Process Optimization and Model Based Control in Pulp and Paper Industry

In this paper we discuss the usage of physical models for diagnostics and process optimization in pulp and paper applications. Specially diagnostics with respect to hang ups and channeling in digesters is described, and how a rigorous model can be used for this, but also for the purpose of MPC, Model predictive control. An example of optimization of a complete pulp mill with respect to mill balances is also presented. An example with respect to optimization of a paper mill with respect to both quality and mill balance is also presented briefly. Introduction: 50 years ago paper machines had very little instrumentation and automated control, as computers did not exist yet. The paper machine speed was low and the paper width normal not very wide, just a few meters. On the other hand the manual control made use of a lot of operators, who normally new their part of the process very well. Today we have a situation where the paper machines run faster and faster, up to more than 1800 m/min, and the width may be up to 11 meters, like the ones at Dagang in China. At the same time we see fewer operators running the production. Instead we have got much more automated control. In the pulp mill we see similar developments, as a fiber-line can have a capacity of 3000 tpd today compared to much less than 1000 tpd 50 years ago typically. In the next step we can see a trend towards even more diagnostics, decision support and process optimization. The process optimization will be from on-line control using Model Predictive Control to dynamic optimization for next hours to days and weeks. To perform these optimizations and controls we need to have a good knowledge over the status in the plant with respect to instruments, performance of equipment and service and maintenance actions needed. This is needed as well as good information on orders and priorities. Process Optimization and control: In this paper we will focus much on the use of adaptive physical models for both on-line control and production planning the next 24 hours. In reality the boarders between statistical models and physical models may not be that clear. If we introduce a number of parameters into a physical model and these have to be tuned by plant data, it is in reality a combined physical and statistical model. The advantage is that we get the robustness of the physical model, but can make use of the statistics really relating to the actual process. In(Wisnewski P.A, Doyle F.J and Kayihan F 1997) we have a good example of a physical model of a digester while in (Aguiar H.C and Filho R.M 1998) we have an example of a neural net model, which is representing a statistical model. Detailed physical model of digester The digester is the major process in the pulp mill. There are two major types of digester: Batch cooking and continuous cooking. This paper will handle the continuous digester and specially the hydraulic digester. There are two types of continuous digesters: hydraulicand vapor phase. The main difference is that the hydraulic digester is full of fluids while the vapor phase digester has a vapor cushion in the top. . In a continuous digester, the wood chips are cooked in an aqueous solution of sodium hydroxide and sodium sulfide called white liquor at elevated temperature and pressure. The objective is to degrade and dissolve away the lignin and leave behind most of the cellulose and hemi-cellulose in the form of intact fibers. In practice, the chemical pulping method is successful in removing most of the lignin but unfortunately we also will get a degradation and dissolution of a certain amount of the hemi-cellulose and cellulose. We have a digester model built on the same principle as the so called Purdue model (Bhartiya et al, 2002 ). The model contains two volume fractions, the volume occupied by the chips with the entrapped liquor, and the volume occupied by the free liquor. This can be divided into two regions: wood and free liquor in each digester section. For digester operations we have been working with physical modeling of the digester in 2-D, including pressure drops inside the digester. The reason for this has been to account for both the aspects detection of channeling and hang ups as well as making use of the model for multi variable control, or Model Predictive Control, MPC. A physical model over a digester has been made taking into consideration the pressure drop inside the digester due to the channels between the wood chips. When there are mostly large chips, the channels between the chips are large, with low pressure drop for fluid flowing in-between. When there are a major amount of fine pins etc, the channels will decrease, and the pressure drop increase. This is what happens in reality if we have different chip size distribution or different packing in the digester. One reason for this may be that the chip size distribution is inhomogeneous or the chip screw in the stack may not feed in a constant way. When it moves up and down the density may be different as it is to switch direction. When we have a lot of flakes these may adhere to the screens and cause hang ups. Aside of causing an increased pressure drop in the screens, also the chips will get different residence times and contact with liquors of different concentration of both chemicals and dissolved organics. This may cause a significant variation in kappa number of the final fibers. By identifying pressure drops, residual concentration of chemicals in the liquors, temperatures and flows and compare actual results to those predicted by the model, we can tune the model to match reality. This is under assumption first of good process performance. Multipurpose use of the model MPC ( production optimization on-line) 3dMPC controller is intended for multivariable feedback control and optimization of an industrial process that has many input and output signals. Inputs are sensor measurements of manipulated variables affecting actuators, and outputs are process variable set points. For processes like pulp production with strong interaction between different signals this technique can offer substantial performance improvement compared with traditional singel-input singel-output control strategies. By using MPC instead of traditional PID to control a bark boiler a pulp mill decreased the NOx from a bark boiler by 50% (Dahlquist et al.2001) Figure 1 Feedback control system. The controller determines the manipulated outputs, OUT, based on actual measurement of process variables, PV, and feed forward signals, FF, and of operator defined parameters, and external inputs. The process variables can be assigned set-points that the target for the feedback control law. The feed forward signals are measurable disturbances acting on the process that can be used for feed forward. It is to prefer a model of lower order, because a system with many inputs and outputs can easily end up with hundreds or even thousands of state variables (Morari et al , 2000). To avoid this we have made some assumptions • Chips are well-steamed (it means that there are no extractives in addition to lignin and carbohydrate) and fully penetrated with the white liquor before entering the digester; • Both chips and liquor in the digester move as plug flow; • The volume of the individual chips remains unchanged during the digestion; • The digester is adiabatic; • The wood chip is isothermal; • The free liquor is homogenous; • Radial gradients in temperature and concentration within phases can be ignored; • Chip pressure and flow resistance are isotropic; • Inertial forces can be neglected. It is important that you do not simplify the models too much. The model must be so close to the real process as possible but still a useful model for control and optimization must calculate fast and give values with a reasonable accuracy. The result of the cook in the digester is measured by the kappa number. The kappa number is controlled by the temperature, residence time and the concentration of the cooking chemicals. The equations that are used in the digester section model are in principle the same that are used and described in (Wisnewski et al 1997) with respect to the chemical reaction, although we have complemented the hydraulics taking into account also pressure flow aspects. The Kappa number is the most important quality variable and is described by: