The Use of Narrowband Low Frequency Air-coupled Transducers for High Speed Detection of Broken Rails

An experimental study was conducted to simulate high-speed continuous monitoring of broken steel rail. A pair of low frequency narrowband ultrasonic transducers configured in pitch-catch mode was used to obtain signals reflected off of a circumferential notch in a spinning aluminum wheel. Due to the relative motion between the impinging ultrasound emitted from the air coupled transducer and the notch, a noticeable Doppler shift occurred. These frequency shifts were clearly observable by using a single-tone continuous sine wave to excite the transmitter. INTRODUCTION To meet the challenging demands for sustaining rail infrastructure, it is important to assess the rail integrity rapidly and accurately with state-of-the-art nondestructive evaluation (NDE) methods [1, 2]. Rail assessment may be approached using “continuous monitoring” and/or “detailed inspection” [3]. The former provides global information regarding the overall health of the rail, whereas the latter focuses on a particular segment to provide a detailed account of the location and/or characteristics of a defect. In dealing with continuous monitoring of broken rail, track circuits are widely used, in which the rail is instrumented as part of an electric circuit so that certain discontinuities such as broken segments may be detected by a change in resistance. Clearly there are many drawbacks and there are many attempts to replace the old technology with more cost-effect and accurate ones. As an alternative, ultrasonic technology employs a rolling wheel mounted with ultrasonic probes that are in contact with the track surface. This technique is advantageous since internal defects are also detectable. It is however functional only at intermediate speeds and quickly degrades at faster speeds comparable to a passenger car due to the constraint of intimate surface contact. This work discussed in this paper is driven by the need to develop improved continuous monitoring methods, with the following objectives in mind: (a) Increase inspection rate over 60 mph (96 km/h), (b) Implement non-contact ultrasonic transducers, (c) Improve accuracy of detection, and (d) Interpret data quickly and easily. 1Send correspondence to Shi-Chang Wooh, Massachusetts Institute of Technology, Room 1-272, 77 Massachusetts Avenue, Cambridge, MA 02139, Tel: (617)253-7134, Fax: (617)253-6044, E-mail: [email protected] 2Graduate students Proceedings of the 7th Annual Research Symposium: Transfer of Emerging NDE Technologies in conjunction with the 9th Asia-Pacific Conference on Nondestructive Testing and 1998 ASNT Spring Conference, American Society for Nondestructive Testing, Anaheim, CA, March 23–27, 1998. DOPPLER EFFECT Our monitoring scenario utilizes a pair of narrowband low-frequency ultrasonic transducer operating in pitch-catch mode coupled through air. The transmitter insonifies the rail with ultrasound continuously at a single tone carrier frequency and the receiver picks up the signal reflected off of the surface and perhaps from the notch of the broken rail segment. Because the reflection coefficient between air and steel is essentially unity due to the high acoustic impedance mismatch, we may expect an excellent signal to noise ratio. This will be particularly advantageous for detecting surface defects and discontinuities. One very fundamental yet unique approach to meet these requirements is to exploit the relative motion between the air-coupled transducer (mounted in the rail car) and the running rail via the Doppler effect. In the existence of a defect, a frequency shift will occur, which is predictable from the known speed of sound propagation in air, speed of the rail car and the angle of incidence formed between the direction of motion and the transducer. It is evident that faster inspection speeds will lead to a more pronounced frequency shift. This is a bonus because not only will it free trains up from excessive inspection time, but also improve the sensitivity and accuracy of the detection of various defects. In this paper, we present the promising results on the processing of the ultrasonic Doppler shifted signal using the Fast Fourier Transform (FFT) and the Short-Time Fourier Transformation (STFT). Both techniques provide a convenient means of identifying surface anomalies with minimum interpretation. HIGH-SPEED MONITORING SIMULATION TEST Since it is impractical to test rail with cars running at high speeds in a laboratory environment, a similar effect was created by using a motor rated at 1700 rpm to rotate a solid aluminum disk with a surface notch on the periphery. Measuring 1/8 in. (3.18 mm) long and 1/8 in. (3.18 mm) deep, the notch was introduced artificially to simulate a surface defect that would be used as the reflector and source of the Doppler-shifted ultrasound. The radius of the disk was chosen to be 6 in. (15.24 cm) in order to create an equivalent linear rail inspection velocity of 60 mph (96 km/h). A pair of 100 kHz non-contact transducers arranged in pitch-catch mode were placed at 45 angles with respect to the vertical centerline of the disk. An HP function generator was used to created a single-tone continuous sine wave of 100 kHz carrier frequency as the transmitter input signal. Figure 1(a) represents the predicted received signal from the notch. Note the interference (manifested in the form of wave packets) between the signals reflected from both the surface and the notch. The corresponding spectrum, shown in Fig. 1(c), takes the form of a delta function indicated by a vertical line at the carrier frequency (100 kHz), accompanied by the predicted Doppler shifted signal at 111.5 kHz. The experimentally acquired waveform and frequency spectrum are shown in Fig. 1(b) and Fig. 1(d), respectively. Note the excellent agreement with the predicted frequency shift caused by wave interference in the region of the notch. As an alternative approach to processing the continuously incoming raw waveform, the Short-time Fourier transformation (STFT) was used [4] for processing the signal in real-time. In contrast to FT, the STFT provides frequency characteristics of local time and is therefore suitable for analyzing signals having time-varying characteristics. The peak magnitude of the STFT spectra within a specified window 0 100 200 300 400 500 Time, t, ( sec) -1.0 -0.5 0.0 0.5 1.0 A m pl itu de ,a rb itr ar y un its 0 50 100 150 200 250 300 350 400 450 500 Time, t, ( sec) -2 -1 0 1 2 A m pl itu de ,a rb itr ar y un its 0 20 40 60 80 100 120 140 160 180 200 Frequency, f, (kHz)