You are on page 1of 7

Chapter 12

A Blending-Based Approach to Mine Planning and Production Scheduling Mark Gershon

Temple University

I. Introduction per role of the operations researcher


Most recent work in the field of is to provide not only useful and ad-
mine production scheduling has focused vantageous approaches, but practical,
on either the computerization of the understandable and implementable ones
traditional methods or the development as well. This is the goal of the paper
of sophisticated mathematical optimi- in describing a new approach, consis-
zation models. Yet it is clear to all tent with the traditional methods and
concerned that the optimizers need to aimed at approximating the results of
be made more practical and the tradi- the optimizers.
tional approach needs to be made more
optimal. This paper, along with an- The method described herein relies
other recent paper (Gershon, 85) open on the best features of both current
a middle ground between these two ap- approaches. As an approximation tech-
proaches. In that recent paper, a nique, it falls into the operations
heuristic procedure for production research classification as a heuristic.
scheduling was introduced. Here, a It may or may not ultimately be the
series of short term optimizations are best heuristic that can be designed for
used in a creative manner to approxi- mine production scheduling, but it has
mate the true long term optimization. been applied successfully to operating
Since these small optimizations are mines and it continues this new area
aimed at blending to meet production of study, heuristics, aimed at improv-
quality specifications, this approach ing it in the future.
is most applicable to commodities where
the blending function is very important Prior to describing the specifics
such as coal, cement, phosphates or of the approach, the scheduling problem
iron, to name a few. itself is discussed along with some of
the current approaches. The heuristic
The traditional and the optimization is then described, followed by a des-
approaches are so far apart that their' cription of the possible implementations
advocates have difficulty identifying and the overall system for production
with the other approach. This is es- scheduling in which it should be used.
pecially true concerning the difficulty
of understanding the optimization 11. The Production Scheduling Problem
models for those not trained in the
field of operations research. This The production scheduling problem,
paper takes the position that the pro- in the mining context, is concerned
120
BLENDING-BASEDAPPROACH TO PLANNING

with the sequence of the removal of the all of the restrictions; one that can
mining blocks within the limits of the be implemented. An excellent descrip-
mine plan. In other words, the problem tion of the details of this approach
is one of block sequencing. The se- can be found in Matheson (1982). This
sequence of mining that yields the lar- approach has traditionally been imple-
gest net present value return is the mented manually, requiring a great
one that should be chosen. deal of calculation and drawing of maps.

While a discussion of the information Today, most companies are incorpor-


system required for production schedul- ating revolutionary changes into this
ing is outside the scope of this paper, same approach. The use of computers for
it is assumed throughout the discussion the calculations as well as the gra-
that the basic information contained in phics is helping to speed up the pro-
most block models is available. That cess. In a way, reducing the time re-
is, the quantity and quality of each quired to produce each production sche-
block have been estimated and are dule is also helping to produce better
readily available to the production production schedules by providing the
scheduler. scheduler with the time to evaluate a
variety of options. Examples of these
The simplistic problem statement approaches are described in Marek (1985)
provided at the beginning of this sec- and Welhener (1985). .
tion is not sufficient once the produc-
tion scheduler considers the many limi- The most sophisticated implementa-
tations on the actual schedule. For a tion of these trial and error appro-
variety of reasons, only a small subset aches involves the use of a complete
of all possible sequences can be im- computer simulation of the mining of
plemented. The rest are not practical. the deposit. While the development of
These reasons include working and final a simulator may require a very large
slope considerations, restrictions on initial investment, it puts in place a
the mobility of equipment, the need to very powerful tool for quickly evalua-
balance strip ratios, and blending con- ting many scheduling options. While
siderations on the qualities. Thus, some companies have implemented simula-
the problem becomes one of finding a tors, most are building their simula-
practical schedule; one that meets all tors piece by piece. This is the com-
of these requirements. puter assisted approach, but the ulti-
mate goal is to use a complete simula-
This is the level at which most pro- tion.
duction scheduling projects are con-
ducted. The difficulty in achieving a The three examples mentioned above,
practical schedule overrides the pro- in actuality, are all different forms
blem of finding the best schedule. With of the same approach, that of "trial
these considerations in mind, the pro- and error". This general approach is
blem definition used in this paper is: to try a schedule, modify it until it
meets all of the specifications, find
Find the schedule (sequence of min- a few more that meet the specifications,
ing blocks) that maximizes the net and choose the best one. But bear in
present value return within the mind that the best of all possible
physical, logistical and contractur- schedules is not found. Only the best
a1 limitations by which the schedu- of those tried is found.
ler is constrained.
In response to this desire for the
111. Current Approaches optimal (best of all possible) schedule,
many approaches based on mathematical
Every mine operates under some kind optimization procedures (Johnson, 1969;
of production schedule, but as has been Smith, 1978; Gershon, 1982, 1983a)
mentioned earlier, most have focused have been developed. These are usually
merely on finding a schedule that meets based on either linear programming or
19th APCOM SYMPOSIUM

integer programming (Hillier and Lieb- fore, the heuristic provides most of the
erman, 1980). While they do find the power of the optimization within the
optimal solution to some mathematical framework of the more traditional sched-
model, the model assumptions are often uling approaches.
such that the resulting schedules may
not be entirely practical. Some mas- Prior to describing the approach, it
saging of the results is then required is necessary to briefly describe the
to achieve a useful and implementable blending problem. In general, we as-
schedule. sume that there are many sources from
which to take the material to make the
Finally, a third group of approaches blend. For the sake of simplicity, as-
attempts to build some level of in- sume that there are three sources.
telligence into a computer model to There will also be a set of quality
guide the scheduling process. Some of specifications that the blend must sat-
these use optimization for certain isfy. For coal, there may be ash, sul-
parts of the analysis (subobtimization) fur and BTU content, among others. Each
such as Lane (1964) or Baafi (1983). commodity would have its own critical
Others, such as Garg (1983) or Gershon qualities. For example, iron mines may
(1985), attempt to build a schedule blend on percent iron, magnetite, lim-
based not on optimization or suboptimi- onite, weight recovery and a grindabil-
zation, but on some practical knowledge ity index. Phosphate operations may
of what factors lead to good schedules. blend on percent phosphate, moisture,
This "heuristic" approach exploits per cent clay and a variety of rare
these factors to attempt to duplicate, .elements. Cement companies have pos-
or at least approximate, the optimiza- sibly the strictest blending require-
tion at a.much lower cost. A new ments and this list could be extended to
heuristic approach, that utilizes the most mining commodities. To complete
suboptimization concept is the subject this problem description, it must be
of the remainder of this paper. assumed that an estimate is available
that describes the actual quality char-
IV. The Blending Approach acteristics of each source.
to Production Scheduling
This blending problem description
The advantage of the use ofanopti- contains all of the concerns that one
mizer for production scheduling is ob- would have in making blends. Most com-
vious. The best possible schedule re- panies are still doing this by intui-
sults. However, there are many advan- tion, but the best way to do it is with
tages to using a heuristic approach for linear programming. The linear model
this same purpose. First, it uses a that solves this problem optimally is
common sense approach that is easy to not developed here, since it is well
understand by the mine engineers who known and has been in use for blending
may have been using a manual or com- in industry for thirty years. However,
puter assisted method. Second, it is it is this simple model that is the
easily implemented within a computer basis for the production scheduling
assisted approach or a simulator. It approach.
can even be used to guide a manually
derived schedule. In fact, the method Let us start with a very bad idea,
described herein has been applied in one that has seriously been suggested
exactly this way. Finally, since the by some in the industry as practical.
optimization methods are so difficult If the optimal blend is mined this week,
to apply to this production scheduling and then the optimal blend is mined the
problem (Gershon, 1983b), a method that next week, and in each succeeding week,
can quickly and efficiently approximate the blending model essentially becomes
the results of the optimizer is needed. a production scheduler. There can be no
The hepristic described here, which argument with this. What can be argued,
makes use of multiple small optimiza- however, is the quality of the resulting
tions, accomplishes this task. There- schedule. Certainly, it is not optimal
BLENDING-BASEDAPPROACH TO PLANNING

in the long run. A much larger linear it has been proven to yield excellent
programming model that solves all of results.
the time periods simultaneously is need-
ed for this. But in the approach de- What is the objective, then, that
veloped here, no optimization is claim- achieves this goal? It depends on the
ed. Therefore, a more serious question property. Since this approach was
is whether the resulting schedule is a first applied to cement properties, let
good one. Again, it is not. Every us consider that first. Limestone is
mining engineer knows the pitfalls of a the most important ingredient, so a
short sighted approach and this is ob- suitable objective could be to minimize
viously short sighted. Mining the op- the amount of limestone used to meet the
timal blend in each time period will specification. This has the effect
soon lead to a situation where .it is that, at the end of each period, the
not possible to meet the specifications most limestone possible remains. In
from the available material. other words, the company has placed it-
self, at the end of each period, in the
As bad as this idea may be, it still best possible position for long term
has a strong appeal. It is simple to mining success. More will be discussed
apply, requires very little computer on this case in the next section.
time (can be micro-based), and gives
the production scheduler full control For coal, it is really heat content
over the development of the schedule. (BTU's) that are bought, so it would
Even more important, there are many seem that saving these for the future
commodities, especially those mentioned is to be the goal. But this is never
previously (coal, iron, phosphite, the case, partially because of the
cement), where a successful production heavy use of incremental contract
schedule is driven by the ability to prices tied to BTU's. In many U.S.
successfully put together a continuing properties, due to strict environmental
sequence of blends. For these commod- laws on the burning of coal, the key
ities, blending does drive the pro- ingredient is sulfur. While it seems
duction scheduling process. odd at first thought, the appropriate
objective to use is to maximize the
The approach taken here is to take sulfur mined, subject to staying within
this bad idea and modify it so that it the sulfur limits. Again, this has the
is a good idea, one that is useful in effect of leaving the least possible
practice. The bad part of the idea sulfur content in the property at any
thus far is that it is very short given time.
sighted. Therefore, our goal is to
modify it so that long term considera- More specifics are provided in the
tions are incorporated. next section. However, there are a few
steps that summarize the procedure.
The key to accomplishing this task These are:
lies in the definition of the "optimal" 1. Model the single period blend-
blend, what the linear programmer calls ing problem;
the objectives. If it means that the 2. Determine the appropriate long
blend is the best in terms of cost or term objective;
profit, somehow the prices must be 3. Select the available sites for
linked to the material quality and mining (the sources);
again this is too short sighted. 4. Optimize the single period
blend for these sources in
A long term objective must be chosen. terms of the long term ob-
We make the assumption that the best jective;
long term schedule is one that allows 5. Return to Step 3 until sched-
the specifications to be met for the ule is complete.
longest period.of time. While this, The idea is to do a series of short
too, is rather simplistic, as an alter- term optimizations using a long term
native to a time long term optimization, objective to approximate the .true long
124 19th APCOM SYMPOSIUM

term optimal schedule. schedulers and mine planners chose to


avoid the problem areas until there was
V. Implementation nothing left with which to blend them.

Since the most critical step in im- This rule leads to a helpful hint
plementing this approach is the selec- toward the selection of a suitable ob-
tion of the appropriate long term ob- jective.
jective, more discussion is needed on Identify the most troubling pro-
this point. In the previous section, blem in mining the deposit and-this
the minimization of limestone usage was will lead to an objective to over-
mentioned as the objective for a cement come it, or at least to minimize
operation. While this is the best ob- the trouble that it causes.
jective across all properties, it has
not been used on any of the three ce- There are some less intuitive ways
ment projects conducted. In one case, in which this approach can prove useful.
it was most appropriate to attempt to For example, a long term (50 year) plan
make maximum use of a large dolomite for a lignite deposit was developed with
deposit available at the same site, so an objective of balancing the BTU pro-
this was chosen as the objective. At duction over time. This resulted in
another site, significant costs result- large fluctuations in the strip ratio
ed from having to track an overburden over time. To overcome this problem,
layer to a waste dump. Here, the ob- another schedule was developed aimed at
jective was to use as much as possible a balanced strip ratio. As one might
of this material in the mix. At a expect, this resulted in large fluctua-
third site, where the most extensive tions in BTU production. By viewing a
work has been done with this approach, combination of strip ratio and BTU's as .
management feared that the operation a "blendt', a more useful and practical
would have to be shut down in a few schedule results that takes both factors
years due to a large amount of a con- into account.
taminant, magnesium, in many areas of
the deposit. The long term objective In addition to the objective, the
chosen was to maximize the magnesium rest of the linear blending model con-
content in each time period (in each sists of the restrictions on the speci-
blend) while staying within the max- fications. Any introductory operations
imum allowed. This resulted in an ex- research text (Gershon, 1987; Hillier
tended useful life for the deposit. and Lieberman, 1980) provides sufficient
information to guide the reader in im-
This last case, where the contamin- plementing this step.
ant magnesium was maximized (mined out
as much as possible), is very similar Another major consideration, with re-
to the case mentioned previously con- gard to the implementation, is the sel-
cerning the sulfur in coal. In each ectionxof the available blocks (sources)
case, there is an element present in for e'ach run. In its most simplistic
the deposit that makes the deposit less sense, we could use all exposed blocks.
desirable. In each case, this approach If this approach were to be used as an
is used to attack the problem and at- automated scheduler, this would be the
tempt to optimize the usefulness of the way to do it. But a skillful engineer
property. The results of this approach can guide the schedule by providing the
point up a rule that seems obvious but model with only sources in selected
is not followed in many cases. That is: areas. The scheduling process then be-
comes one of trial and error, with some
If a deposit has a contaminant, ad- runs responding that the specifications
dress the problem immediately. cannot be met from the sources used.
Don't put it off.
There 'are many cases where a mine was Obviously, a decision made by the
forced to shut down because the respon- engineer early in the scheduling process
sible (irresponsible?) production will have an effect on the long term
BLENDING-BASEDAPPROACH TO PLANNING

schedule that results. This is desir- and, in the other, any cross section.
able since we must presume that this The data base window is used to display
engineer knows the property for which the one reserve model, so that the qual-
helshe has responsibility. On the other ities of the blocks of interest are al-
hand, though, it brings into question ways available to the scheduler. The
the quality (the optimality) of the schedule summary window provides a con-
schedule. As soon as we accept that the tinuously updated summary of the pro-
schedule may not be optimal, it becomes duction schedule, including total ton-
necessary to question how close to op- nage, average qualities, waste-ore
timal it may be. The way to do this (in ratios, etc. While the heuristic is
a qualitative sense) is to use the ap- aimed at approximating the optimization
proach to produce a few different of quality blends, it does not consider
schedules, each based on a different items such as strip ratio explicitly
view of how the engineer sees the long at each step. By using the heuristic
term plan evolving. These different within this environment, the engineer
schedules can then be compared and the can use this summary data to help guide
best one selected. This may even be the scheduling process toward strip
more of a mine planning use of the ap- ratio goals.
proach than it is a production scheduling
use. In either case, it is clearly Finally, the control window is es-
using the approach to analyze various sentially the input window with prompts.
options, while attempting to optimize It is here that the available sources
within each option. Thus, it is using are specified. In addition, the speci-
the approach as a simulation, but with fications can be modified through this
imbedded optimizations. window for any given run. Upon receiv-
ing the command to mine a given blend,
VI. System Design all of the other windows are updated.
The summaries are recalculated and the
The discussion of the implementation graphics windows convert the areas just
in the previous section alluded to the mined to their new status, generally by
fact that this procedure should be com- a change of color in the display.
puterized. In this section, we briefly. Through the control window, it is also
discuss the environment in which the possible to step backward; that is, to
program is used, borrowing heavily from put the blend back and choose another.
many ideas that now constitute a good
computer assisted scheduling system. VII. Summary
Essentially, the blending approach rests
on top of, and becomes the focal point In specific terms, this paper has
of, the computer assisted or simulation presented a new approach aimed at im-
framework. proving the mine production scheduling
process. By combining a short term
An essential point of the system de- optimizer with a long term objective,
sign is that the scheduler must have as what appears to be a simple blending
much information as possible in a readily model is converted into a powerful tool
available form. This may entail a multi- for production scheduling and mine plan-
window screen, multiple terminal screens ning.
or a terminal with a plotter. Assuming
that the multi-window environment is In more general terms, this paper
most ideal, the system should consist of: opens up an area of investigation that
- two graphics windows has not been explored until now. It is
- a data base window conceivable that there are other ways
- a schedule summary window of approximating an optimizer for this
- control window problem. There may also be many ways
of adding intelligence to the tradi-
The graphics windows provide the user tional approach that are less formal
with a quick view of the mine. In one then a true approximation. In either
window any plan view can be displayed case, the optimizations need to become
19th APCOM SYMPOSIUM

more ractical and the traditional ap- optimum cutoff grade, Quarterly of
proacges need to become more optimal. the Colorado School of Mines, Octo-
This paper is one step toward accom- ber.
plishing both of these goals.
ll.Marek, J.M. and H.E. Welhener, 1985.
Cutoff Grade Strategy - A Balancing
References
Act, SME Fall Meeting, Albuguerque,
1. Baafi, E., 1983. Application of October 17-19.
Mathematical Programming Models to
Coal Quality Control, Ph.D. Dis- 12.Matheson, G., 1982. Open Pit Sequen-
sertation, University of Arizona, cing and Scheduling, SME Fall Meet-
Tucson, 106 pp. ing, Honolulu, September 4-9.
2. Garg, 0.. 1983. Personal Communica-
tion, Hanng Mining Company, Cleve- 13.Rose, W.L., 1985. An Approach to
land, OH. Computerized Open-Pit Production
Scheduling, SME Fall Meeting, Albu-
3. Gershon, M.E., 1982. "A Linear guerque, October 17-19.
Programming Approach to Mine Sched-
uling Optimization," Proceedings, 14.Smith, C.E. (1978) The Use of Mixed
17th Application of Computers and Integer Programming in Planning the
Operations Research in the Mineral Depletion of an Alluvial Diamond
Industry Symposium, Denver, Colorado, Deposit, presented to the Operations
April 19-22, pp. 483-493. Research Society of South Africa,
September 28-29, 94-102.
4. Gershon, M..E. 1983a. Mine schedul-
ing optimization with mixed integer
programming, Mining Engineering, 35,
351-4.

5. Gershon, M.E. 1983b. Optimal Mine


Production Scheduling: Evaluation
of Large Scale Mathematical Pro-
gramming Approaches, International
Journal of Mining Engineering,
Vol. 1, pp. 315-329, December.

6. Gershon, M.E., 1985. Developments


in Computerized Mine Production
Scheduling, SME Fall Meeting, Albu-
guerque, October 17-19.

7. Gershon, M.E., 1987. Mining Opera-


tions Research, Chapman-Hall, Lon-
don, to appear.

8. Hillier, F.S. and Lieberman, G.J.


(1980) Introduction to Operations
Research, Holden-Day, Inc., San
Francisco, 829pp.

Johnson, T.B. (1969) Optimum Pro-


duction Scheduling, Proceedings,
8th International Symposium on
Computers and Operations Research,
Salt Lake City, Utah, 539-62.

Lane, K.F. (1964) Choosing the

You might also like