- Email: [email protected]
Simulation of ultrasound imaging inside fully charged pipes
Automation in Construction 15 (2006) 355 – 364 www.elsevier.com/locate/autcon
Simulation of ultrasound imaging inside fully charged pipes Francisco Gomez, Kaspar Althoefer *, Lakmal D. Seneviratne 1 King’s College London, Department of Mechanical Engineering, Strand, London WC2R 2LS, UK Accepted 7 July 2005
Abstract This paper describes an approach for the simulation of ultrasonic scans in water-filled pipes. Comparison of model predictions against experimental results shows the capability and accuracy of the simulation algorithms to generate realistic ultrasound images. The proposed routines were particularly developed for the modeling and simulation of rotating ultrasonic transducers inside tubular structures, but it is expected that this research could be expanded for the ultrasonic-based inspection of other water-filled structures, such as water tanks. The research provides insights into the behavior of the ultrasonic fields inside tubular structures and is a useful aid for the development and design of underwater inspection systems. D 2005 Elsevier B.V. All rights reserved. Keywords: Sewers assessment; Ultrasound; Pipe inspection; Simulation
1. Introduction Pipe inspection in recent years has become a serious issue for water companies around the world because environment offices in many countries have dramatically increased the regulations regarding damaged water pipes and sewers, aiming to reduce the loss of fresh water and pollution of the environment. The most widely used method of collecting information on the internal condition of sewers is closed-circuit television (CCTV) inspection. It is used to detect pipe defects and decreases in flow capacity due to reductions in the cross sectional area of the pipe [4]. Makar stated that CCTV is likely to remain the preferred method of inspection in the near future. Nevertheless, laser and ultrasonic inspection methods will gradually replace conventional CCTV as the preferred method due to their ability to make quantitative
* Corresponding author. Tel.: +44 020 7848 2431. E-mail addresses: [email protected] (F. Gomez), [email protected] (K. Althoefer), [email protected] (L.D. Seneviratne). 1 Tel.: +44 020 7848 2236. 0926-5805/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.autcon.2005.07.001
measurements of sewer damage [8]. As a matter of fact, the quality of inspection of most CCTV methods relies on the skill and experience of the operator, together with the reliability and quality of the TV pictures and can be misleading because vital defects can be concealed by mud and moss [4]. In any case, inspection of sewer pipe assets is expensive, protractive, subjective, and involves working in hostile environment [20]. Automated sewer inspection is a relatively new area. Previously, sewer inspection was a manual task based on visual inspection, limited to pipe sizes which can accommodate a human being. The first commercial attempts at unmanned inspection date back to the early 1990s, using video cameras installed on mobile platforms pulled by wires. As research evolved and new sensor and robot technologies became available, the limitations of simple sensor systems were addressed and new requirements fulfilled. In recent years, research has focused on building multi-sensor systems for pipe inspection [3]. For example, Kurt is a six-wheel autonomous mobile platform with a sensor configuration consisting of two inclinometers and five ultrasound sensors for navigation. It also has a CCTV camera which has an on-board frame grabber with a radio link module to transmit images and relevant
356
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
information to the outside world [6]. Secondly, KARO, a remotely controlled and highly maneuverable robot inspection system, is designed to automatically assess sewer pipes; it is equipped with a 3D optical sensor, ultrasonic sensors in the head for vehicle navigation and in the belly to measure wall thickness. It also has a microwave back scattering sensor for inspection of the soil surrounding the pipe wall [22]. Thirdly, MAKRO is a multi-segment, articulated robot equipped with inclinometers in each member. In the front section, there are four ranging sensors (two ultrasound and two infrared) for obstacles detection, a stereo camera head, two laser projectors, and a lamp [11,23]. Fourthly, Pipe Inspection Real-time Assessment Technique (PIRAT) is a multi-sensor, tele-operated robot able to assess defects in sewers and other pipes. The PIRAT measurement system is composed of a laser scanner which repeatedly measures the pipe radius as its head rotates around the sewer axis to build up the geometry of cross-sections, leading to a 3D sewer surface map [7]. Sewer Scanner and Evaluation Technology (SSET) is an autonomous robot equipped with an optical scanner with a fish-eye lens and a gyroscope, and it has the capacity to log images during the inspection. These images are post-processed by specialized routines to reveal defects and flaws which are finally analyzed by an expert [24]. As can be seen, great advances have been made towards multi-sensor autonomous pipe inspection platforms. But commercial pipe inspection today is still mainly camera based, focusing primarily on the dry sections of pipes [3]. Before inspection takes place, a section of a sewer is commonly drained and blocked off if the whole pipe is to be viewed. Therefore, the inspection of sewers in normal service, where the pipe is partially or fully filled with water, is a research niche that needs to be addressed. Consequently, research efforts are now also focusing on pipe inspection methods, using ultrasound for underwater sections [7,10,17,19]. Tracking the progress of the developments in this particular field, a common feature is the automation of the pipe inspection process. As a matter of fact, it not only can save significant time and money, but also improve accuracy and consistency in diagnostics. Automating these processes can also provide an incentive for checking water and sewer mains regularly as a part of preventive maintenance programs [9]. Based on the above requirement, the Centre for Mechatronics and Manufacturing Systems at Kings College London is developing an autonomous multi-sensor mobile platform for pipe inspection capable of working inside sewers during normal operation, and to be equipped with sensors able to inspect dry and wet sections of pipes simultaneously [18]. The work presented in this publication is part of this multidisciplinary program, focusing on the inspection of the pipe below liquid level. Although, ultrasonic inspection may miss cracking and other small defects, it is capable of providing a quick, quantitative assessment of sewer deformation and other problems,
indicating not just the presence of a problem, but also its extent [8]. Ultrasonic inspection methods have been modestly investigated in the context of sewer inspection, involving mainly heuristic strategies providing results which support the feasibility of ultrasonics for pipe inspection [10,16,18]. Modeling ultrasound fields in medical imaging and in nondestructive testing is a relative mature research area. However, to the best knowledge of the authors no attempt has been made to simulate the behavior of ultrasonic signals inside liquid-filled tubular structures with clean water and sewage. By employing a modeling tool for the simulation of ultrasonic fields inside sewers will substantially benefit the study, design and analysis of ultrasonic inspection systems by shortening the test stages, and therefore, reducing costs and development time. The presented work attempts to bridge this gap using a systematic approach extending an ultrasound model originally developed for medical imaging applications, to simulate the behavior of ultrasound fields generated by rotational ultrasonic scanners inside pipes. Rotating ultrasonic sensors are being used to provide images of inner pipe walls immersed in sewage [16]. However, its capabilities have not been fully exerted. The ultrasonic inspection process consists of scanning a series of consecutive axial views while the sonar moves longitudinally inside the pipe at defined intervals (see Fig. 1). In order to scan a whole section of a sewer, a rotationalultrasonic scanner is mounted on a device which moves along the longitudinal axis of the pipe. The final goal is to generate a detailed three-dimensional profile of the internal pipe surface which can be used to assess the internal condition of it. Conditions such as sediment accumulation, roots intrusion, deformation of the pipe wall and cracks are of great interest to the sewer inspectors. The recovered profiles of the pipe can later be analyzed by an expert or by an automatic diagnostic system to find anomalous conditions that might influence the structural characteristics and mechanical integrity of the pipe. The commonly used pipes for sewer applications have nominal sizes that vary from hundreds to thousands of
z sor Sen US
Pipe
y
x ic aph ogr Tom slices
ng tati Ro am Be
Fig. 1. Ultrasonic inspection of pipes. The ultrasonic head is rotated inside the pipe creating a tomographic slice (B-mode image). To inspect a pipe section, the sensor is moved along the longitudinal axis of the pipe.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
y
millimeters in diameter, being usually made of concrete, clay, or PVC.
Homogenous Medium
2. Modeling methodology
Bvn ðt Þ ThðY r 1; Y r 2 ; t Þ; Bt
ð1Þ
where, p is the incident field at a point in the medium, vector Y r 1 is a point in space where the ultrasonic field is studied, t is the time for the snapshot of the spatial distribution of the pressure, q 0 is the density of the medium, v n is the normal velocity of the vibrating surface and hðY r 1 ; tÞ is the spatial impulse response [25]. The spatial impulse response was an approach developed by Tupholme [15] and Stepanishen [12,13] as an alternative to simplify the complex problem of determining the ultrasound field that reaches a reflector. The spatial impulse response is function of the relative position between the transducer and the observation point. It characterizes the
Electro-mechanical Impulse response (Tx)
Emitted Field (PSF/Eq. 1) Scattered Field (Eq. 3)
Electro-mechanical Impulse response (Rx)
Scatterer
r1
The modeling approach simulates ultrasound fields for different kinds of transducers through the use of linear acoustics. This was achieved by using the ultrasound simulation program developed by Jensen at the Technical University of Denmark, Field II [21]. Field II was created with the aim of simulating ultrasound imaging systems used in medical applications. The novelty of the presented work is the addition of routines to the Field II program for the simulation of ultrasound imaging systems inside tubular structures filled with liquid and its validation through an experimental study. For the sake of showing the capabilities of the modeling tool used, a brief description of the modeling methodology employed by Jensen is provided. Field II is based on linear systems theory (Fig. 2). After the electric pulse is applied to the crystal, a pressure field is irradiated into the medium, propagating in it until reaching the object (phantom). In order to find the echo, the impingent field in the scatterers composing the phantom has to be determined. To simplify the analysis, the centre of the transducer was located at the origin of the coordinate system (Fig. 3). The incident field is determined by convolving the excitation function with the spatial impulse response [5,14]: pðY r 1; Y r 2 ; t Þ ¼ q0
357
Received field (PSF/Eq. 1)
Fig. 2. Block diagram illustrating the simulation process.
r2 z x Fig. 3. Illustration of variables used in the model.
three-dimensional extent of an ultrasonic field for a given transducer, jrY1 Y r 2j Z d t c dS; ð2Þ hðY r 1 ; tÞ ¼ Y Y 2pjr r 2j 1 S where, Y r 2 is a point on the transducer surface (see Fig. 3). Eq. (2) is also described in the literature as the point spread function (PSF) representing the output of an ultrasound system during the scanning of an ideal point target located in space; this term is usually used to refer to two-dimensional representations [26]. The PSF is a function of the relative position between the location of the transducer elements and an observation point in space. Once the scattered field is determined by using the impingent field and the wave equation, reciprocally the received field is determined using the same methodology used for the determination of the impingent field. The field scattered from the phantom is determined by solving a suitable wave equation. The wave equation used in Field II was derived in a similar way as Chernov did in 1960 [1]. In order to solve the wave equation, some assumptions were made: & The instantaneous pressure is equal to the mean pressure of the medium plus the pressure disturbance caused by the ultrasound wave. & The instantaneous density is considered to be equal to the density of the undisturbed medium plus the density change caused by the passing of the wave. & The pressure and density changes created by the ultrasonic wave are considered first-order small quantities. & No heat conduction or conversion of ultrasound energy into thermal energy takes place. The wave equation used by Jensen is: l 2 p1
1 B 2 p1 2Dc B2 p1 1 ¼ þ lðDqÞIlp1 ; 2 3 2 2 q0 c0 Bt c0 Bt
ð3Þ
where, p 1 and q are the pressure and density variations, respectively, caused by the displacement of the ultrasonic
358
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
wave. The density of the medium is q 0 and the speed of the sound in the medium is c 0. l2 is the Laplacian operator, l the gradient operator and I the scalar product. A detailed derivation of this wave equation can be found in [25]. Knowing the emitted field, the scattered field, and the received field, the electro-mechanical coupling of the crystal during transmission and reception has to be determined to find the received signal (Fig. 2). Symbolically, the final expression used by Field II to find the received signal is: pr ðY r 2 ; t Þ ¼ mpe ðt Þ‘fm ðY r 1 Þ ‘ hpe ðY r 1; Y r 2 ; t Þ; r
ð4Þ
where ‘ is time Rconvolution, ‘r is the spatial convolution calculated using V f ðr1 Þd3 r1 and p r is the received signal. The function, v pe(t), describes the transducer excitation and the electro-mechanical impulse response during emission and reception of the pulse, mpe ðt Þ ¼
q0 B3 mðt Þ Em ðt Þ‘ : 2 Bt 3 2c0
ð5Þ
The term, f m, accounts for the inhomogeneities in the medium due to density and propagation velocity perturbations which give rise to the scattered signal, fm ðY r 1Þ ¼
DqðY r 1 Þ 2DcðY r 1Þ : q0 c0
ð6Þ
The modified pulse-echo spatial impulse response, h pe, that relates the transducer geometry to the spatial extent of the pressure fields, hpe ðY r 1; Y r 2 ; t Þ ¼ hðY r 1; Y r 2 ; t Þ‘hðY r 2; Y r 1 ; t Þ:
ð7Þ
The Field II program requires some key parameters in order to simulate an ultrasonic scan. These parameters are: the transducer centre frequency, the transducer radius, the size of the elements the transducer face is going to be divided into, the sampling frequency, the average speed of the sound of the medium, the frequency-independent attenuation and the frequency-dependent attenuation, and finally the phantom (the reflecting object) as an array of points (represented by x, y and z coordinates) (Fig. 4).
The object to be simulated is defined as an array of evenly spaced scatterers, located inside a homogeneous medium, here representing the inner surface of the pipe. The distance between individual scatterers (mesh size) must be chosen appropriately. Reducing distances between scatterers leads to an increase in the number of scatterers per axial section of pipe, which in turn impacts directly on the processing requirements, consequently, demanding more memory and increasing the processing time required to run the model. However, when the distance between scatterers is increased in excess, the resulting image may appear as a discontinuous pipe wall. The optimum value for this parameter to properly represent a pipe is mainly defined by the beam width of the transducer which is function of its frequency and geometry. On the other hand, the optimum scatterer separation to represent a pipe with a given defect is also determined by the size of the defect. The best performance is determined by trial and error varying the distances over a large range and analyzing the response. For the particular case of simulating ultrasonic fields for inspecting 300 mm diameter pipes, a scatterer separation in the range from 0.1 mm to 1.0 mm gives acceptable responses. It was found that, as a rule of thumb, the distance between the scatterers should be at least half of the minimum size of any simulated defect. The investigated transducer used for the validation of the simulations was a single piston transducer with a frequency of 3 MHz. As mentioned beforehand, the centre of the transducer is located at the origin of the coordinate system and its front face is aligned with the x –y plane. Because, the model used by Field II assumes that the transducer is mounted on an infinitely rigid baffle, the ultrasound field is emitted in both z-directions (z and z+). Standard ultrasonic transducers used in pulse-echo applications are mounted on backing material to reduce the undesirable effects caused by the mismatch in the impedance between the transducer and the medium, which creates a long ring-down. The simulation of the interaction between the backing material and the transducer is beyond the scope of this work. Because the
Input parameters: Transducer centre frequency (f ) Transducer radius (r ) Size of mathematical elements (w, h) Sampling frequency (fs) Speed of sound in medium (c0) Phantom (pipe dimensions, defect) Pulse duration Area to be scanned - Initial longitudinal position - Final longitudinal position - Longitudinal step size (long. resolution) - Initial angle - Final angle - Angular step size (angular resolution)
Output: B mode images: - RF lines (θ,t ) - Position (y )
Fig. 4. Input/output parameters for the simulation algorithms.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
ultrasound field is propagating from the front and rear faces of the transducer, the pressure waves propagating from the rear face in the negative z-direction can reflect in the scatterers located in the negative z half-space and return to the transducer, getting mixed with the signal of interest from the front face. In order to overcome this limitation which arises when creating images from pipes immersed in water, a scatterer filter function was implemented. This function analyzes the phantom and the location of the transducer before starting the simulation of the field and removes all the scatterers located in the negative z half-space. The pipe inspection process requires that the transducer rotates around the y-axis (longitudinal axis of the inspected pipe) and translates along it (Fig. 1). However, the original Field II routines do not allow rotation or translation of the transducer. In order to create a 3D profile of an internal section of a pipe, extensions to the Field II program were developed. The developed routines overcome this limitation by rotating and translating the environment to simulate the desired movement of the transducer. It is noted that, instead of rotating and displacing the transducer to scan the object, the object is moved and displaced accordingly, while the transducer is kept still at the origin of the coordinate system.
3. Simulations and experiments In order to validate the developed algorithms for the sewer inspection case, simulations and comparative experiments were conducted. Three experiments were carried out. Firstly, a 0.5-mm cooper wire immersed in clear water was scanned and simulated. The corresponding results were quantitatively compared to validate the accuracy of the simulations through a lateral beam width measurement. Finally, two experiments using Polyvinyl Chloride (PVC-U) pipes, same as the ones found in sewers, with precisely carved defects were carried out. In both cases, experiments and simulations were conducted and the results compared. 3.1. Experimental platform The experimental study was conducted in the Pipe Inspection Lab in the Centre for Mechatronics and Manufacturing Systems at Kings College London. This facility is equipped with a 300-l water tank (100 60 50 cm), a rotational ultrasonic transducer, model SeaKing, manufactured by Tritech International and a SCARA Robot. The sonar system was supplied by one of the project sponsors. The supplied system has the same characteristics as the one that is expected to be used for real pipe inspections. The sonar system houses in the sonar head the electronics responsible for: the generation and receipt of the pulse/echo signal, measurement of the envelope amplitude of the echo, analog to digital conversion, and transmission of the acquired data, via a high speed Arcnet serial link to the host computer. The transducer installed in
359
the sonar is a single piston, unfocussed, piezoelectric transducer made of Lead Zirconate Titanate (PZT). It has a resonant frequency of 3 MHz, a 10-mm diameter and is mounted on a rotating platform. For protection from contact with water, the transducer is enclosed in a rubber cap filled with inhibited mineral insulating oil commercially known as Shell Diala MX. The angular resolution, the duration of the pulses and the sampling rate are user configurable parameters. The angular resolutions are: 1.8-, 0.9-, 0.45- and 0.225-. The minimum pulse duration is 10 As and the fastest sampling rate is 162 kHz. The analogue to digital converter has a resolution of 8 bits. 3.2. Lateral beam width measurement The main objective of this study was to measure the radiation pattern of the transducer and compare it with the predicted one obtained from simulations. The basic measurement to characterize the radiation pattern of a given transducer is the beam width, which is defined as the angular distance between the points in the radiation pattern were the intensity decays 3 dB from its maximum value. The radiation pattern is a 3D characteristic which describes how the energy of the emitted field radiates in the medium. The lateral beam width was measured in water by rotating the sonar head and recording position and intensity of the echoes generated by a thin copper wire of 0.5 mm diameter. The wire was located at a distance of 440 mm and was aligned parallel with the face of the transducer. For the simulation, the centre of the transducer’s face was located at the origin of the coordinate system and the phantom of the wire was positioned parallel to the y-axis. The phantom was created with a mesh size of 0.1 mm, being composed of 24 scatterers. The z-axis was the reference axis for the rotation of the sensor, for instance the 0- position of the transducer lies on the z-axis (Fig. 5). The sensor was rotated around the y-axis, from 10- to 10- about its centre with a step resolution of 0.225-. In order to facilitate the comparison of the experimental and simulated results, the simulated results were downsampled to the match the sampling frequency of the sonar system. The simulator tool generates sampled data at a frequency of 100 MHz, while the sonar system samples at a maximum rate of 162.5 kHz. The excess samples acquired at 100 MHz lying between two 162.5 kHz samples were averaged. In Fig. 6, simulated and experimental signals are shown; the left plots are B-mode images and the right ones are plots of intensity against angular position. From the B-mode images, it can be seen that the shape and intensity profile of the returning echoes are in good agreement with just slight deviations present in the experimental data (Fig. 6(c)) where a shadow is visible at approximately 650 As. This phenomenon could be attributed to the ring-down, i.e. the period of vibration of the crystal after being struck by a short voltage pulse [2]. By closely studying the pulse
360
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
(a)
(b) 441 440.8
400
440.6
300
440.4 200
z [mm]
z [mm]
Wire
100 0 50
440.2 440 439.8
40
439.6 30
20
Front face of sensor 10
0
10
y [mm]
20
30
40
439.4
50
5 0 x [mm] 5
439.2 439 1
0.5
0
0.5
1
x [mm] Fig. 5. (a) 3 MHz transducer and 0.5 mm diameter string phantom. (b) Magnified view of the phantom in the x – z plane.
Fig. 6. 3-D mesh plot, received echoes from 0.5 mm copper wire. Simulated data: (a) axial – lateral plane view; (b) lateral view. Experimental data: (c) axial – lateral plane view; (d) lateral view.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
3.3. Square defect A Polyvinyl Chloride (PVC-U) pipe with a precisely carved squared defect was scanned and simulated. The plastic pipe had a nominal diameter of 300 mm and a measured inside diameter of 296 mm. The carved defect investigated was a carefully cut square hole. The hole dimensions were 10 10 mm. The hole was centred over the x-axis and extended along the y-axis from 70 to 80 mm. The hole and the surrounding area were scanned by moving the transducer along the y-axis 42 mm at 3 mm steps, sweeping from 10.0- to 10.0- at 1- intervals. This results in a rectangular surface of approximately 63 by 42 mm Table 1 Beam width table, experimental and simulated measurements 3 dB beam width
Experimental (-)
Simulated (-)
Difference (%)
2.66
2.2
17
0 5 10 15
Amplitude (db)
generated by the employed sonar system, it was appreciated that even though the electrical pulse which excites the transducer perfectly resembles a square pulse, the emitted pressure pulse showed a main peak which concentrates most of the energy inside it, decreasing monotically, some times accompanied by a tail-peak (ring-down). The duration of the ring-down is about 10 times the duration of the emitted pulse (10 As). This shows that the ultrasound system has a ‘‘high Q’’ and as a consequence a long spatial pulse length which degrades its axial resolution [2]. This ring down effect is reduced by attaching a high density damping material to the rear side of the transducer but this issue has not been simulated as part of this paper. In Table 1, the measured beam widths for the experimental and simulated cases are shown. The difference between them is calculated using the experimental data as reference. There is a reasonable and consistent match, considering the limited precision of the experimental platform. In Fig. 7, the profile of the simulated and experimental beams are compared. Only the main lobe can be compared because the noise level of the experimental platform is around 30 dB, which is so strong that the returning echoes coming from the side lobes are covered. An acceptable match between the two signals can be seen in Fig. 7. The simulated radiation pattern tends to be slightly narrower than the experimental one. The difference between the experimental and simulated results may be due to refraction caused at the interface between the rubber cap filled with transformer oil and the water which was not considered in the simulation. The transformer oil in which the transducer is immersed has lower density than water, making the beam diverge. If the refraction effect created at the boundary of the two media is not considered the simulated beam width will be narrower than the real one, in agreement with the results shown in Fig. 7 and Table 1.
361
20 25 30 35 40 45 50 –10
–8
–6
–4
–2
0
2
4
6
8
10
Transducer position (deg)
Fig. 7. Beam widths comparison: experimental data (continuous line), simulated data (diamond line).
being scanned. The scanning process generated 14 B-mode images, representing tomographic slices of the pipe. To compare the experimental data and the simulation results, an appropriate visualization methodology has to be employed. From the set of consecutive B-mode images acquired during the scanning process, a 3D profile of the inner surface of the pipe was created. The profile was created by measuring the time of flight of the emitted pulse and registering the maximum intensity of the received echo. As a consequence, four variables were obtained: x, y and z indicating the position of the reflector that generated the echo and the echo intensity. A useful way to represent these profiles is through the use of 3D color plots that will be referred to as surface – intensity plots, showing the inner surface of the pipe and the received intensity. In Fig. 8, surface– intensity plots for simulated and experimental data are presented. For the simulated case presented in Fig. 8(a), the removed section can be easily identified as the area where the intensity decreases until reaching a minimum (dark area). Even though, the hole is a perfect square, the resulting image is not a square. It is noted that the shape identified in the surface –intensity plot image is different from the original shape due to constructive and destructive interference occurring at the corners of the square. A closer analysis of the acquired image shows that the contrast between the intensity from echoes returned by the intact wall and the centre of the removed section is around 14 dB, decreasing smoothly from 0 dB to 14 dB. In Fig. 8(b), the experimental results are presented. The experimentally scanned signals are notoriously noisier than the simulated signals, as expected for a real system. In the presented case, the defect is easily recognizable in the surface –intensity plot as an area where the intensities of the returning echoes decrease until reaching a minimum at the centre of the defect. Comparing the echo intensities returning from the surface of the pipe wall and the centre of the hole, a contrast of 14 dB can be appreciated being in perfect
362
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
(a) 105
0
100
-2
95
-4
90
y
-6 85 -8 80 -10
75
-12
70 65
-14 -30
-20
-10
0
10
20
30
x
(b) 105 -52
100
-54
95
(a)
0
-56
80
85
-58
70
80
-60
60
75
-62
70
-64
65
-66
90
-1 -2 -3
y
y
simulated. The carved shape was a triangle with 39 mm base and 60 mm sides. The centre of the triangle base was aligned parallel to the x-axis, going from 19.5 mm to 19.5 mm. The triangle height was 56.74 mm, extending along the y-axis from 20 mm to 76.74 mm. The triangle was scanned moving the transducer along the y-axis from 10 mm to 88 mm at 3 mm steps and rotating the transducer from 11.7- to 11.7- at 0.225- intervals. Hence, the scanned inner surface of the pipe was a 60 mm by 78 mm rectangle. The scanning process generated 27 B-mode images, each one representing a tomographic slice of the pipe. The internal surface of the pipe was profiled from the set of B-mode images by applying a time of flight (TOF) profiling algorithm. A TOF profiling algorithm detects the location of the first reflection inside the acquired RF lines using a pre-fixed threshold. For this particular case, the threshold value was fixed at 10 dB. Results are presented in surface – intensity plots, dashed lines representing the removed section as measured by means of a ruler are overlaid (Fig. 9). In the presented surface –intensity
-30
-20
-10
0
10
20
-4 -5
50
-6
40
-7 30
-8
30 20
x Fig. 8. Surface – intensity plots for square defect. (a) Simulation. (b) Experiment.
-10 -30
-20
-10
0
10
20
30
x
(b)
-55 130 -56 120
-57
110
-58
100
-59
y
agreement with the simulation. In addition, the defect size appreciated in the experimental data is smaller than the one appreciated in the simulated one. This is caused by the difference in the beam width size between them. A wider beam width results in a poorer resolution which can be seen in the smaller dark area in the surface –intensity plot presented in Fig. 8. The deformations appreciated in the plot can be attributed to the differences between the lateral resolution of the stepper motor and the longitudinal resolution of the SCARA robot. From the results of this study case, it can be concluded that the simulation tool is representing the physical phenomenon in an accurate and acceptable way.
10
-9
-60
90
-61
80
-62 70 -63 60 -64
3.4. Triangle defect A PVC-U pipe with a nominal diameter of 300 mm and a measured inside diameter of 296 mm pipe with a precisely carved isosceles triangle was experimentally scanned and
50 -30
-20
-10
0
10
20
30
-65
x Fig. 9. Surface – intensity plots for a triangle defect carved in the pipe wall. (a) Simulation. (b) Experiment.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
plots, the white section in the centre of the triangle shows the segment of the pipe where the reflected echoes are very weak or no echoes are detected. It can be seen that the sides of the triangle delineated by the algorithm in both cases resemble stepped lines. This is due to the 3 mm axial step size used during the scanning process along the y-axis (quantization error) and does not affect the purpose of the study. As was presented in Section 3.2, the measured beam width of the experimental transducer was wider than the simulated one. Therefore, the white area in the experimental case (Fig 9(b)) is smaller than in the simulated case (Fig. 9(a)). During experiments, it was observed that the off centre location of the transducer inside the pipe is a key parameter that affects the quality of the resulting images. In this particular experiment, the transducer was slightly misaligned towards the positive x direction. As a result, the plot in Fig. 9(b) shows a slight asymmetry when representing the triangular hole; the left side of the triangular hole is depicted by a wider grey scale transition along the boundary on the left side (12 mm) when compared to the right side (9 mm). Another indication of the off centre condition of the scanner are the tenuous shadows present at the top-right quadrant of the plot. As has been shown through experiments, the developed simulation tool is able to create ultrasound simulated images inside fully charged pipes in an acceptable way.
4. Conclusions A valuable addition to the Field II routines, usually applied to medical ultrasound imaging modeling, has been developed to simulate ultrasound fields inside fully charged pipes. The presented approach is capable of generating simulated images from given pipes which are constructed from a series of reflectors or scatterers. Defects or anomalies in the pipe wall are represented by removing or adding reflectors which emulate the defect geometry. The presented tool is oriented towards the simulation and visualization of rotating ultrasound transducers providing the creation of two-dimensional (B-mode) and three-dimensional scans inside tubular structures. By means of an experimental study and detailed comparison of real and simulated data, the validity of the developed approach was shown. The presented work can be used to conduct in-depth investigations of ultrasonic based pipe inspection systems, by giving an invaluable insight into the sound wave propagation in fully charged pipes. In addition, it will contribute to facilitate the design of future pipe inspection systems because of having the means to validate the designs before attempting time-consuming measurements in a real scenario. References [1] L.A. Chernov, Wave Propagation in a Random Medium, McGrawHill, New York, 1960.
363
[2] K. Gustav, H. Smith, D. Allison, The Encyclopaedia of Medical Imaging, Volume I. Physics, Techniques and Procedures. NICER Institute. Isis Medical Media (1998). [3] O. Duran, K. Althoefer, L. Seneviratne, State of the art in sensor technologies for sewer inspection, IEEE Sensors Journal 2 (2) (2002) 63. [4] R.A. Fenner, Approaches to sewer maintenance: a review, Urban Water 2 (4) (2000) 343 – 356. [5] J.A. Jensen, A model for the propagation and scattering of ultrasound in tissue, Journal of the Acoustical Society of America 89 (1991) 182 – 191. [6] F. Kirchner, J. Hertzberg, A prototype study of an autonomous robot platform for sewerage system maintenance, Autonomous Robots, vol. 4 (4), Academic Publishers, Netherlands, 1997, pp. 319 – 333 (ISSN: 0929-5593). [7] R. Kirkham, P.D. Kearney, K.J. Rogers, PIRAT—a system for quantitative sewer pipe assessment, The International Journal of Robotics Research 19 (11) (2000 (Nov.)) 1033 – 1053. [8] J.M. Makar, Diagnostic techniques for sewer systems, Journal of Infrastructure Systems 5 (2) (1999 (June)) 69 – 78. [9] O. Moselhi, T. Shehab-Eldeen, Automated detection of surface defects in water and sewer pipes, Automation in Construction 8 (1999) 581 – 588. [10] N.G. Pace, Ultrasonic surveying of fully charged sewage pipes, Electronics & Communication Engineering Journal 6 (2) (1994 (Apr.)) 87 – 92. [11] E. Rome, J. Hertzberg, F. Kirchner, U. Licht, T. Christaller, Towards autonomous sewer robots: the MAKRO project, Urban Water 1 (1999) 57 – 70. [12] P.R. Stepanishen, The time-dependent force and radiation impedance on a piston in a rigid infinite planar baffle, Journal of the Acoustical Society of America 49 (3) (1971) 841 – 849. [13] P.R. Stepanishen, Transient radiation from pistons in a infinite planar baffle, Journal of the Acoustical Society of America 49 (1971) 1627 – 1638. [14] P.R. Stepanishen, Pulsed transmit/receive response of ultrasonic piezoelectric transducers, Journal of the Acoustical Society of America 69 (1981) 1815 – 1827. [15] G.E. Tupholme, Generation of acoustic pulses by baffled plane pistons, Mathematika 16 (1969) 209 – 224. [16] M.E. Andrews, Large Diameter Sewer Condition Assessment Using Combined Sonar and CCTV Equipment, APWA International Public Works Congress, NRCC/CPWA Seminar Series ‘‘Innovations in Urban Infrastructure’’, 1998. [17] R. Browne, G. Knott, Television and Scanning Sonar in Seattle Metro’s Siphons and Brick Sewers, ASCE International Conference on Pipeline Infrastructure II, San Antonio, TX, 1993, pp. 694 – 701. [18] O. Duran, K. Althoefer, L. Seneviratne, Automated sewer pipe inspection through image processing, Proc. 2002 IEEE Int. Conf. on Robotics and Automation (ICRA’2002), Washington, D.C., 11 – 15 May, 2002, pp. 2551 – 2556. [19] F. Gomez, K. Althoefer, L. Seneviratne, Modelling of rotating ultrasonic sensors for pipe inspection, International Conference on Mechatronics, Loughborough, UK, 20032002. [20] R.M. Gooch, T.A. Clarke, T.J. Ellis, A semi-autonomous sewer surveillance and inspection vehicle, Proc. IEEE Intelligent Vehicles, Japan, 1996, pp. 64 – 69. [21] J.A. Jensen, Field: a program for simulating ultrasound systems, Paper Presented at the 10th Nordic – Baltic Conference on Biomedical Imaging Published in Medical and Biological Engineering and Computing, Volume 34, Supplement 1, Part 1, 1996, pp. 351 – 353. [22] H.B. Kuntze, H. Haffner, Experiences with the development of a robot for smart multisensoric pipe inspection, Proceedings of the 1998 IEEE International Conference on Robotics and Automation, Belgium, 1998. [23] E. Rome, H. Surmann, H. Streich, U. Licht, K. Paap, A custom IR scanner for landmark detection with the autonomous sewer robot
364
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
MAKRO, International Symposium on Intelligent Robotic Systems, 2001. [24] CERF, Evaluation of SSET: The Sewer Scanner and Evaluation Technology. The Civil Engineering Research Foundation, Report: #40551, March (2001). [25] J.A. Jensen, Linear description of ultrasound imaging systems, Notes for the International Summer School on Advanced Ultrasound
Imaging, Technical University of Denmark July 5 to July 9, 1999, Technical University of Denmark, June, 1999. [26] M. Anderson, G. Trahey, A seminar on k-space applied to medical ultrasound, Web publication, Department of Biomedical Engineering, Duke University, 2000.
Simulation of ultrasound imaging inside fully charged pipes Francisco Gomez, Kaspar Althoefer *, Lakmal D. Seneviratne 1 King’s College London, Department of Mechanical Engineering, Strand, London WC2R 2LS, UK Accepted 7 July 2005
Abstract This paper describes an approach for the simulation of ultrasonic scans in water-filled pipes. Comparison of model predictions against experimental results shows the capability and accuracy of the simulation algorithms to generate realistic ultrasound images. The proposed routines were particularly developed for the modeling and simulation of rotating ultrasonic transducers inside tubular structures, but it is expected that this research could be expanded for the ultrasonic-based inspection of other water-filled structures, such as water tanks. The research provides insights into the behavior of the ultrasonic fields inside tubular structures and is a useful aid for the development and design of underwater inspection systems. D 2005 Elsevier B.V. All rights reserved. Keywords: Sewers assessment; Ultrasound; Pipe inspection; Simulation
1. Introduction Pipe inspection in recent years has become a serious issue for water companies around the world because environment offices in many countries have dramatically increased the regulations regarding damaged water pipes and sewers, aiming to reduce the loss of fresh water and pollution of the environment. The most widely used method of collecting information on the internal condition of sewers is closed-circuit television (CCTV) inspection. It is used to detect pipe defects and decreases in flow capacity due to reductions in the cross sectional area of the pipe [4]. Makar stated that CCTV is likely to remain the preferred method of inspection in the near future. Nevertheless, laser and ultrasonic inspection methods will gradually replace conventional CCTV as the preferred method due to their ability to make quantitative
* Corresponding author. Tel.: +44 020 7848 2431. E-mail addresses: [email protected] (F. Gomez), [email protected] (K. Althoefer), [email protected] (L.D. Seneviratne). 1 Tel.: +44 020 7848 2236. 0926-5805/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.autcon.2005.07.001
measurements of sewer damage [8]. As a matter of fact, the quality of inspection of most CCTV methods relies on the skill and experience of the operator, together with the reliability and quality of the TV pictures and can be misleading because vital defects can be concealed by mud and moss [4]. In any case, inspection of sewer pipe assets is expensive, protractive, subjective, and involves working in hostile environment [20]. Automated sewer inspection is a relatively new area. Previously, sewer inspection was a manual task based on visual inspection, limited to pipe sizes which can accommodate a human being. The first commercial attempts at unmanned inspection date back to the early 1990s, using video cameras installed on mobile platforms pulled by wires. As research evolved and new sensor and robot technologies became available, the limitations of simple sensor systems were addressed and new requirements fulfilled. In recent years, research has focused on building multi-sensor systems for pipe inspection [3]. For example, Kurt is a six-wheel autonomous mobile platform with a sensor configuration consisting of two inclinometers and five ultrasound sensors for navigation. It also has a CCTV camera which has an on-board frame grabber with a radio link module to transmit images and relevant
356
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
information to the outside world [6]. Secondly, KARO, a remotely controlled and highly maneuverable robot inspection system, is designed to automatically assess sewer pipes; it is equipped with a 3D optical sensor, ultrasonic sensors in the head for vehicle navigation and in the belly to measure wall thickness. It also has a microwave back scattering sensor for inspection of the soil surrounding the pipe wall [22]. Thirdly, MAKRO is a multi-segment, articulated robot equipped with inclinometers in each member. In the front section, there are four ranging sensors (two ultrasound and two infrared) for obstacles detection, a stereo camera head, two laser projectors, and a lamp [11,23]. Fourthly, Pipe Inspection Real-time Assessment Technique (PIRAT) is a multi-sensor, tele-operated robot able to assess defects in sewers and other pipes. The PIRAT measurement system is composed of a laser scanner which repeatedly measures the pipe radius as its head rotates around the sewer axis to build up the geometry of cross-sections, leading to a 3D sewer surface map [7]. Sewer Scanner and Evaluation Technology (SSET) is an autonomous robot equipped with an optical scanner with a fish-eye lens and a gyroscope, and it has the capacity to log images during the inspection. These images are post-processed by specialized routines to reveal defects and flaws which are finally analyzed by an expert [24]. As can be seen, great advances have been made towards multi-sensor autonomous pipe inspection platforms. But commercial pipe inspection today is still mainly camera based, focusing primarily on the dry sections of pipes [3]. Before inspection takes place, a section of a sewer is commonly drained and blocked off if the whole pipe is to be viewed. Therefore, the inspection of sewers in normal service, where the pipe is partially or fully filled with water, is a research niche that needs to be addressed. Consequently, research efforts are now also focusing on pipe inspection methods, using ultrasound for underwater sections [7,10,17,19]. Tracking the progress of the developments in this particular field, a common feature is the automation of the pipe inspection process. As a matter of fact, it not only can save significant time and money, but also improve accuracy and consistency in diagnostics. Automating these processes can also provide an incentive for checking water and sewer mains regularly as a part of preventive maintenance programs [9]. Based on the above requirement, the Centre for Mechatronics and Manufacturing Systems at Kings College London is developing an autonomous multi-sensor mobile platform for pipe inspection capable of working inside sewers during normal operation, and to be equipped with sensors able to inspect dry and wet sections of pipes simultaneously [18]. The work presented in this publication is part of this multidisciplinary program, focusing on the inspection of the pipe below liquid level. Although, ultrasonic inspection may miss cracking and other small defects, it is capable of providing a quick, quantitative assessment of sewer deformation and other problems,
indicating not just the presence of a problem, but also its extent [8]. Ultrasonic inspection methods have been modestly investigated in the context of sewer inspection, involving mainly heuristic strategies providing results which support the feasibility of ultrasonics for pipe inspection [10,16,18]. Modeling ultrasound fields in medical imaging and in nondestructive testing is a relative mature research area. However, to the best knowledge of the authors no attempt has been made to simulate the behavior of ultrasonic signals inside liquid-filled tubular structures with clean water and sewage. By employing a modeling tool for the simulation of ultrasonic fields inside sewers will substantially benefit the study, design and analysis of ultrasonic inspection systems by shortening the test stages, and therefore, reducing costs and development time. The presented work attempts to bridge this gap using a systematic approach extending an ultrasound model originally developed for medical imaging applications, to simulate the behavior of ultrasound fields generated by rotational ultrasonic scanners inside pipes. Rotating ultrasonic sensors are being used to provide images of inner pipe walls immersed in sewage [16]. However, its capabilities have not been fully exerted. The ultrasonic inspection process consists of scanning a series of consecutive axial views while the sonar moves longitudinally inside the pipe at defined intervals (see Fig. 1). In order to scan a whole section of a sewer, a rotationalultrasonic scanner is mounted on a device which moves along the longitudinal axis of the pipe. The final goal is to generate a detailed three-dimensional profile of the internal pipe surface which can be used to assess the internal condition of it. Conditions such as sediment accumulation, roots intrusion, deformation of the pipe wall and cracks are of great interest to the sewer inspectors. The recovered profiles of the pipe can later be analyzed by an expert or by an automatic diagnostic system to find anomalous conditions that might influence the structural characteristics and mechanical integrity of the pipe. The commonly used pipes for sewer applications have nominal sizes that vary from hundreds to thousands of
z sor Sen US
Pipe
y
x ic aph ogr Tom slices
ng tati Ro am Be
Fig. 1. Ultrasonic inspection of pipes. The ultrasonic head is rotated inside the pipe creating a tomographic slice (B-mode image). To inspect a pipe section, the sensor is moved along the longitudinal axis of the pipe.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
y
millimeters in diameter, being usually made of concrete, clay, or PVC.
Homogenous Medium
2. Modeling methodology
Bvn ðt Þ ThðY r 1; Y r 2 ; t Þ; Bt
ð1Þ
where, p is the incident field at a point in the medium, vector Y r 1 is a point in space where the ultrasonic field is studied, t is the time for the snapshot of the spatial distribution of the pressure, q 0 is the density of the medium, v n is the normal velocity of the vibrating surface and hðY r 1 ; tÞ is the spatial impulse response [25]. The spatial impulse response was an approach developed by Tupholme [15] and Stepanishen [12,13] as an alternative to simplify the complex problem of determining the ultrasound field that reaches a reflector. The spatial impulse response is function of the relative position between the transducer and the observation point. It characterizes the
Electro-mechanical Impulse response (Tx)
Emitted Field (PSF/Eq. 1) Scattered Field (Eq. 3)
Electro-mechanical Impulse response (Rx)
Scatterer
r1
The modeling approach simulates ultrasound fields for different kinds of transducers through the use of linear acoustics. This was achieved by using the ultrasound simulation program developed by Jensen at the Technical University of Denmark, Field II [21]. Field II was created with the aim of simulating ultrasound imaging systems used in medical applications. The novelty of the presented work is the addition of routines to the Field II program for the simulation of ultrasound imaging systems inside tubular structures filled with liquid and its validation through an experimental study. For the sake of showing the capabilities of the modeling tool used, a brief description of the modeling methodology employed by Jensen is provided. Field II is based on linear systems theory (Fig. 2). After the electric pulse is applied to the crystal, a pressure field is irradiated into the medium, propagating in it until reaching the object (phantom). In order to find the echo, the impingent field in the scatterers composing the phantom has to be determined. To simplify the analysis, the centre of the transducer was located at the origin of the coordinate system (Fig. 3). The incident field is determined by convolving the excitation function with the spatial impulse response [5,14]: pðY r 1; Y r 2 ; t Þ ¼ q0
357
Received field (PSF/Eq. 1)
Fig. 2. Block diagram illustrating the simulation process.
r2 z x Fig. 3. Illustration of variables used in the model.
three-dimensional extent of an ultrasonic field for a given transducer, jrY1 Y r 2j Z d t c dS; ð2Þ hðY r 1 ; tÞ ¼ Y Y 2pjr r 2j 1 S where, Y r 2 is a point on the transducer surface (see Fig. 3). Eq. (2) is also described in the literature as the point spread function (PSF) representing the output of an ultrasound system during the scanning of an ideal point target located in space; this term is usually used to refer to two-dimensional representations [26]. The PSF is a function of the relative position between the location of the transducer elements and an observation point in space. Once the scattered field is determined by using the impingent field and the wave equation, reciprocally the received field is determined using the same methodology used for the determination of the impingent field. The field scattered from the phantom is determined by solving a suitable wave equation. The wave equation used in Field II was derived in a similar way as Chernov did in 1960 [1]. In order to solve the wave equation, some assumptions were made: & The instantaneous pressure is equal to the mean pressure of the medium plus the pressure disturbance caused by the ultrasound wave. & The instantaneous density is considered to be equal to the density of the undisturbed medium plus the density change caused by the passing of the wave. & The pressure and density changes created by the ultrasonic wave are considered first-order small quantities. & No heat conduction or conversion of ultrasound energy into thermal energy takes place. The wave equation used by Jensen is: l 2 p1
1 B 2 p1 2Dc B2 p1 1 ¼ þ lðDqÞIlp1 ; 2 3 2 2 q0 c0 Bt c0 Bt
ð3Þ
where, p 1 and q are the pressure and density variations, respectively, caused by the displacement of the ultrasonic
358
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
wave. The density of the medium is q 0 and the speed of the sound in the medium is c 0. l2 is the Laplacian operator, l the gradient operator and I the scalar product. A detailed derivation of this wave equation can be found in [25]. Knowing the emitted field, the scattered field, and the received field, the electro-mechanical coupling of the crystal during transmission and reception has to be determined to find the received signal (Fig. 2). Symbolically, the final expression used by Field II to find the received signal is: pr ðY r 2 ; t Þ ¼ mpe ðt Þ‘fm ðY r 1 Þ ‘ hpe ðY r 1; Y r 2 ; t Þ; r
ð4Þ
where ‘ is time Rconvolution, ‘r is the spatial convolution calculated using V f ðr1 Þd3 r1 and p r is the received signal. The function, v pe(t), describes the transducer excitation and the electro-mechanical impulse response during emission and reception of the pulse, mpe ðt Þ ¼
q0 B3 mðt Þ Em ðt Þ‘ : 2 Bt 3 2c0
ð5Þ
The term, f m, accounts for the inhomogeneities in the medium due to density and propagation velocity perturbations which give rise to the scattered signal, fm ðY r 1Þ ¼
DqðY r 1 Þ 2DcðY r 1Þ : q0 c0
ð6Þ
The modified pulse-echo spatial impulse response, h pe, that relates the transducer geometry to the spatial extent of the pressure fields, hpe ðY r 1; Y r 2 ; t Þ ¼ hðY r 1; Y r 2 ; t Þ‘hðY r 2; Y r 1 ; t Þ:
ð7Þ
The Field II program requires some key parameters in order to simulate an ultrasonic scan. These parameters are: the transducer centre frequency, the transducer radius, the size of the elements the transducer face is going to be divided into, the sampling frequency, the average speed of the sound of the medium, the frequency-independent attenuation and the frequency-dependent attenuation, and finally the phantom (the reflecting object) as an array of points (represented by x, y and z coordinates) (Fig. 4).
The object to be simulated is defined as an array of evenly spaced scatterers, located inside a homogeneous medium, here representing the inner surface of the pipe. The distance between individual scatterers (mesh size) must be chosen appropriately. Reducing distances between scatterers leads to an increase in the number of scatterers per axial section of pipe, which in turn impacts directly on the processing requirements, consequently, demanding more memory and increasing the processing time required to run the model. However, when the distance between scatterers is increased in excess, the resulting image may appear as a discontinuous pipe wall. The optimum value for this parameter to properly represent a pipe is mainly defined by the beam width of the transducer which is function of its frequency and geometry. On the other hand, the optimum scatterer separation to represent a pipe with a given defect is also determined by the size of the defect. The best performance is determined by trial and error varying the distances over a large range and analyzing the response. For the particular case of simulating ultrasonic fields for inspecting 300 mm diameter pipes, a scatterer separation in the range from 0.1 mm to 1.0 mm gives acceptable responses. It was found that, as a rule of thumb, the distance between the scatterers should be at least half of the minimum size of any simulated defect. The investigated transducer used for the validation of the simulations was a single piston transducer with a frequency of 3 MHz. As mentioned beforehand, the centre of the transducer is located at the origin of the coordinate system and its front face is aligned with the x –y plane. Because, the model used by Field II assumes that the transducer is mounted on an infinitely rigid baffle, the ultrasound field is emitted in both z-directions (z and z+). Standard ultrasonic transducers used in pulse-echo applications are mounted on backing material to reduce the undesirable effects caused by the mismatch in the impedance between the transducer and the medium, which creates a long ring-down. The simulation of the interaction between the backing material and the transducer is beyond the scope of this work. Because the
Input parameters: Transducer centre frequency (f ) Transducer radius (r ) Size of mathematical elements (w, h) Sampling frequency (fs) Speed of sound in medium (c0) Phantom (pipe dimensions, defect) Pulse duration Area to be scanned - Initial longitudinal position - Final longitudinal position - Longitudinal step size (long. resolution) - Initial angle - Final angle - Angular step size (angular resolution)
Output: B mode images: - RF lines (θ,t ) - Position (y )
Fig. 4. Input/output parameters for the simulation algorithms.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
ultrasound field is propagating from the front and rear faces of the transducer, the pressure waves propagating from the rear face in the negative z-direction can reflect in the scatterers located in the negative z half-space and return to the transducer, getting mixed with the signal of interest from the front face. In order to overcome this limitation which arises when creating images from pipes immersed in water, a scatterer filter function was implemented. This function analyzes the phantom and the location of the transducer before starting the simulation of the field and removes all the scatterers located in the negative z half-space. The pipe inspection process requires that the transducer rotates around the y-axis (longitudinal axis of the inspected pipe) and translates along it (Fig. 1). However, the original Field II routines do not allow rotation or translation of the transducer. In order to create a 3D profile of an internal section of a pipe, extensions to the Field II program were developed. The developed routines overcome this limitation by rotating and translating the environment to simulate the desired movement of the transducer. It is noted that, instead of rotating and displacing the transducer to scan the object, the object is moved and displaced accordingly, while the transducer is kept still at the origin of the coordinate system.
3. Simulations and experiments In order to validate the developed algorithms for the sewer inspection case, simulations and comparative experiments were conducted. Three experiments were carried out. Firstly, a 0.5-mm cooper wire immersed in clear water was scanned and simulated. The corresponding results were quantitatively compared to validate the accuracy of the simulations through a lateral beam width measurement. Finally, two experiments using Polyvinyl Chloride (PVC-U) pipes, same as the ones found in sewers, with precisely carved defects were carried out. In both cases, experiments and simulations were conducted and the results compared. 3.1. Experimental platform The experimental study was conducted in the Pipe Inspection Lab in the Centre for Mechatronics and Manufacturing Systems at Kings College London. This facility is equipped with a 300-l water tank (100 60 50 cm), a rotational ultrasonic transducer, model SeaKing, manufactured by Tritech International and a SCARA Robot. The sonar system was supplied by one of the project sponsors. The supplied system has the same characteristics as the one that is expected to be used for real pipe inspections. The sonar system houses in the sonar head the electronics responsible for: the generation and receipt of the pulse/echo signal, measurement of the envelope amplitude of the echo, analog to digital conversion, and transmission of the acquired data, via a high speed Arcnet serial link to the host computer. The transducer installed in
359
the sonar is a single piston, unfocussed, piezoelectric transducer made of Lead Zirconate Titanate (PZT). It has a resonant frequency of 3 MHz, a 10-mm diameter and is mounted on a rotating platform. For protection from contact with water, the transducer is enclosed in a rubber cap filled with inhibited mineral insulating oil commercially known as Shell Diala MX. The angular resolution, the duration of the pulses and the sampling rate are user configurable parameters. The angular resolutions are: 1.8-, 0.9-, 0.45- and 0.225-. The minimum pulse duration is 10 As and the fastest sampling rate is 162 kHz. The analogue to digital converter has a resolution of 8 bits. 3.2. Lateral beam width measurement The main objective of this study was to measure the radiation pattern of the transducer and compare it with the predicted one obtained from simulations. The basic measurement to characterize the radiation pattern of a given transducer is the beam width, which is defined as the angular distance between the points in the radiation pattern were the intensity decays 3 dB from its maximum value. The radiation pattern is a 3D characteristic which describes how the energy of the emitted field radiates in the medium. The lateral beam width was measured in water by rotating the sonar head and recording position and intensity of the echoes generated by a thin copper wire of 0.5 mm diameter. The wire was located at a distance of 440 mm and was aligned parallel with the face of the transducer. For the simulation, the centre of the transducer’s face was located at the origin of the coordinate system and the phantom of the wire was positioned parallel to the y-axis. The phantom was created with a mesh size of 0.1 mm, being composed of 24 scatterers. The z-axis was the reference axis for the rotation of the sensor, for instance the 0- position of the transducer lies on the z-axis (Fig. 5). The sensor was rotated around the y-axis, from 10- to 10- about its centre with a step resolution of 0.225-. In order to facilitate the comparison of the experimental and simulated results, the simulated results were downsampled to the match the sampling frequency of the sonar system. The simulator tool generates sampled data at a frequency of 100 MHz, while the sonar system samples at a maximum rate of 162.5 kHz. The excess samples acquired at 100 MHz lying between two 162.5 kHz samples were averaged. In Fig. 6, simulated and experimental signals are shown; the left plots are B-mode images and the right ones are plots of intensity against angular position. From the B-mode images, it can be seen that the shape and intensity profile of the returning echoes are in good agreement with just slight deviations present in the experimental data (Fig. 6(c)) where a shadow is visible at approximately 650 As. This phenomenon could be attributed to the ring-down, i.e. the period of vibration of the crystal after being struck by a short voltage pulse [2]. By closely studying the pulse
360
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
(a)
(b) 441 440.8
400
440.6
300
440.4 200
z [mm]
z [mm]
Wire
100 0 50
440.2 440 439.8
40
439.6 30
20
Front face of sensor 10
0
10
y [mm]
20
30
40
439.4
50
5 0 x [mm] 5
439.2 439 1
0.5
0
0.5
1
x [mm] Fig. 5. (a) 3 MHz transducer and 0.5 mm diameter string phantom. (b) Magnified view of the phantom in the x – z plane.
Fig. 6. 3-D mesh plot, received echoes from 0.5 mm copper wire. Simulated data: (a) axial – lateral plane view; (b) lateral view. Experimental data: (c) axial – lateral plane view; (d) lateral view.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
3.3. Square defect A Polyvinyl Chloride (PVC-U) pipe with a precisely carved squared defect was scanned and simulated. The plastic pipe had a nominal diameter of 300 mm and a measured inside diameter of 296 mm. The carved defect investigated was a carefully cut square hole. The hole dimensions were 10 10 mm. The hole was centred over the x-axis and extended along the y-axis from 70 to 80 mm. The hole and the surrounding area were scanned by moving the transducer along the y-axis 42 mm at 3 mm steps, sweeping from 10.0- to 10.0- at 1- intervals. This results in a rectangular surface of approximately 63 by 42 mm Table 1 Beam width table, experimental and simulated measurements 3 dB beam width
Experimental (-)
Simulated (-)
Difference (%)
2.66
2.2
17
0 5 10 15
Amplitude (db)
generated by the employed sonar system, it was appreciated that even though the electrical pulse which excites the transducer perfectly resembles a square pulse, the emitted pressure pulse showed a main peak which concentrates most of the energy inside it, decreasing monotically, some times accompanied by a tail-peak (ring-down). The duration of the ring-down is about 10 times the duration of the emitted pulse (10 As). This shows that the ultrasound system has a ‘‘high Q’’ and as a consequence a long spatial pulse length which degrades its axial resolution [2]. This ring down effect is reduced by attaching a high density damping material to the rear side of the transducer but this issue has not been simulated as part of this paper. In Table 1, the measured beam widths for the experimental and simulated cases are shown. The difference between them is calculated using the experimental data as reference. There is a reasonable and consistent match, considering the limited precision of the experimental platform. In Fig. 7, the profile of the simulated and experimental beams are compared. Only the main lobe can be compared because the noise level of the experimental platform is around 30 dB, which is so strong that the returning echoes coming from the side lobes are covered. An acceptable match between the two signals can be seen in Fig. 7. The simulated radiation pattern tends to be slightly narrower than the experimental one. The difference between the experimental and simulated results may be due to refraction caused at the interface between the rubber cap filled with transformer oil and the water which was not considered in the simulation. The transformer oil in which the transducer is immersed has lower density than water, making the beam diverge. If the refraction effect created at the boundary of the two media is not considered the simulated beam width will be narrower than the real one, in agreement with the results shown in Fig. 7 and Table 1.
361
20 25 30 35 40 45 50 –10
–8
–6
–4
–2
0
2
4
6
8
10
Transducer position (deg)
Fig. 7. Beam widths comparison: experimental data (continuous line), simulated data (diamond line).
being scanned. The scanning process generated 14 B-mode images, representing tomographic slices of the pipe. To compare the experimental data and the simulation results, an appropriate visualization methodology has to be employed. From the set of consecutive B-mode images acquired during the scanning process, a 3D profile of the inner surface of the pipe was created. The profile was created by measuring the time of flight of the emitted pulse and registering the maximum intensity of the received echo. As a consequence, four variables were obtained: x, y and z indicating the position of the reflector that generated the echo and the echo intensity. A useful way to represent these profiles is through the use of 3D color plots that will be referred to as surface – intensity plots, showing the inner surface of the pipe and the received intensity. In Fig. 8, surface– intensity plots for simulated and experimental data are presented. For the simulated case presented in Fig. 8(a), the removed section can be easily identified as the area where the intensity decreases until reaching a minimum (dark area). Even though, the hole is a perfect square, the resulting image is not a square. It is noted that the shape identified in the surface –intensity plot image is different from the original shape due to constructive and destructive interference occurring at the corners of the square. A closer analysis of the acquired image shows that the contrast between the intensity from echoes returned by the intact wall and the centre of the removed section is around 14 dB, decreasing smoothly from 0 dB to 14 dB. In Fig. 8(b), the experimental results are presented. The experimentally scanned signals are notoriously noisier than the simulated signals, as expected for a real system. In the presented case, the defect is easily recognizable in the surface –intensity plot as an area where the intensities of the returning echoes decrease until reaching a minimum at the centre of the defect. Comparing the echo intensities returning from the surface of the pipe wall and the centre of the hole, a contrast of 14 dB can be appreciated being in perfect
362
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
(a) 105
0
100
-2
95
-4
90
y
-6 85 -8 80 -10
75
-12
70 65
-14 -30
-20
-10
0
10
20
30
x
(b) 105 -52
100
-54
95
(a)
0
-56
80
85
-58
70
80
-60
60
75
-62
70
-64
65
-66
90
-1 -2 -3
y
y
simulated. The carved shape was a triangle with 39 mm base and 60 mm sides. The centre of the triangle base was aligned parallel to the x-axis, going from 19.5 mm to 19.5 mm. The triangle height was 56.74 mm, extending along the y-axis from 20 mm to 76.74 mm. The triangle was scanned moving the transducer along the y-axis from 10 mm to 88 mm at 3 mm steps and rotating the transducer from 11.7- to 11.7- at 0.225- intervals. Hence, the scanned inner surface of the pipe was a 60 mm by 78 mm rectangle. The scanning process generated 27 B-mode images, each one representing a tomographic slice of the pipe. The internal surface of the pipe was profiled from the set of B-mode images by applying a time of flight (TOF) profiling algorithm. A TOF profiling algorithm detects the location of the first reflection inside the acquired RF lines using a pre-fixed threshold. For this particular case, the threshold value was fixed at 10 dB. Results are presented in surface – intensity plots, dashed lines representing the removed section as measured by means of a ruler are overlaid (Fig. 9). In the presented surface –intensity
-30
-20
-10
0
10
20
-4 -5
50
-6
40
-7 30
-8
30 20
x Fig. 8. Surface – intensity plots for square defect. (a) Simulation. (b) Experiment.
-10 -30
-20
-10
0
10
20
30
x
(b)
-55 130 -56 120
-57
110
-58
100
-59
y
agreement with the simulation. In addition, the defect size appreciated in the experimental data is smaller than the one appreciated in the simulated one. This is caused by the difference in the beam width size between them. A wider beam width results in a poorer resolution which can be seen in the smaller dark area in the surface –intensity plot presented in Fig. 8. The deformations appreciated in the plot can be attributed to the differences between the lateral resolution of the stepper motor and the longitudinal resolution of the SCARA robot. From the results of this study case, it can be concluded that the simulation tool is representing the physical phenomenon in an accurate and acceptable way.
10
-9
-60
90
-61
80
-62 70 -63 60 -64
3.4. Triangle defect A PVC-U pipe with a nominal diameter of 300 mm and a measured inside diameter of 296 mm pipe with a precisely carved isosceles triangle was experimentally scanned and
50 -30
-20
-10
0
10
20
30
-65
x Fig. 9. Surface – intensity plots for a triangle defect carved in the pipe wall. (a) Simulation. (b) Experiment.
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
plots, the white section in the centre of the triangle shows the segment of the pipe where the reflected echoes are very weak or no echoes are detected. It can be seen that the sides of the triangle delineated by the algorithm in both cases resemble stepped lines. This is due to the 3 mm axial step size used during the scanning process along the y-axis (quantization error) and does not affect the purpose of the study. As was presented in Section 3.2, the measured beam width of the experimental transducer was wider than the simulated one. Therefore, the white area in the experimental case (Fig 9(b)) is smaller than in the simulated case (Fig. 9(a)). During experiments, it was observed that the off centre location of the transducer inside the pipe is a key parameter that affects the quality of the resulting images. In this particular experiment, the transducer was slightly misaligned towards the positive x direction. As a result, the plot in Fig. 9(b) shows a slight asymmetry when representing the triangular hole; the left side of the triangular hole is depicted by a wider grey scale transition along the boundary on the left side (12 mm) when compared to the right side (9 mm). Another indication of the off centre condition of the scanner are the tenuous shadows present at the top-right quadrant of the plot. As has been shown through experiments, the developed simulation tool is able to create ultrasound simulated images inside fully charged pipes in an acceptable way.
4. Conclusions A valuable addition to the Field II routines, usually applied to medical ultrasound imaging modeling, has been developed to simulate ultrasound fields inside fully charged pipes. The presented approach is capable of generating simulated images from given pipes which are constructed from a series of reflectors or scatterers. Defects or anomalies in the pipe wall are represented by removing or adding reflectors which emulate the defect geometry. The presented tool is oriented towards the simulation and visualization of rotating ultrasound transducers providing the creation of two-dimensional (B-mode) and three-dimensional scans inside tubular structures. By means of an experimental study and detailed comparison of real and simulated data, the validity of the developed approach was shown. The presented work can be used to conduct in-depth investigations of ultrasonic based pipe inspection systems, by giving an invaluable insight into the sound wave propagation in fully charged pipes. In addition, it will contribute to facilitate the design of future pipe inspection systems because of having the means to validate the designs before attempting time-consuming measurements in a real scenario. References [1] L.A. Chernov, Wave Propagation in a Random Medium, McGrawHill, New York, 1960.
363
[2] K. Gustav, H. Smith, D. Allison, The Encyclopaedia of Medical Imaging, Volume I. Physics, Techniques and Procedures. NICER Institute. Isis Medical Media (1998). [3] O. Duran, K. Althoefer, L. Seneviratne, State of the art in sensor technologies for sewer inspection, IEEE Sensors Journal 2 (2) (2002) 63. [4] R.A. Fenner, Approaches to sewer maintenance: a review, Urban Water 2 (4) (2000) 343 – 356. [5] J.A. Jensen, A model for the propagation and scattering of ultrasound in tissue, Journal of the Acoustical Society of America 89 (1991) 182 – 191. [6] F. Kirchner, J. Hertzberg, A prototype study of an autonomous robot platform for sewerage system maintenance, Autonomous Robots, vol. 4 (4), Academic Publishers, Netherlands, 1997, pp. 319 – 333 (ISSN: 0929-5593). [7] R. Kirkham, P.D. Kearney, K.J. Rogers, PIRAT—a system for quantitative sewer pipe assessment, The International Journal of Robotics Research 19 (11) (2000 (Nov.)) 1033 – 1053. [8] J.M. Makar, Diagnostic techniques for sewer systems, Journal of Infrastructure Systems 5 (2) (1999 (June)) 69 – 78. [9] O. Moselhi, T. Shehab-Eldeen, Automated detection of surface defects in water and sewer pipes, Automation in Construction 8 (1999) 581 – 588. [10] N.G. Pace, Ultrasonic surveying of fully charged sewage pipes, Electronics & Communication Engineering Journal 6 (2) (1994 (Apr.)) 87 – 92. [11] E. Rome, J. Hertzberg, F. Kirchner, U. Licht, T. Christaller, Towards autonomous sewer robots: the MAKRO project, Urban Water 1 (1999) 57 – 70. [12] P.R. Stepanishen, The time-dependent force and radiation impedance on a piston in a rigid infinite planar baffle, Journal of the Acoustical Society of America 49 (3) (1971) 841 – 849. [13] P.R. Stepanishen, Transient radiation from pistons in a infinite planar baffle, Journal of the Acoustical Society of America 49 (1971) 1627 – 1638. [14] P.R. Stepanishen, Pulsed transmit/receive response of ultrasonic piezoelectric transducers, Journal of the Acoustical Society of America 69 (1981) 1815 – 1827. [15] G.E. Tupholme, Generation of acoustic pulses by baffled plane pistons, Mathematika 16 (1969) 209 – 224. [16] M.E. Andrews, Large Diameter Sewer Condition Assessment Using Combined Sonar and CCTV Equipment, APWA International Public Works Congress, NRCC/CPWA Seminar Series ‘‘Innovations in Urban Infrastructure’’, 1998. [17] R. Browne, G. Knott, Television and Scanning Sonar in Seattle Metro’s Siphons and Brick Sewers, ASCE International Conference on Pipeline Infrastructure II, San Antonio, TX, 1993, pp. 694 – 701. [18] O. Duran, K. Althoefer, L. Seneviratne, Automated sewer pipe inspection through image processing, Proc. 2002 IEEE Int. Conf. on Robotics and Automation (ICRA’2002), Washington, D.C., 11 – 15 May, 2002, pp. 2551 – 2556. [19] F. Gomez, K. Althoefer, L. Seneviratne, Modelling of rotating ultrasonic sensors for pipe inspection, International Conference on Mechatronics, Loughborough, UK, 20032002. [20] R.M. Gooch, T.A. Clarke, T.J. Ellis, A semi-autonomous sewer surveillance and inspection vehicle, Proc. IEEE Intelligent Vehicles, Japan, 1996, pp. 64 – 69. [21] J.A. Jensen, Field: a program for simulating ultrasound systems, Paper Presented at the 10th Nordic – Baltic Conference on Biomedical Imaging Published in Medical and Biological Engineering and Computing, Volume 34, Supplement 1, Part 1, 1996, pp. 351 – 353. [22] H.B. Kuntze, H. Haffner, Experiences with the development of a robot for smart multisensoric pipe inspection, Proceedings of the 1998 IEEE International Conference on Robotics and Automation, Belgium, 1998. [23] E. Rome, H. Surmann, H. Streich, U. Licht, K. Paap, A custom IR scanner for landmark detection with the autonomous sewer robot
364
F. Gomez et al. / Automation in Construction 15 (2006) 355 – 364
MAKRO, International Symposium on Intelligent Robotic Systems, 2001. [24] CERF, Evaluation of SSET: The Sewer Scanner and Evaluation Technology. The Civil Engineering Research Foundation, Report: #40551, March (2001). [25] J.A. Jensen, Linear description of ultrasound imaging systems, Notes for the International Summer School on Advanced Ultrasound
Imaging, Technical University of Denmark July 5 to July 9, 1999, Technical University of Denmark, June, 1999. [26] M. Anderson, G. Trahey, A seminar on k-space applied to medical ultrasound, Web publication, Department of Biomedical Engineering, Duke University, 2000.