Optimization of crushing stage using on-line speed regulation on cone crushers

Erik Hulthn, M.Sc. and C Magnus Evertsson, Ph.D. Department of Applied Mechanics Chalmers University of Technology, SE 41296 Gteborg, Sweden

Abstract
Cone crushers are used in the mineral, mining, and aggregate industry for fragmentation of rock materials. Control systems for cone crusher settings are widely used to compensate for wear and to protect the machines. The eccentric speed in a cone crusher affects the number of compressions the material is exposed to and thus the particle size distribution of the product. Eccentric speed also affects crusher capacity. However, the speed of the crusher is mostly fixed, because speed changes must be made by shifting pulleys, a process which is quite timeconsuming. By applying a frequency converter, which has decreased in cost during recent years, it has become possible to adjust the speed continuously. By also applying mass-flow sensors to the process, the crusher can be run optimally to yield the most sellable products in a given moment. Existing systems normally only protect the machine. The sensors can be mass-flow meters, e.g., conveyor-belt scales. To analyze data from the process and calculate the appropriate value for the speed, an algorithm has been developed. The developed algorithm was tested and evolved at a crushing plant for aggregates that produces around 350 kton a year. The algorithm was implemented in a computer that could communicate with the frequency controller, retrieve data from four mass-flow meters, and also interact with the operator. The computer was reachable over the Internet for the researchers at Chalmers and the algorithm was tuned and improved on-line. The result is a crushing stage that consistently and maximally produces the most desirable product. The performance (yield of required products) increased by 4%. As a bonus, the lifetime of the mantles increased 27% on the evaluated crusher. Key words: cone crusher, speed, crushing plant control, on-line optimization

Introduction
Cone crushers are widely used for size reduction of rock materials, such as aggregate products or ores, into finer fractions. Their main operating principle is the same today as when developed, a century ago. As the mantle and the concave get worn, the distance between them must be adjusted in order to maintain the reduction ratio and control the top size and particle size distribution of the product. Control systems for cone crushers, introduced 20-30 years ago, are widely used to compensate for wear and to protect the machines. The speed of a cone crusher can typically be adjusted by shifting pulleys of the belt drive. This is labor- and time-consuming, and is therefore not done unnecessarily. Besides, most crushing plants have no accurate and reliable way of measuring the changes in production that the speed change implies. Taking belt cuts is one measurement strategy. However, this is very time-consuming, it requires additional sieving, and in the end it is still just a single sample that must be combined with further samples for accurate assessment.

C.1

However, the eccentric speed has a great impact on the product of a cone crusher. The speed affects the number of compressions that the material is exposed to and thus the particle size distribution of the product. Similarly, the speed also affects the shape of the product. However, this issue is beyond the scope of this work. Capacity is also relatively strongly affected by the speed. Figure 1 shows a comparison between different closed side settings (CSS) and speeds. While the CSS moves the product cumulative particle size distribution curve horizontally, the eccentric speed tends to rotate it.

Figure 1. Particle size distribution of the product from a crusher with different parameters, varying a) the closed side settings and b) the speed. By applying a frequency converter, it has become possible to adjust the speed continuously. Frequency converters have decreased in cost during the last years, making them more available for use in standard cone-crushing operation. Using sensors to gain feedback on the process, the crusher can be run optimally at each moment. A conveyor-belt scale, for instance, can easily retrieve and transmit information about the mass flow to a computer. A cost-effective variant which monitor the current capacity by measuring the power draw on a conveyor belt that is performing a lifting work can also be used. Such a sensor can be obtained for about one-tenth of the cost of a traditional belt scale. This principle is described by Hulthn and Evertsson in [3]. By directly monitoring the sellable products after a screen, the control system can control the plant output particle size distribution (Figure 2). Quality aspects, e.g. shape, are beyond the scope of this paper.

C.2

a)

b)
% C Product size

Figure 2. a) A mass-flow meter, e.g. a conveyor-belt scale, is attached to each stream from the screen. The computer then estimates b) the particle size distribution (PSD). The gray line is the real PSD the black is the PSD estimated from the measurements, the red boxes.

Earlier work
In the 1970s, Karra [4] performed a large number of tests on cone crushers with different parameter settings, including various eccentric speeds. However, those tests showed no significant effect from the eccentric speed on the particle size distribution, nor on the capacity. The lack of effect might be attributable to any of a number of reasons, one being that the effects from other parameters were much larger. The lack of an accurate procedure for long-term evaluation is another explanation. However, there are models where the eccentric speed affects both the particle size distribution and the capacity, such as the work by Evertsson [1]. Nonetheless, studies to find the best constant speed for cone crushers are rare. However, regulating the crusher continuously during the operation is even rarer. No reference regarding online speed optimization has been found.

Speed-selection algorithm
Control of eccentric speed is motivated by the non-fixed conditions under which a crusher operates. The crusher can be operated at different eccentric speeds in order to compensate for wear (when there is no crusher setting control), raw material variations, and wear-related crushing chamber changes during its lifetime. The main assumption here is the existence of an optimal speed setting at each moment (Figure 3). Since several factors vary over time, a fair comparison between two different settings at different times cannot be easily conducted. When determining the best setting for the speed, manual step changes are therefore usually performed. The performance of the crushing plant before and after the change can then be compared at fairly equal conditions.

C.3

Figure 3. At each moment, it is assumed that there is an optimal setting of the speed.

A Finite State Machine (FSM) algorithm has been developed. A simple variant of FSM called the Mealy machine [5] is used, in which actions only take place on the entry of a state. The exit of a state is conditioned but brings no actions. The FSM was developed manually (in contrast to a computer-generated algorithm) to find a speed close to the optimal value and to stay at that speed for a time period. The structure of the developed FSM is shown in Figure 4. The introduction of the FSM also enabled high current (amps) to be taken into account when computing the next setpoint.

FSM

1
Operate a short while

2
Reduce Speed

Short Long

Time?

5
Increase Speed

Operate a short while

Operate a short while

Better?

No

No

Better?

Yes

Yes

3
Reduce Speed

4
Increase Speed

7
Reduce Speed

6
Increase Speed

Operate a short while

Operate a short while

Yes

Better?
No

Better?
No

Yes

Figure 4. A Finite State Machine is used in the algorithm to calculate speed setpoints.

C.4

Full-scale tests
The developed speed-control algorithm was tested at a crushing plant in Ludden, 20 km south of Norrkping (Swedish east coast), owned by Sand & Grus AB Jehander, a Swedish subsidiary of Heidelberg Cement Group. The crushing stage, where the tests have been conducted, has a full range of profitable products, namely aggregates 2-5 mm, 5-8 mm, 8-11 mm, 11-16 mm, and 16-22 mm in size and used for tarmac, concrete production, winter road maintenance, drainage, etc. The crushing stage is shown in Figure 5. The tertiary crusher is a Nordberg HP300 equipped with a short head mantle. The feed to the tertiary crusher is a blasted and two-stage pre-crushed granite 32-100 mm in size. The crusher is not equipped with any regulations system for the closed side setting. Instead, this must be adjusted manually. To do this, the operator runs a hydraulic motor that turns the top shell of the crusher, which has threads and is holding the concave.

Figure 5. Tertiary crushing stage at Jehander's plant in Ludden. The plant has six profitable products. The oversize particles are re-circulated to the crusher.

The crusher is driven from a 4-pole asynchronous motor, which normally runs at 50 Hz. At this standard speed on the motor (1500 rpm), the eccentric speed in the crusher is about 340 rpm with the present pulley configuration. The speed of the crusher is controlled by a frequency converter, Control Techniques Unidrive SP, which can supply power up to 200 kW. The drive has a serial interface which allows communication with a computer. The conveyor belts which convey feed to the crusher, product from the crusher, fine products between the two screens, and the circulating load all have mass-flow meters monitoring the electrical power draw. The power is measured with power transducers (Carlo Gavazzi WM-12 DIN) which can deliver information to a computer, where the actual capacities are calculated. An industrial computer is placed near the electrical cabinet in order to monitor the material flows and execute the algorithm. The computer enforces the setpoint speed to the frequency converter. The computer communicates with the power transducers and the frequency converter over an RS-485 wire. The computer has a TCP-IP connection, which makes it possible both to supervise and to control it remotely from the operators maneuver cabin, from the companys head office, and from the university 330 km away, via a web interface. The algorithm was tuned, i.e. some constants were adjusted, on site at the crushing plant in Ludden. The CSS is typically adjusted four times during a day to compensate for wear. The operator uses the current meter to determine when the CSS is sufficiently small. With the configuration described above, the crushing stage has a typical production capacity of about 150-240 tons per hour (tph) depending on the crushing chamber geometry (due to wear).

C.5

At the time of the tests, the total mass output from the plant was the important issue. Therefore, the total mass flow out of the crushing stage is taken as the performance value of the crushing plant. The signals are sampled at approximately 1 Hz (the power transducers themselves sample more frequently to avoid aliasing) and are low-pass filtered to mass-flow values every tenth second. The performance value is calculated every 200th second (t). In order to evaluate the developed algorithm, a comparison between the original speed, the operators choice of speed, and the algorithm speed was set up during a whole mantle lifetime (fall 2007). The operator decided when to use the different methods and selected the speed limits for the algorithm. Hard limits, which were never allowed to be exceeded, were set to 1350 rpm and 1600 rpm, respectively. In order to compare the different methods at fairly equal conditions, the performance values of the methods are compared pair-wise before and after a new method is selected.

Results
The most obvious result, a bit unexpected, was that the lifetime of the mantle increased from about 220 hours to 280 hours (effective operating time). Before the frequency drive was commissioned, liners were changed when the performance value decreased below 150 tph. At the time, no exact measurements were made, but an estimation is shown with a dotted line in Figure 6, where the measured performance from the test period is also shown. The 60 hours of extra liner lifetime is an increase of 27%. However, the comparison between the different methods, below, unfortunately does not show what caused this increase.
300

250

Performance value Poly. (Performance value) Poly. (Original speed -estimated)

200 Performance (tph)

150

100

50

0 0 50 100 150 Mantle lifetime (h) 200 250 300

Figure 6. The performance for the complete mantle lifetime (all three control methods) is measured, while the standard performance (original speed only), is an estimate.

During the test period, the operators switched back and forth between the different modes 26 times. For every switch performance, values were calculated for the method before and after, respectively. These differences are shown in Figure 7.

C.6

Speed reference
210.0 Performance value (tph) 200.0 190.0 180.0 170.0 160.0 150.0 0.0 50.0 100.0 150.0 Mantle lifetime (h)
Figure 7. Comparison of the three different control methods.

1500 rpm Operators' choice Algorithm

200.0

250.0

300.0

To evaluate the complete installation, a comparison between the standard speed and nonstandard speed (both operators choice and algorithm) was made. Switches to and from the original speed were done 15 times. The average absolute improvement was
n

d=

P
i =1 i

Pi c

n
n

= 7.18 tph

(0.1)

The overall relative improvement was

d rel =

P
i =1 i

Pi c

Pi c

= 4.2%

(0.2)

since Pi c = 173 tph. The standard deviation was 6.64 tph. A randomized paired comparison design test, described in [2], was used. According to the null hypothesis, the amount of product produced per hour with the two control methods was the same, 0 = 0 . The standard deviation of the averages was calculated as sd = sd 6.64 = = 1.71 n 15 (0.3)

The value of t0 associated with the null hypothesis S = 0 was

C.7

t0 =

d 0 7.18 0 = = 4.19 sd 1.71

(0.4)

The one-tailed Students t-distribution with 14 degrees of freedom gave

Pr(t 4.19) 0.05%

(0.5)

The test showed the superiority of the crushing operation with either the operators choice of speed or the algorithm speed over operating the crusher at standard speed at a significance level of 0.05%. This means that with 99.95% certainty, the operator/algorithm together was better than the standard speed. The average difference was 4.2%, according to Equation (0.2). Further, the algorithm was 3.7% better than the operators choice, based on 11 comparisons. If a t-test is performed on this comparison, and the null hypothesis is that there is no difference this will be rejected at 0.5% significance level, which means that the algorithm is better with 99.5% certainty.

Conclusions
In this paper, we have shown that several benefits can be gained by installing a frequency converter and by controlling the eccentric speed continuously during operation. First of all, the liner lifetime increased 27%. This could be attributable to several factors, e.g. increased liner wear-hardening and different chamber wear profile. The liners were completely worn out such that the top part of the crusher was turned to its end position, which had never before happened on this specific crusher. Second, the crushing stage total throughput, whose increase was the plants top goal, was increased by 4.2%. A t-test showed the superiority of the crushing operation with either the operators choice of speed or the algorithm speed over crusher operation at a standard speed at a significance level of 0.05%. With an algorithm that automatically chooses the setpoint values, the crusher requires less operator attention. The average improvement when the algorithm chose the speed was 3.7%. However, the limits of the algorithm were set by the operators, who kept the limit range quite narrow. The operators cannot continuously look after the optimal speed, and therefore an algorithm, such as the one presented here, should be used. For the future, this algorithm should be further explored. A crusher with two parameters, speed and CSS, for real-time adjustment would also be of interest.

Acknowledgements
The authors wish to thank the Swedish Mineral Processing Research Association (MinFo), the Ellen, Walter, and Lennart Hesselman Foundation for Scientific Research, and the Development Fund of the Swedish Construction Industry (SBUF) for financial support. Sand & Grus AB Jehander and their personnel in Ludden are gratefully acknowledged for all their support and patience.

C.8

References
[1] [2] [3] Evertsson, C. M., Cone crusher performance, Ph.D. thesis, Chalmers University of Technology, Sweden (2000), pp. 1-49. George E.P. Box, J. S. H., William G. Hunter, Statistics for experimenters: design, innovation and discovery, Wiley-Interscience, Hoboken, N.J., 2005. Hulthn, E. and Evertsson, C. M., A Cost Effective Conveyor Belt Scale, 11th European Symposium on Comminution, The Hungarian Chemical Society (MKE), Budapest, 2006. Karra, V. K., Process performance model for cone crushers, XIV International mineral processing congress, CIM, Montreal, Que, Canada, Toronto, Canada, 1983, pp. III 6.1 - III 6.14. Mealy, G. H., A Method for Synthesizing Sequential Circuits, Bell System Technical Journal, 34 (1955), pp. 10451079.

[4]

[5]

C.9