraction angles, while it has weak effect on the restart for various contraction angles. The bended cowl (with a relatively long bending section) and the coupled configurations based on it can improve the unstart/restart characteristics significantly.