Using real-time particle size analysis for optimisation

Paul Boughton

Alberto Ferrari and David Pugh describe how on-line particle size measurement was used to optimise the development of an innovative milling circuit designed to deliver exceptional performance characteristics for an Italian granulates and powders manufacturer.

On-line analysis is frequently an add-on to an existing plant, a tool that is used to enhance efficiency as the process matures. Historically, there has been little alternative because of the limited availability of reliable technology. For some variables, developing robust process systems remains a challenge, but for others, particle size being a prime example, the situation is very different. Integrated solutions for on-line particle size measurement can now be designed, installed and commissioned in the same way, and at the same time, as other instrumentation.

When particle size is a critical parameter – as it is for many solids handling processes – access to real-time size measurement is of great value when commissioning. Problems can be analysed and solved quickly, and optimal operation rapidly achieved. Use of continuous analysis during the process development cycle is also productive, providing comprehensive information for design and effective scale-up.

Project scope and background

Based in Verona, Italy, Ferrari Granulati is a leading producer of marble granulates and powders, primarily for the construction industry. In the past, the company has focused mainly on the production of marble chips but recently there has been a move towards the grinding of finer powders. The company has exclusive rights to a Greek marble of unsurpassed whiteness which can be milled to form high quality products for the paints and plastics industry (Fig. 1).

One of the white marble powders has a size specification of 0.1-0.4 mm. There is a lucrative market for finer materials produced from the associated 0-0.1 mm ‘waste’ stream, but until relatively recently the company did not have the technology for further processing. In 2004 the decision was made to invest in a milling circuit to produce powders with a Dv50 of 3-8 microns from the 0-0.1 mm stream.

Dv50 is the median particle size based on a volumetric particle size distribution, in other words the particle size below which 50 per cent of the population lies. For the paints and plastics applications of interest, size distribution is also important so Dv98 is specified too. The target range is 15-50 microns since a narrow particle size distribution is preferable. Overall production rates for the process are between 15 and 25 000 tonnes per annum.

The company considered both ball and vertical roller mills when designing the new process. Ball mills are widely used within the minerals processing sector, and are suitable for this application, but energy consumption and footprint are a major disincentive. In the cement industry, vertical roller mills are becoming increasingly popular as they are relatively compact, give tight particle size control and, in comparison with ball mills, use significantly less energy. However, they are not generally used to reduce Dv98 below 50 microns. Selecting a vertical roller mill for this application would mean developing a novel design to extend the technology beyond its normal operating window. The decision was taken to proceed along this route.

Another Italian company, Varese-based STM, was selected as the mill manufacturer and some preliminary trials were carried out at their site. A mill was then specified and installed, with a downstream classifier, to provide a basic system that could be optimised to meet performance targets. The classifier splits the exit stream from the mill to give a closely defined product, and a coarser fraction that is recycled for further grinding. To this extent the system is a small-scale version of those routinely found in the cement industry.

Recognising that real-time particle size measurement would speed up process modification and optimisation, Ferrari Granulati installed a continuous analyser, along with the mill. An Insitec on-line laser diffraction system from Malvern Instruments was selected.

Laser diffraction is a well-established technique for particle size measurement, widely used in a range of industrial sectors. It is a rapid and robust method suited to the process environment. Systems are available for at-, in- and on-line use, to complement conventional off-line laboratory analysis. This means that the same technology can be used in the lab, on a pilot plant, and on full-scale processes, facilitating data comparison and scale-up. The attractions of the Insitec were its measurement rate, excellent reliability record and rugged design – it was clearly a unit designed for the process arena rather than a modified laboratory instrument.

Installation and commissioning

The on-line particle size analyser was quick and easy to install and commission, and worked well from the start. It sits at ground level (Fig. 2), a booster compressor transporting material from the sample point to the measurement zone. Fully automated measurement takes place four times every second and all results are transferred directly into the control system. The measurement range of the instrument is 0.1-1000 microns so it comfortably captures complete particle size distribution data for the product. Instrument reliability is very high while maintenance requirements are minimal.

One issue that did require resolution at the commissioning stage was optimal positioning of the sampling probe. With this system, sample is extracted using a flute, a pipe with holes along its length that catches material across the process line (Fig. 3). This ensures that the sample is representative even if particle size is variable across the pipe diameter. Typically this flute is installed in a process line where material is flowing downwards under gravity. In this case it was initially installed on the line from the classifier to storage, which worked well in terms of sample quality.

However, sampling at the end of the process introduces a time lag when the area of focus is the mill. Experiments were conducted to see if the probe could be installed further upstream to eliminate this minor delay. It is now positioned within the extractor that takes material from the classifier to the filter. On the extractor discharge line there is segregation, but within the extractor flow is turbulent, and the sample is well-mixed. Even so, it is not entirely representative so there is a small but constant offset between off- and on-line data. This is corrected using an algorithm developed for the application, to ensure consistency between the two sets of measurements. The responsiveness of sampling at this point more than compensates for this minor inconvenience. The system tracks changes in particle size very quickly, making it possible to control the mill extremely tightly.

Mill design

Access to real-time particle size data made it easy to see the impact of hardware and process changes, as work was carried out to develop a circuit that would deliver the required product. Innovative modifications were made to tailor mill design to the exacting demands of the application. The mill used has three rollers and a table diameter of one metre.

Air flowing through the milling circuit has two roles. Firstly it lifts powder that falls from the mill table back on for further compaction, and secondly it transports material from the mill to the classifier and filter. Table 1 shows how the air flow required for transporting material to the classifier changes with Dv98.

As the mill is used to produce finer material, throughput drops, as does the air flow required for transport. However, the amount of air needed to recycle material that falls off the table, back onto it, is independent of particle size. This duty consistently requires an air flow rate of around ten thousand cubic metres per hour. When the material being produced has a Dv98 of 50 microns, the air flow required for transport is also around ten thousand cubic metres per hour. For this size of particle then, the air required for the two different functions – mill operation and powder transport – is roughly similar. But, for finer materials this is not the case. As Dv98 decreases, the air flow needed for transport becomes much less than the amount needed to maintain material on the table. For a Dv98 of 10 microns only three thousand cubic metres per hour of air are required for transport.

To avoid the energy implications associated with unnecessarily high air flows, a mechanical powder recycle was developed. Material that falls from the table is collected and fed to the classifier using a screw auger system. Air flow is set on the basis of the amount required for transport to the classifier, minimising energy consumption, particularly for the finer products. For each grade, volumetric air flow, rather than fan power consumption, is maintained at a constant value making the feed to the classifier extremely consistent. This gives the circuit intrinsic stability, making it easier to ensure that particle size remains within the defined specification.

An additional issue associated with processing finer materials is the problem of agglomerate formation. As particle size decreases, particularly below 10 microns, the strength of inter-particle forces increases, encouraging cohesion. In the classifier, any agglomerates formed will be separated out, and returned for further milling, so this problem can result in internal recycling. This has a negative impact on energy consumption and throughput. An additive – a glycol-water mixture – is introduced into the mill with the raw material to reduce electrostatic charge and discourage this behaviour, but this does not completely solve the problem. An in-line pin mill has however proved extremely beneficial in eliminating agglomeration.

The pin mill sits in the line between the mill and the classifier and rotates at around 300-400 rpm. Since the mill has an external diameter of only 500 mm, energy input is relatively modest, but tests show that it is sufficient to break up any agglomerates. This improves energy efficiency, maximises throughput and avoids excessive wear in the mill.

In a vertical roller mill material on the rotating table is crushed as it passes beneath compacting rollers that press on the powder surface. The centrifugal force induced by rotation draws material from the centre of the table out to the edges. For coarser materials this system works without any further control but with finer powders there is a tendency for material to simply flow out from under the rollers without being broken up. This is one of the primary reasons why the technology is not used to reduce Dv98 to less than 50 microns.

To solve this problem, a simple mechanical transducer was installed on the arm of the compaction rollers, to monitor the depth of powder on the table (Fig. 4). As the level of material varies, the transducer moves up and down, so this calibrated system provides reliable indicator that is used to automatically control the speed of rotation of the table. If the depth of powder falls too low then rotational speed is reduced, and vice versa. An optimal set point for powder depth is 15 mm and this usually gives rise to a table rotation speed of between 25 and 30 rpm. If powder depth falls below this level, then although the efficiency of compaction increases, vibration levels unacceptable.

Using this fully optimised milling circuit Ferrari Granulati manufactures three powders with a Dv50 of 3, 5 and 8 microns respectively. The quality of these products is very high thanks to the superior colour of the base material combined with their consistency and extremely tight particle size distribution. The strong focus on energy consumption has paid dividends, and production costs are low. Power consumption per tonne is a prime concern, and has been minimised for each grade. It is estimated that the relatively high cost of the vertical roller mill, compared with a ball mill, will be offset by energy savings alone within 5-8 years. This figure may fall even lower with rising energy costs.

For each product there are defined set points for classifier speed and return flow rate to the mill. Fresh feed rate is automatically controlled from classifier recycle to maintain a consistent total flow. When the finer materials are being produced the recycle from the classifier is higher, so the fresh feed rate will fall. Compaction pressure can be varied but is usually kept at a level of 40-50 bar. Air flow rate through the system is maintained at a constant rate (as described above). The Dv50 and Dv98 of the final product is continuously monitored, an off-line check being carried out once for each silo (every 60 tonnes).

The process is very stable and responsive. It runs every night, when electricity is cheaper, but can operate 24/7 when demand is high. No manual intervention is required other than initial inputting of the set points for the grade. This can be done locally in a dedicated control room or remotely. Fig. 5 shows how quickly product changeover is achieved.

Switching to the production of a coarser material is instantaneous since any residual fines can be tolerated in the new final product. A switch to a finer product takes a little longer since material must be swept from the lines but the transition is still complete within minutes. During this period product remains routed to the silo collecting the coarser material, so there is no waste or re-work.


Continuous on-line analysis is a valuable tool at any time, but particularly during development and commissioning when it can accelerate the learning process. Real-time data allows rapid identification of the nature of problems and promotes the development of effective solutions. It becomes much easier to move swiftly towards fully optimised operation.

Particle size analysers for the process arena are now well-established and can be routinely specified, installed and commissioned in the same way as other process equipment. The data they provide is beneficial for a wide range of powder processors. The case study outlined here shows how real-time analysis provides information for product quality optimisation and variable cost reduction.

Table 1. Impact of Dv98 on production rate and air flow required for transport around the circuit

Dv98 (μm)    Production rate
(t/h)    Air flow required for transport to the classifier (m3/h)
80    6.0    20 000
30    2.0      6000
10    0.5      3000

Alberto Ferrari is Production Manager with Ferrari Granulati and David Pugh is UK & Ireland Sales Manager, Process Systems, Malvern Instruments.

Recent Issues