# Automatic Parameter Tuning of Multiple-Point Statistical Simulations for Lateritic Bauxite Deposits

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Overview of Underlying Methods

#### 2.1. Direct Sampling Algorithm

#### 2.2. Comparing Patterns with Smooth Histograms

#### 2.3. Generalised Simulated Annealing

- Choose a high initial temperature ${T}_{0}$ value, an initial solution ${x}^{0}$ and evaluate the objective function, ${E}^{0}=f({x}^{0})$.
- Propose a new solution ${x}^{i+1}$:
- Generate a candidate solution ${x}^{i+1}$ from the current one (${x}^{i}$) using a predefined visiting distribution.
- Evaluate the change in the objective function for the candidate solution, $\Delta =f({x}^{i+1})-f({x}^{i})$.
- Accept the iteration if it reduces the objective function, $\Delta <0$.
- Otherwise, accept or reject it based on a probability of acceptance criterion.

- Repeat step 2 for L number of iteration times keeping T constant.
- Reduce the temperature to ${T}_{n+1}$ using a cooling function.
- Repeat steps 2–4 until the convergence criteria is satisfied.

## 3. Problem Setup and Methodology

#### Methodology

## 4. Results and Discussion

#### 4.1. Implementation and Analysis of the Tuned Parameters on the Simulations

#### 4.2. Effect of the Optimisation Method and the Initial Parameters on the Final Results

#### 4.3. Effect of the Chosen Parameters on the Pattern Reproduction

#### 4.4. Analysis of the Automatically Tuned Parameters

## 5. Summary and Conclusions

## Author Contributions

## Acknowledgments

## Conflicts of Interest

## Abbreviations

SG | Simulation Grid |

MPS | Multiple-Point Statistics |

TI | Training Image |

DS | Direct Sampling |

SA | Simulated Annealing |

JS | Jensen-Shannon Divergence |

L-BFGS-B | Limited Memory Broyden-Fletcher-Goldfarb-Shanno Optimisation with Box Constraints |

BOBYQA | Bound Optimisation by Quadratic Approximation |

Glossary of Notations | |

${Z}^{*}(x)$ | Simulated value at location x |

${d}_{n}(x,L)$ | Data event constructed at location x of the simulation grid |

${d}_{n}(y,L)$ | Data event constructed at location y of the training image grid |

$Z(y)$ | Value of the attribute at y location of the training image |

t | Acceptance threshold to accept a pattern compatible |

f | Fraction of the training image to be scanned |

$d\{{d}_{n}(x,L),{d}_{n}(y,L)\}$ | Distance between the conditioning data and training image patterns |

${\alpha}_{i}$ | Weighting factor applied to the ith data event node |

$\delta $ | Power for computing the ${\alpha}_{i}$ weighting factor |

w | Special weight attached to the conditioning data events |

T | Search template used to construct the pseudo-histograms |

$pa{t}_{j}^{TI,un}$ | jth unique pattern collected from the training image |

${H}^{d,m}$ | Histogram class of the model image |

${H}^{d,TI}$ | Histogram class of the training image |

$pa{t}_{i}^{m}$ | Pattern i observed in the model image |

${N}^{TI,un}$ | Number of unique pattern categories in the training image |

${d}_{JS}(p,q)$ | Jensen-shannon divergence of p and q density distributions |

$f(x)$ | Objective function |

$\Delta $ | Change in the objective function due to a candidate solution ${x}^{i+1}$ |

$pa{t}_{i}^{C}$ | ith conditioning pattern collected from the simulation grid |

$\overline{pa{t}_{i}^{C}}$ | Mean of the values at the nodes of the pattern $pa{t}_{i}^{C}$ |

${N}^{C}$ | Number of patterns captured from the SG. |

$pa{t}_{i}^{SIM}$ | Pattern i collected from the simulation grid |

$\overline{pa{t}_{i}^{SIM}}$ | Mean value of the nodes in pattern $pa{t}_{i}^{SIM}$ |

${H}_{j}^{{}^{d,SIM}}$ | Value of the jth category of the realisation image pseudo-histogram |

${H}^{d,SIM}$ | Pseudo-histogram of a realisation |

${N}^{SIM}$ | Number of patterns captured in the realisation image generated |

${{H}_{j}}^{{{}^{d}}_{Nor}}$ | Normalised weight in the jth histogram category |

${H}_{j}^{{}^{d,{C}_{Nor}}}$ | Normalised weight of the jth category in the conditioning data pseudo-histogram |

${H}_{j}^{{}^{d,SI{M}_{{}_{Nor}}}}$ | Normalised weight of the jth category in the simulation image pseudo-histogram |

${w}_{hd}$ | Weighting factor attached to the hard data |

${w}_{GPR}$ | Weighting factor attached to the GPR data |

${n}_{hd}$ | Maximum number hard data nodes to be used in the MPS simulations |

${n}_{GPR}$ | Maximum number GPR nodes to be used in the MPS simulations |

${t}_{hd}$ | Acceptance threshold for the hard data |

${t}_{GPR}$ | Acceptance threshold for the GPR data |

$L(\alpha ,\psi )$ | Likelihood as a function of model parameters ($\alpha ,\psi $) |

${L}_{p}(\alpha )$ | Profile of the likelihood as a function of the parameter $\alpha $ |

## References

- Strebelle, S. Conditional simulation of complex geological structures using multiple-point statistics. Math. Geol.
**2002**, 34, 1–21. [Google Scholar] [CrossRef] - Oriani, F.; Straubhaar, J.; Renard, P.; Mariethoz, G. Simulation of rainfall time series from different climatic regions using the direct sampling technique. Hydrol. Earth Syst. Sci.
**2014**, 18, 3015. [Google Scholar] [CrossRef] - Pirot, G.; Straubhaar, J.; Renard, P. Simulation of braided river elevation model time series with multiple-point statistics. Geomorphology
**2014**, 214, 148–156. [Google Scholar] [CrossRef] - Guardiano, F.; Srivastava, R. Borrowing Complex Geometries from Training Images: The Extended Normal Equations Algorithm; Stanford Center for Reservoir Forecasting Report; Stanford University: Stanford, CA, USA, 1992. [Google Scholar]
- Zhang, T.; Switzer, P.; Journel, A. Filter-based classification of training image patterns for spatial simulation. Math. Geol.
**2006**, 38, 63–80. [Google Scholar] [CrossRef] - Arpat, G.B.; Caers, J. Conditional simulation with patterns. Math. Geol.
**2007**, 39, 177–203. [Google Scholar] [CrossRef] - Gloaguen, E.; Dimitrakopoulos, R. Two-dimensional conditional simulations based on the wavelet decomposition of training images. Math. Geosci.
**2009**, 41, 679–701. [Google Scholar] [CrossRef] - Dimitrakopoulos, R.; Mustapha, H.; Gloaguen, E. High-order statistics of spatial random fields: Exploring spatial cumulants for modeling complex non-Gaussian and non-linear phenomena. Math. Geosci.
**2010**, 42, 65–99. [Google Scholar] [CrossRef] - Honarkhah, M.; Caers, J. Stochastic simulation of patterns using distance-based pattern modeling. Math. Geosci.
**2010**, 42, 487–517. [Google Scholar] [CrossRef] - Mariethoz, G.; Renard, P.; Straubhaar, J. The direct sampling method to perform multiple-point geostatistical simulations. Water Resour. Res.
**2010**, 46. [Google Scholar] [CrossRef] - Straubhaar, J.; Renard, P.; Mariethoz, G.; Froidevaux, R.; Besson, O. An improved parallel multiple-point algorithm using a list approach. Math. Geosci.
**2011**, 43, 305–328. [Google Scholar] [CrossRef] - Tahmasebi, P.; Hezarkhani, A.; Sahimi, M. Multiple-point geostatistical modeling based on the cross-correlation functions. Comput. Geosci.
**2012**, 16, 779–797. [Google Scholar] [CrossRef] - Erten, O. Profiling and Mining Control to Mitigate Dilution Effect from SiO
_{2}at the Base of a Bauxite Deposit. Ph.D. Thesis, The University of Queensland, Brisbane, Australia, 2012. [Google Scholar] - Dagasan, Y.; Erten, O.; Topal, E. Accounting for a spatial trend in fine-scale ground-penetrating radar data: A comparative case study. J. S. Afr. Inst. Min. Metall.
**2018**, 118, 173–184. [Google Scholar] - Dagasan, Y.; Renard, P.; Straubhaar, J.; Erten, O.; Topal, E. Pilot Point Optimization of Mining Boundaries for Lateritic Metal Deposits: Finding the Trade-off Between Dilution and Ore Loss. Nat. Resour. Res.
**2018**, 1–19. [Google Scholar] [CrossRef] - Mariethoz, G.; Caers, J. Multiple-Point Geostatistics: Stochastic Modeling with Training Images; John Wiley & Sons: Hoboken, NJ, USA, 2014. [Google Scholar]
- Boucher, A. Algorithm-Driven and Representation-Driven Random Function: A New Formalism for Applied Geostatistics. 2007. Available online: https://pdfs.semanticscholar.org/0d38/a38d3b5c9ed694b788f9c3aab1fad381be7c.pdf (accessed on 21 May 2018).
- Boisvert, J.B.; Pyrcz, M.J.; Deutsch, C.V. Multiple point metrics to assess categorical variable models. Nat. Resour. Res.
**2010**, 19, 165–175. [Google Scholar] [CrossRef] - Liu, Y. Using the Snesim program for multiple-point statistical simulation. Comput. Geosci.
**2006**, 32, 1544–1563. [Google Scholar] [CrossRef] - Meerschman, E.; Pirot, G.; Mariethoz, G.; Straubhaar, J.; Van Meirvenne, M.; Renard, P. A practical guide to performing multiple-point statistical simulations with the Direct Sampling algorithm. Comput. Geosci.
**2013**, 52, 307–324. [Google Scholar] [CrossRef] - Rezaee, H.; Mariethoz, G.; Koneshloo, M.; Asghari, O. Multiple-point geostatistical simulation using the bunch-pasting direct sampling method. Comput. Geosci.
**2013**, 54, 293–308. [Google Scholar] [CrossRef] - Huysmans, M.; Dassargues, A. Direct multiple-point geostatistical simulation of edge properties for modeling thin irregularly shaped surfaces. Math. Geosci.
**2011**, 43, 521. [Google Scholar] [CrossRef] - De Iaco, S.; Maggio, S. Validation techniques for geological patterns simulations based on variogram and multiple-point statistics. Math. Geosci.
**2011**, 43, 483–500. [Google Scholar] [CrossRef] - Tan, X.; Tahmasebi, P.; Caers, J. Comparing training-image based algorithms using an analysis of distance. Math. Geosci.
**2014**, 46, 149–169. [Google Scholar] [CrossRef] - Renard, P.; Allard, D. Connectivity metrics for subsurface flow and transport. Adv. Water Resour.
**2013**, 51, 168–196. [Google Scholar] [CrossRef] - Pérez, C.; Mariethoz, G.; Ortiz, J.M. Verifying the high-order consistency of training images with data for multiple-point geostatistics. Comput. Geosci.
**2014**, 70, 190–205. [Google Scholar] [CrossRef] - Melnikova, Y.; Zunino, A.; Lange, K.; Cordua, K.S.; Mosegaard, K. History matching through a smooth formulation of multiple-point statistics. Math. Geosci.
**2015**, 47, 397–416. [Google Scholar] [CrossRef][Green Version] - Cover, T.M.; Thomas, J.A. Elements of Information Theory; John Wiley & Sons: Hoboken, NJ, USA, 2012. [Google Scholar]
- Kirkpatrick, S.; Gelatt, C.D.; Vecchi, M.P. Optimization by simulated annealing. Science
**1983**, 220, 671–680. [Google Scholar] [CrossRef] [PubMed] - Straubhaar, J. DeeSse User’s Guide; The Centre for Hydrogeology and Geothermics (CHYN), University of Neuchâtel: Neuchâtel, Switzerland, 2016. [Google Scholar]
- Straubhaar, J.; Renard, P.; Mariethoz, G. Conditioning multiple-point statistics simulations to block data. Spat. Stat.
**2016**, 16, 53–71. [Google Scholar] [CrossRef] - Endres, D.M.; Schindelin, J.E. A new metric for probability distributions. IEEE Trans. Inf. Theory
**2003**, 49, 1858–1860. [Google Scholar] [CrossRef][Green Version] - Reeves, C.R. Modern Heuristic Techniques for Combinatorial Problems; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 1993. [Google Scholar]
- Xiang, Y.; Gubian, S.; Martin, F. Generalized Simulated Annealing. In Computational Optimization in Engineering-Paradigms and Applications; InTech: London, UK, 2017. [Google Scholar]
- Metropolis, N.; Rosenbluth, A.W.; Rosenbluth, M.N.; Teller, A.H.; Teller, E. Equation of state calculations by fast computing machines. J. Chem. Phys.
**1953**, 21, 1087–1092. [Google Scholar] [CrossRef] - Sun, N.Z.; Sun, A. Model Calibration and Parameter Estimation: For Environmental and Water Resource Systems; Springer: Berlin, Germany, 2015. [Google Scholar]
- Szu, H.; Hartley, R. Fast simulated annealing. Phys. Lett. A
**1987**, 122, 157–162. [Google Scholar] [CrossRef] - Tsallis, C.; Stariolo, D.A. Generalized simulated annealing. Phys. A Stat. Mech. Appl.
**1996**, 233, 395–406. [Google Scholar] [CrossRef] - Team, R.C. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2016, 2017.
- Xiang, Y.; Gubian, S.; Suomela, B.; Hoeng, J. Generalized Simulated Annealing for Global Optimization: The GenSA Package. R J.
**2013**, 5, 13–28. [Google Scholar] - Zhu, C.; Byrd, R.H.; Lu, P.; Nocedal, J. Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization. ACM Trans. Math. Softw. (TOMS)
**1997**, 23, 550–560. [Google Scholar] [CrossRef] - Powell, M.J. The BOBYQA Algorithm for Bound Constrained Optimization without Derivatives; Cambridge NA Report NA2009/06; University of Cambridge: Cambridge, UK, 2009; pp. 26–46. [Google Scholar]
- Diggle, P.J.; Tawn, J.; Moyeed, R. Model-based geostatistics. J. R. Stat. Soc. Ser. C (Appl. Stat.)
**1998**, 47, 299–350. [Google Scholar] [CrossRef]

**Figure 1.**Simulation and the TI variables: (

**a**) GPR variable used as the secondary simulation variable; (

**b**) borehole data used to condition the simulations (primary variable-black dots overlying the GPR survey); (

**c**) floor survey of a previously extracted mine area (primary TI variable) and (

**d**) GPR data of the previously mined area collected prior to the extraction (Secondary TI variable).

**Figure 5.**(

**a**) Initial average of the realisations; (

**b**) final average of the realisations. Interquartile range maps: (

**c**) before the parameter optimisation and (

**d**) after the parameter optimisation.

**Figure 6.**Evolution of the distributions of the IQR values calculated at each grid node. Mean of the IQR values dropped from 0.60 to 0.47 in automatically found parameters.

**Figure 7.**Variograms of the realisations and the experimental variogram of the borehole data: (

**a**) Variograms before the optimisation; (

**b**) variograms after the optimisation and (

**c**) 99.9% confidence interval of the variograms before and after the optimisation.

**Figure 8.**Convergence comparisons of three different optimisers with three different set of initial DS parameters.

Parameters | ${\mathit{w}}_{\mathit{hd}}$ | ${\mathit{n}}_{\mathit{hd}}$ | ${\mathit{n}}_{\mathit{GPR}}$ | ${\mathit{t}}_{\mathit{hd}}$ | ${\mathit{t}}_{\mathit{GPR}}$ | ${\mathit{d}}_{\mathit{JS}}$ |
---|---|---|---|---|---|---|

Lower Boundaries | 1 | 1 | 1 | 0.0001 | 0.0001 | 0.1520 |

Initial Parameters | 10 | 5 | 20 | 0.010 | 0.10 | 0.1020 |

Automatic Tuning | 13.07 | 14 | 45 | 0.009 | 0.305 | 0.0667 |

Upper Boundaries | 15 | 50 | 50 | 0.5 | 0.5 | 0.1793 |

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Dagasan, Y.; Renard, P.; Straubhaar, J.; Erten, O.; Topal, E. Automatic Parameter Tuning of Multiple-Point Statistical Simulations for Lateritic Bauxite Deposits. *Minerals* **2018**, *8*, 220.
https://doi.org/10.3390/min8050220

**AMA Style**

Dagasan Y, Renard P, Straubhaar J, Erten O, Topal E. Automatic Parameter Tuning of Multiple-Point Statistical Simulations for Lateritic Bauxite Deposits. *Minerals*. 2018; 8(5):220.
https://doi.org/10.3390/min8050220

**Chicago/Turabian Style**

Dagasan, Yasin, Philippe Renard, Julien Straubhaar, Oktay Erten, and Erkan Topal. 2018. "Automatic Parameter Tuning of Multiple-Point Statistical Simulations for Lateritic Bauxite Deposits" *Minerals* 8, no. 5: 220.
https://doi.org/10.3390/min8050220