Frédéric Guerne, Demining Technology Center (DeTeC)
Swiss Federal Institute of Technology Lausanne (EPFL), InF, LAMI, CH-1015 Lausanne Tel.+41 21 693 67 56, Fax. 52 63
Many technologies have been tested for detecting APLs (antipersonnel landmines). Nowadays, no single sensor other than the Metal Detector (MD), which has unfortunately a high false alarm rate, is reliable enough. The Demining Technology Center (DeTeC) in Switzerland has been investigating the problem since the end of 1995 to find the best solutions. The most promising sensor combination includes a Metal Detector (MD) and a Ground Penetrating Radar (GPR).
In order to test the different sensors and to develop our own system, we had to have an area where we could easily bury mines or other objects to be detected, with the constraint of moving precisely and repetitively a sensor above this area. This is the reason why we built a "sandbox" with a Cartesian gantry (Figure 1).
When we built our installation we tried to be as cost effective as possible and therefore tried to choose the technology the less expensive while still reliable enough for such an application.
We hope that this paper will help others who want to set up a similar infrastructure or who want to improve an existing one to develop or test their own sensors.
We built an indoor infrastructure to have reproducible measuring conditions, avoiding especially temperature and humidity dependence.
The "sandbox" is made of concrete bricks, measures 3 x 3 meters and is 1.2 meters high. One third is filled with sand, one third with wet earth covered by grass and the last third with dry earth. The separations between each section are wooden plates.
The Cartesian gantry is able to move a sensor over the full surface. It has two motorized axes (X and Y) while the Z one is manually adjustable or can follow the surface of the soil with a slipping antenna, using springs to reduce the force on it. The control of the displacements is made via a 486 PC compatible computer which can be used as an interface to a specified measuring system.
(Figure 1) Our "sandbox" with the Cartesian gantry
The "sandbox"
Dimensions
To start with, you have to know what kind of sensor you will test. This will influence the dimensions of your box. In our case we test Ground Penetrating Radars (GPR) and Metal Detectors (MD).
The MD used for demining operations are so sensitive that they can be disturbed by a large quantity of metal at a huge distance. For example, at the highest sensitivity, our MDs detect the rebars in the concrete of the soil at more than one meter! This means that you have to be careful not to place your installation near a large piece of metal (or in our case reinforced concrete) or a perturbing device (e.g. a big transformer). In order to reduce the effect of our rebars, the surface of our sand had to be at more than one meter. This was one reason to choose a height of 1.2 meters.
For the GPR, the depth of the box was important in order to avoid reflections from the soil (concrete). The other problem with the radar is its opening angle. Even if you are absolutely vertical with the radar antenna and you are beside the wall of your sandbox at too small a distance, you will see it. This implies that you cannot carry out acquisitions with your GPR on the full surface of your "sandbox", having in the border dead zones whose dimensions depend on the radar type. You have to take this into account to dimension the surface of your box.
Something which is not a detail is the total weight of the sand or earth you will need. In our case, we have more than 9 cubic meters of sand and earth. This means 12.6 tons and gives a pressure of 1.4 tons per square meter. You have to be sure that your soil can support it, especially if you have a cave or another room under it. Another problem which you can meet is the filling of your "sandbox". If you need some cubic meters of sand, it can take a lot of time to handle it by hand. In our case, we asked the sand provider to fill the box himself (Figure 2). The accessibility of the sandbox can also be a problem.
(Figure 2) Filling the sandbox can be a problem
The walls
For most sensors you can not have metal in proximity to your sand or earth. This forbids the presence of metal in the walls of your box. To achieve this, we studied several solutions.
A wooden box could be a good solution as long as you do not use nails or screws. This implies the use of old assembly methods. If you can find somebody used to practice this discipline, it is feasible, but it will certainly be quite expensive. Another problem with wood, if you have long walls (some meters) without big reinforcement, is that it will be deformed due to the non negligible pressure on it.
We also investigated the possibility of buying a plastic or polyester box, like a swimming pool, but this is quite expensive.
The best solution we retained was the standard bricks wall. We did not use any metal. What is interesting is that it does not cost very much and it takes about two weeks for two masons to build. We used standard concrete bricks 13 cm thick (Figure 3).
(Figure 3) Standard concrete bricks were used
To support the Cartesian gantry, we had to put it on four pillars. To fix the pillars we included a tube in each corner in the box walls (Figure 4).
(Figure 4) Four plastic tubes are put in each corner of the "sandbox" to fix the pillars.
Because the use of rebars is forbidden, the masons had to find tricks to make the structure solid enough. They moulded the concrete’s corners with the supporting tubes using an alternating length of bricks to reinforce it (Figure 5 & 6).
(Figure 5) The method of alternate bricks to reinforce the concrete
(Figure 6) When the tube was well positioned it was encased in concrete
Supporting the gantry
Again, the four pillars used to support the Cartesian gantry cannot contain any metal. Another constraint was to be able to modify the height of the gantry. To do this, we built telescopic pillars made with polyester and glassfiber tubes. (Figure 7)
To block the mobile tube we built a PVC ring with a longitudinal gap. Its diameter can be reduced with a screw.
On top of the pillar, we have a mechanical interface made to fix the gantry on the pillar. This is a U-shaped aluminum part fitted in the tube with a cylinder.
(Figure 7) The mechanical plan of the telescopic pillars
The Cartesian gantry
The Main supporting part
You can built yourself this type of gantry if you are equipped enough, but it will take quite long and not be easy. The best solution is to subcontract it to a specialized firm. We worked with the firm ITEM (ITM) which is specialized with this kind of products. Another supplier for such material is BOSCH (BO). They build it from their standard tool kit.
What is really interesting is that you only have to give them your constraints and they build what you need. It does not cost too much, compared with the time you would need if you had to built it from scratch. For example, our Cartesian gantry, without motors and any electronics cost 9561 SFr. (~ US$ 6460).
The specifications of our gantry are as follows:
| Material | Aluminum |
| Dimensions | 3.3 x 3.6 meters |
| Axes | 2 (X & Y) |
| Course | 3 x 3 meters |
| Section of the X profile | 80 x 40 mm light profile |
| Section of the Y profile | 80 x 80 mm light profile |
| Carriage driving | Notched belt 25 mm |
| End of travel | Switches and hydraulic stop |
| Wires guidance | Chain channel |
(Table 1) Gantry characteristics
You can see in Figure 8 a top view of a schematic representation of our Cartesian gantry and in Figures 9 and 10 some details of the construction.
(Figure 8) Top view of the schematic representation of our Cartesian gantry
(Figure 9) Detailed view of the sensor carriage (here without sensor support) and the wires guidance chain channel
(Figure 10) Detail view of the supporting Y axis carriage, the hydraulic stop and the end of travel sensors
Driving features
Still with the aim of reducing the price, we used stepper motors for the motorization. Of course, with such motors, the system is an open loop one. Actually, with our system, the torque on the motor does not change a lot under load because the sensor does not touch the soil or is slightly sliding on it. This is why we can use such motors and why they work well. The specifications of the motors are as follows:
| Manufacturer | Sonceboz SA (SON) |
| Type | 6630R17 |
| Steps per turn | 200 |
| Phases | 2 |
| Current per phase | 6.5 ARMS |
| Holding torque | 10 Nm |
(Table 2) Motors characteristics
To be sure of having enough torque, we added a reduction gear (1:3). It is also used to fix the motor on the gantry.
We put one motor for the Y axis. The X axis has two notch belts, one for each side. To drive it you can use one motor and a transmission shaft. In our case, we used 2 motors (this it is feasible with stepper motors because it is really easy to drive them synchronously). The reason of this choice is that we would like to be able to dismantle our gantry easily to use it outdoors. With a transmission shaft, this would be more difficult.
Sensor fixation
We added to the sensor carriage a non metallic arm (PVC) equipped with a "soil tracking spring based system" (Figure 11). This means that, when you put a sensor which must slide on the soil (e.g. most of the standard GPR antennas), the arm can stretch in or out and springs reduce the force of the antenna on the soil. To move smoothly, even when we have a tangential strength on the sensor sliding on the soil, we had to put rolling bearings on the arm support.
(Figure 11) The sensor supporting arm.
We had to test many kind of sensors (Figure 12). To do that, we simply used a PVC plate fixed at the extremity of our sensor supporting arm. The PVC can be easily drilled for any configuration.
1.
2.
3.
4.
(Figure 12) Some examples of sensors fixed to the PVC plate.
1: A Metal Detector fixed with a plastic screw. This sys. permits to rotate easily by 90°.
2: A prototype system with a VHF antenna and a dipole, fixed with wires and special VHF plastic material.
3: A 500 MHz GPR antenna not touching the soil, fixed with aluminum strips and screws.
4: A 1 GHz GPR antenna sliding on the soil, with a "inverted hat" to help slipping on the rugged surface. The fixation is made of PVC adapted pieces and plastic screws.
Electric driving
The power driving of the motors is made through three power drivers (SDM 741) by Selectron (SEL).
The interface between power drivers and the PC compatible computer is realized by small controllers through RS232. The controllers we used are obsolete ones. For this reason, we would like not to mention the type, but you can find compatible devices by Selectron (SEL).
To give power to the stepper motors drivers and the controllers, we built a quite powerful power supply whose main characteristics are :
| Power | 630 VA |
| Input voltage | 220 V |
| Output voltage | 42 V |
| Voltage regulation | none |
(Table 3) Power specifications
We put the power supply, the serials controllers and some electronic interfaces in an aluminum box (Figure 13) provided by Item (ITM) which can be fixed easily to the gantry.
All the cables going out of this box are plugged through connectors. This simplifies dramatically the dismantling operations.
(Figure 13) The aluminum box containing power supply, motor controllers and electronic interfaces. We can see in the rear the power drivers mounted on DIN rails.
The "intelligent" interface
We use a PC compatible computer as gantry controller (Figure 1). It can be manipulated by people or by another computer for specific applications. We wrote a simple program in Visual Basic to work with it.
Total cost of our infrastructure
A the list of the total material costs follows:
| Material |
Price $US |
| Cartesian gantry |
6460.- |
| 3 Motors + 3 motor controllers |
2235.- |
| 3 Power motor drivers |
1100.- |
| Small material (cables, boxes, connectors, power supply, small electronics) |
1730.- |
| 8 Glassfiber tubes for 4 pillars |
340.- |
| Material for the sensor supporting arm |
280.- |
| Sand and earth with transport |
1100.- |
| Concrete bricks and other material |
900.- |
| Total |
14145.- |
(Table 4) Total material costs of our infrastructure
Conclusions
It took about 3 months, from the beginning to the end, and cost quite a lot to built it from scratch. But if we did not have it, we could not have done our investigations in sensors testing and building. This "sandbox" might have seemed superfluous at the beginning but it quickly became indispensable.
Addresses
(ITM) ITEM Industrietechnik und Maschinenbau GmbH
Friedenstrasse 107-109
D-42699 Solingen
Germany
Tel. 0212 / 65 80 0
Fax. 0212 / 62 04 9
(BO) Robert Bosch GmbH
Industrieausrüstung
Verkaufsbüro IA3/VKF3-Hr
Ulmer Strasse 4
D-30880 Laatzen
Germany
Tel. (05 11) 86 06 - 2 15
Fax. (05 11) 86 06 - 4 19
(SON) Sonceboz SA
CH-2605 Sonceboz
Switzerland
Tel. +41 32 488 11 11
Fax. +41 32 488 11 00
(SEL) Selectron Lyss AG
Industrielle Elektronik
Bernstrasse 70
CH-3250 Lyss
Tel. +41 32 384 61 61
Fax. +41 32 384 48 20
http://diwww.epfl.ch/lami/detec/susdem_sandbox.html| Last modified 7.10.1997