Sie sind auf Seite 1von 12

Geophysical Prospecting, 2007, 55, 737748

Modelling magnetic fields due to steel drum accumulations


Peter Furness
Ilkley Rd., Ilkley Q4554, Australia

Received June 2006, revision accepted January 2007

ABSTRACT Modelling the magnetic fields produced by accumulations of steel drums is a problem that is relevant to the detection and evaluation of disposal sites containing materials that are potentially hazardous to the environment. Accurate modelling is possible with existing integral equation techniques but these are numerically intensive due to the need to solve very large systems of linear equations. Use of an approximate iterative technique for the solution of the equations (system iteration) allows the integral equation technique to be extended to modelling the magnetic effect of substantially large accumulations, comprising up to several hundred drums, on very moderate computing facilities. However, even this process remains time-consuming and suggests the use of more rapid, if less accurate, modifications. Several are available. Surprisingly, quite reasonable results can also be achieved with a very basic approximation that represents each drum by a discrete dipole located at its centroid. The dipole moments are found from the magnetic behaviour of single drums exposed to a uniform inducing field, which can be conveniently defined by a dyadic drum apparent susceptibility. The basic discrete dipole model for drum accumulations can be substantially improved by using a first-order accommodation of the depolarizing effect produced by the shape of the accumulation. All of the above modelling techniques require details of individual drum locations and orientation. This information is generally unavailable to geophysical practitioners involved in environmental surveys and so prompts the idea of models that represent drum accumulations as a continuous distribution of magnetization. The convenience of neglecting details of drum location and orientation comes at the cost of some loss in accuracy of the modelled responses. However, for accumulations buried sufficiently deep and in which the drums are uniformly distributed, the total field magnetic anomaly is found to be reasonably approximated by the effect of a continuous magnetization, expressible in terms of an effective isotropic susceptibility. Again, the basic model can be improved by the accommodation of demagnetization effects due to the shape of the accumulation.

INTRODUCTION A trend over recent decades has been the increasing use of geophysical surveys for environmental monitoring and remedial purposes. A popular technique involves the use of magnetics to locate ferromagnetic bodies of cultural origin, including steel

E-mail: furness@powerup.com.au

drums containing various hazardous waste materials. The interest here is due to the inevitable corrosion of the containers over time and the consequent leakage of the contained fluids into the groundwater regime. Clearly, the location of disposal sites and their evaluation are increasingly important objectives in environmental investigations. Typical waste disposal sites involve many containers buried at shallow depths. Consequently the problem of location is relatively simple. The main interest lies in determining the

2007 European Association of Geoscientists & Engineers

737

738 P. Furness

horizontal position and lateral extent of dumps and these features are generally obvious, due to the magnitude of the magnetic response measured immediately above the buried containers. However, the second objective of evaluating the observed magnetic anomalies in terms of the number of containers is not so easily achieved. The present paper represents a step towards this second objective. It addresses the problem of computing the magnetic fields of accumulations of thin-walled ferromagnetic containers. The topic does not seem to have attracted any serious attention in the geophysical literature to date. The only relevant work appears to be the field studies conducted by Emerson et al. (1992) over a rectangular 3 3 array of vertical 205litre drums. Emerson et al. (1992) also reported the results of some first-order modelling experiments for drum accumulations. They achieved satisfactory fits to the measured response by modelling procedures using an equivalent dipole for each drum as well as a single equivalent dipole for the entire array.

P R E L I M I N A RY C O N S I D E R AT I O N S It is practical to restrict the present discussion to situations typically encountered in environmental investigations. The location and orientation of individual drums in a waste disposal site are to some extent random, so for practical reasons accumulations will be represented here by rectangular arrays of drums aligned parallel to the geographic northerly, easterly and vertical directions. The individual drums considered will be of the standard closed-head 205-litre type described by Furness (2002) with an outside radius of 287 mm, a height of 849 mm, a wall thickness of 1 mm and an intrinsic susceptibility of 350 (SI). The magnetic effect of the drum accumulations will be assumed to be due only to induced magnetization. This neglect of remanent magnetization should not be taken to suggest that it does not exist in thin-walled steel containers. Indeed, Eskola, Puranen and Soininen (1999) and others have reported ample evidence of its presence. However, since it is probable that the predominant component of any remanent magnetization is acquired prior to fabrication, its effect in the completed drum will probably be diminished by destructive cancellation of the individual effects in the three basic components of any drum. This argument for neglecting the effect of remanent magnetization clearly becomes stronger in the case of multiple drums. In fact, it seems possible that accumulations of steel drums constitute one of the few source bodies in magnetic exploration where the neglect of remanent magnetization may be fully justified. The above comments are not relevant to viscous

remanent magnetization acquired over the period since the drums were dumped. For old disposal sites, the effect of this phenomenon on magnetic surveys is indistinguishable from an increase in the intrinsic susceptibility of the drum walls and, as such, is automatically accommodated in the present models. It is also assumed that the drums are buried at shallow depths in a non-magnetic earth and exposed to a primary geomagnetic field BP of 50 000 nT magnitude with a declination of 0 and an inclination of 50 , conditions typical of middle latitudes in the southern hemisphere. Sensors of contemporary total field magnetometers are generally deployed at a height of a few metres above the ground level in order to minimize noise deriving from surficial sources. Consequently, even for dumps with very shallow depths of burial, the sensor can normally be expected to clear any magnetic material by at least a couple of metres. For this reason, all source bodies modelled will lie at a depth of two metres below the sensor. Discussions of the various modelling techniques will be limited to deriving the secondary magnetic H fields due to drum accumulations. Unless otherwise indicated, the total field anomalies are then computed as measured by modern nuclear magnetometers. Since the environments considered are non-magnetic, secondary H fields are first converted to the measurable B fields by the equation, BS = 0 HS . The total magnetic B field is then found by addition of the primary (geomagnetic) field, i.e. B = BP + B S , and the total field magnetic anomaly is computed as the difference between the measured and geomagnetic field amplitudes, i.e. T = B BP Finally in the discussion that follows reference will be made to a conventional right rectangular coordinate system with the x-axis directed (geographic) northwards, the y-axis directed eastwards and the z-axis directed vertically downwards. In all the profiles presented, the central station is located vertically above the centre of the drum accumulation.

A N E X A C T I N T E G R A L E Q U AT I O N M O D E L An exact integral equation model for a steel drum with susceptibility k in a non-magnetic environment exposed to an incident magnetic field HP was used by Furness (2002). The

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Modelling magnetic fields due to steel drum accumulations 739

technique was fully described by Furness (1999). It uses a surface integral representation of the secondary H field due to the drum at an external field point, which is conveniently expressed in dyadic form as HS (r) = where 1 1 3 , I (r r )(r r ) = |r r | |r r |5 |r r |3 k 4 (r )n
S

1 ds , |r r |

(1)

is the unit dyadic. is the scalar magnetic potential on and I S, the (inside and outside) surface of the drum walls, and is found by solving the Fredholm integral equation of the second kind, (r) + k 2 (k + 2) (r )n
S

2 1 ds = P (r), |r r | k+ 2

(2)

where S denotes the surface S minus the singular point at r. p is the primary potential due to the incident magnetic field HP , i.e. HP = P . The model can be immediately extended to multiple bodies with a constant susceptibility by interpreting S as the total bounding surface of all drums, i.e. in the case of N drums, by letting
N

S=
i =1

Si ,

where Si is the (inside and outside) surface of the ith drum. With this notation, equation (1) can be conveniently written in the form, HS (r) = k 4
N

(r )n
i =1 Si

1 ds , |r r |

(3)

where the potential on the ith drum is given by i (r) + k 2 (k + 2)


Si N

(r )n

1 ds |r r | = 2 P (r). k+ 2 (4)

+
j =1 j =i S j

(r )n

1 ds |r r |

constant, the present approach requires the solution of N M linear equations for the surface potentials. In order to extend the technique to the maximum possible number of drums, it is clearly necessary to find the minimum number of surface elements required to represent each drum adequately. In order to achieve a high degree of accuracy in the determination of the dipole moments of magnetized drums, Furness (2002) represented a standard steel drum by a hollow polygonal prism defined by 1440 planar surface elements with maximum dimensions of approximately 60 mm. Happily, it appears that such a fine discretization is not required for the present purpose of computing the magnetic fields due to buried drums. Indeed, the present technique is relatively insensitive to the size of the surface elements used, provided field points are sufficiently distant from the surface (i.e. for distances in excess of the maximum dimension of the elements). On the other hand, there are limitations on the size of planar surface elements used to represent a drum, which are imposed by the need to maintain the basic drum geometry. Clearly, the choice of a suitable surface discretization for the present purpose involves a compromise between minimizing the number of elements while still preserving the overall drum geometry and metal volume. With this in mind, several surface discretizations were used to model the total field anomaly measured along a northsouth profile over the centre of a square array of 2 2 close-spaced and vertical standard steel drums buried at a depth of 2 m below the sensor. The drums were arranged with their centres separated by 0.6 m in the x- and y-directions so as to achieve a spacing of 26 mm in these directions. The results of these experiments are shown in Fig. 1. Profiles resulting from all discretizations are remarkably similar; however, the coarser discretizations using 48 and 128 elements produce profiles that differ visibly from those using the finest discretization involving 1536 elements. Convergence of the results appears to be essentially complete for an intermediate discretization involving 448 surface elements, comprising two closed surfaces each with 16 elements around the circumference, 3 along the radius and 8 along the height. The results suggest that this discretization involving elements with a maximum dimension of just over 100 mm is the most appropriate for the present purpose. A perspective view of a single drum represented in this manner is shown in Fig. 2.

The obvious problem with this numerically intensive approach is the practical limitation on the number of drums imposed by the execution speed and memory of contemporary computers. If each of the N drums is approximated by M surface elements over which the potential is assumed to be

T H E T E C H N I Q U E O F S Y S T E M I T E R AT I O N The basic problem with the above integral equation technique concerns the difficulty of solving the very large systems

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

740 P. Furness

600

400

T (nT)

200

-200 0 2 4 6 8 10

x (m)

Figure 1 Total field magnetic anomaly along a northsouth profile over a rectangular array of 2 2 close-spaced vertical standard drums computed with the exact integral equation technique, showing the influence of the surface discretization. Each drum is represented by an internal and an external surface with a discretization denoted by nc nr nh where nc , nr and nh are the number of elements around the circumference, along the radius and along the height, respectively: 32 5 14 resulting in a total of 1536 elements (continuous line), 16 3 8 resulting in 448 elements (crosses), 8 2 4 resulting in 128 elements (circles) and 6 1 2 resulting in 48 elements (triangles).

Figure 2 Perspective view of a standard steel drum represented by 448 (224 internal and 224 external) planar surface elements.

of linear equations that result from the discretized form of equation (4) for multiple drums. As an example, a moderatesized accumulation of 100 drums, each represented by 448 surface elements in the manner described above, would require the solution of 44 800 equations. A simple approximate technique for the solution of equation (4) is possible by considering each drum to be influenced by a local primary field, comprising the geomagnetic field plus the perturbing fields of all other drums. If the latter are considered to be known, equation (4) for the potential on the ith drum can be written as i (r) + k 2 (k + 2)
Si

(r )n
N

2 1 ds = P (r) |r r | k+ 2 1 ds |r r | (5)

k 2 (k + 2)

(r )n
j =1 j =i S j

which suggests an iterative technique for the solution of the surface potentials. Initially, the surface potentials on all drums are provided by some suitable first approximation. An obvi-

ous choice is to neglect the perturbing effect of the drums entirely and assume that the surface potentials are equal to the potential of the applied field. Equation (5) can then be solved individually for the potential on each drum and the result used immediately to update the solution so as to allow a more accurate evaluation of the local field for the next drum. After each iteration, the surface potentials on all N bodies will have been updated and a convergence criterion can be tested. Convergence of the solution can be judged to be complete in a practical sense when the relative changes in all the unknown surface potentials over one iteration have an absolute value less than an acceptably small number. Equation (3) is then used to compute the required magnetic field using the resulting surface potentials. This method was first described in the geophysical literature by Xiong (1992) as the technique of system iteration. It was originally designed to solve large 3D electromagnetic problems using a volume integral equation formulation. The same type

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Modelling magnetic fields due to steel drum accumulations 741

of body-wise iterative solution technique was independently developed by Kikawada (1994) to solve large groundwaterflow problems involving multiple bodies using surface integral equations. The technique is particularly efficient for the present purpose where all bodies are identical. Consequently, the matrix equivalent of equation (5) remains the same for all bodies except for the right-hand-side vector, which depends on the local primary field. As a result, the coefficients of the matrix need be computed for one body only. Moreover, the matrix needs be decomposed only once during the whole modelling operation. The linear system can then be repeatedly solved for any right hand side (i.e. for any body in the group) by forwardand back-substitution using the appropriate right-hand-side vector. The whole process has shown itself to be quite robust, and convergence is very rapid. Convergence can be reckoned to be essentially complete in a practical sense when the solution potentials for all surface elements change by less than 0.1% between successive iterations. Typically this is achieved in 10 or fewer iterations. The technique has been routinely used to solve problems involving up to 50 000 unknowns. The performance of the technique can be judged by a perusal of Fig. 3. This compares the total field magnetic anomaly for

the situation previously considered in Fig. 1, computed by solving equation (4) directly and by the present method of system iteration (equation (5)). In both cases, individual drums were represented by 448 surface elements. The correspondence between the two results is seen to be very satisfactory and, in view of this result, system iteration will be used in all further demonstrations involving substantially larger accumulations, as a basis for evaluating the performance of more approximate modelling techniques.

MODIFIED TECHNIQUES OF SYSTEM I T E R AT I O N The technique of system iteration effectively removes the requirement for the storage and solution of large systems of linear equations. However, it remains numerically quite demanding and can be very time-consuming for large numbers of drums. This suggests the use of a more rapid, if approximate, modification of the original technique. Analysis of the algorithm shows that the time-consuming process involves the evaluation of the integral term on the right-hand side of equation (5), the contribution of which is regarded as part of the primary potential for the ith drum: I (r) = k 2 (k + 2)
N

(r )n
j =1 j =i S j

1 ds . |r r |

(6)

600

In practice, this is evaluated as a summation of partial integrals over the M individual surface elements of each drum as follows:
400

I (r) =
T (nT)

k 2 (k + 2)

(r )n
j =1 k=1 S jk j =i

1 ds . |r r |

200

Now, if the surface element Sjk is sufficiently removed from the field point that the gradient of the inverse distance remains essentially constant over its extent, then this last equation can be written in the form, I (r) = k 2 (k + 2)
N M

ds (r )n
j =1 k=1 S jk j =i

1 , |r r jk|

-200 0 2 4 6 8 10

x (m)

Figure 3 Total field magnetic anomaly along a northsouth profile over a rectangular array of 2 2 close-spaced vertical standard drums computed by the integral equation technique: direct solution of linear equations (continuous line), method of system iteration (crosses).

where rjk is the position vector locating the centre of the kth element of the jth drum. This equates each surface element to a point dipole located at its centroid with moment m jk = 2k k+ 2 ds (r )n

S jk

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

742 P. Furness

or in the case of constant planar surface elements, 2k jk, jk Ajkn m jk = k+ 2 where Ajk is the area of the kth element of the jth drum. Equation (5) then becomes i (r) + k 2 (k + 2)
Si

3000

2000

(r )n

2 1 ds = p (r) |r r | k+ 2
T (nT)

1000

1 4

m jk
j =1 k=1 j =i

1 . r r jk

(7)

Alternatively, a more rapid, if less accurate, approach can be made if the gradient term in equation (6) is assumed to remain constant over each drum so that, to a good approximation, we have I (r) = k 2 (k + 2)
N

-1000

ds (r )n
j =1 j =i S j

1 , |r r j |

-2000 0 4 8 12 16 20

x (m)

where rj is now the position vector locating the centroid of the jth drum. This equates the effect of the jth drum to that of a point dipole with moment, mj = 2k k+ 2 ds , (r )n
Sj

Figure 4 Total field magnetic anomaly along a northsouth profile over a rectangular array of 10 10 close-spaced vertical standard drums computed using various integral equation and discrete dipole approximations based on a dyadic susceptibility: system iteration (continuous line), modified system iteration (crosses), discrete dipoles neglecting depolarization (circles), discrete dipoles with depolarization (triangles).

and in the case of constant planar elements, mj = 2k k+ 2


M

jk. jk Ajkn
k=1

Equation (5) may then be written in the form, i (r) + k 2 (k + 2)


Si

(r )n 1 , r rj

1 2 ds = p (r) |r r | k+ 2

1 4

mj
j =1 j =i

(8)

This latter modification of system iteration employing equivalent drum dipoles is particularly appealing due to its efficiency in time. Dipole moments for each drum are computed just once per iteration immediately after the potentials for a particular drum are updated. Computation of the primary potentials expressed by the right-hand side of the last equation is then trivial. The accuracy of this modification for a more substantial array of 10 10 close-spaced vertical standard drums is demonstrated in Fig. 4. The agreement with results computed by the original technique of system iteration is seen to be reasonably satisfactory, bearing in mind that the latter took several hours

to compute while the former required a time of the order of only 10 seconds. While the nature of the present modification is quite extreme, the results of Fig. 4 suggest that it retains a fair degree of accuracy. If greater accuracy is required, a range of possibilities exist. A more cautious approach would be to retain the right-hand side of equation (5) for the computation of the primary potentials due to drums in the immediate neighbourhood of the field point and to use the equivalent dipole representations only for sufficiently distant elements (equation (7)) and eventually with increasing distance for entire drums (equation (8)).

D I P O L E A P P R O X I M AT I O N S U S I N G A D YA D I C D R U M S U S C E P T I B I L I T Y The integral equation techniques discussed to date clearly provide the most accurate indication of magnetic fields due to drum accumulations. However, the computations are numerically intensive and require significant computing power. The modified technique of system iteration allows the rapid computation of magnetic fields on very moderate

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Modelling magnetic fields due to steel drum accumulations 743

computing equipment but remains quite complicated to implement. It is therefore worthwhile investigating the performance of some simpler techniques. The most obvious of these techniques treats each drum in terms of an equivalent magnetic dipole located at its centre, so that the magnetic H field due to an accumulation of N drums is given by HS (r) = 1 4
N

equation (9), this can be written as H(ri ) = HP + 1 4


N

m j
j =1 j =i

1 , i = 1, 2, 3 . . . N, |ri r j |

where ri locates the centroid of the ith drum. The component of il ) then becomes the field in the drums lth principal direction (u Hil = HPil + 1 il u 4
N

m j
j =1

1 , r rj

(9)

m j
j =1 j =i

1 , i = 1, 2, 3 . . . N, |ri r j |

where rj locates the centre of the jth drum. According to Furness (2002), the component of the dipole moment induced in a drum in each of its three principal directions is given by m = k V H where V is the volume of metal contained in the drum, H is the component of the inducing H field in a particular direction and k is the apparent susceptibility in this direction. For a standard steel drum, this was found to vary with the susceptibility k as follows: ak k = , (10) 1 + bk where a = 0.740176, b = 0.00157453 for the longitudinal direction and a = 0.627583, b = 0.00205086 for the transverse directions. This behaviour suggests that the dipole moment induced in a drum can be conveniently expressed in the form, H, m = Vk is a dyadic apparent susceptibility defined by where k = k1 u 1u 1 + k2 u 2u 2 + k3 u 3u 3, k (12) (11)

l = 1, 2, 3, where il HP . HPil = u This induces a component of the dipole moment in the lth direction of the ith drum, given by mil = kil V Hil or mil = Vkil il u 4
N

m j
j =1 j =i

1 + Vkil HPil , |ri r j |

i = 1, 2, 3 . . . N,

l = 1, 2, 3.

Finally, by substituting
3

m=
k=1

k, mku

this becomes mil Vkil il u 4


N 3

and the unit vectors define the three orthogonal principal directions of the drum. The simplest method of implementing these equations is to neglect the demagnetizing field due to the accumulation and assume that each drum is influenced by the primary (geomagnetic) field HP alone. Use of this field in equation (11) and substitution of the resulting dipole moment in equation (9) then yields HS (r) = V 4
N j =1

jk m jku
j =1 k=1 j =i

1 = Vkil HPil , |ri r j | (14)

i = 1, 2, 3 . . . N,

l = 1, 2, 3.

j HP ) (k

1 , |r r j |

Equation (14) expresses the component of the ith dipole moment in the lth principal direction in terms of the component of the jth dipole moment in the kth principal direction. If m and n are now used to denote these two directions, i.e. if m indicates the lth principal direction of the ith drum and n indicates the kth principal direction of the jth drum, so that m = 3(i 1) + l , and n = 3( j 1) + k, j = 1, 2, 3 . . . . N, k = 1, 2, 3, i = 1, 2, 3 . . . . N, l = 1, 2, 3,

(13)

which provides a very crude approximation for the anomalous magnetic field due to an accumulation of steel drums. Alternatively, a first-order accommodation of the depolarizing effect of the accumulation can be made by assuming the ith drum to be exposed to a uniform field comprising the primary field plus the secondary fields of all other drums. Using

then equation (14) can be written in matrix form as Ax = b,

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

744 P. Furness

where A is a 3N 3N matrix with coefficients, amn = Vkil 1 jk il u u , 4 |ri r j |

1600

1200

when j = i. When j = i and k = l, then amn = 1.


800

amn = 0. x is the 3N vector of unknown dipole moment components, xm = mil and b is the 3N vector of known primary field terms, bm = kil V HPil . The solution of the linear system then yields the 3N components of the dipole moments associated with the N drums. These are used in equation (9) to find the anomalous H field due to the drum accumulation. The basic assumption used in this approach is that each drum experiences an essentially uniform field comprising the geomagnetic field plus the effects of all other drums in the assemblage. Clearly this will be best satisfied when the drums are sparsely distributed. When more densely packed, the influence of other drums cannot be expected to be uniform over the dimensions of a particular drum. However, the approach provides an easily implemented first-order accommodation of the demagnetizing effect due to the shape of the drum accumulation. Figure 4 demonstrates the accuracy of the two discrete dipole approximations for the 10 10 array of close-spaced vertical drums described previously. The simple approach neglecting demagnetization is seen to provide a reasonable indication of the anomaly as defined by the more accurate integral equation solutions. However, not surprisingly, it overestimates the anomaly amplitude. The more refined approach accommodating demagnetization produces a significantly better fit. Figure 5 demonstrates the performance of all approximate techniques discussed to date in computing the total field anomaly over what might be considered to be a more realistic representation of a drum accumulation comprising a 3D rectangular array of 16 2 3 horizontal standard drums with axes in the y-direction. Drum centres are arranged at intervals of 0.6 m in the x- and z-directions, providing spacings of 26 mm, and at intervals of 0.88 m in the y-direction, with spacings of 31 mm. The depth to the top of the accumulation below the sensor is maintained at 2 m.

T (nT)
400

Otherwise, when j = i and k = l,

-400 0 4 8 y (m) 12 16

Figure 5 Total field magnetic anomaly along an eastwest profile over a rectangular array of 16 2 3 close-spaced horizontal standard drums computed using various integral equation and discrete dipole approximations based on a dyadic susceptibility: system iteration (continuous line), modified system iteration (crosses), discrete dipoles neglecting depolarization (circles), discrete dipoles with depolarization (triangles).

The relative performances of the various approximations appear to be similar to those demonstrated in Fig. 4 for the array of vertical drums. The integral equation techniques provide the best accuracy. The discrete dipole approach accommodating depolarization produces results with reasonable accuracy, while the simpler approach neglecting depolarization substantially overestimates the anomaly amplitude. D I P O L E A P P R O X I M AT I O N S U S I N G A SCALAR DRUM SUSCEPTIBILITY The techniques discussed to date offer a range of possibilities for computing the magnetic fields produced by collections of steel drums. However, they all require information on the orientation of the individual drums. In many modelling applications, this type of detailed information is unavailable. For this reason, the use of an average scalar (isotropic) apparent susceptibility used in conjunction with discrete dipoles may be attractive. If the apparent susceptibility of a drum can be assumed to be a scalar quantity, the induced dipole moment is given by m = k V H, (15)

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Modelling magnetic fields due to steel drum accumulations 745

k =

3k(2k + 3) , (2k + 3)(k + 3) 2k2 (a /b)3

(16)

T (nT)
400

and is always in the direction of the inducing field H. It remains only to define a suitable scalar apparent susceptibility for a steel drum. Two possibilities exist. In a numerical study of isolated steel drums, Furness (2002) showed that a reasonable value for the scalar apparent susceptibility of a steel drum is provided by an equivalent spherical shell with the same wall thickness and metal volume. The apparent susceptibility of such a shell is given by

1600

1200

800

where a and b are the shells inner and outer radii, respectively. The radii of a spherical shell equivalent to a drum with wall thickness t and surface area A are conveniently found from b= and a = b t. The work of Furness (2002) would suggest that the use of a scalar susceptibility found from an equivalent spherical shell is applicable to most closed containers with roughly equal dimensions. An alternative approach, used here for drums with standard dimensions, employs the average value of the apparent susceptibilities of a standard drum in its three principal directions. This is easily computed from equation (10) with the constants provided. As before, equation (15) can be implemented immediately by neglecting the depolarizing field. Substitution of the resulting dipole moment in equation (9) then yields HS (r) = Vk 4
N

A 4

1/2
-400 0 4 8 y (m) 12 16

Figure 6 Total field magnetic anomaly along an eastwest profile over a rectangular array of 16 2 3 close-spaced horizontal standard drums computed by the technique of system iteration (continuous line) compared with the results using a scalar drum apparent susceptibility to define discrete dipoles (circles) and discrete dipoles with depolarization (triangles).

HP
j =1

1 . |r r j |

(17)

Alternatively, the accuracy can be improved by incorporating demagnetization. The components of the unknown dipole moments located at the centroids of each drum are found by solving a rather simplified version of equation (14), i.e. mil Vk il u 4
N 3

jk m jk u
j =1 k=1 j =i

1 |ri r j |

= Vk HPil ,

i = 1, 2, 3 . . . N,

l = 1, 2, 3,

horizontal drums considered in the previous figure. Results computed with the discrete dipole approaches, using a scalar apparent susceptibility in the manner described, are seen to provide reasonable indications of anomaly form but to overestimate anomaly amplitudes. The problem is seen to be reduced by acknowledging the influence of depolarizing fields. It is also relevant to note that the present test represents an extreme case in which all drums are orientated transversely to the inducing field. In this situation, a more appropriate value of the drum susceptibility would be the transverse value rather than the substantially larger average drum susceptibility used here. Undoubtedly this overestimation of the susceptibility contributes to the excessive amplitudes seen in Fig. 6. A comparison of Figs 5 and 6 shows the present results to be inferior to those of the corresponding discrete dipole models employing a dyadic apparent susceptibility. Nevertheless the present results appear to be useful, particularly as drum orientations are not required for the model. C O N T I N U O U S M A G N E T I Z AT I O N A P P R O X I M AT I O N S The above considerations suggest taking the use of a scalar susceptibility further so that detailed drum locations also

and the resulting dipole moments are used in equation (9) to compute the secondary H field due to the accumulation. The result of using a scalar apparent susceptibility can be judged by a perusal of Fig. 6. It shows the total field anomaly along an eastwest profile over the array of close-spaced

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

746 P. Furness

become unnecessary for modelling. Accordingly, if the distribution of drums in the accumulation is sufficiently uniform and provided the magnetic sensor is sufficiently remote from the accumulation that the effects of individual drums are not apparent on the profile, it seems reasonable to represent the accumulation as a whole by a continuous magnetization defined by M= Nm . VA

where the unit vectors define the three principal directions. The apparent susceptibilities in these directions are given by ki = k , 1 + Ni k i = 1, 2, 3, (21)

where the Ni are the relevant depolarizing factors (Clark and Emerson 1999). The use of a continuous magnetization to represent the accumulation as a whole now allows the anomalous magnetic field to be expressed in the familiar form, HS (r) = 1 4 M(r )
V

Here, N is the number of drums in the accumulation, m is the dipole moment of an individual drum given by equation (15) and V A is the volume of the accumulation. Substitution for the dipole moment allows the magnetization to be written in the form, M = Dk V H, where D is the volume density of drums, V is the volume of metal in a drum and H is the magnetic field existing inside the accumulation. This last result can be written in the form, M = kH, (18)

1 , |r r |

(22)

where V is the region occupied by the accumulation and the magnetization is given by equation (19) or (20) depending on whether depolarizing effects are recognized. Figure 7 allows an evaluation of the continuous magnetization approach for computing the total field anomaly over the close-spaced array of horizontal drums considered previously. Again, the total field anomaly, computed with the integral equation approach using system iteration, is included for comparison. Both profiles using a continuous magnetization were computed by using the subroutine mbox provided by

where k is an effective susceptibility for the accumulation, defined by


1600

k = Dk V. Again, there is a choice in the implementation of equation (18), depending on the interpretation of the field acting inside the accumulation. If the demagnetizing field is assumed to be negligible, the internal field is equated to the primary geomagnetic field and M = kHP . (19)
1200

800

T (nT)
400

Alternatively, if the demagnetizing field is accommodated, then


0

M = k(HP + HS ), which is conventionally, if approximately, implemented by appeal to the behaviour of an equivalent triaxial ellipsoid. This allows the definition of apparent susceptibilities relating components of the magnetization to components of the primary field in the three principal directions of the accumulation. Accordingly, the magnetization is given by
3
-400 0 4 8 y (m) 12 16

M=
i

i, ki HPi u

(20)

Figure 7 Total field magnetic anomaly along an eastwest profile over a rectangular array of 16 2 3 close-spaced horizontal standard drums computed by the technique of system iteration (continuous line) compared with the results obtained by using a continuous magnetization (circles) and a continuous magnetization with depolarization (triangles).

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Modelling magnetic fields due to steel drum accumulations 747

Blakely (1995), which approximates the total field anomaly by the component of the anomalous B field in the direction of the geomagnetic field. The accumulation was approximated by a rectangular prism with dimensions 9.6 1.76 1.8 m. Equations (19) and (20) were used to compute the magnetizations without and with depolarization, respectively. In the latter case, the apparent susceptibilities in the three principal directions of the accumulation were found from equation (21) by using approximate values for the depolarizing factors of a rectangular prism given by Clark and Emerson (1999). The quality of the present result neglecting demagnetization is seen to be similar to that produced by the corresponding discrete dipole approximation of the previous figure and again the same comments apply concerning the excessive amplitude. The present result incorporating demagnetization is somewhat different from that of Fig. 6, which probably reflects the difference in the approach used to accommodate depolarizing effects (including, amongst other things, the use of very approximate values for the depolarizing factors). While the anomaly magnitudes are somewhat in error, the anomaly shape is seen to be reasonably predicted by both continuous magnetization models. Overall, it must be conceded that the results are surprisingly good, bearing in mind the approximations involved and the ease of computation. It is suggested that there are many practical applications where the quality of these results would be quite acceptable.

600

400

T (nT)

200

-200 0 10 20 30

d (m)

Figure 8 Total field magnetic anomaly measured along a profile bearing 50 magnetic over a 3 3 array of close-spaced vertical drums without bases except for the central drum: field data from Emerson et al. (1992) (crosses) and computed result using system iteration with 384 surface elements per drum (continuous line).

A FIELD EXAMPLE It may be interesting to see an application of the present theory to some field data. Emerson et al. (1992) published total field profiles over the centre of a rectangular 3 3 array of closespaced vertical 205-litre drums, all without bases except for the central drum. Data plotted with crosses in Fig. 8 show the total field anomaly measured over the centre of the drum array taken graphically from Emerson et al. (1992). The profile advances in a north-easterly direction bearing 50 magnetic with a sensor height of 2.7 m above the top of the drums. The results modelled with the present integral equation technique using system iteration are shown as a continuous line in Fig. 8. The integral equation approach employed a 3 3 array of standard steel drums with a susceptibility of 350 SI, all without bases. The base of the central drum was neglected for the sake of convenience and all drums were represented by 368 planar surface elements with maximum dimensions of a little over 100 mm. Remanent magnetization was ignored.

The fit of the two data sets is remarkable and quite unexpected, bearing in mind the variability in the drums used in the field experiment. Emerson et al. (1992) measured profiles over individual drums that showed good evidence of remanent magnetization as well as substantial variation in susceptibility and/or remanence between drums. Moreover, the standard drum (without base) used in the present model has significantly less mass than many of the drums used in the field experiment. However, it is probable that much of this difference in mass is due to seams, welds, rolled joints, etc., which, being more equi-dimensional than the walls, are unlikely to contribute significantly to the magnetic anomaly. The possibility of thicker walls contributing to the increase in mass is disputed by the data published by Eskola et al. (1999). Figure 8 demonstrates that, in spite of the variation in the properties of individual drums, their combined magnetic effect is satisfactorily explained by an array of standard drums. This tends to support the idea that a standard drum with susceptibility in the vicinity of 350 SI offers a reasonable average representation of available 205-litre drums. Moreover, it lends substantial weight to the argument that while individual drums may possess significant remanent magnetization, the

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

748 P. Furness

magnetic fields produced by their accumulations show little evidence of this phenomenon.

Table 1 Summary of approximations used and relevant figures demonstrating their application Approximation used Relevant figures 3 4, 5 4,5 6 7

CONCLUSIONS A range of possibilities have been presented for the approximate computation of magnetic fields produced by accumulations of steel drums. It may clarify the discussion to summarize the approximations considered. Accordingly, Table 1 lists all the approximate techniques described and the figures demonstrating their application. It should be noted that the last three approximations listed all admit two variations in which the depolarization fields due to the accumulation as a whole is either neglected or accommodated. These demonstrations suggest that the magnetic field produced by large accumulations of thin-walled steel drums can be accurately modelled by a surface integral equation technique used in conjunction with system iteration for the solution of the linear equations. Even so, the process is numerically intensive. The modified technique offers huge economies in computing time with little loss of accuracy, and so suggests itself as a first choice for practical applications where drum orientations and locations are known. The various models using discrete dipole approximations have the advantage of simplicity and, in the cases examined to date, retain a surprising degree of accuracy. In general however, they require the use of a dyadic drum susceptibility, as well as the accommodation of depolarizing effects due to the accumulation as a whole, to yield reasonably accurate results. Of all the techniques considered, the use of a scalar susceptibility to represent drum accumulations produces the least accurate results. However, it is fair to note that the present tests of these techniques are somewhat exceptional in that all drums are orientated transversely to the inducing field. It is unlikely that a situation would arise in practice in which all the drums in an accumulation would be aligned transversely (or longitudinally) to the geomagnetic field. The use of a scalar susceptibility has several advantages that may be appreciated in some practical applications: it does not require any detailed information on the orientation and location of individual drums and it can be easily implemented using

Integral equation with system iteration Integral equation with modified system iteration Discrete dipoles using a dyadic susceptibility Discrete dipoles using a scalar susceptibility Continuous magnetization

existing magnetic modelling software. It should be seriously considered as a technique for producing indicative anomalies when only the accumulation dimensions and drum concentration are available. ACKNOWLEDGEMENTS The author is grateful for the very detailed comments provided by Horst Holstein and another anonymous reviewer. These have significantly contributed to a clearer presentation. REFERENCES
Blakely R.J. 1995. Potential Theory in Gravity and Magnetic Applications, p. 379. Cambridge University Press, ISBN 0-521-41508-X. Clark D.A. and Emerson D.W. 1999. Self-demagnetization. Preview 79, 2225. Emerson D.W., Reid J.E., Clark D.A., Hallett M.S.C. and Manning P.B. 1992. The geophysical responses of buried drums field tests in weathered Hawkesbury Sandstone, Sydney Basin, NSW. Exploration Geophysics 23, 589617. Eskola L., Puranen R. and Soininen H. 1999. Measurement of magnetic properties of steel sheets. Geophysical Prospecting 47, 593602. Furness P. 1999. A versatile integral equation technique for magnetic modelling. Journal of Applied Geophysics 41, 345357. Furness P. 2002. The magnetic fields of steel drums. Journal of Applied Geophysics 51, 6374. Kikawada A. 1994. An integral equation technique for groundwater modelling. PhD thesis, University of Queensland. Xiong Z. 1992. Electromagnetic modelling of 3-D structures by the method of system iteration using integral equations. Geophysics 57, 15561561.

2007 European Association of Geoscientists & Engineers, Geophysical Prospecting, 55, 737748

Das könnte Ihnen auch gefallen