You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/345884045

Hogg & Fuerstenau Power Model for Grinding Mills Adjusted with Discrete
Element Method and Classical Mechanics Approach

Method · November 2020


DOI: 10.13140/RG.2.2.23286.83527

CITATIONS READS

0 773

1 author:

Mario Cerda

14 PUBLICATIONS 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mario Cerda on 14 November 2020.

The user has requested enhancement of the downloaded file.


Hogg & Fuerstenau Power Model for Grinding Mills
Adjusted with Discrete Element Method and Classical
Mechanics Approach

Mario Cerda
1. Mechanical Engineer, Chile

ABSTRACT

From simulations with the discrete element method (DEM), the power associated with each particle
is obtained, according to the development of an innovative approach to classical mechanics.

The power determination was validated industrially through operation data in SAG and balls mills.
This methodology serves as a basis for adjust of general Hogg & Fuerstenau power model method,
where the unknown characteristic of the angle for the center of mass correlates with the actual
position of the center of mass calculated from the method of discrete elements and classical
mechanics with approximately 230 million records and 40 billion mathematical operations. Such
correlation, obtained through a multivariable non-linear regression on 4 dimensions, provides a
surface that depends only on the percentage of the critical speed, the level of load inside the mill
and the angle of the lifter, which naturally varies according to the height of the lifter and the rate
of wear.

Two validations are carried out with industrial grind out data for a 36-foot SAG mill and the
correlation coefficient obtained was 0. 718 with average daily data and 0.996 with crash stop data.

1
INTRODUCTION

Historically, predicting the energy demand of grinding mills in mining operations, have faced the
difficulty of obtaining a simple and reliable model. Multiple models have been developed to meet
this goal and most have failed to stay over time because of the level of error obtained, especially
for large equipment.

The use over time of the methodology developed by Hogg and Fuerstenau for this purpose is
highlighted, despite the little correlation that this methodology has with the real behavior of the
load inside a mill and even more so with the actual position of the center of mass of the load, which
has relegated the method to a purely operational adjustment use..

This study takes advantage of the development of a novel approach closer to the reality of
multiparticle system kinetics and the creation of specific applications capable of processing trillions
of data over time, in order to obtain an equation for the angle of the center of mass in the Model
of Hogg and Fuerstenau, through the multivariate regression technique and surface of responses ,
allowing to convert the overall power model to take a relevant role in the predictive development
of the operation and design of mills.

Two validations were performed with the obtained function. Both with operational data of a Sag
mill recorded during 1 year of operation and the second with crash stop values, delivering
correlation coefficients of 0.718 and 0.996 respectively.

METHODOLOGY

The methodology developed can be summarized in the following steps:

● Calibration of the discrete element.


● Development of simulations with discrete elements, where the semi-angle of the lifter (s),
the load level (Jc) and the percentage of critical speed (PCS) are varied. Given the linearity
and proportionality of the variables, such as the nominal diameter of the mill (D), the
length (L) and the apparent density of the load (s/he is not necessary to sensitize the
study with such variables, nor with the use of the ball level variable (Jb) and the height of
lifters (h), since it is implicit in the apparent density and the nominal diameter (d),
respectively. ρ

2
● Since the power of the System is not constant over time and experiences sine behavior
identical to that observed in different operations, it is defined to use the 99% reliability
level in power determination, in permanent regime of the load path within a Mill.
● Data processing with an own application that uses equation (5).
● Multivariate statistical analysis for tables with three dependent variables and application
of the response surface method for tables with two dependent variables.
● Determining the function for the Angle.
● Determination of correlation of the overall power model of Hogg & Fuerstenau adjusted
with real values.

As explained above, the Hogg & Fuerstenau power model of the reference [1], is based on
approximations to the actual behavior of the load, as shown in Figure 1. However, it has been
revealed through simulations with numerical methods and operational observations, that the load
inside the mills has a different behavior, as shown in Figure 2. For this reason, it is relevant to obtain
a transfer function between the power model functions of equation (2) and the power model
equation (3).

Figure 1. Load distribution for Hogg and Fuerstenau general power model

.
𝑃 = 0.238 ∗ 𝐷 ∗ 𝐿 ∗ 𝑁𝑜𝑝 ∗ 𝜌 ∗ (𝐽𝑐 − 1.065 ∗ 𝐽𝑐 ) ∗ 𝑠𝑖𝑛 (𝛽)

(1)
𝑃 = 𝐹 𝐷, 𝐿, 𝑁𝑜𝑝, 𝐽𝑐, 𝜌 ∗ 𝑠𝑖𝑛 (𝛽)

3
(2)

This transfer function therefore turns out to be the inverse function indicated in the equation (5).

Figure 2. Load distribution for real power model

𝑷𝑪 = 𝑮 𝑫, 𝑳, 𝑵𝒐𝒑, 𝑱𝒄, 𝝆𝒂𝒑

(3)
𝑭 𝑫, 𝑳, 𝑵𝒐𝒑, 𝑱𝒄, 𝝆𝒂𝒑 ∗ 𝒔𝒊𝒏 (𝜷) = 𝑮 𝑫, 𝑳, 𝑵𝒐𝒑, 𝑱𝒄, 𝝆𝒂𝒑

(4)
𝑮 𝑫, 𝑳, 𝑵𝒐𝒑, 𝑱𝒄, 𝝆𝒂𝒑
𝒔𝒊𝒏(𝜷) =
𝑭 𝑫, 𝑳, 𝑵𝒐𝒑, 𝑱𝒄, 𝝆𝒂𝒑

(5)

Hogg & Fuerstenau General Power Model Adjusted

The classic Hogg & Fuerstenau power model for mills is not only needed to know the angle β, for
the determination of energy demand, but also an adjustment to the effective design length is
required, which is otherwise , only a geometric variable, as can be deduced with the support of
figure 3.

4
Figure 3. Load distribution for real power model

Determination of the Angle and Long Effective Le

The equation for the angle β of the model is finally represented in equation (6), with the highest
2
coefficient of determination (r = 0.994), and the smallest standard or residual error (Yi-Y = 0.75
degrees) over a total of 87 functions developed for the study from multivariable regressions.

Equation (7) corresponds to the effective length necessary to include and improve the Hogg &
Fuerstenau model, thereby shortening the use of losses to those of the electrical system.

𝟏𝟔𝟔. 𝟔𝟐 ∗ 𝑱𝒄𝟏.𝟑𝟐 ∗ 𝑷𝑪𝑺 𝟎.𝟏𝟓


∗𝜶 𝟎.𝟎𝟒
𝜷= 𝟎.𝟎𝟎𝟎𝟏𝑱𝒄 ∗𝟏.𝟑𝟑𝑷𝑪𝑺
𝟎. 𝟎𝟑

(6)
𝑳𝒆 = 𝑳𝒔 + 𝑱𝒄 ∗ 𝑳𝒉

(7)

5
RESULTS AND DISCUSSION

The proposed equations were used with measurements for a complete liner campaign in the same
SAG Mill, but based on daily average values, understanding that there is an important error in the
form of measurement, a significant correlation coefficient of 71.8% as shown in Figure 4 could still
be observed. All calculations were adjusted in the nominal diameter and angle of the lifter according
to reference constant wear rates adopted.

Figure 4 Daily average observed vs predicted power draw using β and L Adjusted on Hogg & Fuerstenau
model

CONCLUSION

The equations presented allow to improve the general power model of Hogg & Fuerstenau, thus
transforming it, into a tool that facilitates the control and management of the operational assets of
comminution of a plant, without forgetting that the present model only corresponds to a
deformation of the behavior of the load inside a mill.

NOMENCLATURE

PHF Hogg & Fuerstenau general power model, kW


PC Cerda’s power model, kW
Jc Total load level (from 0.1 to 0.5)

6
View publication stats

Jb Total ball level (from 0.1 to 0.5)


3
ρap apparent density of mill internal load, kg/m
PCS Critical speed percentage (from 0.4 to 0.8)
α lifter half angle, (from 20 degree)
β Center of mass angle at Hogg&Fuerstenau model, (from 20 degree)
D effective internal diameter of the mill, m
Ls Internal shell length of the mill, m
Lh Internal head length of the mill, m
Le Effective internal length of the mill, m
Nop Mill operating speed, rpm

REFERENCES

[1] Pontificia Universidad Católica de Chile Manual of SIMULA v1.0, School of Engineering – Mining
Center

You might also like