Grinding and classification performance evaluation and modelling of an industrial-scale horizontal roller mill process

1 Introduction

Cement is conventionally produced in multi-compartment type ball mills which is a energy intensive process [3]. Typical energy consumption of the grinding stage of cement production is about 40 kWh/t in a cement plant [4, 5]. In this respect, grinding technologies such as vertical roller mills, high pressure grinding roll crushers (HPGR) and the Horomill are used in finish grinding of cement in cement plants in the recent years as they are known to be energy efficient systems. Grinding is achieved by high compression in these systems. The energy efficiency of these systems was compared with ball mill technology in the literature [6]. Research results have shown that, high compression grinding mill circuits are more energy efficient than conventional single stage multi-compartment ball mill circuits [6]. Among the high compression milling technologies, Horomill (Horizontal roller mill) is a ring roller mill which is a joint development by the French plant manufacturer FCB and the Italian cement producer Fratelli Buzzi [7]. The Horomill mechanically combines the elements of a ball mill such as a cylindrical shell on hydrodynamic shoes, drive gear rim and components from HPGR such as roller and bearings but operates at much lower grinding pressures than a HPGR.

The operating principle of the Horomill is shown in Figure 1.The Horomill consists of a horizontal shell equipped with a grinding track in which a roller exerts grinding force. The shell rotates faster than the critical speed which leads to centrifuging of the material. The cylinder is rotated about 1.5 times the critical speed. The main feature is the roller inside the shell which is rotated by the material on its shaft without a drive. Material is fed to the mill by gravity. There are scrapers located in the upper part of the shell. Scrapers cover the entire length of the mill and scrape off the material which falls onto the adjustable panel of the material advance system. Grinding pressures change within a range of 500 to 800 bars. Concave and convex geometries of the grinding surfaces lead to angles of nip two or three times higher than in roller presses which resulted in a thicker layer of ground material [8]. The Horomill mainly consists of three zones: feeding, grinding, and discharging. The cylindrical roller transfers the grinding power onto the material in the grinding zone. The material bed in the mill is generated by the centrifugal effect.

Horomills are used for both finish grinding and for pre-grinding of cement or ores. The energy saving is similar to that of high pressure grinding rolls (HPGR) but 30-50 % lower when compared with a ball mill [10]. 50 % energy reduction and grinding up to 20 % moisture content could be achieved in Horomills as compared to ball mills in raw material grinding [11]. The Horomill consumes 15 % less energy as compared to a process combination of a hybrid HPGR and a multi-compartment ball mill grinding system in cement production [12].

TSV high efficiency dynamic air separators (air classifiers) are operated in closed circuit with the mill in Horomill grinding circuits. TSV type air classifiers are used in classifying cements, cement raw meals, coals and minerals. They can be used in closed circuit with a ball mill, a vertical mill or a Horomill to attain the control of the fineness of the finished product. They are characterized with handling of high loads [13].

In this study, grinding and classification performance of the Horomill process was determined and evaluated. Size reduction performance was compared with the size reduction performance of an industrial-scale two-compartment ball mill operating with a HPGR pre-crusher unit in the circuit. The Horomill size reduction performance was found to be higher than the open circuit HPGR plus the two-compartment ball mill process with the same cement production type. Horomill grinding was modelled by using the perfect mixing modelling approach. Ratio of specific discharge rate to specific breakage rate function was determined to characterise the grinding performance of the mill using the compressed bed breakage function. Modelling results provided a characteristic perfect mixing mill model parameter variation to predict mill product size distribution. Classification performace of the TSV air separator was characterised by using the efficicency curve approach [2] and the classification performance was found to be sufficiently high.

2 Methods

2.1 Plant-site studies

An extensive sampling study was conducted around the Horomill circuit given in Figure 2 at the steady-state conditions during CPP-30R (puzzolanic Portland cement) type cement production. Mass balanced tonnage flowrates and fineness as 45 µm residue % were also shown in Figure 2. Design parameters of the Horomill and TSV air separator operating in the circuit are given in ­Table 1. Steady state conditions were verified by examining the variations in the values of control room variables. Important variables of the operation were recorded every 5 minutes in the control room and average values of the control room data were used in the mass balance calculations (Table 2). Sufficient and representative samples were collected from the sampling locations around the circuit when the steady state conditions were achieved. Sampling locations are given in Figure 2. Collected sample amounts are summarized in Table 3.

2.2 Laboratory studies

Samples from the plant-site were conveyed to the laboratory for particle size distribution analysis. Representative samples were collected by using a riffler for the size analysis. Samples were screened from the top particle size down to 149 µm by using a ro-tap. A sub-sieve sample of 149 µm was sampled for size analysis by using a Sympatec laser diffractometer in dry mode. Size analysis was conducted from the top size which is 76.1 mm down to 1.8 µm. Ro-tap and laser sizing results were combined mathematically to define the whole particle size distributions.

3 Results and discussions

3.1 Mass-balancing

Mass-balanced tonnage and particle size distributions were estimated by using the JKSimMet Steady State Mineral Processing Simulator. Humidity measurement was taken and the dry tonnage value for the puzzolan feed was calculated and used in the mass balance calculations. Humidity content of the sample was determined as 16.61 % for the puzzolan feed. Mass-balance results indicated that, the measured and calculated values were in a good fit, thus the data are reliable to use in performance evaluation studies. Statistically corrected particle size distributions as compared to the experimental particle size distributions are given in Figure 3 and Figure 4. Sufficient agreement between the mass balanced and experimental particle size distributions also indicated the success of the sampling study. Thus, mass-balanced tonnage and particle size distributions were used in performance analysis and modelling study.

Typical operational results from the investigated Horomill grinding process are compared with the industrial scale two-compartment ball mill grinding process and results are given in Table 4. Horomill production configuration has resulted in energy savings of 50 % as compared to the ball mill which is operating with a High Pressure Grinding Rolls (HPGR) pre-crusher in open circuit at the same production type. Horomill size reduction performance was also found to be higher than the two-compartment ball mill (Table 4).

3.2 Compressed bed breakage function

Piston-Die test was used to estimate mill feed breakage characteristics. The Piston-Die breakage set-up is shown in Figure 5. Breakage of particles was achieved by the compression mechanism in the Piston-Die set-up. Particle size fractions of -19+16mm-11.2+9.5mm, -9.5+8mm were prepared and certain amounts were placed in the die for compression loading. Breakage of different particle size fractions were achieved under different specific energy consumption levels. t₁₀ – t_n family curve approach [14] was used to determine breakage function of mill feed. t₁₀ breakage index values were determined from the particle size distributions of the compressed bed breakage products to characterize the breakage product fineness. t₁₀ was the particle size distribution parameter and defined as the cumulative passing % amount of 1/10^th of the broken particle size fraction. Other particle size distribution parameters were defined by the t₂, t₄, t₂₅, t₅₀ and t₇₅ values. Piston-Die test conditions and experimental results are tabulated in Table 5. tn values as a function of t₁₀ values were plotted to obtain t₁₀ - t_n family curves which are given in Figure 6. Regression relations of t₁₀– t_n curves were used to reconstruct the breakage function of mill feed required in the estimation of the mill product and specific breakage rates. Regression equations for the estimation of breakage functions (appearance functions) were given in Table 6. Estimated compressed bed breakage function is graphically shown in Figure 7.

3.2.1 Horomill modelling approach

Size reduction models are based on the concepts of [15]:

Probability of breakage which is called the selection or breakage rate function.

Characteristic size distribution after breakage which is called breakage distribution or appearance function.

Differential movement of particles through or out of a continuous mill which is generally size dependent and called classification or discharge rate function.

The perfect mixing modelling approach was developed by Whiten [16] and a limited number of researchers have used this model in modelling of cement mills [3, 17, 18, 19]. If a suitable breakage function can be defined, the perfect mixing mill model can be used to describe any grinding device as stated in the literature [2]. According to the perfect mixing modelling approach, the discharge of i^th size fraction from the mill can be given by Equation 1.

d_i=  p_i––    s_{i  ⇥}(1)

s_i is the mass of size fraction i (tons) in the mill hold-up (S) as tons, p_i is the mass flow rate of particle fraction i out of the mill as product. It is difficult to measure mill hold up (S) in wet grinding mills. If S could be the determined discharge rate, d_i could be directly calculated from Equation 1. Thus, ball mills used in the wet and dry grinding modes were modelled by back calculating the ratio of breakage rate to discharge rate (r/d) in the literature [2, 17, 18, 19, 20, 21, 22]. Conventionally, the ratio of (r/d) defined as the perfect mixing mill model parameter and was back calculated from Equation 2.

    _i      r_j         r_i

f_i + S a_ij p_j(–– ) – pi (–– )– p_i = 0 ⇥(2)

    _j=1       d_j         d_i 

f_i: mass flowrates (t/h) of size fraction i in mill feed

p_i: mass flowrates (t/h) of size fraction i in mill product

a_ij: breakage distribution function (appearance function)

r_i: specific breakage rate of size fraction i (ton broken per hour per ton in the mill which is h^-1)

d_i: specific discharge rate of size fraction (i) (ton discharged per hour per ton in the mill which is h^-1)

In this study, the Horomill was considered as a perfectly mixed single tank. Mass balanced feed and product size distributions in addition to the mill feed tonnage, and experimentally measured compressed bed breakage function were used to estimate r/d model parameter. Estimated typical variation for the model parameter r/d function for the prediction of mill product size distribution is given in Figure 8. Typical variation of r/d function indicated a characteristic shape from the top size down to 0.0018 µm depending on the breakage behaviour of particles with different breakage characteristics and textural properties [23, 24, 25].

3.3 High efficiency TSV air separator performance

Classification performance of the high efficiency air separator in the circuit was evaluated by using the efficiency curve approach [2]. The efficiency curve describes the proportion of a given size of solids which reports to the coarse product. Actual and corrected efficiency (Tromp) curves of the air separator are given in Figure 9. Efficiency curve parameters are tabulated in Table 7.

The d₅₀ size corresponds to 50 % of the feed passing to the coarse stream. It is therefore, the size which has the equal probability of passing to either coarse or fine streams. When this size is decreased, the fineness of the product increases. The operational parameters that affect the cut size are rotor speed and separator air velocity. The percentage of the lowest point on the tromp curve is referred as the by-pass. It is the part of the feed which directly passes to the coarse stream (separator reject) without being classified. By-pass value is a function of the separator ventilation and separator feed tonnage. Fish hook effect (b) is the portion of fines returning back into separator reject stream. When there is incomplete feed dispersion at the separator entry, or even within the classification zone, aggregates of fine particles may be classified as coarse particles and thus report to the coarse stream. The sharpness of separation (k) was defined as d₂₅/d₇₅ (d₇₅: particle size whose 75 % is reported to the separator reject; d₂₅: particle size whose 25 % is reported to the separator reject). The range of this parameter k (acuity) depends on the type of separator. Usually for TSV it is between 0.55 to 0.7. When the normal range for sharpness (k) parameter is considered, it can be said that, it is not in the normal range. Corrected tromp defines only the classified part of the tromp curve. d_50c is the corrected cut size and it is defined as the separating size in the classified fraction. The global by-pass value for the TSV high efficiency air separator is 0 – 20 % [13]. The by-pass value was determined to be quite high. It can be concluded that, TSV separator is classifying efficiently. However, performance of TSV could be improved further by adjusting the operational parameters.

4 Conclusions

In this study, the grinding and classification performance of the Horomill process was determined. Horomill was modelled by using the perfect mixing modelling approach using compressed bed breakage function. Grinding behaviour was characterised by the estimated ratio of the specific discharge rate to the specific breakage rate function. Classification performance of the TSV air separator was found to be sufficiently high. 50 % energy saving could be achieved as compared to conventional two-compartment ball mill grinding where the HPGR is operated in open circuit as a pre-crusher unit with the same cement production type. Research results provided some basic information for model building studies.

Nomenclature

i: particle size fraction i

j: particle size fraction j

x: particle size fraction

f_i: mass flowrate of size fraction i in mill feed (t/h)

p_i: mass flowrate of size fraction i in mill discharge (t/h)

r_i: specific breakage rate of size fraction i (h^-1)

d_i: specific discharge rate of size fraction i (h^-1)

a: single column step triangular breakage function matrix

s_i: mass of size fraction i (t)

r/d: ratio of specific breakage rate to specific discharge rate

Acknowledgements

The author would like to gratefully acknowledge the staff of the Moctezuma Tepetzingo Cement Plant (Mexico), Prof. A. Lopez Valdiviesso, Dr. Juan Luis Bahena, Julio Alberto Palacios Resendiz, Dr. Cristobal Alberto Perez Alonso from the Instituto de Metalurgia at the Autonomous University of San Luis Potosi (UASLP), and Prof. A. Hakan ­Benzer from Hacettepe University.