When Every Penny Counts

Following a major kiln-area upgrade, management at Capitol Aggregates' San Antonio, Texas, cement production plant discovered kiln throughput now exceeded downstream processing capacity. Capitol Aggregates' choices for eliminating the bottleneck were to spend more money to upgrade or replace major pieces of processing equipment, or invest in technology that offered improved operational efficien...

By Dave Harrold November 1, 2003

Sidebars: Optimize non-linear business constraints

Following a major kiln-area upgrade, management at Capitol Aggregates’ San Antonio, Texas, cement production plant discovered kiln throughput now exceeded downstream processing capacity.

About cement

Cement has a low value-to-weight ratio, thus cement plants are generally located close to raw material sources and the product is sold to markets within close proximity to where it’s produced.

For a cement producer to extend its market range and attract new customers, it must lower its production costs enough to offset increased transportation costs.

Kilns and mills are the heart of cement production.

Kilns are horizontally sloped steel tubes over 160 feet (50 meters) in length and 25 feet in diameter, turning at one to four revolutions per minute. The kiln’s interior is lined with firebrick to withstand processing temperatures that can reach 3,400 °F (1,870 °C).

Raw material enters the kiln and tumbles through progressively hotter zones until it reaches the flame. Intense heat triggers chemical and physical changes, converting calcium and silicon oxides into calcium silicate. At the end of the kiln, red-hot calcium silicate nodules, collectively called clinker, emerge.

Following cooling, clinker is conveyed to storage where it becomes feed stock for the milling process—the bottleneck area at Capitol Aggregates. (See “Cement ball mill operation” diagram.)

Cement ball mills are horizontal steel tubes filled with steel balls. As the tube rotates, steel balls tumble and crush clinker into a super-fine powder called Portland cement. Gypsum is added as a hydration retardant—a common method to control how quickly cement “sets.”

Mills are designed with large through-flow areas of ventilated air. This large air volume helps maintain a low-pressure drop across the mill and a correspondingly low-power consumption by the mill’s ventilation fan—an especially important feature for mills using moist additives.

In 1999, Capitol Aggregates invested millions of dollars to upgrade its San Antonio plant’s kiln system from 1,500 to 2,000 tons per day. The project was successful, but, as mentioned, after the upgrade, kiln production exceeded capacity of the milling operations. Possible remedies included:

Mechanical upgrades to the existing ball mills;

Replace the existing ball mills with newer milling technology; and/or

Optimize existing ball mill operation.

To preserve Capitol Aggregates’ investment, plant manager Tom Spaits decided it was best to first ensure the existing ball mill operation was optimized.

APC solution

“We had tried fuzzy logic advanced control tools,” Spaits explains, “but we had found these solutions required more ongoing attention than we were able to provide. Still, after talking it over with our process manager, Tom Giuliani, we agreed that if we could find an advanced process control [APC] solution that didn’t require a lot of upkeep, it might help us reduce the milling bottleneck.”

“I started conducting an APC search and read about Pavilion Technologies just as it was beginning to develop a cement industry focus,” says Giuliani. “Pavilion was headquartered in nearby Austin and had added cement industry experience to its staff, but they didn’t have quantified and proven cement industry successes, so we were reluctant to write a check and see what we got.”

Like many industries, the cement industry is accustomed to purchasing proven, turnkey solutions that come with guaranteed results. However, further APC investigations by Capitol Aggregates failed to locate an APC provider with quantified, third-party results.

“Pavilion was quite confident it could meet our expectations, but we wanted to limit our risk,” says Spaits. “Eventually we struck a 30-day deal. Pavilion would implement the APC and, if results failed to meet our agreed upon expectations, we would pay only for work hours expended. If results did meet expectations, we would pay Pavilion’s license fees and implementation costs.”

Multiple models

Cement producers use a measurement known as Blaine number to gauge the quality and strength of the final product. Blaine numbers represent a particle’s surface area and are usually expressed as cm2/gram. In cement production, the smaller the surface area, the finer the particle, the better the quality.

By applying laser diffraction technology to a light scattering methodology, manufacturers have created on-line instrumentation for measuring cement particle size. However, initial and life-cycle costs often place on-line Blaine instrumentation out of reach for some cement producers.

As an alternative to a physical on-line instrument, Pavilion created a Blaine virtual on-line analyzer (VOA) for Capitol Aggregates.

Every 30 seconds, Pavilion’s VOA uses multiple-inputs to calculate a single output that represents the Blaine number. When laboratory sample results become available (once every four hours at Capitol Aggregates) an auto-bias feature corrects the VOA model. The VOA-calculated Blaine value is stored in the plant’s data historian and is provided as an input to Pavilion’s Process Perfecter model predictive controller (MPC).

“Experienced operators often have a ‘feel’—a sixth-sense—about the quality of the products they are producing. Being able to compare ball-mill operating parameters, VOA-calculated Blaine numbers, and lab samples are helping our operators and laboratory personnel ensure production remains on-spec,” says Spaits.

Cement ball mill control has traditionally used rule-based control solutions. However, with few exceptions, rule-based controls apply linear control strategies to ball mills that demonstrate highly non-linear behavior.

The key manipulated variables in a cement ball-mill include new material feed rate and separator speed.

Control and constraint variables include mill level or elevator load, return flow from the separator, and cement fineness.

Disturbance variables that can influence the MPC include mill fan speed, water spray, ball wear, and clinker grindability (hardness).

Using three months of ball-mill historical data, Pavilion engineers developed the VOA and ball-mill controller models; completed step testing; and had the VOA and MPC ready to go online in less than two weeks.

Pavilion’s Process Perfecter uses cascaded models and integrated trajectories that allow models to share data and calculate future control moves. This feature allows operators to see how the MPC has controlled the process in the past and how it’s expected to control the process in the near future.

“Because our operators could see what the controller was anticipating it would do next, they quickly become comfortable that the MPC was operating the process in a safe and responsible manner. Almost from the first day, operators felt confident the MPC was going to work,” says Giuliani.

After 10 days of operation with Pavilion’s MPC solution, results demonstrated this was going to be very successful. By the end of the 30-day trial period Pavilion’s solution exceeded Capitol Aggregates’ operational expectations, and demonstrated that the solution would not require a lot of attention.

“In just a few short weeks, Pavilion’s solution allowed us to increase our production by more than 10%. The financial benefits we expect to realize encouraged us to extend the application throughout our facility,” says Max Frailey, Capitol Aggregates vice president.

Extending the application

Capitol Aggregates first applied Pavilion’s Process Perfecter to one of two ball-mills. Since then the second ball-mill has been added and investigations are underway to add the kiln.

“The operational improvements Process Perfecter has already provided makes us wish we had done this sooner,” says Spaits. “Once we complete the optimization of our production processes, we intend to investigate ways for Process Perfecter to help optimize other business entities, such as energy and emission predictions.”

Rarely does a small, commodity-producing company consider and then commit to use technology to optimize its processes and its margins. By investing in this technology, Capitol Aggregates has demonstrated an industry leadership position.

Comments? Email dharrold@reedbusiness.com

Optimize non-linear business constraints

Most processes contain varying amounts of non-linearity. Sometimes the non-linearity is inconsequential, other times it occurs in regions outside the control range. In those instances, traditional control algorithms provide suitable solutions.

However, when non-linearity is sizable and/or occurs in the control region, non-linear algorithms are required.

When control requirements include economic, quality, energy, rate, and environmental constraints, considerable amounts of non-linearity are likely to be present.

However, when all the constraints are simultaneously under control, business operations become optimized.

How well a business is optimized is, at least in part, demonstrated by how agile the business can adapt to product grade changes.

Depending on product value and the frequency of grade changes, hundreds of thousands of dollars can be saved by decreasing product grade change transition times. (See “Turning waste into profits” diagram.)

Pavilion’s Process Perfecter uses artificial neural networks, cascaded models, and time-based vectors to provide a better understanding of past, current, and future events to help optimize business processes including decreasing product grade change transition times.