**DOI :**

**10.17577/IJERTV11IS020118**

**Open Access****Authors :**P. Musenge , E. K. Chanda, Bunda. B , J. K Kaunde K, F. Bokwala B , B. Fyama. M, S. Kanke. M, Sebastian. A**Paper ID :**IJERTV11IS020118**Volume & Issue :**Volume 11, Issue 02 (February 2022)**Published (First Online):**24-02-2022**ISSN (Online) :**2278-0181**Publisher Name :**IJERT**License:**This work is licensed under a Creative Commons Attribution 4.0 International License

#### Ultimate Pit Limits using Maximum Flow Algorithm Ford and Fulkerson: The Case of Study of North Mutoshi Project, DRC

P. Musenge a,c, E. K. Chanda a, Bunda. B a, J. K Kaunde K b, F. Bokwala B c, B. Fyama. Mc ,

S. Kanke. M c, Sebastian. A d.

a Departement of mining Engineering, school of mines , University of Zambia P.O. Box 32379 Lusaka, Zambia

b Departement of mines and geologie, Mapon University 01 kosakosa,Kindu, D.R Congo

c Departement of mines, Polytechnic, University of Lubumbash P.B.1825 Lubumbashi, D.R C

d Department of Mining, Predictive Geometallurgy and Geostatistics Lab Queens University in Kingston, Ontario, K7L 3N6 Canada

Abstract:- The determination of an ultimate pit limit of an open- pit mine is an important part of the planning of any mine. This is a phase which is related to the pit optimisation. New tools to solve mathematical programming problems have emerged in recent times prompting researchers to revisit the determination of ultimate open pit limits. Application of these tools such as Python could result in considerable savings in the cost of technical computing for mining companies in this study, a proposed algorithm for Ford and Fulkerson model was developed and coded in Python programming environment to solve the ultimate pit problem and the results are compared to the Lerchs-Grossman algorithm. The research demonstrated that the Ford and Fulkerson model run in Python, can be effectively used to find an ultimate pit limit in 3-dimensions through 2-dimension cross-sections. The ultimate pit limits obtained using the Ford and Fulkerson algorithm programmed in Python found optimal pit design in record time. To validate the proposed Ford and Fulkerson model, a numerical result comparison was done against the best-known LG. The FFA determines the ultimate pit limit by finding the maximum net value from blocks extracted which represent the minimum cuts. This research provides a framework for running the Ford and Fulkerson Algorithm to the mining domain using Python along with a brief description of the method and its application to a real copper project to exemplify its use. The optimum pit values obtained from the FFA and the LG are USD 881 870 000 and USD 880 210 100, respectively. These results gave the adaptation model with the regression of 99,8 % and the net maximum value of USD 891 949 000 compared to the FFA model using python Although the proposed Ford and Fulkerson algorithm demonstrated its efficiency and applicability to deal with the ultimate pit limits.

Keywords: Ultimate pits, optimisation, maximum flow, Graph, Ford and Fulkerson

INTRODUCTION

The mining activities have high economic risks and the determination of the optimum ultimate pit limit of an ore body greatly affects the economic feasibility of an open-pit mine, many researchers proposed different algorithms and heuristic methods that maximize the economic value of a mine while satisfying operational and extraction sequence constraints. In open-pit mining, the long-term planning problem for the exploitation of the reserve is often divided into two major tasks; the first is the determination of the ultimate pit limit and the production scheduling (Epstein et

al., 2012). Optimization techniques applied to the production scheduling problems are based on the use of operations research (OR), which have been introduced in the mining industry since the work of Lerchs and Grossmann in the early 1960s (Lerchs and Grossmann, 1965). Johnson (1968) developed exact algorithms to optimize the long-term production plan in open-pit mining. The optimal solution provided by these exact algorithms may maximize the 2 objective functions but fail to define the mining extraction sequence of a large-scale mine whereby blocks must be mined according to an established time horizon. The mine production schedule optimization must comply with a set of physical, technical, and economic constraints (Khan and Niemann-Delius, 2015). This study focusses on the determination of ultimate contour limits of the pit using the maximum flow algorithm developed by Ford Fulkerson. In the last decades, there have been two major approaches to determining the final pit limit. The first approach is undiscounted profit maximization and the other is maximizing the NPV of the ultimate pit limit. For each approach, some methods and algorithms are presented. In the first approach, initially, the final pit outline to maximize undiscounted profit is determined. Then achievement of the highest net present value (NPV) is planned for the pit production scheduling. Heuristic algorithms such as the floating cone (Pana, 1965) and its improved methods (Wright, 1999) and Korobov (David, Dowd, & Korobov, 1974) were presented for this purpose. The Lerchs- Grossman (1965) algorithm (LG) based on graph theory and the network flow algorithm (Johnson & Barnes, 1988; Yegulalp & Arias, 1992) also determine the final pit through a mathematical approach. Of these, the LG method is the most widely used when designing the final pit limit. Currently, a variety of software programs are being used to solve the problem, each of which has challenges in terms of acquisition, learning and application, all of which require a high degree of mining knowledge, as well as vast experience and high-level skills in computer applications. As a result, existing mathematical methods for handling the mining problem that requires less dependence on software have been introduced and are being enhanced. (Meisam Saleki & Kakaie, 2019).

BACKGROUND AND FORD AND FULKERSON ALGORITHM

ULTIMATE CONTOURS

It is not so long ago that operations research has been used to solve the problems of planning operations in surface mines. One of the first problems which one tried to solve in this field was the determination of the ultimate contours of a pit operation. The ultimate contours of an open-pit mine represent the geometric contours of the mine after its operation or they represent the appearance of the mine at the end of its economic life. Optimal contours are those that maximize total profits regardless of time. This is a long-term planning problem and allows, among other things, to estimate the quantity of ore that it will be possible to extract from the deposit, to estimate the economic duration of the exploitation, to plan the size of the surface installations and the machinery, and to plan the capacity of production. Knowing the final contours also makes it possible to plan and organize short, medium, and long-term operating sequences that will maximize profits.

Several people have tried to design an algorithm that would optimize the final contours of an open-pit mine. The heuristic methods were the first methods developed to determine the final contours of the pit operations. From heuristics developed for the problem of ultimate contours, those which use the moving cone method or, floating cones, are the best known and have been the most widely used in

Figure 1: Example of deposit in 2Dimension (E. Chanda, 2021)

Each of the blocks in figure 1 becomes a node of the graph. Each node of the graph is associated with a value represented by the cost or the profit generated by the extraction of the corresponding block. We can associate with this deposit the graph of precedence then must connect all the nodes to an artificial node named S. Thus, a tree is formed where the artificial node s is the root of the tree. Figure 2 shows the initial tree thus created.

The second part of the algorithm consists in finding the maximum closure on this graph. Even though Lerchs and Grossmann’s algorithm has been proven for several decdes now and is the benchmark algorithm from which all newly developed algorithms are ranked and judged, it was hardly ever used in the industry before the mid-1980s. The main reasons for this phenomenon were practicality, over years, however, it has been remodeled and modified frequently to make it more efficient, Stuart (L992), Zhao and Kim (1992). They have all, at some level, succeeded in overcoming certain limitations mentioned above such as incorporating variable slopes during calculations and reducing resolution times; with the advent of commercial software Whittle, the

industry because of their simplicity both in terms of understanding and implementation. It was Pana, in 1965, who first proposed this method and it remained for a long time the effective tool of design for open pit mines. Up to everything recently, several authors have been interested in this method. The second type of heuristics is based on the concept of dynamic programming. These methods originated from the 2-D algorithm of Lerchs and Grossmann and were then developed to deal with the three-dimensional problem.

In 1965 Lerchs and Grossmann were the first to present an optimal method for the problem of ultimate pit contours. Based initially on graph theory, Lerchs and Grossmann’s algorithm was taken up first by Johnson (1968) then by Picard (1976) who demonstrated that we could model the problem as that of a maximum flow problem in a graph.

Lerchs and Grossmann 3-D graph theory algorithm

Lerchs and Grossmann developed the first optimal three- dimensional method for determining the ultimate contours of an open-pit mine, this is the second part of their 1965 article. The Lerchs and Grossmann 3D algorithm is separated into two parts.

The first part is used to build the graph and the second part is used to calculate the maximum weight closure on the graph. During the first part, you build the initial graph by drawing oriented arcs between each of the nodes and the 9 nodes which are above it. Suppose, the two-dimensional deposit presented in Figure 1:

Figure 2: Graph and The initial tree graph representing the constraints of precedence between blocks (E. Chanda, 2018)

Lerchs and Grossmann method has become increasingly popular in the industry, from the mid-1980s until now, many years after the release of the original article in 1965. For a long time, the different methods developed over the years to determine the ultimate contours were not favored in the industry because of the implementation problems that they presented, such as their inability to take into account variable slopes for different parts of the pit.

MAXIMUM FLOW ALGORITHM

In 1965, when Lerchs and Grossmann proposed their algorithm, they specified that the problem could be dealt with in several different ways, either by using the dynamic programming, or using graph theory (maximum closure of a directed graph), either as a flow problem or as an analogy hydrostatic (project management problem). In 1968, as part of his doctoral thesis, Johnson was the first to recognize the relation between the problem of the ultimate contours and the maximum flow. In 1976, Picard demonstrated mathematically that the method presented by Lerchs and Grossmann, which consisted of finding the maximal closure

on a graph, was equivalent to finding a maximum flow on a graph adapted to the problem of contours ultimate. The new graph is obtained by adding a source node s and a node sink

t. The source node is connected to all the nodes i which have a positive value by an arc oriented (s, i) of capacity equal to the value of the node. The sink node is connected to all nodes

j which have a negative or zero value by an oriented arc (0,

i) of capacity equal to the absolute value of the node. In this new graph, the arcs of precedence have an infinite capacity. Figure 3 presents the Picard graph constructed from the deposit presented in Figure 2 in the section on Lerchs and Grossmann 3D.

Figure 3: Graph for the calculation of ultimate contours with maximum flow formulation for Ford and Fulkerson Algorithm Chanda, 2019)

THE FORMULATION OF THE MAXFLOW ALGORITHM AND METHODOLOGY

firstly we need to have the economic block value, The equations to compute the value of a block are given as follows (Kubra, 2019): = ( ) (1)

) for the ore blocks (2)

for waste bocks

where R is revenue, P is price, S is selling cost, g is grade, t is the tonnage of the block, EBV is economic block value, MC is mining cost, and PC is processing cost.

CONCEPTS OF THE MAXIMUM FLOW ALGORITHM FORD AND FULKERSON

The ij blocks are the blocks that make up the economic block value for the Mutoshi deposit, and five primary sections can be explained as follows: the first step concerns the mineral resources modeling and calculation; the next step concerns block aggregation then the software surpac was used for providing the economic block model then translate the problem to optimized with max flow algorithm Ford and Fulkerson using python programming. To carry out the research of this work, two types of methodology, theory and practice. The theory part includes some very advanced theories and The practical part requires knowledge of the several tools and mining software

Figure 4: Flowchart of the proposed maxflow algorithm for solving the ultimate pit limit problem

FORD AND FULKERSON ALGORITHM IN THE CONTEXT OF THE UPLO

According to (Goldberg & Tarjan, 2014)The FF algorithm was developed to improve the augmenting paths work and proved to give better results experimentally when compared to other algorithms methods. With the augmenting paths, a new breath-first search is normally started from the source

- to sink (t) paths once all the pre-examined paths are exhausted. This search in UPLO can be achieved by scanning the blocks (vertices) within the block model. Since it can be very costly to repeat the whole search process every time, the FF algorithm builds two search trees, where one starts from the source and the other from the sink. The algorithm also reuses these trees in their search instead of starting afresh, thus saving on time. The algorithm works by maintaining two search trees T and S to give a min-cut, which are rooted at the sink node (t) and the source node (s), The algorithm iterates three main steps:
Growth step: this is the step where the tree grows by linking the active nodes in the tree to the free nodes whose edges are not saturated until an augmenting path is found. In UPLO, the blocks will be linked depending on the set of blocks that are required to be mined to pave the way for mining a certain ore block. This is achieved by linking the active nodes of the two sets of trees S and T.

Augmentation step: this step enhances the augmenting paths found in the growth step by trying to push maximum flow through the edges such that some edges become saturated.

Adoption step: with the creation of a forest of trees from the augmenting step, the adoption step restores the original setup of S and T trees. This is achieved by looking for new

parents with non-saturated edges for the visit from the same tree they have come from. If the visited do not get new valid parents, they become free nodes. This step ends when all the visited cease to exist and the original S and T trees are left. Once the adoption step ends, the algorithm starts again at the growth step until the time when all the active nodes phase- out, thus achieving a maximum flow. Then the min-cut will be created with all saturated nodes and that will be the maximum closure.

APPLICATION OF MAX FLOW ALGORITHM FOR THE CASE STUDY

Formulation of the network max-flow model for Mutoshi deposit

Let = (, ) be a direct graph network with = | | the number of nodes and with = | | the number of arcs. Let with (, ) A be a direct arc from to , then and

represents the arc capacity (non-negative real number) and arc flow, respetively. By setting a lower bound capacity to zero, a Pseudoflow f assigns to each arc (, ) a flow so that 0 . An , -graph = ( , ), corresponds to an extension of with two additional nodes: a source and a sink , = , {, }. The set of arcs now includes source-to-node arcs A(s) and node-to-sink arcs A(t), = A A(s) A(t). A closure graph is an s,t-graph whose arcs with finite capacities are only the arcs adjacent

to the sources and sink nodes, residual graph, inflow(u), outflow(u), and excess, are required to provide the step-by- step operation of the generic Pseudoflow algorithm that maximizes the flow. (Sebastian Avalos, 2020), so the set of nodes involved in the maximum flow solution corresponds to the blocks in the resulting ultimate pit limit. (Bai, 2017). Figure 5 illustrates all details about the formulation network of the case study deposit

Figure 5.: Graph representing the Mutoshi block model and block dependencies (cross section 3)

Figure 6: Graph showing the augmenting paths from source (s) to sink (t).

Assuming a maximum slope angle of 45Â°, the active nodes link the free neighboring nodes, which in this case are the waste blocks in blue colored that have to be mined for the ore block to be mined. The growth continues until the two trees join, as shown in Figure 6. Since the Ford Fulkerson maximum flow algorithm is an augmenting path method, it makes sure that the flow-conservation constraints and the capacity constraints are adhered to, such that: (Akisa David Mwangi, 2020), Chanda, 2018)

Adjacency Matrix for Mutoshi deposit in 2 Dimensions

The adjacency matrix of a graph = (, ) is an n Ã— n matrix = ( ), where = { 1 , 2 , , , } is the vertex set, is the edge set of G and is the number of edges between the vertices and . In the adjacency matrix of a directed graph, equals the number of arcs from the vertex to in the current case n = 14 and 13

Table 1: Description of the Mutoshi Adjacency matrix for one section

Mutoshi Adjacency matrix on one cross-section N 182 blocks from with source node 0 and sink node 183 = { 1 , 2 , , , } 184 vertices represent the blocks 662 arcs, 480 arcs (internals) and 182 arcs (externals) 662 edges (rows)= 13 vertices (columns)= 14 vertices

() 14(columns) = 168 arcs () 13(rows) = 312 arcs () 1 = 480 arcs (externals)

Development of the model

We begin by importing four python packages (or libraries):

NumPy; NetworkX; pseudoflow2 and time. the full code is

given in the figure below

# Python program for finding min-cut which represents the ultimate pit limit for the open pit in 2D in the given graph which is a cross-section from 3D

from collections import defaultdict class Graph:

def init (self, graph):

self.graph = graph # residual graph self.org_graph = [i[:] for i in graph] self.ROW = len(graph)

self.COL = len(graph[0])

def BFS(self, s, t, parent): visited = [False] * (self.ROW) queue = []

queue.append(s) visited[s] = True

while queue:

u = queue.pop(0)

# Python program for finding min-cut which represents the ultimate pit limit for the open pit in 2D in the given graph which is a cross-section from 3D

from collections import defaultdict class Graph:

def init (self, graph):

self.graph = graph # residual graph self.org_graph = [i[:] for i in graph] self.ROW = len(graph)

self.COL = len(graph[0])

def BFS(self, s, t, parent): visited = [False] * (self.ROW) queue = []

queue.append(s) visited[s] = True

while queue:

u = queue.pop(0)

for ind, val in enumerate(self.graph[u]): if visited[ind] == False and val > 0:

queue.append(ind) visited[ind] = True parent[ind] = u

return True if visited[t] else False

def minCut(self, source, sink): parent = [-1] * (self.ROW) max_flow = 0

while self.BFS(source, sink, parent): path_flow = float(“Inf”)

s = sink

while (s != source):

path_flow = min(path_flow, self.graph[parent[s]][s]) s = parent[s]

max_flow += path_flow v = sink

while (v != source): u = parent[v]

self.graph[u][v] -= path_flow self.graph[v][u] += path_flow v = parent[v]

for i in range(self.ROW): for j in range(self.COL):

if self.graph[i][j] == 0 and self.org_graph[i][j] > 0: print(str(i) + ” – ” + str(j))

graph = [[adjacency matrix]] g = Graph(graph)

source = 0; sink = N g.minCut(source, sink)

# This code is contributed by pathie Musenge for the mining domain

for ind, val in enumerate(self.graph[u]): if visited[ind] == False and val > 0:

queue.append(ind) visited[ind] = True parent[ind] = u

return True if visited[t] else False

def minCut(self, source, sink): parent = [-1] * (self.ROW) max_flow = 0

while self.BFS(source, sink, parent): path_flow = float(“Inf”)

s = sink

while (s != source):

path_flow = min(path_flow, self.graph[parent[s]][s]) s = parent[s]

max_flow += path_flow v = sink

while (v != source): u = parent[v]

self.graph[u][v] -= path_flow self.graph[v][u] += path_flow v = parent[v]

for i in range(self.ROW): for j in range(self.COL):

if self.graph[i][j] == 0 and self.org_graph[i][j] > 0: print(str(i) + ” – ” + str(j))

graph = [[adjacency matrix]] g = Graph(graph)

source = 0; sink = N g.minCut(source, sink)

# This code is contributed by pathie Musenge for the mining domain

Figure 7: Proposed maximum flow algorithm of the Ford and Fulkerson algorithm

Following are steps to print all edges of the minimum cut: Run the Ford-Fulkerson algorithm and consider the final residual graph; Find the set of vertices that are reachable from the source in the residual graph; All edges which are from a reachable vertex to a non-reachable vertex are minimum cut edges. Print all such edges. Solution of the max-flow model for Mutoshi deposit using Python Program The maximum pit value can be realized by separating the trees S and T using the invalid parts created by the saturated

edges linking the two trees. The saturated edges from the source node to the ore blocks are normally broken to avoid the support of the ore blocks to the overlying waste blocks, the other saturated edges from the waste blocks to the sink node are also broken to avoid the support of these waste blocks from the underlying ore blocks. This also makes sure that there are no outgoing arcs from the maximum closure. In this case, the ultimate pit contains the blocks in table 2 with a maximum close value of USD 3203 and his ultimate pit value as shown in Figure 8

0 85 72 183 117 183 135 183 151 183 167 183 0 86 75 183 118 183 136 183 152 183 169 183 0 87 89 183 119 183 137 183 155 183 170 183 0 88 92 183 120 183 138 183 156 183 171 183 0 90 99 183 121 183 141 183 157 183 172 183 0 86 100 183 122 183 142 183 158 183 173 183 0 87 101 183 127 183 143 183 159 13 174 183 0 88 102 183 128 183 144 183 160 183 175 183 0 90 103 183 129 183 145 183 161 183 176 183 0 104 106 183 130 183 146 183 162 183 177 183 0 105 113 183 131 183 147 183 163 183 178 183 0 107 114 183 132 183 148 183 164 183 179 183 0 123 115 183 133 183 149 183 165 183 180 183 61 183 116 183 134 183 150 183 166 183 181 183 182 183 Figure 5. 1: Ultimate pit limit using Maxflow algorithm Ford and Fulkerson on section 3

Ultimate pit limits for each 2-dimensional section (13 cross-sections)

pit value

pit value

To demonstrate the computation speed of the maximum flow algorithm, a series testing of 13 cross-sections through block models was used. The 45-degree slope angle is adopted for the block model. The number of arcs created for this slope setting and the economic value is listed in Table 2 Note that, the actual number of blocks and arcs used in the optimization is called active blocks and active arcs, in f, the active blocks are colored in red, blue, orange and yellow. (Active blocks represent the blocks that contain the ultimate pit limits in each section, and all the precedent blocks linked to the active blocks need to be removed to access the ore.

Maximum pit value

90000

80000

70000

60000

50000

40000

30000

20000

10000

0

cross sections

Maximum pit value

90000

80000

70000

60000

50000

40000

30000

20000

10000

0

cross sections

sec3

sec4 sec5 sec6 sec7 sec8 sec9 sec10 sec11 sec12 sec13 sec14 sec15

tot amount

sec3

sec4 sec5 sec6 sec7 sec8 sec9 sec10 sec11 sec12 sec13 sec14 sec15

tot amount

Figure 9: Maximum pit value for all 13 cross sections

FINAL ULTIMATE PIT LIMITS FOR THE DEPOSIT

The limit of the final pit is made up of 13 cross-sections which were mineralized, and of these sections, we have a set of blocks which represents the min-cut found during programming with the python application with their respective limits and the figure 10 is the Mutoshi ultimate pit limits, the pit representation of all the sections in 3 faces is and the construction of the pit shell was carried out in windows excel, In figure 10:

COMPARISON OF MAX-FLOW RESULTS WITH LERCHS AND GROSSMANN

Study case Ultimate pit limits using the Lerchs and Grossmann algorithm concept

Figure 8: Ultimate pit limit using Maxflow algorithm Ford and Fulkerson on section 3

- the blocks in blue, yellow, and orange blocks represent the active blocks which are joined by the active
- the blocks in black represent the set of blocks that constitute the blocks that must not be mined and will remain in our mines
- the blocks in blue represent the set of all the wastes blocks that continue within the limit to be exploited and those in yellow color are the ore blocks also contained within the limit of the blocks to be mined
- all the blocks in red represent the ultimate limit and its blocks will also be extracted

Figure 10: a representation of Mutoshi ultimate pit limit using maximum flow algorithm

=

The first step is to

,

The essence of the LG algorithm is to split the economic block model into parallel vertical sections or 2D space, then determine the contour of the pit on each section that yields the maximum profit. The general configuration of the pit contour on a cross-section (2D) consists of three sides: two walls inclined at a certain slope angle and the bottom level of the pit. Analytically, the LG seeks to maximize the objective function Z shown in Equation 6.1 to design an

compute the cumulative profits realized after the extraction of a single column of blocks having at its base the block . The calculation of within a column is independent of other columns as shown in Equation 3 where the index identifies all blocks involved.

=

optimum pit: (musema, 2020)

=,

+ max{ +1,1}

Table 1: Introduction of a dummy row on a cross-section block model of ij values

Cross-section of Mij | |||||||||||||||

0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 |

0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |

0 | -22 | -29 | -36 | -36 | -36 | -31 | -28 | -28 | -30 | -33 | -39 | -37 | -39 | -39 | 0 |

0 | -44 | -58 | -72 | -72 | -72 | -62 | -56 | -56 | -61 | -66 | -79 | -74 | -78 | -78 | 0 |

0 | -66 | -87 | -108 | -108 | -108 | -94 | -85 | -85 | -93 | -101 | -121 | -111 | -117 | -117 | 0 |

0 | -88 | -117 | -144 | -144 | -144 | -126 | -114 | -117 | -131 | -139 | -163 | -149 | -156 | -157 | 0 |

0 | -110 | -147 | -180 | -180 | -180 | -159 | -146 | -154 | -170 | -175 | -97 | -189 | -196 | -198 | 0 |

0 | -132 | -177 | -217 | -217 | -217 | 174 | -109 | -193 | -9 | -211 | -102 | -231 | -237 | -240 | 0 |

0 | -127 | 131 | -178 | 18 | -252 | 762 | 1052 | -229 | -19 | -248 | -143 | -274 | -279 | -282 | 0 |

0 | -149 | 100 | 6 | 530 | -292 | 1054 | 1877 | -63 | -54 | -284 | -183 | -318 | -323 | -314 | 0 |

0 | -171 | 69 | 183 | 1024 | -332 | 1014 | 1839 | -297 | -88 | -320 | -223 | -362 | -366 | -336 | 0 |

0 | -193 | 39 | 143 | 1224 | -373 | 974 | 1801 | -332 | -122 | -356 | -263 | -406 | -399 | -358 | 0 |

0 | -215 | 10 | 102 | 1183 | -414 | 933 | 1763 | -367 | -156 | -392 | -304 | -448 | -422 | -380 | 0 |

0 | -237 | -16 | 61 | 1142 | -455 | 892 | 1725 | -402 | -190 | -427 | -333 | -471 | -444 | -402 | 0 |

0 | -259 | -38 | 21 | 1101 | -497 | 851 | 1688 | -436 | -221 | -453 | -355 | -493 | -466 | -424 | 0 |

The process continues until the last column generating overall cumulative values on each block. The optimum

pit contour on the section is that which yields the maximum cumulative value and in the current case, the net value for the cross-section is USD 3175.

Figure 11: ultimate pit limit using LG for Mutoshi block model cross-section 3

COMPARISON OF OPTIMAL PIT VALUE

The ultimate pit limit for this deposit was calculated using the FFA and LG. The FFA maximum flow algorithm gave a maximum pit value of USD 881 780 000 in 25s with 1261 as some active blocks. The LG gave the maximum pit value of USD 880 211 but still with a long time of 45s with 1443 number of active blocks, table 6.3 gave the percentage square around 99,9 This, therefore, shows that the FFA

maximum flow can be applied in the mining industry ultimate pit optimization since the results shown the on others sections the same values with LG model, which is already being applied. Table 5.13 gives a summary of the comparison of maximum flow algorithm Results for Ford and Fulkerson and Lechrs and Grossmann. The two final ultimate pits obtained from FFA and LG for Mutoshi deposit for this deposit are shown in Fgure 12.

Table 5: Introduction of a dummy row on a cross-section block model of ij values

Comparison | ||

Sections | Maxflow FF maximum pit value | LG maximum pit value |

sec3 | 3203 | 3175 |

sec4 | 11288 | 11301 |

sec5 | 12218 | 12242.4 |

sec6 | 13828 | 13852 |

sec7 | 10938 | 10920.5 |

sec8 | 12681 | 12177.2 |

sec9 | 5873 | 5931.4 |

sec10 | 739 | 814.08 |

sec11 | 2607 | 2606.9 |

sec12 | 6361 | 6334.01 |

sec13 | 6270 | 6256.367 |

sec14 | 1696 | 1933.4 |

sec15 | 476 | 476.8 |

Total value | 88178 | 88021.057 |

Active blocks | 1261 | 1443 |

Total time /seconds | 25 | 45 |

Difference between the total amount | 156.947 |

Figure 11: The finals ultimate pits obtained from FFA and LG for Mutoshi

comparaison maximum pit value FFA and LG

15000

10000

5000

0

comparaison Maxflow econimic value

comparaison maximum pit value FFA and LG

15000

10000

5000

0

comparaison Maxflow econimic value

Table 6: Table regression statistics

SUMMARY OUTPUT | |

Regression Statistics | |

Multiple R | 0.999537 |

R Square | 0.999075 |

Adjusted R Square | 0.99899 |

Standard Error | 154.9037 |

Observations | 13 |

sec3

sec4 sec5 sec6 sec7 sec8 sec9 sec10 sec11 sec12 sec13 sec14 sec15

tot

difference

sec3

sec4 sec5 sec6 sec7 sec8 sec9 sec10 sec11 sec12 sec13 sec14 sec15

tot

difference

Figure13: comparison maximum pit value for FFA and LG

COMPARISON OF MODEL PERFORMANCE

Y

Y

The coefficient of determination (R-squared) was used. R- squared shows how well the regression model fits the observed data, the correlation coefficient interprets the R

15000

X Variable 1 Line Fit Plot

y = 1.0143x – 84.824

15000

X Variable 1 Line Fit Plot

y = 1.0143x – 84.824

10000

10000

Y

Y

5000

5000

0

0

0

X Variable 1

15000

0

X Variable 1

15000

5000

5000

10000

10000

Figure 14: correlation between FFA model and LG model

values as 0 < R < 30% implies weak correlation, 30% < R < 70% implies moderate correlation and R > 70% implies strong correlation. (Amankwah, 2011) figure 13 compare results between FFA and LG using regression analysis

Predictecd and Maxflow found

1 2 3 4 5 6 7 8 9 10 11 12 13 14

model found maxflow

Predictecd and Maxflow found

1 2 3 4 5 6 7 8 9 10 11 12 13 14

model found maxflow

Figure 15: predicted model and FFA model

Application of the model to Maxflow algorithm Ford and Fulkerson, the model found y = 1.0143x – 84.824

Table 7: results FFA and regression results

<tdSec12

data | |||

sections | Maxflow Ford Fulkerson Algorithm | Model found | |

1 | Sec3 | 3203 | 3135.5785 |

2 | Sec4 | 11288 | 11377.7803 |

3 | Sec5 | 12218 | 12332.64232 |

4 | Sec6 | 13828 | 13965.2596 |

5 | Sec7 | 10938 | 10991.83915 |

6 | Sec8 | 12681 | 12266.50996 |

7 | Sec9 | 5873 | 5931.39502 |

8 | Sec10 | 739 | 740.897344 |

9 | Sec11 | 2607 | 2559.35467 |

10 | 6361 | 6339.762343 | |

11 | Sec13 | 6270 | 6261.009048 |

12 | Sec14 | 1696 | 1876.22362 |

13 | Sec15 | 476 | 398.79424 |

Total value | 88178 | 89194.93412 |

CONCLUSIONS AND RECOMMENDATIONS

Ultimate pit limit optimization (UPLO), playing a major role in the mining industry and algorithms for solving UPLO, have been developed and improved by various researchers since the 1960s. This research project aimed at finding the ultimate pit limit for Mutoshi deposit, a Ford and Fulkerson algorithm model was developed and coded in Python programming language to be run on a block model through 13 cross-sections in two- dimensions. The application of the proposed Ford and Fulkerson algorithm in this research determined the best results and applicability in the mining domain compared to results obtained using Lerchs and Grossmann with the same data set. However,

The implementation of the Ford and Fulkerson algorithm to solve the ultimate pit limits problem in an open-pit mine has shown encouraging results and its applicability compared to Lerchs and Grossmann when applied to the case study. It was demonstrated that the Ford and Fulkerson Algorithm fits well in solving the ultimate pit limits problem as those two methods consider Lerchs and

Grossman as the base model already found and proved through several authors as the first mathematical method to solve the ultimate pit limits problem. From two methods we found a model for validation and compare the results from this one the applicability of the Ford and Fulkerson algorithm has been proved to be used in the mining domain.

To test and validate the proposed FFA model, this study used also LG which is a known method for solving the ultimate pit limits problem. The deposit has 3,304 blocks, with 60 in X, 60 in Y, and 20 in the Z direction as user blocks, the FFA used 2366 as blocks and LG used 2366 Blocks. When applied the FF model, the FFA provided an ultimate pit limit with 4424 arcs and 1261 active blocks, and the results in excel provide 1443 active blocks for LG.

ACKNOWLEDGEMENTS

This paper presents part of work conducted by the first author in partial fulfilment of the requirements for an master research suty degree at the school of mines, University of Zambia, Lusaka, Republic of Zambia

REFERENCES

- Abzalov, 2006. Localised uniform conditioning (LUC): A new approach for direct modeling of small blocks.Mathematical Geology. Volume II, pp. 27-45.
- Akisa David Mwangi, Z. J., 2020. Ultimate Pit Limit Optimization using Boykov-Kolmogorov Maximum. Journal of Mining and Environment (JME), pp. 4-7.
- Amankwah, H., 2011. Mathematical Optimization Models and Methods for Open-Pit Mining. In: Linkoping: LiU-Tryck, Linkoping, pp. 23-30.
- AUGER, G., 2000. STUDY AND PROGRAMMING OF THE MAXIMUM FLOW ALGORITHM. In: Montreal: POLYTECHNIC SCHOOL OF MONTREAL, pp. 23-45.
- Ayisi, M. O., 2015. 3D Block Modeling and Reserve Estimation of a Garnet Deposit. pp. 13-22.
- Bai, Y., 2017. pseudoflow method for pit optimization. technical report.
- Chanda, E., 2018. advances in applied strategic mine planning. In:
r. Dimitrakopoulos, ed. Montreal: s.n.

- Goldberg, R. E. & Tarjan, 2014. Efficient Maximum Flow Algorithms. Magazine Archive, Volume vol. 57.
- HEUREUX, 2011. MODEL OF OPTIMIZATION FOR MEDIUM PLANNING. pp. 23-30.
- HOCHBAUM, C., 1998. Performance analysis and best implementations of old and new algorithms for the open-pit mining problem. to appear in naval research quartly, p. 40.
- Hustrulid & Kuchta, M., 2016. Open pit mine planning and design.. Volume Volume 1-fundamentals.
- KABEZYA, 2017. Apllication of goetechnical model in the short term planning of open pit case of the North Mutoshi project.. pp. 4-41.
- KIM, Y., 1978. Ultimate pit limit design methodologies using conputer models – The state of the art. Mining Engineering. pp. 1454-1459.
- KÃœBRA, A. F., 2019. A NEW APPROACH TO 3- DIMENSIONAL OPTIMIZATION OF ULTIMATE PIT GEOMETRY FOR OPEN PIT MINES WITH VARIABLE OVERALL SLOPE ANGLES. pp. 8-25.
- Kubra, F., 2019. A NEW APPROACH TO 3-DIMENSIONAL OPTIMIZATION OF ULTIMATE. le caire: MIDDLE EAST TECHNICAL UNIVERSITY.
- kwiri J, G., 2017. Mine planning and optimisation techniques used in surface mining. pp. 200-220.
- Meisam Saleki, M. A. & Kakaie, R., 2019. Mathematical relationship between ultimate pit limits generated by discounted and undiscounted block value maximization in open pit mining. Journal of Sustainable Mining Elsevier, pp. 94-96.
- Muke, P. M., 2019. Open-pit mine production scheduling using the genetic algorithm. research gates, pp. 15-24.
- musema, m., 2020. o p e n p i t m i n e p r o d u c t i o n s c h e d u l i n g u s i n g t h e g e n e t i c a l g o r i t h m. pp. 20-26.
- Osanloo, M. G. J. K., 2008. Long-term open-pit mine production planning : a review of models and algorithms. International Journal of Mining, Reclamation and Environment, pp. 3-25.
- Sebastian Avalos, J. M. O., 2020. A guide for pit optimization with Pseudoflow in python 1. Annual Report, pp. 186-187.
- Shiv Prakash, H. A., 2015. Simulation and Optimization in Open Pit Mining. Applications of Cumputers and Operations Research in Mineral Industry, Alsaka, pp. 534-540.
- ww.developers.google.com/optimization/flow/maxflow, 2020.

Netwok flows. [Online]

Available at:

ww.developers.google.com/optimization/flow/maxflow