Acoustic signals for fuel tank leak detection More info about this article: http://www.ndt.net/?id=22865 Wenbo Duan 1, Siamak Tavakoli 1, Jamil Kanfoud 1, Tat-Hean Gan 1, Dilhan Ocakbaşı 2, Celal Beysel 2 1 Department of Mechanical, Aerospace and Civil Engineering, Brunel University London, Uxbridge, Middlesex, UK Wenbo.Duan@brunel.ac.uk Tat-Hean.Gan@brunel.ac.uk Abstract 2 Floteks Plastik Inc. Co. Demirtaş Sanayi Bölgesi Fulya Sk. No: 7 Osmangazi, Turkey Plastic and composite fuel tanks are widely used in the automotive industry. Quality control systems normally require a non-destructive testing method to ensure they are leak-free during the production stage in the factory before delivery to market. In this paper an acoustic method to inspect the tanks non-destructively and efficiently is proposed. To initiate possible acoustic signals due to leaks, a compressor is connected to the tank which is then sealed and immersed in water. Air is pumped into the tank at a pressure of 1.3bar. If there is a leak, the pressurised air will squeeze out through the leak and bubbles will appear in the water. Four hydrophones are placed in the water surrounding the tank to detect the acoustic signal generated by the bubbles, revealing the presence of leaks in the tanks. The technique is proposed for application in the factory environment where environmental noise may be very high. The bubble signal is found to be very weak and immersed in the noise floor. To improve the signal to noise ratio, a band-pass filter is used where the signal is band-pass filtered around 4100Hz. A number of parameters, such as root-mean-square (rms), standard deviation, etc are extracted after filtering. These parameters are used for automatic inspection of fuel tanks. 1. Introduction Plastic and composite fuel tanks are widely used in the modern automotive industry. They offer a number of advantages over conventional steel tanks, such as lower weight, higher corrosion resistance, better crash performance, etc. Such tanks are mass produced, and quality control systems require non-destructive testing techniques to inspect their integrity during the production stage in the factory environment. This paper studies the inspection of fuel tanks using acoustic signals. The technique represents a significant improvement over conventional visual inspection by removing the need for a labour-intensive human operator/inspector. In the UK, current standard procedure for tank leak detection mandates manual inspection. This requires the tank to be filled with Creative Commons CC-BY-NC licence https://creativecommons.org/licenses/by-nc/4.0/
test fluid and gradually pressurised, while an inspector visually inspects any leaks [1]. The test duration must exceed 5 minutes each time. This method is thus labour-intensive and not suitable for inspection in mass production. A new acoustic inspection technique is proposed. This requires the tank to be filled with pressurised air at 1.3 bar and immersed in water, when bubbles will appear if the tank is leaky. Bubble sound propagates in the water, and four hydrophone sensors are used to detect the bubble sound automatically. A set of parameters are then extracted to determine if the tank is leaky. Acoustic signals are widely used to detect blockages or defects in pipelines [2]. In these applications, an active sound source injects a well-designed sound wave into the pipeline which will be scattered by blockages or defects. Acoustic sensors are used to receive both the incident and scattered signals and positions of the blockages or defects are determined based on the flight times. Acoustic signals are also used for leak detection in pipes [3,4]. This requires the pipe to be filled with water when the inspection is carried out. Any leak will produce noise signals which can be picked up by two (or more) remote sensors. Cross correlation of the received signals reveals the relative position of the leak. This is a passive technique, and the strength of the acoustic signal is mainly determined by water pressure inside the pipe and the size of the leakage. A similar principle to leak detection in pipes could also be used for leak detection in tanks. To develop an automatic leak detection system, it is more convenient to fill the tank with pressurised air, and in this case, the object is to detect bubble signals from leaks when the tank is immersed in water. Leifer and Tang [5] studied the emission of acoustic signals from single bubbles emitted from a marine hydrocarbon seep. The measurements were carried out in the natural undersea environment, and individual bubbles could be observed. The Minnaert formula was used to link the size of the bubbles with the peak frequency of the bubble signal. Leighton and White [6] further proposed a method to estimate the power spectral density of a bubble cloud emitted from undersea methane seeps or gas pipeline leaks. An inverse formulae was used to estimate the generation rate of bubbles of given sizes. It is thus possible to detect bubbles using hydrophones in a natural undersea environment, and this forms the basis of the current leak detection method. However, one of the difficulties of bubble detection is that the signal to noise ratio is relatively low in a natural undersea environment [5] and this becomes even more problematic in an industrial manufacturing environment. Frequently the noise level is so high that the bubble signal is totally immersed in the noise floor. In this present paper, an appropriate filter is designed to extract the bubble signal from the noise floor. Furthermore, a number of parameters, such as rms, deviation, etc. of the filtered signal are used to infer the presence of bubbles. This is required by the automotive industry so that a leak condition can be determined by an automated system, rather than a human operator. The ultimate goal is to make leak detection of mass production tanks efficient and robust. The structure of the paper is as follows: an introduction is given in section I; the inspection technique is presented in section II; results and discussions are given in section III; and finally conclusions are drawn in section IV. 2
2. Inspection technique The leak detection system is shown in Fig. 1. A water pool has been built in an industrial plant room in Turkey for the detection of bubbles. The water pool is 2150mm long, 1350mm wide, and 1100mm high. The fuel tank is connected to an air hose, and sealed. The air hose allow pressurised air to be pumped into the tank. The compressor provides stabilised air pressure of 1.3 bar. The difference between the pressurised air inside the tank and atmospheric pressure outside will produce air bubbles if the tank has a leak that is larger than a critical size. The water pressure difference between the water pool surface and the leak location is less than 0.11 bar, and this doesn t make a significant difference. The bubble strength and bubble generation rate is directly related to the air pressure difference inside and outside of the tank. The bubble signal could be enhanced by increasing the air pressure inside the tank. However, the automotive industry limits the maximum pressure to be used inside fuel tanks, and 1.3 bar is considered to be a representative air pressure which avoids deformation and damage of the tank. In this paper, the air pressure inside the tank is thus fixed at 1.3bar. Figure 1. Structure of the leak detection system. Four hydrophones were placed at four corners of the water pool. The hydrophones are Brüel & Kjær type 8103 hydrophones. The frequency range is between 0.1Hz and 180kHz. A National Instrument data acquisition system is used to record the data received by the hydrophones. The maximum sampling frequency is 100kHz. A preamplifier is used between the hydrophones and the NI data acquisition system, and the pre-amplifier enhances the signal by 30dB. To test the noise level inside the plant room, an intact plastic fuel tank of size 450mm long, 200mm wide and 180mm high is immersed in the pool. No bubbles were generated, and the four hydrophones measure the background noise level which is stored as a reference signal. Then the tank to be inspected is immersed in the pool at the same position ensuring that the structure of the system remains roughly the same, and sound waves will be scattered in roughly the same way in the pool. The signal is recorded, and compared to the background 3
reference. The difference between the test signal and the reference signal determine whether the tank is leaky or not. The strength of the bubble signal is related to the size of the hole in the tank, which was not possible to quantify accurately in a plastic tank. To study the relationship between the size of the tank hole and the emitted acoustic signal, a set of additional tests were carried out using laser produced holes in a metal disk. These holes are circular, and the size of the holes can be precisely controlled. The metal disk is connected to the air hose, and the bubble signal generated from the calibrated holes investigated. The effect of the hydrophone positioning within the tank was also investigated. A number of parameters, such as rms, deviation, etc., are used to discriminate the leaky disk from the reference tank under different test conditions. The results are presented in the next section of the paper. 3. Results and discussions The tests were carried out in Turkey in a plant room where noise levels are high. The reference signal without bubbles is measured first. An intact tank is immersed into the water pool, and no bubbles are present in this case. The measured signal indicates the noise level inside the plant room. The raw signal recorded by one of the hydrophones and its frequency spectrum are presented in Figs 2 (a) and (b) respectively. The signal is continuously recorded for 120s for all the four hydrophones. The signals recorded by the remaining three hydrophones are very similar to the signal presented here, and are not shown. It can be seen that the noise level is very high. The origin of the noise could be electronic and hardware generated noise, environmental noise transmitted from air to the water, as well as vibrational noise transmitted through the ground from other running machinery in the plant room. The noise spectrum is very wide, spanning from zero frequency to 50kHz. Note that a pre-amplifier is normally used when the signal is very small, and in this case the noise level could be increased as well. Next a leaky plastic tank is immersed in the pool. A hole is manually cut in the plastic tank whose size is measured to be roughly 220µm. The raw signal is shown in Fig. 3(a). It is clear that the bubble signal is embedded in the noise floor. The leak signal is almost the same size as the reference signal shown in Fig. 2(a). To detect the presence of bubbles, a Butterworth band pass filter is applied. Fig. 2(b) shows that the noise spectrum covers the entire frequency range, however, the bubble spectrum is normally in the frequency range below 10kHz [5, 6]. The challenge here is to find a frequency bandwidth for filtering that produces the optimal signal to noise ratio. A number of post-processing experiments were carried out by varying the bandwidth of the filter. A suitable frequency range is found to be from 3.6kHz to 4.6kHz for the pass band, and from 2.8kHz to 5.2kHz for the stop band. The stop band attenuation is 20dB. This filter is applied to both the reference signal and the leak signal. The presence of bubbles is very clear in Fig. 3 (b). To automate the algorithm processing, a number of parameters are extracted. A Hilbert transformation is applied to the filtered time domain signal to extract its wave amplitude. A set of parameters, including rms, mean, standard deviation, etc. of the Hilbert transformed signal are extracted. The rms value of the reference signal is 0.00268, and the rms of the 4
leaky signal is 0.00494. This shows a difference of 84% from the reference signal. It is thus possible to set a threshold value above the reference signal, above which the signal can be identified as a leak. Amplitude (mv) Time (s) Frequency (khz) Figure 2. (a) Time history of the reference signal; (b) spectrum of the reference signal. Amplitude (mv) Spectrum Time (s) Amplitude (mv) Bubbles Time (s) Figure 3. (a) Leaky tank signal before filtering; (b) leaky tank signal after filtering. 5
4. Conclusions In this paper, a fuel tank leak detection method is proposed which makes use of acoustic signals. The fuel tank is pressurised by an air pump, sealed, and immersed in a water pool. Four hydrophones are used to measure the sound wave inside the pool. A Butterworth filter is applied to determine if the tank is leaky. To automate the leak detection algorithm, a set of parameters, including rms, deviation, etc., are extracted and analysed. Acknowledgements The authors would like to thank EU Horizon 2020 project Leakfree for sponsoring the work reported here (Project reference: 673155). References 1. Vehicle certification agency, Procedures for inpection bodies-testing and inspection of UK tanks, 2017. https://www.dft.gov.uk/vca/additional/files/dangerous-goods/conformityassessment-bodies/testing-inspecting-uk-tanks.pdf 2. W. Duan, R. Kirby, J. Prisutova, K. V. Horoshenkov, On the use of power reflection ratio and phase change to determine the geometry of a blockage in a pipe, Appl. Acoust. 87, 190-197. 2015. 3. O. Hunaidi, W. T. Chu, Acoustical characteristics of leak signals in plastic water distribution pipes, Appl. Acoust. 58, 235-254, 1999. 4. Y. Gao, M. J. Brennan, P. F. Joseph, J. M. Muggleton, O. Hunaidi, A model of the correlation function of leak noise in buried plastic pipes, J. Sound Vib.. 277, 133-148, 2004. 5. I. Leifer, D. Tang, The acoustic signature of marine seep bubbles, J. Acoust. Soc. Am. 121, EL35-40, 2007. 6. T. G. Leighton, P. R. White, Quantification of undersea gas leaks from carbon capture and storage facilities, from pipelines and from methane seeps, by their acoustic emissions, Proc. R. Soc. A. 468, 485-510, 2012 6