SME Annual Meeting Feb.

23-25, Denver, Colorado

Preprint 04-91
EVALUATION OF A NEW METHOD FOR WORK INDEX ESTIMATION USING SINGLE PARTICLE IMPACT TESTS K. S. Free Agapito Assocs., Inc. Grand Junction, CO M. K. McCarter R. P. King Univ. of Utah Salt Lake City, UT

ABSTRACT
Mine plans typically consider the profits and costs associated with extracting and processing minerals into marketable products. Interrelationships between mine and mill costs are often overlooked during the process of mine plan optimization. A simple and quick method for predicting milling costs for a given block would improve the planning process, especially if the procedure could be incorporated as part of routine ore grade analysis of drill core or cuttings. Incorporating milling costs as part of the mine planning process could improve overall optimization of the mine/mill combination. A method developed by the Utah Comminution Center (UCC) at the University of Utah provides a possible answer to the need for quick and reliable grinding cost estimation (King et al., 1996). Using results from single-particle impact tests, a work index, called the Utah Load Cell Work Index (ULCWI), is calculated. The potential for routine use of the new method is explored by comparing ULCWI values with corresponding Bond Work Index (BWI) values for eight mineral materials (Free, 2000).

accurately measured. By empirical means, the measured energy can be converted to ULCWI. By this methodology, a relatively quick and simple estimation of grinding cost is obtained. At present, the industry standard for forecasting grinding costs is to use empirical relationships presented by Bond (1952) to relate laboratory milling test results to predict industrial scale energy consumption in terms of a Bond Work Index (BWI). The empirically based BWI, although widely accepted, has limitations and a more theoretical method of estimating work indices would be a welcomed advance in the industry. The ULCWI has some key advantages over the BWI, but whether or not it is as reliable as the BWI, has not yet been established. This study investigates the validity of the ULCWI values by comparing them with corresponding BWI values determined from laboratory tests on various mineral materials. This paper presents: 1) the theory behind calculating BWI and ULCWI values, 2) results from laboratory tests, 3) a discussion of results, and 4) conclusions.

INTRODUCTION
In most mining operations, blasting, crushing, and grinding are necessary parts of the comminution process. Quick and accurate methods for estimating the respective costs for each step of the comminution process are needed to properly optimize the entire comminution process and the overall mining operation. This paper will focus on the portion of the comminution process which deals with the prediction of the energy consumption due to the grinding portion of the comminution process. The ball mill is a common piece of equipment in grinding circuits. Modeling the grinding process in a ball mill is a complex task. Balls and individual particles of varying sizes and geometries collide at various angles and magnitudes. The process is further complicated by the production of countless offspring particles. To simplify the task of modeling this process, the many collisions occurring in the mill can be considered a collection of single impacts. Viewing the milling process as the sum of individual impact events leads to the study of comminution by means of single-particle impacts. A method for predicting grinding energy costs based on singleparticle impact tests has been developed by the UCC at the University of Utah. Using a drop weight apparatus called the Utah Load Cell (ULC), the energy absorbed by a particle up to the point of brittle fracture can be

PROCEDURE AND THEORY

The BWI is based on Bonds theory that the energy required to break a spherical particle is inversely proportional to the square root of the particle diameter (d) or proportional to d -1/2. Using extensive industrial and laboratory data, Bond further developed his size-energy relationship into the empirical relationship referred to as the BWI equation 1,

(1)
where U1 is the passing size in m of the test sieve, Gbp is the ball mill grindability, and U80 and F80 are the product and feed 80% passing sizes (Bond 1961). The BWI obtained by this equation is a measure of a materials resistance to breakage and is given in units of kW-hr per short ton. The ULCWI is based on measured parameters obtained from single-particle impact tests using a comminution device called the Utah Load Cell (ULC). This apparatus consists of a drop-weight that impacts

Copyright 2004 by SME

SME Annual Meeting Feb. 23-25, Denver, Colorado

particles placed on the surface of a long steel rod. Strain gauges attached to the rod allow the compressive force through the particle and rod to be recorded with an oscilloscope at time intervals as small as 0.5 microseconds. Using this force-time record together with the mass and initial velocity of the ball, the energy absorbed by a particle during the impact can be calculated. The total energy absorbed by a particle from first contact until brittle failure is defined as the fracture energy. The fracture energy of many individual particles over a narrow size range are plotted as a cumulative probability distribution as shown in Figure 1. This distribution represents the probability of breakage for particles of a particular size class and type. A method of estimating the probability of breakage over a range of particle size classes is obtained by an empirical relationship referred to in this work as the EM50 model.

Figure 2. Fracture Energy Model (EM50 Model) for Calcite In addition to knowing P(E) or the breakage probability, the progeny size probability (bij) is also needed to determine the work index. A mathematical model known as the t10 model is used to predict progeny sizes from a given impact energy. The t10 concept was originally developed by the Julius Kruttschnitt Mineral Research Center (Narayanan and Whiten, 1988) and subsequently modified by the Utah Comminution Center. The model, as modified by Milin (1994), is given as

(4) Figure 1. Particle Fracture Energy Distribution of 2.8 mm (2.36 to 3.36 mm) Taconite Particles For each of several particle sizes, the fracture energy distribution is plotted and the median mass-specific fracture energy (EM50) is determined. The EM50 represents the energy of an impact causing breakage 50% of the time. A relationship between EM50 and particle size exists and can be described by equation 2 where dpo is the particle diameter associated with a fracture energy distribution, and E, dp, and v are parameters determined by a best fit of the laboratory data to the EM50 function (King et al., 1996).

(2)

where t10 is the mass fraction passing one tenth the parent size. A10 and B10 are material specific constants and E* is the relative impact energy defined by dividing the mass-specific impact energy (EI) by EM50. The t10 values are determined graphically by plotting progeny size distributions on a log-log plot and interpreting the mass fraction passing 1/10th of the parent particle size. The t10 value can then be plotted against E* values, again using a log-log coordinate system. A least-sum-of-squares, logarithmic regression is used to find the A10 and B10 values that yield the best fit to the model. A graphical representation of the t10 model fit to actual data points is shown in Figure 3. From the t10 model, the entire breakage function (tn) can be estimated by applying some complex empirical relationships that express tn as a function of An, Bn, and tn where the subscript n represents 1/nth of the parent particle size (Milin, 1994). Other models defining the breakage function are also available (King, 2001). The breakage function tn is often expressed as bij or the mass fraction of particles breaking in any parent size class j that end up in progeny size class i (King, 2001). The size distribution resulting from a single-particle impact of known energy can be estimated by equation 5

The EM50 model is shown graphically in Figure 2. Through laboratory experience in single-particle impacts using the ULC, mass-specific fracture energy distributions have been found to demonstrate log-normal distributions (King and Bourgeois, 1993). Therefore, the breakage probability can be expressed as where erf is the error function, EM is the mass-specific fracture energy, and ( E ) 2 is the logarithmic variance of the distribution (Tavares, 1997).

(5)

(3)

where pi represents the mass fraction in size class i, and the superscript n is the number of stages or impact events. Pi (E) is the probability that a particle in size class i will break during an impact event of energy E and is calculated using equation 3. The milling operation is simulated by evaluating equation 5 n times until the desired product size distribution is obtained. The energy of each impact required to produce the desired

Copyright 2004 by SME

SME Annual Meeting Feb. 23-25, Denver, Colorado

product size distribution is totaled and used in the non-empirical version of Bonds Work Index relationship

RESULTS
BWI values at the 20 mesh test size (850 m) for the eight materials investigated in this work are shown with corresponding ULCWI values in Figure 5. The impact energy used in determining the ULCWI by equation 6 is assumed to be the median fracture energy of the 80% passing product size (EM80).

(6)

where ET is the summation of n EIs totaled during the simulation. In this study, the Wi determined by equation 6 is referred to as the ULCWI. An example of graphical output from the simulated impacts described above is shown in Figure 4.

Figure 5. Work Indices as Calculated by Bonds Method as well as Work Indices Calculated by ULC Method for Selected Mineral Materials (20 Mesh)

Figure 3. Experimental Data Points That Have Been Logarithmically Regressed to Fit the t10 Model

Torque measurements of the work imparted to the load of a ball mill provide a means to calculate the Operating Work Index (OWI) of a given sample. Two of the eight mineral samples were available for this experimentation, and resulting OWI values are displayed with corresponding BWI and ULCWI values in Figure 6.

Figure 4. Simulated Comminution of Quartzite Using a Sequence of Single Impacts (the data points are taken from Bond Grindability Test, while the lines are computer generated by the EN_CALC Fortran routine)

Figure 6. Work Indices for Tests on Fluorite and Basalt

Copyright 2004 by SME

SME Annual Meeting Feb. 23-25, Denver, Colorado

DISCUSSION
By observation of Figure 5 alone, it is difficult to make the claim that a correlation between the BWI and the ULCWI exists. Plotting the BWI versus corresponding ULCWI values as in Figure 7 gives a more easily quantified basis for comparison. The regressed line with an R2 value of 0.77 in Figure 7 shows that a general relationship between BWI and ULCWI exists. A statistical t-test leads to the conclusion that for a 95% confidence level, a linear relationship between the BWI and ULCWI does exist. The conclusion is that for a size reduction from about 3 mm to 0.7 mm (80% passing sizes) the BWI and ULCWI are correlated. ULCWI is determined. The EM determined from ULC impacts is sensitive to particle size. The probability of fracture used in equation 5 comes from the size-sensitive EM. With only two data points defining the BWI, it is difficult to make any defensible conclusions, but it is still interesting to note that the empirical nature of the BWI may cause it to be less size-sensitive than the OWI or ULCWI. Another observation from the same figure is that the OWI in all cases is of higher magnitude than the corresponding BWI or ULCWI. Stated in another way, the observed energy consumption is greater than that predicted by either BWI or ULCWI methodology.

CONCLUSIONS
BWI and ULCWI values are found to have a general correlation for feed particle size distributions with 80% passing sizes ranging from 3.0 mm to 0.7 mm. A correlation is also likely for particle sizes below this range, but more data is needed to defend this claim (Free, 2000). Comparisons of OWI and ULCWI values indicate the possibility that the ULCWI could be more accurate in predicting grinding energy consumption than the BWI. The effect of particle size on the energy required for comminution is more accurately accounted for by the ULCWI than the BWI. Also, the ULCWI has a stronger theoretical basis than the BWI. The BWI and OWI are different in that the BWI is the energy per unit of mass required for a given size reduction, while the OWI is the actual energy per unit of mass consumed for a given size reduction. Thus the OWI includes the mill efficiency but the BWI does not. A grinding mills efficiency is affected by particle size, and the size-sensitive methodology of the ULCWI could be the reason for the apparent correlation between ULCWI and OWI. Figure 7. BWI Plotted Versus ULCWI for Selected Mineral Materials Using a 20 Mesh Test Size There are several physical differences between mill impacts and single-particle impacts that may prevent a stronger correlation of BWI with ULCWI. Slippage between balls with other balls or walls is one example of energy consuming events in the mill not present in single particle testing. Also, fine material in the mill causes an impact to share energy with many particles, often resulting in impact energies too small to cause breakage. Other physical differences between mill impacts and single-particle impacts also exist, but are not mentioned here (Free, 2000). Another contributing cause for the difference between BWI and ULCWI values is that the BWI is based on a correlation of laboratoryscale with full-scale equipment data. Included in this empirical relationship is the mechanical efficiency of the milling machinery. The ULCWI does not account for such energy losses. Also of interest to this study is the comparison of OWI values with BWI and ULCWI values. Referring to Figure 6, it is noticeable that the trend of the OWI is quite similar to that of the ULCWI for each of the two materials shown. This observation seems to indicate that the BWI is less sensitive to particle size than the OWI or ULCWI. A possible explanation for the size-sensitivity of the ULCWI lies in the method by which the At present it is unlikely that ULCWI methods will replace the established BWI as a preferred grinding cost estimator. However, this study has shown that improvements to the EN_CALC routine might produce more accurate ULCWI values and increase the size range over which ULCWI values can be reasonably established (Free, 2000). Among these improvements would be the ability to handle size distributions with bimodal characteristics and the ability to account for particle interactions. The ULCWI approach is well suited for mine to mill planning in that it provides a relatively simple method for predicting milling costs. Drill cuttings or core from exploration or production drilling could provide a sufficient sample for ULC testing, thus allowing mill cost information to be readily incorporated in ore body models. With some modifications, it is possible that the ULCWI could become the preferred method for predicting grinding energy consumption.

ACKNOWLEDGMENTS
This study has been made possible through the financial support of the William C. Browning Graduate Scholarship granted through the University of Utahs Mining Engineering Department. The Generic Mineral Technology Center in Comminution at the University of Utah is also acknowledged for the use of laboratory facilities and technical assistance.

Copyright 2004 by SME

SME Annual Meeting Feb. 23-25, Denver, Colorado

REFERENCES
Bond, F.C., 1952, The Third Theory of Comminution, SME-AIME, Vol. 193, pp. 484494. Bond, F.C., 1961, Crushing and Grinding Calculations, Allis Chalmers Publication 07R923B. Free, K.S., 2000, Relationship Between Bond Work Index and Single Particle Comminution Parameters, Masters Thesis, University of Utah. King, R.P., 2001, Modeling and Simulation of Mineral Processing Systems, Butteworth-Heinemann, Oxford. King, R.P. and F. Bourgeois, 1993, Measurement of Fracture Energy During Single-particle Fracture, ComminutionTheory and Practice, Kawatra, S.K., Editor SME, Littleton, Colorado. King, R.P., Tavares, L.M., and Middlemiss, S., 1996, Establishing the Energy Efficiency of a Ball Mill, Submission for Publication Milin, L., 1994, Incomplete Beta Function Modeling of the t10 Procedure. Internal Report, Comminution Center, University of Utah, pp. 43. Narayanan, S.S., and Whiten, W.J., 1988, Determination of Comminution Characteristics from Single-particle Breakage Tests and its Application to Ball-mill Scale-up, Trans. Instn. Min. Metall., Sec. C, Vol. 97, pp.115123. Tavares, L.M., 1997, Microscale Investigation of Particle Breakage Applied to the Study of Thermal and Mechanical Predamage, PhD Thesis, University of Utah.

Copyright 2004 by SME