Real-Time Blaine Cement Measurement Leads to Process Optimization

At Ash Grove Cement (Leamington plant, Utah), Insitec is used to measure the size of the finished product. The instrument location is at the product outlet of the milling circuit, in a position which allowed easy access facilitating both installation and routine maintenance. As both separator feed and separator rejects are located directly above the instrument, the Insitec may in the future be used for circulating load determination. For instrument-PC communication, a one-inch conduit cable run of 400 feet was installed from the instrument to the control building. Since the instrument was put in service in November 1998, it has been trouble free with a 100% run time. Considerable data was reviewed before complete acceptance by production. In June 1999, it was put in an automatic PID loop to control the cement fineness by adjusting the separator speed.

A new strength predictor

It is well known that early strength is dependent on both chemistry and fineness of the cement. With that in mind a multi-variable regression analysis was performed, correlating particle size variables and chemical analysis to 1day strengths (see Figure 1). As the general cement literature indicates, it is possible to predict 1-day strength using the measured Blaine number and a measure of clinker chemistry, (namely S_iO₂). S_iO₂ is the main component in two of the main cement compounds, C3S and C2S. Although the standard error of this correlation is not great (R2 = 0.59), this is a significant improvement in correlation using particle fineness alone (R2 = 0.38). Furthermore, the laboratory analysis is measuring secondary oxides S_iO₂, Al2O3, Fe2O3, CaO, SO3 and not the "real" strength producing components: C3S, C2S, C3A and C4AF. The potential composition of the cement compounds in the product (and clinker) is estimated using the Bogue formulas [3], presuming specific heating and cooling conditions in the kiln which might or might not be present at all times during production. Our model is not complete, but gives a simplified tool that describes the interaction between fineness, a chemical indicator (S_iO₂) and the 1-day strength. Despite its simplicity and current limitations, even this crude correlation will enable the plant to improve prediction of the early product strength because it includes a figure of merit for both the effects of chemistry and particle fineness. Changes in product chemistry can thus be compensated by product fineness, allowing optimization of the milling process.

Figure 1: Correlation of measured early strength with predicted strength (Blaine/chemistry factors) using traditional Blaine measurements

To verify the robustness of the correlation model, it was applied to a subset of the data for which we have Insitec measurements. Figure 2 shows the result of applying the model on a data set consisting of 44 days of production from two different periods (Sep-Oct '98 and June '99). The correlation coefficient R2 = 0.69, and the standard error is 107 psi, which is on the order of the uncertainty (75 psi) of the strength measurement itself. While the correlation model is based entirely on laboratory data, we find that we also obtain satisfactory results with on-line Insitec data. As more Insitec data becomes available the correlation model will be further confirmed and the confidence level will improve.


Figure 2: Correlation of measured early strength with predicted strength (Blaine/chemistry factors) using Insitec Blaine measurements

How can this new on-line particle size information be used to improve production and impact the bottom line? Real time feed back control improves the consistency of the final product. Having better control over the milling process makes it possible to mill product closer to specifications thereby increasing throughput and lowering manufacturing cost. For example: our cement customer produced 14 days of type I cement. The average 1-day strength was 2280psi, 480psi above the ASTM specification. Using the model derived earlier, (see figure 1), we want to define a target Blaine # that will yield the specification 1-day strength (plus a safety margin). A conservative margin would be to use three times the standard error of the prediction model. Statistically, this will result in less than 1% rejects. The current rate of recycle is approximately 5%, therefore this margin would reduce the recycle rate by a factor of 5. From Figure 3, the 3-sigma target for 1-day strength = 1800+3*107=2121 psi (159 psi lower than the month average). It is important to note that any improvement in the standard error of the correlation model for strength will allow even further improvement.


Figure 3: Correlation between off-line Blaine measurement and on-line Insitec SSA measurement

From the 1 day strength prediction model we can define a non-dimensional correlation relationship:

S/S0 = A*B/B0 + D*C/C0 + F ...........Eqn. (1)

where:
S = 1 day strength (psi) and the subscript refers to a reference value of 2000 psi.
B = Blaine Number (cm2/g) and the subscript refers to a reference value of 4000 cm²/g.
C = Chemistry parameter (S_iO₂) and the subscript refers to a reference value of 21%.
A = Fineness coefficient = 1.18
D = Chemistry coefficient = -2.92
F = Constant = 2.88


Figure 4: Comparison of Insitec SSA with traditional Blaine measurement during 12 days of production

In addition, we have determined a correlation (Figure 4) that gives a relation between product throughput and one-day strength:

P/P0 = -0.83 B/B0 + 1.83 ................Eqn. (2)

where:
P = Production rate in Ton/hr and the subscript refers to a reference value of 100 Tons/hr:
Using Equations 1 and 2 we can eliminate the Blaine number and show that dP*/dS* = -0.7.

Thus a reduction in average strength of 159 psi or dS* = -8% gives a 5.6% (5.4 ton/hr) increased production. This equals 1800 tons lost production for the 14 production days in June of 1999 or at a sales price of $70 /ton, a lost revenue value of $126,000.00 for the period.

The above example assumes that the production is at capacity and that the market can absorb the increased production. If production were below capacity, the savings would be in reduced milling costs, still substantial savings. From Peray [4], the savings in theoretical milling is on the order of 3 % per 100 cm²/g decrease in fineness. Our current measured correlation is about 2% per 100 cm²/g decrease in fineness, in reasonable agreement with the theory.

Conclusions

We have developed a quantitative correlation between the Insitec Blaine # and the traditional air permeability Blaine measurement for the Ashgrove Leamington plant. An early strength (1day strength) prediction model has been established using a combination of fineness and chemical composition (Blaine # and S_iO₂). The standard error of the model is on the order of the expected error of the strength measurement. This correlation model shows that process throughput can be increased by 5%, while at the same time meeting strength requirements.
The year of 1999 saw an all time record year for cement production at Leamington to which Insitec was a major contributor. Due to tighter control of particle size, Ashgrove Cement were able to ensure the quality at slightly lower fineness: 3870 cm²/gm in 1998 vs. 3770 cm²/gm in 1999.