摘要:
激波管是生成强爆炸冲击波环境的理想实验平台。激波管能够产生不同强度的冲击波,通过在等径激波管驱动段内加入不同漏空率的带孔挡板,可以使得驱动段中产生几乎连续的稀疏波系,实现冲击波衰减历程调控,同时延长冲击波正压作用时间。挡板设计参数多、设计复杂,难以通过简单的“试凑”方式,实现对任意峰值压力和正压作用时间冲击波的快速精确模拟。为此,采用正交试验方法,分析驱动段挡板参数对冲击波波形的影响;并以正交试验结果为指导,设计反向传播(Back Propagation, BP)神经网络的训练样本,建立快速准确预测冲击波峰值压力、正压作用时间和压力衰减历程的代理模型;采用基于代理模型的随机搜索方法,实现样本范围内冲击波任意设计指标到激波管驱动参数的反演设计。研究结果表明,驱动段内加入挡板后,冲击波的正压作用时间显著提升;采用“V”型挡板排列方式并适当减小挡板间距,可以有效提高模拟冲击波与标准指数衰减冲击波的相似度;训练得到的BP神经网络在测试集上具有良好的预测性能,压力峰值、正压作用时间的平均绝对百分比误差分别为0.07%和0.95%,压力衰减历程的平均相似度系数为0.997;寻优算法可以在35 s内给出驱动参数优化方案,生成超压、动压衰减历程与标准指数衰减冲击波的相似度系数均超过0.95的目标冲击波,验证了基于驱动段挡板设计的等径驱动激波管调控冲击波波形的可行性。
Abstract:
The shock tube is an ideal experimental platform for generating strong blast shock waves. By incorporating perforated baffles with different porosities into the driver section of a constant-diameter shock tube, nearly continuous rarefaction wave trains can be generated in the driver section, enabling the regulation of the shock wave attenuation history while extending the positive pressure duration. The design of baffles involves numerous parameters and is highly complex, making it difficult to achieve rapid and accurate simulation of shock waves with arbitrary peak pressures and positive pressure durations through simple trial-and-error methods. A parameterized driver-section baffle configuration was first constructed for a constant-diameter shock tube. Baffle annular width, installation position, baffle thickness and baffle spacing were treated as design variables. An orthogonal experimental design was employed to analyze the influence of driver-section baffle design variables on the shock wave profiles. Guided by the orthogonal experimental results, sample set for a back propagation (BP) neural network was designed and divided into training, validation, and test subsets. The trained BP network acted as a surrogate model to rapidly and accurately predict the peak overpressure, positive pressure duration, and pressure decay history of the shock wave. On this basis, a surrogate-model-based random search method was developed for the inverse design of driver-section baffles. The inverse design objective was to minimize the discrepancy between the predicted and target shock-wave indicators, including peak overpressure, positive pressure duration, and pressure decay-history similarity, subject to the driver parameters being within the sample range. Orthogonal test results indicate that increasing the number of baffles can prolong the positive pressure duration, and adopting a V-shaped baffle arrangement combined with an appropriate reduction in baffle spacing effectively improves the similarity between the simulated shock wave and the standard exponentially decaying shock wave. The trained BP neural network exhibits excellent predictive performance on the test set, with mean absolute percentage errors of 0.07% for peak overpressure and 0.95% for positive pressure duration, and an average similarity coefficient of 0.997 for the pressure decay history. The optimization algorithm can provide a driver parameter optimization scheme within 35 seconds, generating target shock waves with similarity coefficients exceeding 0.95 for both overpressure and dynamic pressure decay histories compared to the standard exponentially decaying shock wave. These results demonstrate the feasibility of regulating shock wave profiles in a constant-diameter shock tube through driver-section baffle design. The proposed method provides a practical optimization design tool for shock wave waveform regulation in shock tubes and offers a basis for the rapid design and regulation of shock wave profiles.