Artificial neural networks for optimization of gold-bearing slime smelting

doi:10.1016/j.eswa.2009.03.016

Expert Systems with Applications

Volume 36, Issue 9, November 2009, Pages 11671-11674

https://doi.org/10.1016/j.eswa.2009.03.016 Get rights and content

Abstract

Pyrometallurgy is often used in the industrial process for treating gold-bearing slime. Slag compositions have remarkable influences on the recovery of gold and the gold content in slag. A method for determining optimum flux compounding with neural networks is studied in this paper, and the neural network model for estimating the gold contents with different slag compositions is presented. On the basis of the neural network model, an algorithm for searching the optimum flux compounding in the gold-slime smelting process is proposed, and the optimum flux compositions are obtained accordingly.

Introduction

In the gold manufactory industry, pyrometallurgy (Marsden & House, 2006) is commonly used to process gold-bearing slime or gold slime for short. In this process, appropriate fluxing agents (i.e. slag compositions) are added at high temperature into the smelting furnace, which results in separating precious metals (gold and sliver) from oxides and gangue with the latter go into the remains called slag. In this way, gold–silver alloys are obtained by pyrometallurgy. In the smelting process, slag compositions essentially determine the amounts of gold left in slag, hence, the recovery of gold. To maximum the recovery of gold, it is of great importance to optimize the slag composition so as to minimise the gold content in slag. In fact, research outcomes on determining ‘best’ slag compositions have been reported (Yuan, Yao, Qiu, & Li, 1995) based on the nonlinear regression technique. However, smelting gold slime is a complicated process which involves chemical reactions of multi-phases. Therefore, it is usually hard to describe the relationship between slag compositions and gold content in slag explicitly. The application of nonlinear regression method requires that some presumptions be made about the form of distributions of data or the functional relations among the parameters concerned. Therefore, human errors are likely introduced to the problem. It is the above difficulty that has motivated us to search for a new method.

In the last two decades, artificial neural network (ANN) has revealed its huge potential in many areas of science and engineering, with the rapid development in its learning algorithms. Its exceptional function of self-organising, self-study, fault tolerance and high robustness has paved the way for its wide applications such as pattern recognition, pattern classification, tendency analysis, prediction and nonlinear functions. The neural network has also proven to be a powerful tool in many areas including industrial processes (Schlang, Lang, Poppe, Runkler, & Weinzierl, 2001), prediction of materials properties such as steel (Bahrami et al., 2005, Capdevila et al., 2006, Guo and Sha, 2004), etc. In addition, there are many other reports that the neural network approach has been used in material science based research as discussed by Sha and Edwards (2007). Artificial neural networks are now well established, and prominent in the literature. However, its application to pyrometallurgy industries has not been examined thoroughly.

As such, the ANN approach is adopted in this paper as the substitute for nonlinear regression to identify the optimum slag compositions for the process of gold recovery in pyrometallurgy. We has constructed a neural network model for estimating gold content in slag for the ternary system B₂O₃–SiO₂–Na₂O in our previous studies (Liu, Yuan, & Liao, 2009), and aim at developing a algorithm by which the neural network model can be used to predict the optimum slag compositions in this study. For the continuity of this paper, we will describe the model briefly in Section 2, following which, in Section 3, the algorithm for optimization of slag compositions will be discussed.

Section snippets

The neural network model

As mentioned above, an artificial neural network model was developed for estimating or predicting gold content in slag in pyrometallurgy. As known, artificial neural network is a network with nodes or neurons analogous to the biological neurons. The nodes are interconnected to the weighted links and organised in layers. The performance of a neural network depends mainly on the weights of its connections. The knowledge is represented and stored by the weights (strength) of the connections

Optimizing flux composition using the neural network

In the section above, we have described the established neural network, which is capable of calculating the gold contents in slag for the pyrometallurgical experiments. Because the trained network has already ‘stored’ the nonlinear relationships between slag compositions and gold content in slag, we can now adopt an optimization procedure, which determines the optimum slag compounding that minimises the gold content in slag. The problem can be interpreted mathematically as the follows:

Given V = {V

Conclusions

We use artificial neural networks to optimize the slag compositions in pyrometallurgical processes of gold slime. (i) The paper has demonstrated that artificial neural network can be used to determine the relationships between slag compositions and gold content in slag. Compared with the traditional regression method, it makes no functional assumptions on the relationships, hence, eliminates the errors brought in by Man. (ii) Further, the algorithm for optimizing the slag compositions in the

References (11)

A. Bahrami et al.
Prediction of mechanical properties of DP steels using neural network model
Journal of Alloys and Compounds
(2005)
C. Capdevila et al.
Neural network analysis of the influence of processing on strength and ductility of automotive low carbon sheet steels
Computational Materials Science
(2006)
Z. Guo et al.
Modeling the correlation between processing parameters and properties of maraging steels using artificial neural network
Computational Materials Science
(2004)
V. Kurkova
Kolmogorov’s theorem and multilayer neural networks
Neural Networks
(1992)
D. Liu et al.
Artificial neural network vs. nonlinear regression for gold content estimation in pyrometallurgy
Expert Systems with Applications
(2009)

There are more references available in the full text version of this article.

Cited by (5)

Methods for Petroleum Well Optimization: Automation and Data Solutions
2021, Methods for Petroleum Well Optimization: Automation and Data Solutions
A hybrid data mining approach for identifying the temporal effects of variables associated with breast cancer survival
2020, Expert Systems with Applications
Citation Excerpt :
ANNs are inspired by their biological counterparts, whereby a highly interconnected network of simplistic neurons generates a learning model that can learn arbitrarily complex non-linear functions. As such, these algorithms have been widely used in machine learning contexts, such as classification (Cetinic, Lipic & Grgic, 2018; Melin, Miramontes & Prado-Arechiga, 2018), regression (Genc & Dag, 2016; Zhou, Zhou, Yang & Yang, 2019), optimization (Liu, Yuan & Liao, 2009; Yazdi, Khorasani & Faraji, 2011), and pattern recognition (El-Midany, El-Baz & Abd-Elwahed, 2010; Patterson, 1996). The artificial neurons mimic the signal integration and activation of their biological counterparts using mathematical functions, such as sigmoidal activation functions.
Predicting breast cancer survival is crucial for practitioners to determine possible outcomes and make better treatment plans for the patients. In this study, a hybrid data mining based methodology was constructed to differentiate the variables whose importance for survival change over time. Therefore, the importance of variables was determined for three different time periods (i.e. one, five, and ten years). To conduct such an analysis, the most parsimonious models were constructed by employing one regression analysis method—Least Absolute Shrinkage and Selection Operator (LASSO), and one metaheuristic optimization method, namely a Genetic Algorithm (GA). Due to the high imbalance between the number of survivals and deaths, two well-known resampling procedures—Random Under-sampling (RUS) and Synthetic Minority Over-sampling Technique (SMOTE)—were applied to increase the performance of the classification models. In the final stage, two data mining models, namely Artificial Neural Networks (ANNs) and Logistic Regression (LR), were utilized along with 10-fold cross-validation. Sensitivity analysis (SA) was conducted for each model to identify the importance of each variable for a certain model and time period. The obtained results revealed that certain variables lose their importance over time, while others gain importance. This information can assist medical practitioners in identifying specific subsets of variables to focus on in different periods, which will in turn lead to a more effective and efficient cancer care. Moreover, the study findings indicate that extremely parsimonious models can be developed by adopting a purely data-driven approach, rather than eliminating the variables manually. Such methodology can also be applied in treating other types of cancer.
A population-feedback control based algorithm for well trajectory optimization using proxy model
2017, Journal of Rock Mechanics and Geotechnical Engineering
Citation Excerpt :
The performance of a neural network depends mainly on the weights of its connections. If correct weights can be trained, then an ANN can do an exceptional job (Liu et al., 2009). The neural model used in this study consists of an input layer with three nodes (mud weight, azimuth and inclination), an output layer with one node (NYZA), and one hidden layer with seven nodes.
Wellbore instability is one of the concerns in the field of drilling engineering. This phenomenon is affected by several factors such as azimuth, inclination angle, in-situ stress, mud weight, and rock strength parameters. Among these factors, azimuth, inclination angle, and mud weight are controllable. The objective of this paper is to introduce a new procedure based on elastoplastic theory in wellbore stability solution to determine the optimum well trajectory and global minimum mud pressure required (GMMPR). Genetic algorithm (GA) was applied as a main optimization engine that employs proportional feedback controller to obtain the minimum mud pressure required (MMPR). The feedback function repeatedly calculated and updated the error between the simulated and set point of normalized yielded zone area (NYZA). To reduce computation expenses, an artificial neural network (ANN) was used as a proxy (surrogate model) to approximate the behavior of the actual wellbore model. The methodology was applied to a directional well in southwestern Iranian oilfield. The results demonstrated that the error between the predicted GMMPR and practical safe mud pressure was 4% for elastoplastic method, and 22% for conventional elastic solution.
Exploring the risk factors of preterm birth using data mining
2011, Expert Systems with Applications
Citation Excerpt :
The results of our study can provide the information for medical staff or pregnant women to prevent the incidence of preterm. A neural network model is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information (Liu, Yuan, & Liao, 2009; Rajan, Ramalingam, Ganesan, Palanivel, & Palaniappan, 2009). Neural network is composed of artificial neurons or nodes.
Preterm birth is the leading cause of perinatal morbidity and mortality, but a precise mechanism is still unknown. Hence, the goal of this study is to explore the risk factors of preterm using data mining with neural network and decision tree C5.0. The original medical data were collected from a prospective pregnancy cohort by a professional research group in National Taiwan University. Using the nest case-control study design, a total of 910 mother–child dyads were recruited from 14,551 in the original data. Thousands of variables are examined in this data including basic characteristics, medical history, environment, and occupation factors of parents, and variables related to infants. The results indicate that multiple birth, hemorrhage during pregnancy, age, disease, previous preterm history, body weight before pregnancy and height of pregnant women, and paternal life style risk factors related to drinking and smoking are the important risk factors of preterm birth. Hence, the findings of our study will be useful for parents, medical staff, and public health workers in attempting to detect high risk pregnant women and provide intervention early to reduce and prevent preterm birth.
Modelling and simulation of the cyanidation process of Aghdareh gold ore using artificial neural network and multiple linear regression
2016, International Journal of Mining and Mineral Engineering

View full text