*Article* **Variability of the Ball Mill Bond's Standard Test in a Ta Ore Due to the Lack of Standardization**

**Gloria González García 1 , Alfredo L. Coello-Velázquez 2 , Begoña Fernández Pérez <sup>1</sup> and Juan M. Menéndez-Aguado 1, \***


**Abstract:** There is no doubt about the practical interest of Fred Bond's methodology in the field of comminution, not only in tumbling mills design and operation but also in mineral raw materials grindability characterization. Increasing energy efficiency in comminution operations globally is considered a significant challenge involving several Sustainable Development Goals (SDGs). In particular, the Bond work index (*w<sup>i</sup>* ) is considered a critical parameter at an industrial scale, provided that power consumption in comminution operations accounts for up to 40% of operational costs. Despite this, the variability of *w<sup>i</sup>* when performing the ball mill Bond's standard test is not always understood enough. This study shows the results of a variability analysis (a 3 3 factorial design) performed to elucidate the influence on *w<sup>i</sup>* of several parameters obtained from the particle size distribution (PSD) in feed and product. Results showed a clear variability in the work and grindability indexes with some of the variables considered.

**Keywords:** comminution; grindability; work index; energy efficiency

#### **1. Introduction**

There is no doubt about the importance of Fred Bond's methodology [1–5] and its practical value in the field of comminution, not only in tumbling mills design and operation but also in the characterization of mineral raw materials grindability. The Third Law of Comminution, also known as the Bond's Law, is summarized in Equation (1) [5].

$$\mathcal{W} = 10 \cdot w\_{\bar{l}} \cdot \left( \frac{1}{\sqrt{P\_{80}}} - \frac{1}{\sqrt{F\_{80}}} \right) \tag{1}$$

wherein:

*W* is the specific power consumption [kWh/t];

*wi* is the Bond work index [kWh/t];

*P*<sup>80</sup> is 80% passing size in the grinding product particle size distribution (PSD); *F*<sup>80</sup> is 80% passing size in the feed PSD.

Increasing energy efficiency in comminution operations globally is considered a significant challenge involving several SDGs, especially goals 7 (affordable and clean energy), 9 (industry innovation and infrastructure), 12 (responsible consumption and production) and 13 (climate action), since the increasing energy efficiency reduces waste and emissions production and increases energy availability. In particular, the Bond work index (*w<sup>i</sup>* ) is considered a critical parameter at an industrial scale, for power consumption in comminution operations accounts for up to 40% of operational costs [6–8]. Moreover, *w<sup>i</sup>* should be one of the key parameters to consider in a potential process plant digitalization action, using adequate measurable parameters correlation. Despite this, the variability of *w<sup>i</sup>* when performing the ball mill Bond's standard test is not always considered or

**Citation:** García, G.G.; Coello-Velázquez, A.L.; Pérez, B.F.; Menéndez-Aguado, J.M. Variability of the Ball Mill Bond's Standard Test in a Ta Ore Due to the Lack of Standardization. *Metals* **2021**, *11*, 1606. https://doi.org/10.3390/ met11101606

Academic Editors: Giovanni Principi and Antoni Roca

Received: 12 August 2021 Accepted: 5 October 2021 Published: 9 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

understood at an industrial scale [9–13]. In the study presented by Mosher and Tague [9], they addressed the variability of Bond test results independent of sampling or procedural variation. They discussed test sensitivity and detailed test procedures to maximize the accuracy and precision of the test, concluding that the Bond tests within one laboratory showed repeatability of less than ±4% at two standard deviations. They also recommended not to report Bond work indices beyond 0.1 kWh/t, based on the precision of the test and suggested that determination of the reproducibility of *w<sup>i</sup>* can be improved significantly by accurate determination of the fresh feed and product PSD. Rodríguez et al. [11] studied this extent, showing that the methodology used for *F*<sup>80</sup> and *P*<sup>80</sup> determination by interpolation significantly affects *w<sup>i</sup>* calculation.

In the case of the research presented in [10], the results of this research, carried out on a porphyry copper ore, concluded that the Bond work index values differ with different Bond ball mills and with different grinding ball charge distributions, but variations were higher when comparing different Bond ball mills than when comparing different ball charges in the same mill. Maximum variations of 8.6% with different mills and 6.2% with different grinding ball charges were measured.

The authors could not find a precedent comprising a variability study on the Bond standard test itself; mineral processing engineers sometimes attribute the *w<sup>i</sup>* variations to ore grindability changes, while the reason can yield in feed PSD variations. Recently, it has been evidenced that, for a given ore, the grindability function (variation of the Maxon index, *gbp*, with *P*100) can present a regular shape while the *w<sup>i</sup>* function with *P*<sup>100</sup> can be pretty erratic [14]. Some lack of standardization in the so-called standard test can be the most probable cause of *w<sup>i</sup>* variability. This work presents the result of a careful experimental design defined to elucidate the influence of several parameters obtained from the particle size distribution (PSD) in feed and product on *w<sup>i</sup>* determination.

#### **2. Materials and Methods**

#### *2.1. Materials*

In order to carry out the series of tests, a 400 kg Ta-Nb-Sn ore sample from the tailings deposit of former mining activities in the Penouta mine (Orense, Spain) was received. A detailed characterization of this ore sample can be found in previous research works [15–17]. The sample was fully sieved in the following size intervals (µm): 3150/2500; 2500/2000; 2000/1600; 1600/1250; 1250/800; 800/500; 500/400; 400/200; 200/160; 160/100. With adequate blending, using the aforementioned size intervals, nine composite feed samples were prepared to fulfil the requirements posed by the multivariate design. In each case, the composite sample was homogenized and divided, checking by PSD analysis that aliquots verified the requirements in each case (Figures S1–S27 at the Supplementary Materials).

#### *2.2. Methods*

#### 2.2.1. Bond Ball Mill Standard Test

The procedure to carry out the Bond grindability test [1,18] is described below. The test is performed in the so-called Bond's standard ball mill, a laboratory mill 12′′ × 12′′ , running at 70 rpm (BICO, San Francisco, CA, USA) with rounded inner edges and without lifters. The grinding charge is comprised of a steel balls distribution; Table 1 shows the distribution proposed by Bond in 1961 [5] and that proposed in 1999 [19]; the latter was selected for this test.


**Table 1.** Evolution of the ball grinding charge distributions proposed by Bond.

The mill feed must be prepared by controlled crushing to 100% passing 6 Tyler mesh (3.35 mm). The first grinding cycle feed must be 700 cm<sup>3</sup> , and this volume's weight is fixed as the mill charge in all subsequent cycles. Fresh feed PSD is obtained to calculate the 80% passing size (*F*80) and undersize weight already present in the feed. The test procedure consists of performing several dry grinding cycles to simulate a continuous closed-circuit operation with a 250% circulating load. The circuit is closed by a sieve (*P*100) selected according to the industrial grinding size target, always between 28 and 325 Tyler mesh (600–45 microns). The detailed grinding cycles procedure can be found in [5,18].

Once finished the grinding cycles, a minimum of five, the ball mill Bond's work index *wi* [kWh/sht] can be calculated using Equation (2). In order to express it in metric tons, the corresponding conversion factor must be used.

$$w\_i = \frac{44.5}{P\_{100}^{0.23} \cdot gbp^{0.82} \cdot \left(\frac{10}{\sqrt{P\_{80}}} - \frac{10}{\sqrt{F\_{80}}}\right)}\tag{2}$$

where:

*wi* is the ball mill Bond's work index [kWh/sht];

*P*<sup>100</sup> is the mesh size used to close the grinding circuit [µm];

*gbp* is the grindability index [g/rev].

It has been recently proposed *gbp* be renamed as the Maxson index [14]. Walter Maxson led the first research in which *gbp* was named as the grindability index [1], and was also Fred Bond's mentor at the beginning of his successful career.

#### 2.2.2. Multivariate Experimental Design

The standard test states tight conditions to some test parameters, while others can rest in broad validity ranges. For instance, *F*<sup>80</sup> and *P*<sup>100</sup> only limitations are being less than 3.35 mm and 600 microns, respectively. Moreover, the undersize content in the ore feed sample is considered by some authors as a variability source. Accordingly, with the same ore, minor differences under correct sampling procedures or even internal procedures in different laboratories could lead to different *w<sup>i</sup>* values. Following the considerations above, the selected variables to perform a variability analysis on the Bond's ball mill standard test were the following:


It is important to notice that *F*<sup>80</sup> and the undersize percentage in the feed (% < *P*100) variations could occur easily due to changes in material preparation; changes in *P*<sup>100</sup> should be justified due to changes in the ore liberation size, which is not a strange event in mine operations over time.

Table 2 shows the variables coding (D, C, F) and their values (level 1, 2 or 3) in each case. A total of 27 combinations of variables and levels defined the conditions of the 27 Bond standard tests. Enough ore feed was carefully prepared to fulfil D and F requirements (nine different feed samples prepared), and the Bond standard test was carried out at C value of *P*<sup>100</sup> (three levels). It must be understood that, with the same ore and with no further specifications, each of the 27 possibilities fulfils the standard test requirements and the corresponding *w<sup>i</sup>* should be considered with the same validity. The basis and practical use of the ANOVA (SPSS, IBM, Amonk NY, USA) test can be found in Navidi [20].


**Table 2.** Three levels multivariate experimental design.

#### **3. Results and Discussion**

Table 3 collects the results of Bond work index, *w<sup>i</sup>* determination after performing the resulting 27 Bond standard tests; the Mosher and Tague repeatability estimation was considered adequate [9], lower than ±4% at two standard deviations, after checking it with preliminary tests. In Table 3 the *gbp* value obtained in each test is also included. Full details of the performed tests can be found in the spreadsheet file provided as Supplementary Materials.


**Table 3.** Experimental results of *w<sup>i</sup>* and *gbp*.

The first glance at Table 3 evidences a variability in both *w<sup>i</sup>* and *gbp* values; this variability should be explained due to the sole effect of variables combination in each test. It must be highlighted again that feed preparation was performed carefully, and feed variations among synthetic feeds and a naturally taken feed could be similar to those produced in the field sampling process. In all cases, test conditions fulfilled the Bond standard test requirements (which, in passing, are very open; the only limitation is that feed top size must be under 3.35 mm). Therefore, in summary, the different nine synthetic feeds could be the result of different sampling procedures performed on the same deposit without enough representativity, provided that a tailings pond could show differences in the spatial distribution of particle sizes. Results are also depicted in the Supplementary Materials Figures S28–S30 in the case of *w<sup>i</sup>* , and Figures S31–S33 in the case of *gbp*.

A formal analysis of results was carried out employing the ANOVA test [20], both on *w<sup>i</sup>* and *gbp*. Table 4 garners the ANOVA test results in the case of *w<sup>i</sup>* .


**Table 4.** Analysis of variance (ANOVA) test results on *w<sup>i</sup>* .

Table 4 breaks down the variability of *w<sup>i</sup>* into contributions due to individual variables effects and the binary interactions among them. Considering the sum of squares values and *p*-values in the case of individual variables and binary interactions, variable D (*F*80) is identified as the primary source of variability among the studied ones. The second source of variability stems from C and D interaction, that is, *F*<sup>80</sup> and *P*<sup>100</sup> combined effect, which surprisingly has more significant influence than C alone effect. From a *w<sup>i</sup>* variability point of view, F (undersize feed content) was identified as the third variable in importance. In the case of D and F interaction, the *p*-value is not less than 0.05, so this combination does not have a statistically significant effect on *w<sup>i</sup>* , at the 95.0% confidence level.

Similarly, another ANOVA test was carried out on Maxson grindability index values, and the results are provided in Table 5. In this case, variable C (*P*100) is identified as the most relevant source of variability; despite D, F and C and F having a *p*-value more than 0.05 (in consequence, they have a statistically significant effect on *gbp*, at the 95.0% confidence level), the difference in the sum of squares values lets us affirm that C can be considered as almost the only source of variability in this case.


**Table 5.** ANOVA test results on gbp.

Results suggest that, under the conditions considered in the multivariate design described, the Maxson grindability index, *gbp*, represents more robustly the intrinsic grindability properties of the ore, being its source of variation the Bond standard test condition, *P*100. This result reinforces the concept, first proposed by Maxson et al. [1] and subsequently adopted and disseminated by Bond [3–5], that *gbp* was the best index in characterizing the ore comminution amenability. This fact also justifies the proposal of renaming *gbp* as the Maxson grindability index.

On the other side, Bond work index variability has a more profound influence from feed PSD conditions (mainly *F*<sup>80</sup> value), even to a far greater extent than *P*<sup>100</sup> values. As the standard test established relatively frugal recommendations about feed PSD conditions (maximum feed size, *F*100, less than 3.35 mm), it can be qualified as a worrying source of *w<sup>i</sup>* variation, and the following additional recommendations should be taken into account:

• To establish desirable Bond test conditions, always consider performing feed preparation according to the planned/expected industrial conditions (for instance, by product size estimation on the previous comminution stage—fine crushing or coarse grinding); • When reporting *w<sup>i</sup>* results, *P*<sup>100</sup> and *F*<sup>80</sup> values in the test should always be indicated, especially *F*80, which seems more responsible for *w<sup>i</sup>* variability than *P*<sup>100</sup> itself.

#### **4. Conclusions**

The following conclusions were derived from this research work and considering the tested ore:


**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/met11101606/s1, Figure S1: Feed PSD, test C1-D1-F1, Figure S2: Feed PSD, test C1-D1-F2, Figure S3: Feed PSD, test C1-D1-F3, Figure S4: Feed PSD, test C1-D2-F1, Figure S5: Feed PSD, test C1-D2-F2, Figure S6: Feed PSD, test C1-D2-F3, Figure S7: Feed PSD, test C1-D3-F1, Figure S8: Feed PSD, test C1-D3-F2, Figure S9: Feed PSD, test C1-D3-F3, Figure S10: Feed PSD, test C2-D1- F1, Figure S11: Feed PSD, test C2-D1-F2, Figure S12: Feed PSD, test C2-D1-F3, Figure S13: Feed PSD, test C2-D2-F1, Figure S14: Feed PSD, test C2-D2-F2, Figure S15: Feed PSD, test C2-D2-F3, Figure S16: Feed PSD, test C2-D3-F1, Figure S17: Feed PSD, test C2-D3-F2, Figure S18: Feed PSD, test C2-D3-F3, Figure S19: Feed PSD, test C3-D1-F1, Figure S20: Feed PSD, test C3-D1-F2, Figure S21: Feed PSD, test C3-D1-F3, Figure S22: Feed PSD, test C3-D2-F1, Figure S23: Feed PSD, test C3-D2- F2, Figure S24: Feed PSD, test C3-D2-F3, Figure S25: Feed PSD, test C3-D3-F1, Figure S26: Feed PSD, test C3-D3-F2, Figure S27: Feed PSD, test C3-D3-F3, Figure S28: Variability of *w<sup>i</sup>* [kWh/t] (*P*<sup>100</sup> = 500 µm), Figure S29: Variability of *w<sup>i</sup>* [kWh/t] (*P*<sup>100</sup> = 400 µm), Figure S30: Variability of *w<sup>i</sup>* [kWh/t] (*P*<sup>100</sup> = 200 µm), Figure S31: Variability of *gbp* [g/rev] (*P*<sup>100</sup> = 500 µm), Figure S32: Variability of *gbp* [g/rev] (*P*<sup>100</sup> = 400 µm), Figure S33: Variability of *gbp* [g/rev] (*P*<sup>100</sup> = 200 µm).

**Author Contributions:** Conceptualization, J.M.M.-A.; methodology, G.G.G. and J.M.M.-A.; validation, G.G.G. and B.F.P.; formal analysis, G.G.G. and J.M.M.-A.; investigation, G.G.G.; resources, J.M.M.-A.; data curation, A.L.C.-V.; writing—original draft preparation, G.G.G. and J.M.M.-A.; writing—review and editing, B.F.P. and J.M.M.-A.; visualization, G.G.G. and J.M.M.-A.; supervision, A.L.C.-V. and J.M.M.-A.; funding acquisition, G.G.G. and J.M.M.-A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is part of the OPTIMORE project funded by the European Union Horizon 2020 Research and Innovation Programme under grant agreement No. 642201.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

