A Machine Learning Based Intelligent Decision Support System for Flotation Dynamics in Simulation of Clay Minerals

Chen, Wanying; Li, Zhihong; Hao, Bo; Gao, Jianchuan; Fan, Minqiang

doi:https://doi.org/10.1155/2022/7893940

Scientific Programming

On this page

Abstract Introduction Materials and Methods Results and Discussion Conclusions Data Availability Conflicts of Interest Acknowledgments References Copyright Related Articles

Special Issue

Decision Support System for Developing Smart and Intelligent Applications

View this Special Issue

Research Article | Open Access

Volume 2022 | Article ID 7893940 | https://doi.org/10.1155/2022/7893940

A Machine Learning Based Intelligent Decision Support System for Flotation Dynamics in Simulation of Clay Minerals

Wanying Chen,¹Zhihong Li,¹Bo Hao,¹Jianchuan Gao,²and Minqiang Fan¹

Academic Editor: Muhammad Zakarya

Received05 Feb 2022

Revised15 Feb 2022

Accepted22 Feb 2022

Published23 Mar 2022

Abstract

In coal mines, process management is related with planning and flow of coal minerals from their mining points to journey’s end. The process is dependent on the operational decisions, which should be completed during coal production. In these systems, simulation modelling is assumed a powerful tool for decision making. The simulation modelling can further be enhanced through applying artificial intelligence (AI) and machine learning (ML) methods. The communication of water with clay minerals upholds the water adsorption on the clay surface, which makes them complex systems. Therefore, evading the water absorption from the clay turns out to be a hard job. The computational trainings of clay minerals are needed to comprehend the dynamics of water distribution. Tradition of coal slurry treatment is completed by adding medicament of its mineral flocculation sedimentation. As a result of coal, slime water contains a lot of clay minerals that are rich in kaolinite versatility and it is difficult to settle. The flotation will be one of the kaolinite recycling. In this paper, clay minerals containing a variety of minerals were taken as samples, and sodium dodecyl sulfate and sodium oleate were used as collectors to explore the flotation effect through test and molecular dynamics simulation. A machine learning based intelligent decision support system is designed to improve the outputs of the simulation model. The results show that when the pH value is 8 and the amount of collector and sec-octanol are 150 g/t and 250 g/t, respectively, the flotation rate of fine mineral can reach 63.25%. According to the molecular dynamics simulation results, the addition of the collector can reduce the hydrophilicity of the kaolinite surface, and the physical adsorption of SDS only occurs on the (001) surface.

1. Introduction

The coal slurry preparation plant contains a large number of clay minerals, resulting in fine particle size, high viscosity, and complex interaction between particles. The slime water is difficult to settle [1], and the difficulty of treatment increases. Common clay minerals in coal include kaolinite, illite, and aemon mixed layer [2], among which kaolinite is widely used [3]. If it is enriched from tail coal for reuse, the utilization rate of tail coal can be improved, the difficult problems of environmental and slime water treatment can be solved, and considerable economic efficiency can be brought.

Flotation is the main method of fine mineral separation. Ionic collector is commonly used in kaolinite flotation. When dodecylamine was used as a collector [4], the good flotation effect under acidic conditions was due to the weakened adsorption difference between the cleavage planes and the occurrence of flocculation of mineral particles, which increased the adsorption density of dodecylamine [5]. The research results show that [6], due to the differences in the electrostatic forces between different tertiary amine cations and kaolinite surface, substituent effect and space effect of tertiary amine substituent group, and N atom bonding, the flotation results are different, and the results are similar to those of molecular dynamics simulation [7]. In the quaternary ammonium 123C system, metal ions and pH have a great influence on the floatability of kaolinite [8]. The results showed [9] that the adsorption capacity of CTAC, DTAC, and TTAC on kaolinite increased with the increase of dosage of reagents. Aminoamide has a good collecting effect on pure kaolinite under acidic conditions, and the yield increases with the increase of dosage [10]. It is found that [11] flotation efficiency of flotation reagent varies with different concentration, pH, and temperature. Sodium oleate is a collector, and kaolinite with small particle size has better flotation efficiency [12]. Among them, reagents and kaolinite are mainly chemical adsorption, and pH and metal ions have great influence on the flotation effect of kaolinite [13–15].

The element composition of different surfaces of kaolinite is different, and the interaction between different reagents and different surfaces of the kaolinite can be well understood through molecular dynamics simulation, so as to explain the difference of the flotation efficiency between reagents [16–21]. Kaolinite is widely used in clay minerals. In this paper, the flotation performance and action mechanism of coal measure kaolinite by sodium dodecyl sulfate (SDS) and sodium oleate (NaOL) were investigated by flotation test and molecular dynamics simulation. The interaction between Na+ in SDS and O atoms in kaolinite, O in SDS anionic group, and H atoms in kaolinite belongs to ion-dipole interaction. The carboxylic acid ion of sodium oleate adsorbed with kaolinite (001) surface, and the hydrophobic alkane chain of oleate ion adsorbed with (00-1) surface [22, 23].

The communication of water with clay minerals upholds the water adsorption on the clay surface, which makes complex systems. Therefore, evading the water absorption from the clay turns out to be a hard job. The computational trainings of clay minerals are needed to comprehend the dynamics of water distribution. Moreover, process management in coal mines is related with planning, and flow of coal minerals from their mining points to journey’s end. The process is dependent on the operational decisions, which should be completed during coal production. In these systems, simulation modelling is assumed a powerful tool for decision making. The simulation modelling can further be enhanced through applying artificial intelligence and machine learning methods [24–27]. Following are the major contributions of this paper.(i)We explore the flotation effect through test and molecular dynamics simulation in clay minerals.(ii)A machine learning based intelligent decision support system is designed and integrated in the simulation model.(iii)A data aggregation technique is presented to improve the model training time.(iv)The simulation modelling can further be enhanced through applying artificial intelligence and machine learning methods.

The remainder of the paper is structured as follows. In Section 2, we present an outline of the materials and methods. A brief debate over the machine learning, decision support system, and data aggregation is also included. Section 3 is related to the datasets and assessment metrics. In addition, experimental details and results are deliberated. Finally, Section 4 concludes this paper and offers several directions for further research and investigation.

2. Materials and Methods

2.1. Test Samples

The test samples were collected from clay minerals in Malan Coal Mine of Shanxi Coking Coal Group. The sample was in block and was ground to −0.074 mm by laboratory prototype, bagged, sealed, and reserved. Table 1 and Figure 1 show the chemical composition and mineral composition of the samples, respectively.

As can be seen from Table 1, SiO₂ and Al₂O₃ contents are the highest in the chemical composition of the samples, corresponding to quartz minerals and kaolinite and other clay minerals. It can also be seen from the diffraction peak and content percentage in Figure 1 that the main minerals in the sample are kaolinite and quartz.

2.2. The Reagents

Flotation collector is sodium dodecyl sulfate and sodium oleate (SDS, NaOL, 5% aqueous solution with water), regulator is HCl solution and NaOH solution, and foaming agent is sec-octanol; all agents are analytically pure.

2.3. Test Method

(1)Flotation: the XFG ii laboratory hanging cell flotation machine was used for the test, with the rotation speed of 1800 r/min. Moreover, periodically 6 g samples were taken each time, and put into 60 mL flotation cell. Next, by adding 50 mL tap water and mixing slurry for 2 min, we adjusted pH with HCL or NaOH. Subsequently, we added flotation collector and stirring for 3 min, and mixed foaming agent for 2 min and then blowing bubbles for 5 min. The flotation concentrate and tailings were dried and weighed to calculate the yield, and the element analysis was carried out to calculate the Al/Si ratio.(2)X-ray diffraction analysis: Rigaku MiniFlex600 X-ray diffractometer was used for testing with a scanning range of 5°–85°.(3)Molecular dynamics simulation: the molecular model of kaolinite and SDS was constructed and optimized by Material Studio software. The size of the simulated box was 44.67 × 51.5 × 120 Å, and 30 pharmaceutical molecules were constructed. The thickness of the vacuum layer was 100 Å. The charge of kaolinite was calculated by QEq method, and the charge force field parameters of other molecules were allocated. Temperature is controlled at 298 K by Nose. The simulation time step is 1FS and the simulation time is 1000 ps. The electrostatic force was truncated by Ewald addition method, and the Van der Waals force was truncated by atom based method with a truncation radius of 12.5 Å. The balance part of 100 ps was used for subsequent analysis.

2.4. Decision Support System Framework

A decision support system (DSS) is an interactive, computer-based system that assists a user in making more effective decisions while solving semistructured geographical data. Satellite remote sensing and digital elevation modelling provide a systematic, rational framework for advancing scientific knowledge of geophysical phenomena, which frequently lead to the detection of natural disasters or mineral wealth. Figure 2 illustrates the process of the simulation model and Figure 3 shows a decision support system framework for coal mines. As stated in Figure 2, after deciding the problem and objectives, the collected data will help to develop the simulation model. The model, as such, can offer a solution to the problem. The dataset is divided into training and testing parts, where the latter one is used to validate the solution against a set of objectives. The DSS system comprises several parts including machine learning, IoT sensors, and data aggregation module. These are discussed in subsequent sections.

2.4.1. Machine Learning

The mining industry’s interest in using AI approaches in fields like geology and minerals processing has grown in recent years. In the realm of scientific inquiry, this pattern is repeated. Soft computing has been employed in the modelling, design, and optimization of mining operations, among other things. They can also help with the design, development, and operation of intelligent systems capable of adapting, learning, and operating autonomously in uncertain and imprecise environments. To extract useful models from empirical data, machine learning methods are used. These algorithms may be applied to metallurgical process optimization and design. On a larger scale, these approaches are used as soft sensors for the prediction of difficult-to-measure data. Modeling of metallurgical reactions or subprocesses engaged in integral processes such as flotation is one of the uses of these technologies. Machine vision is the study of ways for extracting useful information from high-dimensional pictures, and it has been utilized nearly exclusively in mining. In this paper, we used support vector machine (SVM) and long short-term memory (LSTM) approaches to further enhance the simulation model and improve decisions making.

2.4.2. Internet of Things

Internet of things (IoT) sensors are placed in the coal mineral site which are used to collect data. These sensors are distributed across the mining sites and required data is collected. The data is used to monitor the entire site and it can also be sent to the DSS for taking appropriate decisions. In addition, the data can also be utilized for safety purposes. Various IoT sensors that can measure physical surroundings and capture real-time environment changes make up the IoT sensor data layer. Temperature, pressure, humidity, level, accelerometer, gas, gyroscopes, motion sensors, image, optical sensors, radiofrequency identifier (RFID) sensors, and infrared (IR) sensors are some of the most prevalent IoT sensors. Similarly, management in coal mines is related with planning, and flow of coal minerals from their mining points to journey’s end. The process is dependent on the operational decisions, which should be completed during the coal production.

2.4.3. Data Preprocessing and Aggregation

The data is processed and certain methods are applied to remove redundancy. We use the well-known Euclidian distance equation to design an aggregation approach, as shown in Algorithm 1. The suggested algorithm compares two data points; in case they belong to the overlapping regions, one point is discarded and the other one is taken into account. This could be achieved through computing the Euclidian distance and then comparing the difference with a predefined threshold value. If the difference is less than or equal to the threshold value, then one point is discarded. Aggregation is a data preprocessing technique that reduces the burden on the decisions support system, as shown in Figure 3. The processed data is saved to a centralized database that can be used in the prediction and simulation of the coal mineral. Approaches like aggregation might be useful in removing duplicate data that, subsequently, reduces the training time. After preprocessing, the clean data is saved in the database for other tasks including prediction and monitoring. Different AI and ML algorithms are used in the prediction process.

	Input: Original dataset denoted by points D_i and D_j
	Output: Refined dataset denoted by two points and
	for each element i, j in D_i, D_j
	if j ≤ i or i D_i or j D_j
	exit()
	end if
	E_d(D_i, D_j) = (d_ik − min_jk)² where d_ik є D_i and d_jk є D_j
	if E_d(D_i, D_j) ≤ T_dthen
	Discard either D_i or D_j, to form or
	end if

3. Results and Discussion

3.1. Flotation Performance of Clay Minerals by Collectors

It should be noted that, in this study, NaOL and SDS were used as collectors. Moreover, due to their foaming properties and characteristics, the effects of adding and not adding foaming agents under different pH conditions were examined in the test. The relevant flotation test results are shown in Figure 4. Among them, the amount of collector without foaming agent is 500 g/t, the amount of collector with foaming agent is 150 g/t, and the amount of octanol in foaming agent is 250 g/t.

As can be seen from Figure 4, the flotation effect of SDS under the same conditions is significantly better than that of NaOL. When the pH value of NaOL is lower than 9, NaOL has no collecting effect on the sample, but when the pH value is higher than 10∼11, NaOL has collecting effect, but the flotation rate of fine mineral is less than 20%. When SDS was used as collector, the flotation concentrate mineral yield was about 40%, significantly higher than that of NaOL collector, but the increase of fine mineral yield was slow when pH value was more than 7. The highest yields of NaOL and SDS were 52.89% and 64.51%, respectively, after the addition of sec-octanol foaming agent. The effects of two kinds of collecting agents were significantly improved. In particular, NaOL collector can obtain good flotation results under neutral and weak alkaline conditions.

Figure 5 shows the influence of different collector dosage on flotation results of clay samples when pH = 8 (pH value of laboratory tap water) and the dosage of sec-octanol 250 g/t. It can be seen that when pH = 8, the dosage of foaming agent is unchanged, and the two collectors have the same flotation rule for samples; that is, with the increase of the dosage of collector, the flotation fine mineral rate gradually increases, and the fine mineral rate tends to be stable when the dosage of reagent is ≥ 150 g/t. The maximum yield of SDS is about 7% higher than that of the NaOL.

Tables 2 and 3 show the chemical composition analysis results of the flotation products, for SDS and NaOL, respectively; under the above conditions when pH = 8, the foaming agent dosage is approximately 250 g/t, and the collector dosage is 150 g/t.

As can be seen from Tables 2 and 3, under the action of two kinds of collectors, Al₂O₃ content in flotation concentrate is higher than tailings, SiO₂ content is lower than tailings. Except when SDS is used as collector, the Na₂O content in flotation concentrate is 0.2% lower than that in tailings. In other cases, the Na₂O and K₂O contents in concentrate are higher than those in tailings, while Na and K mainly exist in clay minerals.

Comparing the effects of the two collectors, it can be seen that the Al/Si ratio of the two reagents concentrate is greater than that of the raw ore, and the Al/Si ratio of the tailings is smaller than that of the raw ore, indicating that the two reagents preferentially enrich clay minerals such as kaolinite, and the difference of the Al/Si ratio of the fine tailings when SDS is the collector is greater than that when NaOL is the collector, indicating that the selectivity of SDS is stronger than that of NaOL.

The phase analysis results of XRD spectra of flotation concentrate and tailings are shown in Table 4. It can be seen from Table 4 that the kaolinite collection ability of SODIUM SDS is greater than that of NaOL, but the quartz content in concentrate is also higher. The content difference between kaolinite and quartz in tailings of sodium lauryl sulfate is greater than that of NaOL and the difference is large, which indicates that the separation effect of SDS is greater than that of NaOL.

3.2. Molecular Dynamics Simulation of the Interaction between Collector and Kaolinite

Figure 6 shows the initial configuration and final adsorption model of two different collectors in water and kaolinite surface (001) and surface (00-1).

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Figure 6

Initial configuration and final adsorption model of collector on kaolinite crystal surface (001) and (00-1). (a) Initial configuration of SDS-(001) crystal surface; (b) final adsorption model of SDS-(001) crystal surface; (c) initial configuration of SDS-(00-1) crystal surface; (d) final adsorption model of SDS-(00-1) crystal surface; (e) initial configuration of NaOL-(001) crystal surface; (f) final adsorption model of NaOL-(001) crystal surface; (g) initial configuration of NaOL-(00-1) crystal surface; (h) final adsorption model of NaOL-(00-1) crystal surface.

As can be seen from Figure 6, one end of sulfate ion in SDS interacts with kaolinite (001) crystal face and forms adsorption, but it is almost not adsorbed on kaolinite (00-1) crystal face. NaOL is adsorbed with its carboxylic ion to kaolinite (001) crystal face, while the hydrophobic alkane chain of oleate ion is adsorbed with the (00-1) crystal face, and the hydrophilic carboxylic ion head group is oriented towards the water phase. Tables 5 and 6 show the calculated energy of interaction between the kaolinite (001) crystal face and kaolinite (00-1) crystal face with each molecule, respectively.

The formula for calculating the interaction energy between the agent and kaolinite is

EA + K = (E_total–EA–EK + W–EK–EA + W + EW + EA + K)/2 [28].

Agent is A; kaolinite is K; water is W.

As can be seen from Table 5, the total adsorption energy of kaolinite (001) crystal face and water is −1702.12 kCal/mol, in which the electrostatic interaction energy is a large negative value, indicating that kaolinite (001) crystal face and water have a strong mutual attraction, while its van der Waals action energy is relatively small positive, playing a repulsion effect. The overall effect is strong adsorption dominated by electrostatic force. After SDS and NaOL adsorption, the total adsorption energy of kaolinite (001) crystal face with water is −1613.70 kCal/mol and −1669.59 kCal/mol, respectively, which are lower than the total adsorption energy of the original sample with water, indicating that the addition of agents can weaken the hydrophilicity of kaolinite (001) crystal face. The improvement effect of the SDS is better than the earlier approaches, i.e., NaOL for coal measure kaolinite.

It can be seen from Table 6 that the total adsorption energy of kaolinite (00-1) crystal face with water is −465.26 kCal/mol, which is more hydrophobic than that of (001) surface. Van der Waals force is the main force and electrostatic force is weak. The interaction force between kaolinite (00-1) crystal face and SDS is not strong, corresponding to the phenomenon that the agent molecules are almost not adsorbed on the surface of (00-1) after equilibrium, and the total adsorption energy of (00-1) crystal face and water changes very little, indicating that SDS is not adsorbed on this surface, nor does it change the hydrophilicity of this surface. Compared to kaolinite (00-1) crystal face with NaOL interaction, van der Waals interaction energy is stronger and has the effect of attracting, so as to make the simulation after balance, oleate ions of hydrophobic chain alkanes and (00-1) surface adsorption, and hydrophilic carboxylic acid ions head towards the water phase; the adsorption configuration cannot obviously improve the feeling on the surface of the water. Although the interaction energy between the surface and water decreased after the adsorption of NaOL agent, the hydrophobicity of the surface changed little under the adsorption configuration of NaOL agent. Figure 7 shows the relative concentration distribution curves of the kaolinite (001) crystal face-water, kaolinite (00-1) crystal face-water, kaolinite (001) crystal face-water-SDS, kaolinite (00-1) crystal face-water-SDS, kaolinite (001) crystal face-water-NaOL, and kaolinite (00-1) crystal face-water-NaOL.

(a)

(b)

(c)

(d)

(e)

(f)

Figure 7

The relative concentration distribution of kaolinite, water, SDS, and NaOL. (a) Relative concentration distribution of kaolinite (001) crystal face-water; (b) relative concentration distribution of kaolinite (00-1) crystal face-water; (c) relative concentration distribution of kaolinite (001) crystal face-SDS-water; (d) relative concentration distribution of kaolinite (00-1) crystal face-SDS-water; (e) relative concentration distribution of kaolinite (001) crystal face-NaOL-water; (f) relative concentration distribution of kaolinite (00-1) crystal face-NaOL-water.

In Figure 7(a), due to intermolecular interaction, water molecules migrate and are tightly adsorbed on the surface of kaolinite (001) crystal face [29], indicating that (001) surface is strongly hydrophilic. After the interaction with SDS and NaOL, it can be seen from (c) and (e) that Na and O plasma and atoms in the agent interact with the kaolinite surface and adsorb on the kaolinite surface, reducing the hydrophilicity of (001) crystal face. It can be seen from the radial distance in Figure 7(b) that when water molecules are adsorbed on the surface of (00-1), a thick drainage zone of water molecules is generated, indicating that this surface is a hydrophobic surface. As can be seen from the peak strength of agent molecules in Figures 7(d) and 7(f), NaOL is better adsorbed on the surface of (00-1) than SDS. In addition, the peaks of water molecules in kaolinite (00-1)-water-SDS system are still sharp and the peak strength is almost unchanged, indicating that the addition of SDS has a weak effect on the (00-1) surface, while the peak strength of water molecules in kaolinite (00-1)-water-NaOL system is greatly weakened, further indicating that the adsorption of NaOL on (00-1) is stronger. However, the surface (00-1) is a hydrophobic surface, and the adsorption of agents on this surface cannot improve the hydrophilicity of kaolinite, and the waste of agents will be caused under the same dosage of agents. Therefore, compared with SDS, the flotation effect of NaOL is relatively poor, which is the same as the flotation test result. Figure 6 shows the RDF of kaolinite (001) surface, SDS, and water.

As can be seen from Figure 8(a), the first peak of O atom in kaolinite and H atom in water appears at 0.155 nm, and the first peak of H atom in kaolinite and O atom in water appears at 0.203 nm, both within the length range of hydrogen bond, so kaolinite (001) crystal face is mainly adsorbed with water in the form of hydrogen bond. In Figure 8(b), the distance between hydroxyl H on kaolinite (001) crystal face and sulfate group in SDS is 0.161 nm, the distance between hydroxyl O on kaolinite (001) crystal face and Na+ in SDS is 0.211 nm, the distance between Na+ in kaolinite and O atom in kaolinite. The interaction of O in reagent anion group and H atom in kaolinite is ion-dipole interaction, which is the main source of adsorption force. Although Na + repels H atom on kaolinite surface (because H atom in hydroxyl group has certain positive charge in polarization), Na+ and THE O atom in kaolinite and the O in the reagent anion group have strong electrostatic interaction force of adsorption, which can be said to bridge the relationship between kaolinite and the reagent anion, making their adsorption firmer. Similarly, Figure 8(c) is the radial distribution function of kaolinite (00-1) crystal face and water. It can be seen that the first peak of O atom on kaolinite (00-1) crystal face and H atom in water is 0.175 nm and the peak strength is very weak, indicating that the probability of forming hydrogen bond between (00-1) crystal face and water is very low, so the interaction strength is very weak [30]. These results and findings are the same as those of relative concentration and adsorption energy. In addition, sulfate base group and carboxylic acid ions are not adsorbed on this surface, so the radial distribution function of (00-1) crystal face and reagent has no peak formation, so it is no longer displayed, indicating that the hydrophilic groups of SDS and NaOL have no obvious adsorption on (00-1) crystal face.

(a)

(b)

(c)

The outcomes of the machine learning approaches and accuracy with and without aggregation approach are shown in Figures 9 and 10, respectively. In Figure 9, the aggregation approach significantly decreases the model training (42% to 47%) and prediction time (7.4% to 7.8%). The time is denoted in seconds. However, both SVM and LSTM are not significant to reduce the prediction times and are comparable. In Figure 10, the aggregation approach almost produced comparable results in terms of accuracy, i.e., RMSE and MAPE values. This means that our model has learned from the less data very well. The RMSE should be read as a number and the MAPE value is given in percentage. The lesser values represent a more accurate system and vice versa. In conclusion, the addition of the two agents can reduce the hydrophilicity of the kaolinite surface. The main source of the adsorption force is the interaction between Na+ in the agent and the O atom in kaolinite and the interaction between THE O atom in the anionic group of the agent and the H atom in kaolinite, which belongs to ion - dipole interaction.

4. Conclusions and Future Work

In this paper, clay minerals containing a variety of minerals were taken as samples, and sodium dodecyl sulfate and sodium oleate were used as collectors to explore the flotation effect through test and molecular dynamics simulation. A machine learning based intelligent decision support system is designed to improve the outputs of the simulation model. The flotation effect of SDS is better than that of NaOL for coal measure kaolinite. The addition of foaming agent can greatly reduce the dosage and improve the yield. When pH = 8 and the dosage of sec-octanol is 250 g/t and that of SDS is 150 g/t, the flotation rate of refined minerals by SDS can reach 63.25%. According to the molecular dynamics simulation, kaolinite (001) crystal face is hydrophilic, and (00-1) crystal face is hydrophobic. The interaction energy between kaolinite and water decreases after the interaction with the agent, indicating that the addition of the agent can weaken the hydrophilic kaolinite crystal surface. The sodium oleate is adsorbed with its carboxylic ion to kaolinite (001) crystal face, while the hydrophobic alkane chain of oleate ion is adsorbed with (00-1) crystal face, and the hydrophilic carboxylic ion head group is oriented towards water phase. One end of sulfate ion in SDS interacts with the crystal face of kaolinite (001) and forms adsorption, but it almost does not adsorb on the crystal face of kaolinite (00-1).

Furthermore, the coal mineral simulation model can be further improved through integrating a decision support system and machine learning technique. We observed that decision making is significantly enhanced while reducing the model training time when a data aggregation mechanism is applied. In the future, we will further investigate the impact of other machine learning approaches and, in particular, how the proposed decision system will be integrated in a real world scenario.

Data Availability

The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

This work was supported by Study on Coordinated Dewatering and Stable Clean Combustion Mechanism of Urban Sludge-Coal Slime Based on Mild Oxidation (U1910214).

References

F. F. Min, J. Chen, and L. Y. Liu, “State-of-the-art and developing trend of the new technologies for treating difficult-to-settle coal slurry,” Journal of Coal Preparation Technology, vol. 5, pp. 4–9, 2018.
View at: Google Scholar
J. T. Liu, M. Q. Zhang, and Y. Zeng, “Effects of different type clays on the dispersion of fine particles in coal slurry,” Journal of China University of Mining & Technology, vol. 39, no. 1, pp. 59–63, 2010.
View at: Google Scholar
L. S. Yi, Q. R. Li, L. N. Qi, and H. Li, “Study on technological mineralogy of the kaolin from Ningxiang,” Multipurpose Utilization of Mineral Resources, vol. 2, pp. 78–80, 2016.
View at: Google Scholar
F. Q. Lu, B. H. Hou, and M. X. Zhang, “Study on flotation characteristics of coal series minerals in dodecylamine system,” Journal of Chifeng University (Natural Science Edition), vol. 36, no. 5, pp. 39–41, 2020.
View at: Google Scholar
H. P. Li, Y. H. Hu, D. Z. Wang, and J. Xu, “Mechanism of interaction between cationic surfactant and kaolinite,” Journal of Central South University (Science and Technology), vol. 2, pp. 228–233, 2004.
View at: Google Scholar
X. F. Cao, C. M. Liu, and Y. H. Hu, “Research on the flotation of kaolinite using a series of dodecyl tertiary amines,” Strategic Study of CAE, vol. 13, no. 1, pp. 93–97, 2011.
View at: Google Scholar
C.-m. Liu, A.-s. Feng, Z.-x. Guo, X.-f. Cao, and Y.-h. Hu, “Dynamics simulation of tertiary amines adsorbing on kaolinite (001) plane,” Transactions of Nonferrous Metals Society of China, vol. 21, no. 8, pp. 1874–1879, 2011.
View at: Publisher Site | Google Scholar
Y. L. Zhou, Y. H. Wang, Y. H. Hu, D. X. Sun, and M. J. Yu, “Influence of metal ions on floatability of diaspore and kaolinite,” Journal of Central South University (Science and Technology), vol. 40, no. 2, pp. 268–274, 2009.
View at: Google Scholar
H. Jiang, G. Y. Xiang, S. Ahmed Khoso, J. H. Xie, K. Huang, and L. H. Xu, “Comparative studies of quaternary ammonium salts on the aggregation and dispersion behavior of kaolinite and quartz,” Minerals, vol. 9, no. 8, 2019.
View at: Publisher Site | Google Scholar
S. M. Zhao, D. Z. Wang, Y. H. Hu, J. Xu, and X. L. Zhao, “A series of aminoamides used for flotation of kaolinite,” Journal of University of Science and Technology Beijing (English Edition), vol. 3, pp. 208–212, 2005.
View at: Google Scholar
M. Fan and X. Y. Wang, “The surface properties of kaolinite and flotation characteristics,” Modern Chemical Research, vol. 9, no. 3, pp. 52–57, 2012.
View at: Google Scholar
J. Guo, M. Li, Q. M. Fang, and Y. P. Hu, “The influence of particle size on separation of diaspore and kaolinite,” Multipurpose Utilization of Mineral Resources, vol. 5, pp. 17–21, 2003.
View at: Google Scholar
G. F. Zhang, Q. M. Feng, Y. P. Lu, and L. M. Ou, “Mechanism on diaspore and kaolinite collected by sodium oleate,” The Chinese Journal of Nonferrous Metals, vol. 2, pp. 298–301, 2001.
View at: Google Scholar
L. H. Xu, F. Q. Dong, H. Q. Wu, H. Jiang, Z. Wang, and J. H. Xiao, “Solution chemistry mechanism of sodium oleate on kaolinite,” Journal of wuhan university of thehnology, vol. 34, no. 12, pp. 119–123, 2012.
View at: Google Scholar
B. H. Hou and L. Y. Liu, “Effect of Ca2+ on kaolinite and quartz flotation at different pH values in the presence of sodium oleate,” Journal of Coal Engineering, vol. 52, no. 4, pp. 149–154, 2020.
View at: Google Scholar
L. Xu, Y. Hu, F. Dong, Z. Gao, H. Wu, and Z. Wang, “Anisotropic adsorption of oleate on diaspore and kaolinite crystals: implications for their flotation separation,” Applied Surface Science, vol. 321, pp. 331–338, 2014.
View at: Publisher Site | Google Scholar
L. H. Xu, H. Jiang, F. Q. Dong, H. Q. Wu, Z. Wang, and J. H. Xiao, “The effect of DTAC on flotation of kaolinite with different particle size,” Journal of China University of Mining & Technology, vol. 42, no. 5, pp. 832–837, 2013.
View at: Google Scholar
B. Liu, M. Dai, I. Ali et al., “Molecular insights on the influence of temperature and metal ions on the hydration of kaolinite (001) surface,” Molecular Simulation, vol. 47, no. 12, pp. 1029–1036, 2021.
View at: Publisher Site | Google Scholar
T. Yue, W. Sun, and P. Chen, “Mechanism of reverse flotation desilication forbauxite by quaternary ammonium salt collector,” The Chinese Journal of Nonferrous Metals, vol. 24, no. 11, pp. 2872–2878, 2014.
View at: Google Scholar
P. Chen, W. Sun, and T. Yue, “Dynamics simulation of tributyltetradecylphosphonium chloride on kaolinite (001) plane,” Journal of China University of Mining & Technology, vol. 43, no. 2, pp. 294–299, 2014.
View at: Google Scholar
L. N. Guo, Z. H. Li, Z. L. Zhu, Y. L. Liu, and M. Q. Fan, “Study and dynamics simulation of cationic collecting performance on kaolinite,” China Mining Magazing, vol. 26, no. 5, pp. 112–116, 2017.
View at: Google Scholar
R. Wilson, P. H. J. Mercier, B. Patarachao, and A. Navarra, “Partial least squares regression of oil sands processing variables within discrete event simulation digital twin,” Minerals, vol. 11, no. 7, p. 689, 2021.
View at: Publisher Site | Google Scholar
S. Bandera, C. O’Sullivana, P. Tangney, and S. Angioletti-Uberti, “Coarse-grained molecular dynamics simulations of clay compression,” Computers and Geotechnics, vol. 138, Article ID 104333, 2021.
View at: Publisher Site | Google Scholar
M. Shishvan and J. Benndorf, “Operational decision support for material management in continuous mining systems: from simulation concept to practical full-scale implementations,” Minerals, vol. 7, no. 7, p. 116, 2017.
View at: Publisher Site | Google Scholar
M. Kozielski, M. Sikora, and Ł. Wróbel, “DISESOR-decision support system for mining industry,” in Proceedings of the Federated Conference on Computer Science and Information Systems (FedCSIS), IEEE, Lodz, Poland, September 2015.
View at: Publisher Site | Google Scholar
Y. Zhang and C. Xiao, “Molecular dynamics simulation of clay hydration inhibition of deep shale,” Processes, vol. 9, no. 6, p. 1069, 2021.
View at: Publisher Site | Google Scholar
L. A. Cisternas, F. A. Lucay, and Y. L. Botero, “Trends in modeling, design, and optimization of multiphase systems in minerals processing,” Minerals, vol. 10, no. 1, p. 22, 2020.
View at: Publisher Site | Google Scholar
X. Lyu, X. You, M. He et al., “Adsorption and molecular dynamics simulations of nonionic surfactant on the low rank coal surface,” Fuel, vol. 211, pp. 529–534, 2018.
View at: Publisher Site | Google Scholar
M. Yuan, W. Nie, H. Yu et al., “Experimental and molecular dynamics simulation study of the effect of different surfactants on the wettability of low-rank coal,” Journal of Environmental Chemical Engineering, vol. 9, no. 5, Article ID 105986, 2021.
View at: Publisher Site | Google Scholar
C. Qi, H. Manzano, D. Spagnoli, Q. Chen, and A. Fourie, “Initial hydration process of calcium silicates in Portland cement: a comprehensive comparison from molecular dynamics simulations,” Cement and Concrete Research, vol. 149, Article ID 106576, 2021.
View at: Publisher Site | Google Scholar

Copyright

Copyright © 2022 Wanying Chen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies