Using a mobile NQR detector with a mini-helicopter.

Using a mobile NQR detector with a mini-helicopter..

USE OF A MOBILE NQR DETECTOR USING A MINI-HELICOPTER  

The article discusses an improved NQR mine detection system, characterized by increased phase stability and designed to detect NQR signals with high multiplicity and spectrum width. A program for spectral evaluation of broadband (up to 100 kHz) NQR signals with a low signal-to-noise ratio is also proposed. The mobile NQR detector is integrated with a mini-helicopter.

The first experiment on detecting the domestic TM-62P mine was carried out near Kaliningrad (KVIUIV) in 1984 [1] at a frequency of 3410 kHz (hexogen) at a distance of 7 cm, with an additional rise of the plane of the coil surface above the ground by 2 cm. The pulse power was 0.5 kW, for a SORC pulse sequence with a detuning of 1 kHz. A synchronous detector with a pure absorption signal was used [2]. In 1985, a new device (Fig. 1) was designed for a frequency of 5192 kHz (RDX) [3], while detecting TS-2.5, TS-6 and M-14 mines (for tetryl 5290 kHz). During mine production, molten tetryl solidifies directly in the mine body and is a solid solution, therefore the 14N NQR signal in TNT has a more complex multiplet structure than the signals from RDX and requires that the detecting device have high sensitivity and resolution with a wide band (up to 100 kHz) of the NQR signal. The use of a synchronous detector allows using digital signal accumulation to increase the signal-to-noise ratio. The quartz generator synchronizes the operation of RF generators 1 and 2, as well as the pulse generator. The sinusoidal signal from the generator output goes to the input of the corresponding digital signal formers (FCS 1 and FCS 2), from the output of which a meander of the corresponding frequency is taken. Then these signals are modulated by the pulse generator signal, mixed and fed to the power amplifier. The use of a synchronizing quartz generator allows maintaining the signal phase constant in all circuits of the circuit. The coil system consists of a single-turn, flat irradiating coil and 4 receiving ferrite antennas. The Q-switch is used to suppress the signal of the receiving circuit during the action of the irradiating pulses. After amplification of the received echo signal of the mine, it was mixed in a two-channel synchronous detector of the Schuster type, with the signals of the master oscillators 1 and 2. After digitizing the NQR signal from the TNT with a sample of no more than 5 μs, it is accumulated and recorded as a file, which allows for its further digital processing by spectral evaluation methods in order to improve the resolution and signal-to-noise ratio.

Fig. 1 Schematic diagram of the mine detection device

A digital phase detector on a J,K trigger was also tested. The resonant frequencies of the irradiating and receiving circuits and the degree of their coupling were adjusted so that the maxima of the frequency response of the receiving and irradiating system coincided with the frequencies of the master oscillators.

In addition to explosive solutions, a solid solution of paranitrotoluene and trinitrotoluene, which is often encountered in practice, was also studied. The presence of paranitrotoluene gives TNT a characteristic yellow color and an unpleasant odor, by which dogs find mines. When 7% TNT was added to paranitrotoluene, the NQR line at a frequency of 1144 kHz broadened 10 times and a sequence of stimulated echoes had to be used to observe it, whereas for TS-2.5 and TS-6 mines a sequence of equally spaced pulses with detuning (SORC) was used. We used the same sequence to detect tetryl mines (M-14). Fig. 2 shows the tuning of a surface coil in the form of a ring, 25 cm in diameter, and four ferrite coils on M-61 ferrites for the n- and n+ frequencies of TNT. The partial Q-factors of the curves reached 460, and with the use of a Q-factor multiplier – up to 4600, which allows detecting NQR spectra in dual-frequency mode for the TS-2.5 mine at a distance of up to 8 cm. To suppress piezoelectric effects, the flat ring was raised 2 cm above the soil surface. The contours are adjusted using variable capacitors and by changing the coupling coefficient between the flat and ferrite coils by changing the distance between the ring and the perpendicular ferrites. In this case, the Q-factor of the partial resonance curves can also be adjusted within a wide range. Typically, work is carried out at partial Q-factors of 450 – 500.

Fig. 2. Tuning to trinitrotoluene frequencies

In fact, the system regulates each circuit for tuning to two frequencies. If the coupling coefficient was made less than one, then the quality factor sharply increased, and the detection range of the NQR signal increased. If the quality factors of the coupled circuits (a flat ring and four ferrites with coils) are different, then the partial quality factors also differ greatly from each other. It is believed that both circuits can be tuned to different frequencies and have different quality factors. Changing the coupling coefficient does not affect the partial quality factors, but has a strong effect on the partial frequencies. A computer program was used for more accurate and convenient tuning. Not only solid solutions of TNT were studied, but also its complexes with naphthalene and paranitrotoluene, as well as some TNT isomers that can be found in mine fillers. Double NQR was used for this [4, 5]. Although NQR can be used for mine detection, reports at the International Conference on Mine Action in Ljubljana (2000) once again showed a number of serious problems in detecting TNT, as opposed to RDX and tetryl [1, 2]. System [3] was tested only in laboratory conditions, as it required complex adjustment, while system [2] turned out to be the simplest and most reliable system, having been tested at the KVIUIV in Borisov, as well as in Nakhabino and Novosibirsk. A similar system included a surface coil with a diameter of 20 cm (coaxial with the TNT coil) and coupling coils for RDX and tetryl were wound on each of the four M61 ferrites with adjustments as in Fig. 3.

Fig.3. Tuning to RDX and tetryl frequencies

Fig. 5. Induction signal spectrum after processing by the MPM method

In [12], a comparison of signal-to-noise ratio improvement methods was made: the matrix MPM method and the linear prediction method (LPSVD) for the NQR signal from RDX. Random noise was superimposed on the original signal and an attempt was made to detect the NQR signal using the above methods. The results are presented in Table 1.

Table 1.

The table shows that both methods are approximately equal in their capabilities. Both methods can extract a useful signal with a signal-to-noise ratio of at least 0.5, which is much better than the standard Fourier transform, which does not work under noise. The LSPD method works more accurately, but the computation costs are more than 2 times higher than the MPM method. The amount of data in the calculations was N = 575. MatLab V5 was used for the calculations. At N > 2000, there were problems with LSPD, but this was not observed with the MPM method.

To digitize the signals, a program for automatic data extraction from graphs and generation of noise of the required level was written in C++ Builder V5.

Photo 1. Power amplifier of the mobile NQR detector

The device uses AR-347 –1kW power amplifiers after modification. The weight of the mobile NQR detector was about 20 kg together with a mini-computer and an LG7030 phone, which, via the LGInternetKit program, allowed Internet access using Data and GPRS Call.

The mini-helicopter is shown in photo 2. It weighs 40 kg, has a speed of 90 km/h, and has a flight time of 10 hours. The terrain was filmed using a Digimax 101 digital camera. A thermal imager was installed on the mini-helicopter platform, and its data was transmitted to the NQR detector monitor and used for preliminary mine detection. An image of a digital map of the terrain with GPS data was also transmitted to the NQR detector, and the mini-helicopter was controlled from a special remote control. One person controls the entire system. The system is brought to the work site in the trunk of a car. The NQR detector and control system are moved on a small cart. The mini-helicopter engine is 15 hp.

Detection depth is 25 cm for tetryl and hexogen, 8 cm for TNT.

Photo 2. Mini-helicopter NQR detector

The NQR signal consists of a mixture of the absorption signal V(fp) and the dispersion signal u(fp), which have different dependences on the detuning fp. The quadrature detector can isolate the pure absorption signal if j = 0 for one channel and j = p/2 for the other, but in the general case V2(fp)cos2Dj + u2(fp)sin2Dj № 1, i.e. the phase error is not compensated, and a distorted signal is observed in both channels. Working with the detuning in the SQRC program [5] becomes difficult, since the detuning cannot be made optimal. The pure absorption signal of 14N, obtained using a synchronous detector at a frequency of 5192 kHz in RDX, is shown in Fig. 4 (450 g sample at a distance of 20 cm from the coil surface). Fig. 5 shows the NQR signal in RDX at an SNR ratio of 0.5. The obtained parameters are presented in the table, which demonstrates the advantage of parametric methods when working under noise. In local NQR, SNR decreases as 1/r, where r is the distance from the sample to the surface coil, i.e. at a distance of 25 cm, SNR decreases by a factor of 25 and is 0.2, so that for the application of the LPSVD and ITMPM methods (SNR = 0.5), preliminary accumulation of the signal (10 accumulations) is required, but at 12 cm there is no need to use an accumulator, which significantly simplifies the equipment. Such an experiment was carried out.

To reduce the role of acoustic resonances occurring in various plastic parts, the quantum beat method was used, when the transitions n0 = 1782 kHz and n- = 3410 kHz were excited using two crossed semi-toroidal coils, and the radiation at the frequency n+ (5192 kHz) was received by a ring with a diameter of 20 cm, located perpendicular to the fields of the semi-toroidal coils, which significantly reduced the dead time at the n+ transition, as well as interference from acoustic resonances responsible for false alarms in the process of detecting a substance [2]. This method works well at distances of up to 10 cm, but it meets the requirements for the probability of reliable detection of 99.6%.

If we add the signal with noise S(t) + u(t) n times, then the signal at SNR > 1 is added incoherently, and the noise has dispersion

and then we get
.
From here

However, this ideal case is violated at SNR <1 , since in the first approximation 
for the signal itself – it is also subject to dispersion.

If SNR = 0.5, then

and the phase is dispersed by 390, which is much less than the noise dispersion. The SNR ratio falls as 1/r, where r is the distance from the surface coil to the sample. We assume that the pulse power for each distance is optimal, i.e.

, where ;

P is the pulse power;
g – gyromagnetic ratio of 14N nuclei;
Q – quality factor of the surface coil;
n – frequency;
V – volume of the pr-field of the coil;
tw – pulse duration.

If P = 2 kW, Q = 450; n = 5192 kHz, V = 5 ns, then the pulse duration at 20 cm will be 150 μs, and SNR = 0.1, i.e. dispersion
,
and at 35 cm
.

At large distances, the dispersion of noise and signal become the same and then
will give
,
i.e. further accumulation can no longer give anything [1]. However, the ITMPM method works at SNR = 0.5, which allows observing the NQR signal at 10 cm with a single sweep, i.e. RDX can be detected on a person in 1 s. The ratio

we find by using a large number of accumulations (10,000 — 40,000) at distances of 25 — 35 cm for RDX. If the distance increased, then

,
after which the accumulation no longer worked. Taking into account nonlinearity increases the signal dispersion by 20 cm to
,
and at 30 cm to
.
Since dispersion is the second moment of the line, then at 20 cm the NQR line in RDX broadens by 4.2 times, and at 30 cm by 5.5 times and becomes 1.6 kHz, i.e. it is necessary to reduce the duration of the exciting pulse by 5 times, increasing its power by 5 times, which was done during the experiments.

It is easy to show that at distances greater than 10 cm this error can no longer be ignored and the phase of the reference voltage needs to be adjusted. The use of parametric methods of spectral estimation and phase adjustment with increasing distance to the substance allowed us to increase the probability of reliable detection to 99.6%, which meets the UN requirements. Similar effects can be observed when rotating CCL3 groups in NQR, where the Fourier transform cannot be used under noise, and working with detuning leads to an increase in dispersion.

A special role would be given to the so-called “colored” noise. All the above results seem amazing, but it should be taken into account that the noise we modeled and summed with the signals is “white”. Its power is evenly distributed across the spectrum. In reality, we have to work with “colored noise”. Their spectral power density is modulated by various functions. They are less amenable to statistical accumulation than white” noise.

Photo 3. View of the area from a mini-helicopter

The results of the research have been discussed many times at scientific seminars in Kaliningrad, Aachen, Dortmund, Darmstadt (Germany) and Ljubljana (Slovenia), Seoul.

The use of a mini-helicopter makes it possible to reduce the time required to clear mines from an area, since the thermal imager on the mini-helicopter already provides an approximate location of the mines, where the operator is directed with a quadrupole detector, which makes it possible to accurately determine the position of the mine with a probability of 97%.

Literature

1. V.S. Grechishkin, Appl.Phys. A 55, 505 – 507, 1992.
2. V.S. Grechishkin, Appl.Phys. A 58, 63 – 66, 1994.
3. V.S. Grechishkin, N.Ya. Sinyavsky et al., Izvestiya vuzov, Physics, No. 7,58 – 61, 1992.
4. V.P. Anferov et al, Rev.Sci.Instr. 71(4), 1656 – 1659, 2000.
5. V.S. Grechishkin, Izvestiya vuzov, Physics, No. 12, 107 109, 1994.
6. T.N. Rudakov et al, Meas.Sci.Techn. 8, 444 – 448, 1997.
7. M.D. Rowe, J.A.S. Smith, Eurel International Conference, The detection of abandoned land mines, 7 – 9 October, p. 62 – 66, 1996.
8. E.O. Azizov, V.S. Grechishkin, Yu.M. Lugansky and G.I. Luganskaya. Dual-frequency pulsed NQR spectrometer 14-N. Izvestiya AN SSSR, Physical Series, Vol. 42, 10, 1978.
9. Yung-Ya Lin, P.Hodgkinson, M.Ernst, and A.Pines, A Novel Detection-Estimation Scheme for Noisy NMR Signals: Application to Delayed Acquisition Data, J.Magn.Res. 128, 30 41, 1997.
10. S. Marple, Digital Spectral Analysis with Applications, Prentice-Hall, 1987.
11. Hua Y., and Sarkar T.K., Matrix Pencil Method for Estimating Parameters of Exponentially Damped/Undamped Sinusoids in Noise, IEEE Trans.Acoustics, Speech and Signal Processing, vol.3 8, num.5, pp.814 – 824, May 1990.
12. V.S. Grechishkin, Vestnik KSU, 3, 86 – 95, 2003.