|
A s the pulp and paper industry faces the challenge of reducing operating costs and improving quality, many companies have adopted philosophies that equate to “more, better, faster, for less.” Many companies are turning to advanced control tools to reduce operating costs. These projects tend to have the high rate of return required today, with a relatively smaller capital investment.
In 1997, Gulf States Paper in Demopolis, Ala., was approached by one of its suppliers to implement an advanced control solution that would reduce energy consumption and improve quality for the paperboard mill’s No. 3 lime kiln. This was thought by many at the mill to be almost unachievable on the unit, which was installed in 1995 and operated at 50% turndown. Because of the complexity of the No. 3 lime kiln’s operation and the potential for improvement, the proposal was ultimately accepted at Gulf States and success criteria was set as:
1. Reduce energy usage by 4%
2. Improve residual carbonate control
3. Stabilize operation of the kiln
Total time from project kick off to startup of the system was approximated at 25 weeks. The quick turnaround for this type of optimization project was extremely desirable in order to prove results from the capital expenditure. The project timeline included step testing for model development, hardware installation, operator training, and control tuning.
Gulf States’ No. 3 kiln, viewed here from the feed end, is a 14 ft-by-400 ft rotary kiln that burns fuel oil and/or natural gas.
LIME KILN CONTROL OBJECTIVES. The recausticizing process produces white liquor from the addition of lime—principally calcium oxide (CaO)—to green liquor. The waste from the process is lime mud—the main constituent of which is calcium carbonate (CaCO3). Calcium carbonate can be converted to lime through a process known as calcination, where high temperature is used to remove carbon dioxide from the calcium carbonate forming calcium oxide.
When this is performed in a lime kiln, the process can be broken down into three unit operations. First, the moisture in the lime mud fed to the kiln is evaporated. This occurs in the first third of the kiln, or the drying zone, where the chain section is located to increase heat transfer and to prevent the formation of large balls due to the rotation of the lime kiln.
After the drying zone, the material enters the heating zone, where calcining temperature (at least 800ºC) is reached. A discharge dam, located at the end of this section, maintains a constant bed level, stabilizing the retention time in the first two zones. In the final third of the kiln, the calcining zone, the temperature of the material is increased to around 1,000ºC and calcium carbonate is converted to calcium oxide.1 Another dam is located at the end of this section to stabilize retention time. The sum of the retention times for each individual zone is referred to as the filling degrees of the kiln.
Control of these zones is essential for good lime kiln and caustic room operation in order to prevent underburning or overburning of the lime. Lime that has been underburned has a high residual carbonate content, increasing deadload in the recausticizing system and reducing causticizing efficiency. Overburned lime has been calcined too much and slakes poorly. This is as undesirable as underburning, because, when the overburned lime is mixed with green liquor in the slaker, the reaction is difficult to control and more grit is produced.2
The keys to controlling the reaction zones in the lime kiln are:
1. Maintain good control of the temperature profile in the kiln
2. Minimize the effect of disturbances in the kiln
3. Maximize combustion efficiency without damaging refractory
LIME KILN CONTROL ISSUES. Successfully controlling a lime kiln presented a unique challenge for Gulf States. In general, long transport delays (three hours to five hours) and long response times characterize lime kiln systems. Additionally, the kiln is riddled with non-linear process responses that change as the operating conditions of the process change. Process measurement is also inherently noisy due to the harsh environment. Added to these problems is the fact that the kiln is a multivariable system where only two manipulated variables can be utilized to control at least three control variables. All these factors combine to make control of the system a major challenge.
Traditional lime kiln control schemes may control fuel flow in a single loop with the operator attempting to anticipate the effects of kiln disturbances on cold and hot end temperatures. Temperature control must be accomplished while maintaining complete combustion, providing a stable primary and secondary air profile, and controlling total reduced sulfur (TRS) emissions. This is a difficult task for even the most experienced operator.
Alternatively, two proportional integral derivative (PID) controllers can be used to control the temperature profile in the kiln, but this technique has also proven inadequate. PID controllers do not handle processes with long dead times very well, and the interaction between the two controllers can further destabilize the system.3
CONTROL ARCHITECTURE AND APPLICATION. The No. 3 lime kiln is a conventional 14 ft-by-400 ft rotary kiln, with tube coolers, that burns fuel oil and/or natural gas. The kiln is fed by two lime mud precoat filters, which discharge lime mud at 70% solids to a screw feeder. Environmental controls are provided by a dry bottom electrostatic precipitator, which recycles dust directly to the kiln feed screw.
Multivariable predictive control. Targeted for installation on the No. 3 lime kiln was Honeywell’s robust multivariable predictive control technology (RMPCT), which is implemented as a multiple input/ multiple output controller. This is the heart of the advanced control system. The RMPCT control system is designed to anticipate the effects of a disturbance on the system based upon an unbiased model prediction. Using the predictions of the model, the controller then makes a discrete change to the manipulated variables to keep the system within operating tolerances. The operating tolerances for the system are set by the operator as range control versus traditional methods of single setpoint control.
RMPCT also applies an optimizer to the control system. The application of the optimizer to the system results in increased operational flexibility for the kiln and gives the controls the freedom to manipulate the temperature profile, within the boundaries set by the operator, to operate as efficiently as possible.
Control architecture. In Gulf States’ lime kiln control architecture, the controls communicate to the Honeywell DCS through an open systems interface known as the real time application environment. This application environment runs in tandem with a Honeywell communications interface client/server, which maintains the connection between the application environment and the DCS. Both systems operate under the Windows NT environment and extensively use Windows NT functionality.
Security is integrated into the application with two operating modes. The primary mode of operation is operator mode. Operator mode is used to make setpoint and range changes to the kiln through the real time application environment. A second mode, developer mode, allows changes to control models and instrument error tolerance. Developer mode is an engineering mode only and is password protected.
In addition to the functional features of the system, a graphical depiction of the filling degrees and statistical performance of the process is integrated into the system. Trending capabilities are also included for the unbiased models and qualitative testing, as well as for the system variables. The optimization system has proven a user-friendly system and the operators quickly became comfortable with the system.
Training. Both Gulf States and its control system supplier felt that operator training was key to project success. The optimization system was presented to the operators as the master controller of the kiln. Although the optimization system would be running the kiln, the operator is still the "master of the master controller" and is ultimately responsible for operation of the kiln.
CONTROL MODEL DEVELOPMENT. Development of the control models for Gulf States’ new lime kiln control system was an integral part of the project. Model development required approximately two weeks of process run time due to the long dead time of the kiln. The disturbance and manipulated variables had to be modeled versus the control variables to develop a control model matrix for the process (a three-by-three matrix for this project). A typical model development run involved three to six step tests for each variable that could last as long as 10 hours, depending on the variable tested. Table 1 describes each variable that was modeled.
The step tests were performed over a wide range of operating conditions to collect as much operating data as possible. The data from the step tests were processed using advanced RMPCT regression analysis tools in order to develop a mathematical transfer function relationship as an approximation of the dynamic behavior of the process. Since robust multivariable predictive control is dependent upon the accuracy of the transfer functions, the range and quality of the step tests ultimately determined the success of the project.
To add further flexibility to the optimization system, the transfer functions were not restricted to linear approximations of the process. Rather, the use of more complex second and third order transfer functions to model the process compensates for the degree of oscillation, non-linearity, and the effects of long time delays in the process.4 This is necessary to accurately model the process response due to the dynamic characteristics exhibited by the lime kiln and to meet the constraints of the control ranges set by the operator.
| TABLE 1:For Gulf States’ new lime kiln control system, disturbance and manipulated variables had to be modeled versus the control variables to develop a control model matrix for the process. |
| Control Variables |
Manipulated Variables |
Disturbance Variables |
| Excess Oxygen |
Energy Flow |
Production Rate |
| Cold End Temperature |
Induced Draft Fan Speed |
| >Hot End Temperature |
ADVANCED CONTROLS FUNCTIONALITY. Gulf States’ advanced lime kiln control system is broken down into three parts: production rate control, air-to-fuel control, and RMPCT control. In addition, the optimizer provides another important means of achieving efficient lime kiln operation. Figure 3 shows a schematic of the lime kiln controls.
Production rate control. The production rate controller sends a remote setpoint to control lime mud flow and density of the slurry to the precoat filters through the DCS. All changes to the production rate are made through the real time application environment. Production rate changes are ramped at a rate of six tons per hour (tph) to reduce the effects a disturbance will have on the filling degrees of the kiln.
Air-to-fuel control. The second component of the control system is a feed forward ratio controller that maintains proper primary air distribution based on the fuel-firing rate. This controller is designed to optimize flame shape and maximize heat transfer without causing damage to kiln refractory.
RMPCT control. RMPCT controls energy flow and induced draft fan speed. Using the models developed during step testing, these two variables are manipulated to keep the control variables within range. All three levels of control maintain the temperature profile and maximize combustion efficiency without wasting heat generated in the kiln.
Optimizer. The optimizer for the control system is set in the developer mode of the control package. The optimizer can be assigned a value between zero and one. The larger the value, the more the controls will attempt to optimize the system. For this application, the optimizer was given complete freedom and Gulf States has been pleased with the results.
DISTURBANCE / MANIPULATED / CONTROL VARIABLES. The disturbance variable for the kiln system is production rate. The production rate disturbance includes the lime mud flow to the precoat filters, the lime mud density to the precoat filters, and the rotation speed of the lime kiln. The production rate disturbance effects are determined by modeling the rate of disturbance versus integral production rate. The model then recalculates the filling degrees in the kiln and compensates for dead time changes to the process through the RMPCT control.
As mentioned previously, RMPCT allows for range control of the control variables. This part of the optimization system must be tuned with operational experience and trial and error. Once the controls were implemented, Gulf States found that it was surprisingly easy to dial the system in. All manipulated variables are controlled by RMPCT.
Excess oxygen range was considered to be the most important variable to maintain in control. Excess oxygen must be controlled in a balance where complete combustion is occurring without wasting energy by pulling heat out of the kiln. This control is also vital to TRS control. Too little oxygen will starve the system and unacceptable TRS levels will result. It was found that this range needed to be set at approximately 1.2% difference between high and low setpoint.
The cold end temperature or exhaust temperature required significant tuning to find the optimum range. Cold end temperature was given second highest control priority because Gulf States felt that stabilization of the mud drying zone was critical to production of quality lime. The problem was that the mill also desired cold end temperature to be as low as possible without inhibiting heat transfer. Various ranges were attempted for maximum efficiency, but 10% of low setpoint was the optimum range.
The hot end temperature range was maintained at 50ºF between high and low setpoints. Like most large kilns, hot end measurement is a problem due to dusting. It was the general belief that this would be the weakest link of the control system. The implementation of air-to-fuel control to better control the flame shape has allowed this instrument to become a relatively good indication of lime quality and has not compromised operation of the advanced control system. It is also thought that the amount of process stabilization done at the back end of the kiln has made the hot end temperature measurement less critical.
Although the temperature ranges sound large and rather out of control, the control system has dropped the standard deviation of excess oxygen, cold temperature, and hot end temperature by significant amounts.
RESIDUAL CARBONATE CONTROL. Gulf States wanted residual carbonate control to complement the advanced control system. Control of temperature profile was one level of control, but the mill felt that the residual carbonate control offered the opportunity to maximize energy efficiency of the system. The residual carbonate control was installed as a proportional integral controller with amplitude proportional gain (APG).
The APG controller increases or decreases the high and low range for the cold and hot end temperatures in order to fine-tune lime quality control. As the residual carbonate tests deviate from setpoint, the controller proportionally raises or lowers the temperature range at both ends of the kiln to insure that the lime is not overburned or underburned.
The APG controller will only change high and low setpoints at a maximum of 1.6% for either temperature range. Although this might appear a minute change, a higher rate of change could cause the controller to continuously overshoot and oscillate. This controller gives the advanced control system further flexibility by allowing the mill to decide what carbonate target is best and by giving the controls the freedom to fine tune independently. This has proven very effective for Gulf States and has further enhanced the advanced controls.
LIME KILN PROJECT RESULTS. Gulf States has seen a variety of product, quality, and operational improvements resulting from the lime kiln optimization project. Jimmy Edmonds, the pulp, power, and recovery superintendent, has commented that the project “greatly exceeded expectations.” Some of the notable improvements as a result of the project, which was accomplished within 25 weeks as scheduled, are as follows:
Production
• Kiln energy usage was reduced by 8%, exceeding the project target
Quality
• Residual carbonate variability was reduced by 35%
• Lime availability variability was reduced by 47%
• Lime reactivity variability was reduced by 43%
Kiln operation stabilization
• Cold temperature standard deviation reduced by 65%
• Excess oxygen standard deviation reduced by 60%
• Hot end temperature standard deviation reduced by 10%
As would be expected, additional results point to improved white liquor quality and decreased recausticizing variability.
Johnnie Pearson is process engineer for Gulf States Paper Corp., Demopolis, Ala. Michel Dion is a principal control engineer for Honeywell Inc., Phoenix, Ariz.
REFERENCES
1. Adams T. in Alkaline Pulping (T. Grace, and E. Malcolm Eds.), Joint Textbook Committee of The Pulp and Paper Industry, TAPPI Press, Atlanta, CPPA, Montreal, 1989, Chap 22.
2. Valiquette J., Savoie M., LeClerc M, TAPPI 82(5):130(1999)
3. Osmond D., Tessier P., and Savoie M., TAPPI 77(2):187(1994)
4. Seborg D., Edgar T., Mellichamp D, Process Dynamics and Control, John Wiley and Sons, New York, 1989, pp. 130-157, 443-469
|
