Short-time Proper Orthogonal Decomposition (POD) is proposed as an image-based technique to study the transient jet screech characteristics of moderately under-expanded supersonic jets emanating from a circular baseline and two bevelled nozzles. Time-resolved schlieren imaging of turbulent flow structures were performed with an ultrahigh-speed schlieren setup. Short-time POD was performed by systematically sampling image-series with a short time delay, performing PODs and applying spectral analyses on the first POD mode coefficients, and plotting the peak frequencies from the resulting PSDs into a peak frequency-occurrence count histogram. The results are in good agreement with the near-field noise spectra and wavelet transform analysis of the microphone measurements, which revealed intermittent jet screech occurrences at St=0.25 for both baseline and 30° bevelled jets, while none was detected for the 60° bevelled jet. In particular, the occurrence counts of the frequency bins is proposed as a suitable parameter to characterize the intermittent nature of jet screech, with the frequency bin revealing the jet screech frequency if present. The present study demonstrates the advantage of short-time POD analysis on time-resolved schlieren images over traditional image-based POD methods, which includes computational gains from parallelization, the ability to handle much larger datasets and revealing insights into a transient flow and noise phenomenon.

短时适当的正交分解（POD）是一种基于图像的技术，用于研究由圆形基线和两个倾斜喷嘴发出的中等膨胀不足的超声速射流的瞬态射流尖叫特性。用超高速schlieren装置对湍流结构进行时间分辨的schlieren成像。短时POD通过以下方式进行：对短时延的图像系列进行系统采样，执行POD并将光谱分析应用于第一POD模式系数，然后将所得PSD的峰值频率绘制成峰值出现频率计数直方图。结果与麦克风测量的近场噪声谱和小波变换分析非常吻合，这表明在St = 0时出现间歇性的喷鸣声。基线和30°斜角喷嘴均为25，而60°斜角喷嘴均未检测到。特别地，频率箱的出现次数被建议为表征喷射声的间歇性的合适参数，频率箱揭示了喷射声的频率（如果存在）。本研究表明，与基于图像的传统POD方法相比，对时间分辨的纹影图像进行短时间POD分析具有优势，其中包括并行化的计算收益，处理更大数据集的能力以及揭示对瞬态流动和噪声现象的洞察力。频率仓显示出射流尖叫声频率（如果存在）。本研究表明，与基于图像的传统POD方法相比，对时间分辨的纹影图像进行短时间POD分析具有优势，其中包括并行化的计算收益，处理更大数据集的能力以及揭示对瞬态流动和噪声现象的洞察力。频率仓显示出射流尖叫声频率（如果存在）。本研究表明，与基于图像的传统POD方法相比，对时间分辨的纹影图像进行短时间POD分析具有优势，其中包括并行化的计算收益，处理更大数据集的能力以及揭示对瞬态流动和噪声现象的洞察力。