Cement manufacturers historically have been interested in implementing process-control strategies to aid in the production process. Real-time measurement of Blaine number is of particular interest, since this is the primary measurement that cement producers use to gauge quality and strength of the final product. The three major hindrances to widespread application of in-line Blaine number measurement instrumentation in the past have been:

* Fragility of available instrumentation;

* Inability of existing instrumentation to resolve small changes in material properties; and

* Lack of an effective means to directly measure Blaine number or functional properties of the Blaine number.

Malvern Instruments' Insitec line of in-process particle size analyzers enables active process control in production environments.

Light scattering methodology The process particle measurement technology is based on the now classical ensemble laser diffraction technology, which is widely used in laboratory measurement instruments. The Insitec system consists of a ruggedized sensor head, electronics box, and remote computer. The dimensions of the particle flow access region define the particle-sensing region or sample volume. Particles may pass anywhere along the length of the exposed laser beam.

Particle velocity does not affect the measurement. As particles pass through the laser beam, light scattered in the forward direction is collected by the receiver lens and focused onto a log-scaled annular ring detector. The detector is scanned at high speed by the interface card, which records the signal levels on each ring. These signals are sent over a digital interface and stored for analysis.

Each ring on the detector measures total signal intensity. Each particle scatters light on all the rings of the detector. Therefore, the scattered light is the summation of all the light scattered from all the particles. Once a significant number of detector scans is acquired, the software uses a non-linear inversion technique to solve for the relative particle concentration. The instrument matrix for the solution is defined by the theoretical scattering dependent on the refractive index of the particles and carrier. No assumptions are required about the shape of the size distribution, obtained directly from the experimental scattering information.

For cement applications, a standard process instrument (Figure 1) is hardened for the corrosive and abrasive nature of the material. The original application of this technology was in process environments involving high-value products, such as toners and pharmaceuticals. The primary issue that had to be addressed in developing an instrument specifically for the cement industry was mechanical wear of the surfaces that were exposed to this highly abrasive material.

After testing different designs under a range of operating conditions, acceptable component wear life was achieved by incorporating abrasion-resistant ceramics in the high-velocity sections of the flow system coupled with a combination of hardened and tungsten-coated steel in other, less abrasive locations. This configuration has eliminated mechanical wear in the regions that could not tolerate abrasion, and has provided acceptable lifetimes (greater than two years) for the low-cost, replaceable components.

The instrument is affixed to the process stream via an auger feeder, which moves a representative fraction of the primary flow from the main process stream to a secondary stream, and then to a venturi educator driven by compressed air that moves material from the secondary flow to the sensor head. The venturi functions as both a pump and a dispersion device to ensure that the powder is completely de-aggregated. The material passes through the optical sensor and is returned to the process stream. Installed, the instrument is integrated as part of the process line.

This method for interfacing the device to the process flow achieves a uniform and representative particle flow to the detector. Automatic valves, controlled by the instrument software are installed at the inlet to the sensor allowing for automatic background measurements and maintenance of the equipment without interrupting the primary process flow.

Aside from the abrasion- and corrosion-resistance requirements, a critical requirement for operation of the Insitec is long-term maintenance of clean sensor windows by means of a pneumatic purge system. The fluid-dynamic design of this interface is important for providing long-term operation without dust fouling. The critical elements of this design require avoidance of particulate recirculation near the windows (which are recessed from the powder flow).

Once fine particles contact the windows, it is not possible to clean them by air-purging, even at high velocities because of strong electrostatic charging of small particles. Purge velocities are chosen to exceed primary flow velocity and must be maintained at all times. However, if the windows do become contaminated, access is provided to clean the windows with lens paper, an operation that requires about 15 minutes. The purge flow velocity designs have evolved through a combination of intuition and testing. Current cement industry users report cleaning maintenance periods of one month or longer.

Range and accuracy The in-line instrument gives the user the same degree of measurement accuracy in-line as is achievable using available off-line laboratory particle-size analyzers. Figure 2 illustrates measurements of a single type of cement under three different operating conditions. The laser attenuation values of 20%, 50%, and 70% represent increasing particle concentrations that would be expected in typical cement production processes.

The data analysis uses a patented algorithm capable of measurements over the range of 5% to 95% light attenuation.(Footnote: United States Patent #5,619,324),(Footnote: T.L. Harvill, J.H. Hoog, and D.J. Holve. In-Process Particle Size Distribution Measurements and Control, Particles & Particle Systems Characterization, 12(1995), pp. 309-313.) Typical laboratory instrumentation is limited to light attenuation less than 50%. In general, this limitation is not critical for laboratory applications since one can dilute the sample appropriately. However, a broader range of adaptability is important for process applications, where particle concentrations can vary.

Quality assurance tool The laser diffraction particle size measurement technique employed by the process instrument varies greatly from the packed-bed pressure-drop method that the typical Blaine number measurement device utilizes. However, a recent measurement sensitivity analysis has verified that the optical instrument was able to resolve the 7% variations in Blaine numbers of the samples tested.

In addition, a good correlation exists between the "laser-diffraction" Blaine and "pressure-drop" Blaine measurement. To develop this correlation, various samples of a particular type of cement that represented the high-acceptable, low-acceptable, and average-acceptable Blaine characteristics (i.e. the control range of interest) were measured.

Blaine apparatus measurements were then performed on these samples by the produce manufacturer, and compared with the analysis performed at Insitec using optical technique. The results indicate that the Blaine number, or specific surface area (SSA), yielded by the instrument can be used as a surrogate for a pressure-drop Blaine analysis. The optical instrument determines the specific surface area by effectively integrating the light scattering from individual particles. Note that there is not an exact numerical comparison between the two techniques, and that no error-bar uncertainties have been assigned to the measured Blaine numbers.

Allen(Footnote: Allen, Terence, "Particle Size Measurement," Vols. I & II, 5th Edition, Chapman and Hall, 1997.) has analyzed the reliability of the Blaine con cept and states: "The assumptions made in deriving the Carman-Kozeny equation (on which the Blaine number is based) are so sweeping that it cannot be argued that the determined parameter is a surface," and "The determined surface areas are usually lower than those obtained by other measuring techniques." This last comment is obviously consistent with the current optical results, which give higher values than the Blaine number. Allen points out a range of other limitations for permeametry techniques. Thus we conclude that the precision and accuracy of the Blaine number is limited.

From an operational point of view, the Blaine test takes time, which delays optimal mill/classifier adjustment. Over-grinding increases energy consumption, while under-grinding reduces product quality. The ideal goal is to meet quality requirements and maximize production rates (use of capital equipment) with minimum energy consumption. A current optical instrument user has found a three-fold reduction in product variation from 8% to 10% down to 3%. Although some customers still require Blaine tests, these measurements are more a confirmation of the optical instrument results.(Footnote: "On-line Particle Analyzer Tracks Changes in Cement Production," Powder and Bulk Engineering, March 1998, pp 36-39.)

Optical SSA and Blaine measurements These initial successful comparisons of Blaine and SSA measurements encouraged more comprehensive test comparisons. Figure 3 shows the results of a two-week test series comparing independent Blaine results and the optical SSA. The SSA values were normalized to the Blaine number results by a constant determined in a previous test series. On this graph, the Blaine number error bars are indicated for one standard deviation (about 1.5% as estimated by ASTM for an individual technician). For multiple technicians, it is known that different techniques may be used and the standard deviation would be of the order of 2.5%.

Note that the two different measurement techniques agree quite well, with a few data points at the beginning and end of the test series showing deviations exceeding 1.5%. An error bar has not been applied to the optical instrument measurements, but it can be assumed that it is of the same order, namely 1% to 1.5%. The error bars of the two sets of measurements would overlap in all cases, indicating that the two techniques are statistically in agreement. That is, there is no statistical discrimination for choosing one method as being more accurate or precise than the other. However, there remains a natural prejudice that the traditional Blaine number is the de facto standard when a new technique is in disagreement. How can this question be resolved?

One of the production parameters that can be correlated with these measurements is the production rate, which is a measure of the mill throughput of clinker. It is well known that a higher throughput of material leads to less milling and thus a smaller SSA, i.e. the material is not ground as finely. The data for production rate has been modified to delete large excursions in the feed rate for better comparison with the two surface-area measurements.

An independent correlation of the productivity with each surface area can then be performed. A comparison with the Blaine measurement, giving R2 value of 0.362 and given the Blaine measurement uncertainty of 1.5%, shows a weak correlation.

A similar correlation of production rate with the optical SSA measurements shows a much tighter correlation with an R[superscript]2 value of 0.746. This stronger correlation gives evidence that the SSA is providing superior precision when compared to the Blaine results.

Conclusions For many decades, the Blaine measurement has provided a reasonable industry standard for predicting ultimate strength of cements. However, improving production efficiency while maintaining even tighter quality control continues to motivate development of better measurement techniques.