This report builds on theoretical analysis of jet engine infrared signatures and their potential relationships to jet engine acoustic emissions in order to identify the region of a jet engine plume most likely to emit both in temporal infrared and in acoustic domains. As a means of verifying initial assessments that infrared and acoustic emissions are connected via physics-based mechanisms, a field campaign to collect relevant data was carried out with a bank of infrared instruments imaging a T700 turboshaft engine undergoing routine operational testing, while microphones collected simultaneous acoustic data. In-depth mathematical analyses using principal component analysis and power spectral density analyses were conducted on the infrared and acoustic data to determine whether any relationship exists between infrared emissions and acoustic emissions.

Technical Approach

Hypertemporal imaging uses fast-framed image data (typically hundreds or thousands of frames per second), together with time-domain analysis, to extract dim fluctuating signals, especially against bright backgrounds. Principal component analysis is commonly used as a means of dimensional reduction, transforming observations into correlated subsets called principal components, which are linearly uncorrelated to each other. Here PCA can reveal spatial patterns of temporal correlation.

To collect data for this effort, a mid-wave infrared (MWIR) camera (model FLIR SC6701, bandwidth 3-5 μm, with a 50 mm f/2.5 lens) was set to stare at a jet engine plume, about 5-6 jet diameters away from the exhaust plane. Acoustic sensors were also emplaced around the plume, including on a line directly between the instruments and the expected maximum-emissions zone. The engine was run at a careful throttle profile, and the MWIR camera collected data (10 seconds @ 120 Hz – max capture rate) at each throttle setting. Throttle settings were held for 10-30 seconds, and then a ramp up or down to a new setting was completed. A given engine run would incorporate from two to four different throttle settings, and would last approximately ten minutes. Altogether, three engine runs were completed, and data was captured from at least one throttle setting.

Data were analyzed via three methods. To determine spatial regions of temporal correlation, principal components analysis was performed on the frames of the infrared imagery data in order to separate sources of IR fluctuations. This produced a single image showing the RMS AC signal of each pixel time history. These images illustrated plume structure and the fluctuating regions of IR energy transfer in the plume. Comparing figures reveals that plume structures change in size as throttle levels rise; additional features make it clear highly complex fluid flow is being observed.

To determine differences in temporal PSD signature, a PSD of the time history of the maximum pixel in each specific principal component was generated to show the power spectral density. This was to search for a spike in the PSD at a frequency known to be the same as, or perhaps a harmonic of, an acoustic frequency with strong emissions – no such obvious spikes were uncovered, although further analysis, including: 1) calculations of the correlation with acoustic data (cf. Section 5.4); 2) interrogation of the spectral phase indicated by each principal component; 3) reliance upon higher frame rate MWIR imagers would be desirable.

To perform acoustic analysis, pressure spectra (sound pressure levels across frequencies, with a total SPL listed at the top) on the left, and spectrograms (sound levels over a thin slice of time during that zone era, at various frequencies, with darker red indicating more power) were used. Spikes appear at below 200 Hz, about 350 Hz, transiently near 650 Hz, and very broadly at just over 1000 Hz.

In general, there is some apparent structure in the acoustic noise, and there is also evidence of features of this structure changing as the throttle settings change. While all this is to be expected, any correlation detected between this and the hypertemporal structure detected in the plume may show how the two are related. Unfortunately, all of the most prominent acoustic features lie outside the range accessible by the imagers employed, although some correlations were observable.

The spike data is then used to determine the correlation relationship between IR and Acoustic data and plot it spatially. Low-pass filtered acoustic data, extracted to match timing of infrared imagery data, is covariance-mapped in its time history to the time history of the infrared imagery data.


Analysis of simultaneously acquired acoustic and IR measurements of a T700 engine under test yielded several notable conclusions relevant to using hypertemporal imaging techniques to aid in mitigating jet engine noise generation.

An analysis of MWIR data alone indicates the specific structure of correlated energy transfer processes within the engine plume down to significantly small spatial scales and at significantly small AC contributions. Such fine resolution may be very useful by itself as truth data for refining computational fluid dynamics models used when designing jet engines. Such additional refinement may lead to greater success in mitigating jet engine noise via design.

Direct analysis of the acoustic data indicated that the most prominent noise features are present at frequencies beyond the frame rate capabilities of the IR imagers used in this analysis. However, IR imagers with adequate frame rates (>500 Hz) may be acquired for future work based on this effort.

Of most significance here, the analysis incorporating both MWIR and acoustic data indicates regions of high correlation with clear structure. In this way, at least some of the dynamics of the engine plume that are responsible for acoustic generation are clearly captured by IR observations. Viewed in light of the principal components images of Section 5.1, striking similarities make it appear that the regions of highest correlation coincide with the energy transfer processes captured by principal components between five and 25. If these processes can be understood and separated from non-noise generating flow, mitigation of such processes could be prioritized when designing quiet jet engines.


Jet engine noise can be both a health hazard and an environmental pollutant, particularly affecting personnel working in close proximity to jet engines, such as airline baggage handlers and mechanics. Mitigating exposure to jet engine noise could reduce the potential for hearing loss in runway workers; however, there exists a very complex relationship between jet engine design parameters, operating conditions, and resultant noise power levels. This is demonstrated with the initial results highlighting the utility of high-speed imaging (also called hypertemporal imaging) in correlating the infrared signatures of jet aircraft engines with acoustic noise from the jet engines. This detection enables the use of a new domain in characterizing jet engine noise, which may in turn enable new methods of predicting or mitigating jet engine noise, which could lead to socioeconomic benefits for airlines and other operators of large numbers of jet engines.