GCI TECH NOTES ©

Introduction

In order to better understand the effects of process variables on emissions at cement plants it is necessary to obtain representative samples of process inputs and outputs during emissions testing. Representative process sampling is also required for BIF Compliance Tests and Trial Burns. EPA guidance on Trial Burns suggests (but does not require) sampling each process stream every fifteen minutes. This sampling frequency is clearly based on EPA's experience with hazardous waste incinerators. EPA has never demonstrated that a fifteen minute sampling frequency can improve the accuracy of process sample determinations for a cement plant. Process streams at cement plants are substantially more uniform based on the cement production requirements and methods. A simple grab sample is frequently considered representative of an entire day's worth of feed or production. Somewhere between those two extremes would seem to be a conservative and technically defensible sampling frequency.

Process Sample Types

Process inputs include kiln feed and fuels. Kiln feed is ground and stored in large batches prior to feeding to the kiln. It is stored as either a fine dry powder or wet slurry. Dry powder kiln feeds are stored in blend silos. Slurries are stored in slurry tanks which maintain a continuous agitation system. With either of these systems, batch sizes are frequently twenty-four hours worth of feed or larger. However, some dry feed systems have storage for as little as ten hours worth of raw material. Fuel is fed in a variety of forms into cement kilns. Coal and coke is ground to a fine powder prior to being blown into the kiln. The large quantities of these fuels purchased and stockpiled at a cement plant result in a very uniform feed. Waste fuels include solvents, containerized solids/sludges, ground solids, tires, (both whole and chipped), as well as other miscellaneous materials. Liquids are accumulated in large batch tanks which are continuously agitated. Depending on feed rates and tank sizes, batches will last ten hours or more. Other fuels are accumulated and fed in large enough quantities to maintain uniform feedstream characteristics just as with the coal and coke.

Process outputs other than emissions include cement clinker and cement kiln dust. Cement clinker by virtue of its product quality requirements is a uniform process output. Cement kiln dust is similarly uniform. Changes in process outputs that may occur as a result of sudden input changes, metal spiking as an example, typically take four to ten hours to have full effect. The large residence time for materials in the kiln combined with internal recirculating loads will necessarily result in any changes occurring over a number of hours.

Data From 1992 COC Tests

During the 1992 COC Tests performed to comply with the BIF regulations, two plants (Holnam - Clarksville, MO and Holly Hill, SC) with three kilns analyzed both run by run and daily composites of most process inputs and outputs. All samples were taken on an hourly basis. Both run composites (three to four hourly samples) and daily composites (ten to fourteen hourly samples) were made and analyzed. The remarkable consistency in the data demonstrates the uniformity of cement plant feed streams. Bear in mind that the typical analysis of duplicate samples can vary 10 to 25% for these types of determinations. Results close to the detection limit can vary even more. The data is consistent enough that problems with sample digestion, such as the

coal/coke kiln 2 run 2 results, and other analytical problems are easily spotted. Tables 1 and 2 demonstrate the consistency of the slurry (kiln feed) and coal/coke. Table 3 provides similar information for the liquid hazardous waste fuel at both plants. Tables 4 and 5 demonstrate the consistency of the kiln outputs - clinker and cement kiln dust. Clearly, had there been significant variation in these process streams that would require more frequent sampling, such as every 15 minutes, these results could not be as consistent.

Conclusion and Recommendations

For the majority of cement plant process streams, hourly sampling is a conservative sampling strategy. More frequent sampling is unnecessary during normal/stable operations. Samples can be composited on a run by run or daily basis depending on the data quality objectives and how the data is to be utilized.