JOINT TIME-FREQUENCY ANALYSIS OF RADAR MICRO-DOPPLER SIGNATURES FROM AIRCRAFT ENGINE MODELS

Transcription

1 J. of Electromagn. Waves and Appl., Vol. 25, , 2011 JOINT TIME-FREQUENCY ANALYSIS OF RADAR MICRO-DOPPLER SIGNATURES FROM AIRCRAFT ENGINE MODELS H. Lim and J. H. Park Department of Electrical Engineering Korea Advanced Institute of Science and Technology (KAIST) Guseong-dong, Yuseong-gu, Daejeon , Korea J. H. Yoo, C. H. Kim, and K. I. Kwon Agency for Defense Development Yuseong P. O. Box 35-3, Daejeon , Korea N. H. Myung Department of Electrical Engineering Korea Advanced Institute of Science and Technology (KAIST) Guseong-dong, Yuseong-gu, Daejeon , Korea Abstract Micro-Doppler signatures from four CAD aircraft engine models and one experimental engine model are analyzed using joint time-frequency analysis (JTFA). The signatures are obtained from shooting and bouncing rays (SBR) simulation for the CAD models and from measurement for the experimental model. The JTFA results give additional useful physical feature information regarding the respective engines, such as whether the number of blades is odd or even and the length of blades, in addition to the number of blades, which can be extracted by traditional spectrum analysis. 1. INTRODUCTION Jet engine modulation (JEM) has been widely used as an outstanding non-cooperative target recognition (NCTR) method. JEM is the Doppler characteristics of radar signals scattered from rotating compressors in an engine. The features of a jet engine for the purpose Received 25 January 2011 Accepted 3 March 2011, Scheduled 12 March 2011 Corresponding author: Ho Lim (limho@kaist.ac.kr).

2 1070 Lim et al. of recognition can be extracted from the Doppler characteristics. Bell developed a parametric model based on periodic modulation of the scattered return using autocorrelation, JEM spectrum, and cepstrum [1]. The parametric model analysis could give just the number of blades in each stage. In other published studies, timeor frequency-only analysis techniques have been applied to JEM analysis [2], while joint time-frequency analysis (JTFA) has been applied to human gait or the rotors of a helicopter [3, 4]. However, the application of JTFA to aircraft engine models has not been reported so far, although it can give more information than time- or frequency-only analysis. We have investigated various techniques of JTFA [5 10] and present JEM analysis using the JTFA technique, which is applied to time-varying radar cross section (RCS) data, which is acquired from the commercial software package, virtual aircraft framework (VIRAF, by IDS inc. in Italy) using the shooting and bouncing rays (SBR) technique [11 14] for four types of aircraft engine CAD models. In addition, we fabricated one engine model and measured it by the instrumentation radar. The measured data are analyzed to validate the possibility of JEM analysis using JTFA on numerically yielded time-varying RCS data. The JTFA results yield additional useful physical features of the respective engines, such as odd/even check and the length of blades in addition to the number of blades from traditional time- or frequencyonly analysis. This paper describes micro-doppler signature [15, 16] analysis using the JTFA technique for aircraft engine models and additional extracted information in detail. The analysis results for aircraft engine CAD models are given in Section 2, and the analysis results of the fabricated engine model are given in Section 3. In Section 4, summary and conclusions are presented. 2. AIRCRAFT ENGINE CAD MODELS AND ANALYSIS 2.1. Aircraft Engine CAD Models Four aircraft engine 3D CAD models imitating real engines were designed in detail using the commercial software CATIA for electromagnetic simulation and to obtain JEM data. The complete CAD models are shown in Fig. 1, and the related dimensions and the numbers of blades are given in Tables 1 and 2, respectively. In Table 2, models A and C are high bypass types, which mean that the ratios of bypassed airflow to combustor airflow are high for fuel efficiency. Aircraft engines related to models A and C are used in the C-5 Galaxy military transport aircraft and the Boeing 787 commercial

3 Joint time-frequency analysis of radar micro-doppler signatures 1071 (a) (b) (c) (d) Figure 1. Aircraft engine CAD models: (a) model A (TF 39), (b) model B (F136), (c) model C (Trent 1000), (d) model D (AE3007). Table 1. Dimensions of jet engine structures, where D and L indicate overall diameter and length, respectively, and a, b, andc indicate length of first, second, and third stage blades, respectively. model dimensions [cm] D L a b c A B C D fabricated

4 1072 Lim et al. Table 2. Numbers of blades. model stage A B C D fabricated airliner. Models B and D are low bypass types, which mean that ratios of bypassed airflow to combustor airflow are low for high exhaust velocities. Aircraft engines related with models B and D are used in the F-35 Joint Strike Fighter and the Global Hawk unmanned aerial vehicle, respectively Analysis Radar micro-doppler signatures were extracted from aircraft engine CAD models described in the previous subsection using JTFA by the following procedures: 1. JEM data extraction using VIRAF. 2. JTFA application to JEM data. 3. Radar micro-doppler signatures extraction and analysis. In stage 1, each model is simulated using SBR-based VIRAF simulation packages, and the simulation parameters shown in Table 3 are used. VIRAF is specialized in radar signature prediction and management for aircraft platforms, and in particular, it is possible for VIRAF to simulate JEM analysis using SBR. Fig. 3 shows the simulated JEM data of model A when the incident angle is 10. The number of blades can be extracted by traditional time or frequency analysis from Fig. 3 [1]. That is, the spool period can be extracted from the time domain signal using autocorrelation or cepstrum, and the number of blades can be extracted from the outstanding spectrum in the frequency domain JEM data by dividing the spool period, which was obtained from the time domain. In stage 2, short time Fourier transform (STFT) is adopted as representative for JTFA, and it is defined as in [17] as STFT(t, f) = s(t )w ( t t ) e j2πft dt, (1)

5 Joint time-frequency analysis of radar micro-doppler signatures 1073 Table 3. Simulation parameters. parameter value frequency 10 GHz pulse repetition frequency (PRF) 80 khz polarization HH number of bursts 1600 rotation speed 6000 RPM SBR ray density λ/10 incident (elevation) angle 10 (a) Figure 2. A fabricated aircraft engine model: (a) isometric view and (b) a rotating part. (b) where s(t ) is the signal to be transformed, and w(t ) is the window function. Figure 4 shows a spectrogram of model A when the incident angle is 10. The spectrogram includes a variety of information. Time or frequency resolution can be adjusted by controlling the window size. With a relatively small window, JEM signal components, appearing as horizontal lines in Fig. 4(a), can be extracted using high frequency resolution. With a relatively large window, more information can be extracted, as seen in Fig. 4(b). Blade period indicates the period when a blade rotates from the original position to the adjacent position as seen in Fig. 4(c). Outer components may indicate maximum Doppler

6 1074 Lim et al. (a) Figure 3. Simulated JEM data from the jet engine model A when incident angle is 10 :(a)timevs.jemand(b)jemspectrum. (b) shift frequency defined by (2) in Fig. 4(c): 2ωL sin α f d,max = + C, (2) λ where f d,max is the maximum Doppler shift frequency, ω is the angular speed, L is the length of blade, α is the aspect angle defined as the angle between the hinge direction and the radar line of sight (LOS), and C is the constant considering cavity effect. If the aspect angle is predicted using radar tracking information, the length of blade information can be estimated with some error from the cavity effect and the STFT window effect. Although the maximum Doppler-shift frequency in the spectrogram can be varied according to window size, the relative length information of each engine under the same window condition will not change; thus, it is possible to extract real blade length in case of suitable compensation with respect to window size. This length information is additional useful feature that can facilitate aircraft engine recognition. The vertical lines in Fig. 4(c) indicate that the number of blades is even because the Doppler components of approaching and receding occur at the same time in the case of an even number. The relative length of blades and odd/even check information cannot be obtained by the traditional technique; thus, the proposed technique increases the possibility of high-quality target recognition. If the signal-to-noise ratio (SNR) is not enough high for the parametric JEM model, the number of blade would be easily under- or over-estimated by one. In this case the proposed odd/even check provides a means to avoid this error.

7 Joint time-frequency analysis of radar micro-doppler signatures 1075 (a) (b) (c) Figure 4. Spectrogram from the jet engine model A when incident angle is 10 : (a) using STFT with a large window of 512, (b) using STFT with a small window of 32, and (c) expanded spectrogram related to (b). Figure 5 shows a spectrogram of model D when the incident angle is 10, and unique radar micro-doppler signatures can be extracted and explained in the same way as in Fig. 4. In this case, the outer component is smaller than model A because the blade length is shorted than that of model A. The slanted lines in Fig. 5(b) indicate that the number of blades is odd because the Doppler components of approaching and receding occur alternately. The estimated information for four engine models is tabulated as shownintable4.

8 1076 Lim et al. (a) Figure 5. Spectrogram from jet engine model D when incident angle is 10 : (a) using STFT with small window of 32 and (b) expanded spectrogram related to (a). Table 4. Estimated information. model odd/even length [cm] real estimated real estimated error [%] A even even B even even C odd odd D odd odd (b) It can be observed that the odd/even check is available for all cases, and the length of blades can be estimated with small errors from cavity effects. In model A case, the length of second stage blades are estimated because second stage blades are longer than first stage blades unlike models B, C and D cases. In addition to the information of the number of blades, this information from JTFA can increase the possibility of high-quality NCTR by complementing the traditional technique. 3. A FABRICATED AIRCRAFT ENGINE MODEL AND ANALYSIS An aircraft engine model was designed and fabricated. The complete CAD model and rotating part are shown in Fig. 2, and the related

9 Joint time-frequency analysis of radar micro-doppler signatures 1077 (a) (b) (c) Figure 6. Spectrogram from measured data of jet engine model when incident angle is 10 : (a) using STFT with small window of 32, (b) spectrogram after IIR filtering, and (c) expanded spectrogram related to (b). dimensions are described in Tables 1 and 2. The JEM data was acquired from measurement of the fabricated jet engine models using an instrumentation radar system. The radar system and jet engine parameters are shown in Table 5. Figure 6 shows a spectrogram of the fabricated aircraft engine model when the incident angle is 10. The spectrogram with measured data is more complicated and irregular than those for simulations because of noise and the environmental effects of measurement. In Fig. 6(a), period indicates the spool period. It can be seen that

10 1078 Lim et al. Table 5. Simulation parameters. parameter value frequency 10 GHz pulse repetition frequency (PRF) 1.8 khz polarization HH number of bursts 2048 rotation speed 60.1 RPM incident (elevation) angle 10 the blade period of first-stage blade is not correct in the spectrogram due to noise and effect of second- and third-stage blades. Therefore, infinite impulse response (IIR) filters are used to filter out unwanted components and focus on the first harmonic of first-stage blade. Spectrogram data of the fabricated aircraft engine models through IIR filtering and STFT is shown in Figs. 6(b) and (c). The spectrogram is more regular and the blade period is clearer, and further it can be observed that the odd/even check is available from vertical and slant components. It can be observed that the length of blade information can be estimated with some errors from the STFT window effect, cavity effect, noise and environmental effects of measurement. In this case, the STFT window size of 64 is more suitable than the size of 32 to estimate accurately the length of blade information. The techniques described in this paper should be developed and extended to make suitable compensation with respect to window size, but the fundamental principles can still be followed. 4. CONCLUSION We acquired JEM data from electromagnetic simulation by VIRAF and measurement by instrumentation radar. We made a detailed analysis of jet engine modulation signals of aircraft engine models systematically using JTFA for the first time. It was shown that unique spectrograms related to JEM signals from jet engine structures can be obtained. The obtained spectrograms showed additional information beyond traditional analysis such as odd/even check of the number of blades and the length of blades. These results can be very useful for the analysis of jet engine modulation and to enable high-quality NCTR by complementing and assisting traditional analysis.

11 Joint time-frequency analysis of radar micro-doppler signatures 1079 ACKNOWLEDGMENT This research was supported by the Agency for Defense Development (ADD), Korea. REFERENCES 1. Bell, M. R. and R. A. Grubbs, JEM modeling and measurement for radar target identification, IEEE Trans. Aerospace and Electronic Systems, Vol. 29, No. 1, 73 87, Cuomo, S., P. F. Pellegrini, and E. Piazza, Model validation for jet engine modulation phenomenon, Electronics Lett., Vol. 30, No. 24, , Chen, V. C., F. Li, S. S. Ho, and H. Wechsler, Micro-Doppler effect in radar: Phenomenon, model, and simulation study, IEEE Trans. Aerospace and Electronic Systems, Vol. 42, No. 1, 2 21, Chao, N. and H. Jing, Micro-motion simulation and micro- Doppler extraction, PIERS Proceedings, , Xi an, China, Mar , Chen,V.C.andH.Ling,Time-frequency Transforms for Radar Imaging and Signal Analysis, Boston, MA, Artech House, Narayanan, R. M., M. C. Shastry, P. H. Chen, and M. Levi, Through-the-wall detection of stationary human targets using Doppler radar, Progress In Electromagnetics Research B, Vol. 20, , Zhang, Q., Y. Luo, D. L. Hu, B. F. Luo, and Y. S. Zeng, Avian detection and monitoring using frequency-stepped chirp signal radar, PIERS Online, Vol. 4, No. 1, 51 55, Lazaro, A., D. Girbau, and R. Villarino, Wavelet-based breast tumor localization technique using a UWB radar, Progress In Electromagnetics Research, Vol. 98, 75 95, Yu, X. and G. G. Bi, Performance of multiband complex wavelet based multicarrier Ds-CDMA system with multi-antenna receiver over Nakagami-m fading channel, Progress In Electromagnetics Research, Vol. 98, , Han, S. K., H. T. Kim, S. H. Park, and K. T. Kim, Efficient radar target recognition using a combination of range profile and time-frequency analysis, Progress In Electromagnetics Research, Vol. 108, , Ling, H., R. C. Chou, and S. W. Lee, Shooting and bouncing

12 1080 Lim et al. rays: Calculating the RCS of an arbitrarily shaped cavity, IEEE Trans. Antennas Propag., Vol. 37, No. 2, , Kim, B. C., K. K. Park, and H. T. Kim, Efficient RCS prediction method using angular division algorithm, Journal of Electromagnetic Waves and Applications, Vol. 23, No. 1, 65 74, Park, K. K. and H. T. Kim, RCS prediction acceleration and reduction of table size for the angular division algorithm, Journal of Electromagnetic Waves and Applications, Vol. 23, No , , Gao, P. C., Y. B. Tao, and H. Lin, Fast RCS prediction using multiresolution shooting and bouncing ray method on the GPU, Progress In Electromagnetics Research, Vol. 107, , Guo, K. Y. and X. Q. Sheng, A precise recognition approach of ballistic missile warhead and decoy, Journal of Electromagnetic Waves and Applications, Vol. 23, No , , Guo, K. Y., Q. Li, and X. Q. Sheng, A precise recognition method of missile warhead and decoy in multi-target scene, Journal of Electromagnetic Waves and Applications, Vol. 24, No. 5 6, , Aldirmaz, S. and L. Durak, Broadband interference excision in spread spectrum communication systems based on short-time Fourier transformation, Progress In Electromagnetics Research B, Vol. 7, , 2008.