Operators in a modern cement plant could be forgiven if they feel like Charlie Chaplin in the classic movie “Modern Times,” racing around the factory trying to keep everything running smoothly. The problem is that cement plants have become more automated than ever before, so they have dramatically reduced the number of people that run them. At the same time, their operations have grown far more complex. The result is that operators have so many tasks to perform that it has become more difficult for them to manually maintain optimized process conditions.

For example, sensors in the kiln determine operating temperatures, and chemical analysis of exhaust gases shows whether overburning or underburning is occurring. Based on this type of information, operators are constantly expected to make decisions such as increasing or reducing the feed rate of fuel and raw material into the kiln. But operators have so many tasks to perform that in many cases by the time they are free to address this type of issue, a major swing may have occurred and a large step may be required to bring the process under control. This creates variation in the quality of the finished product and also runs the risk of moving the process too far in the opposite direction, in the worst case causing an upset that may result in a production loss.

THE BIG MOVE

Lafarge Group, Paris, France, the world's largest cement producer, has overcome this problem and in the process increased production by 4% to 5% in 50 cement mills by implementing a rule engine application that deploys the knowledge of Lafarge's experts to optimize operating parameters such as feed rates. The application, which is also known as an expert system, makes smaller and more frequent changes than operators were able to do, thus increasing output consistency, making it possible to run closer to operating limits. The production increase is most beneficial in situations where plants are sold out but the company also has realized quality improvements and reduced operating and maintenance costs in plants running below capacity.

The application was developed by an in-house team based on the G2 real-time rule engine platform from Gensym Corp., Burlington, Mass. A senior engineer with the team said that the expert system, which Lafarge calls Lucie, for Lafarge Universal Cement Inference Engine, operates mills and kilns in closed loop mode for approximately 90% of the time, while operators handle the most complicated situations that make up the other 10%.

The kiln is the area where capital costs are highest, fuel demands are largest, and process control is the most crucial. The large electric power demands of finish milling mean that improving the efficiency of grinding and separation of the ground clinker can yield significant energy efficiency improvements and cost savings.

LUCIE'S MISSION

“There simply aren't enough people in a modern cement plant to achieve true process optimization using conventional manual control methods,” the engineer said. “Our goal in developing Lucie was to provide an automated method of optimizing the process during the vast majority of the time when the plant is operating under standard conditions. Extremely complex situations, which only occur occasionally, are difficult to parameterize, so we feel they are better handled by the operators.”

The first efforts at automated process control began in the late 1960s. A major step came in 1989 with the decision to use G2 as the basis for expert system development. G2 applications transform operations data into automated decisions and actions in real time. The G2 platform uniquely combines real-time reasoning technologies, including rules, procedures, object modeling, simulation, and graphics, in a single development and deployment environment. The platform supports the entire application lifecycle by enabling developers to design using prototyping and simulation, develop iteratively, deploy, and continuously adapt their reasoning-driven solutions.

Lucie has grown over the years to become a reliable and stable means of controlling most of the operating equipment in a cement plant. The system's control strategies are based on closed-loop sets of calculations that are kept generic to facilitate maintenance. The generic software release that forms the basis of each plant's control system provides strategies for different types of kilns, grate coolers, conventional ball mills, and vertical spindle mills. A fuel manager oversees the feeding of alternate energy sources to the kiln system. Lafarge is continually interviewing engineers and operators to improve the Lucie system. A graphical rule language was developed to simplify the entry of knowledge.

Lucie collects sensor signals from the process controllers. Sensor signals are subjected to validity and trustworthiness tests; signals that fail the tests may be subjected to more rigorous filtering or thrown out. Treated sensor values are mathematically processed to generate short-term actions based on the goals of increasing production rates, improving product quality, and reducing energy consumption. Short-term actions are modeled based on the expertise of the company's most experienced operators. For example, abrupt action is taken to recover from an upset condition. After conditions have returned to normal, these actions are reversed to maintain long-term stability.

Lucie has judgment functions, which apply expertise in the form of a reasoning technique known as fuzzy logic to judge what the process conditions are in real time. For example, an oven judgment function might conclude whether it is too cold, too hot, or just right. Or using measurements for oven temperature, kiln torque, and gas analysis, Lucie will judge if the material is at correct burn or quality level. With these judgments, there is no need to wait for lab results on quality, which can take one half-hour to run and report back to operations. With Lucie, the goal is to act immediately based on the judgments.

Lafarge is working to implement Lucie in Lafarge's many plants located around the world. The implementation process begins with a workshop that teaches plant personnel how to apply Lucie to finish mills, which are relatively easy to control. An engineer or information technology person is assigned the job of connecting Lucy to the plants sensors and control systems. Once the expert system is up and running, this person can go back to their normal job. After the plant has experience with running Lucie to control their finish mill, they begin using the expert system to run other areas of the plant such as kilns. Using this basic process, Lucie has already been implemented in 50 out of Lafarge's 120 kilns around the world. Often, Lucie plant champions will make improvements to the generic code and if these improvements provide advantages to most plants they will be included in the next worldwide release.

QUANTIFIABLE BENEFITS

The Lafarge engineer said that since the company has the most experience with closed-loop control of finish mills, this is also where the benefits are most easily quantified. A consistent approach to mill circuit control has increased mill production by an average of 6% in United States plants. The variability of quality also was reduced 16%. The first mill where Lucie was installed in Greece recorded an immediate increase in production of 2.3% compared to the others that were still operating manually. “If the plant is sold out, these production increases can immediately be converted to revenues,” the Lafarge engineer said. “If not, mills can still improve quality and reduce operating costs.” For example, two cement mills in Malaysia operating below full capacity were able to reduce power consumption 4.1% and 5.1%, and reduce quality variability 40%.

Some plants also are starting to see improvements in the more difficult kiln area. In one plant, kiln production climbed to record levels after Lucie was implemented. A study of kiln control in France showed that Lucie was able to reduce clinker quality variation nearly 40%. This study also showed that the number of human interventions per hour dropped from 5.83 to 0.249 after installing Lucie, providing operators with much more time to focus on proactive tasks. In the eight installations of the Lucie fuel manager in Eastern Europe, cases were reported where the fuel manager recovered from the sudden failure of one of several fuel feeders through compensation with the remaining feeders within one minute without an upset. In the past when manual control was used, the recovery process took 10 minutes or longer depending on operator availability and nearly always involved an upset.

“Our plants throughout the world that have implemented Lucie have achieved production increases, quality improvements, and operating cost reductions,” the engineer concluded. “The key to Lucie's success is its ability to stabilize the process at or near optimal levels by making frequent but small steps to maintain optimal operating conditions while avoiding upsets. Having seen the benefits that can be achieved, we are working as quickly as possible to implement the expert system in our remaining plants in order to maximize the competitive advantage that we are achieving from real-time closed-loop supervisory control.”

This article was adapted from materials provided by Gensym Corp., (+1) 781-265-7100; Fax: (+1) 781-265-7101; e-mail: info@gensym.com; http://www.gensym.com