Enhancing Grade Changes Using Dynamic Simulation

A dynamic simulation model of a 3-ply paperboard production process was developed. The simulation model consists of stock preparation and proportioning, short circulations, wire and press sections, and 75 drying cylinders including the steam and condensate system. Functionality of the automatic grade change program used on the machine was modeled as well. Physical, first principles models were used whenever possible. Simulator was extensively validated using historical data. The simulator has been used as a test bench in order to find better values for automatic grade change parameters. First set of the new simulator tested parameters has been carried out to the real machine with good results. INTRODUCTION It is common that paper and board machines produce tens of grades to meet different customer specific needs. To avoid large storage, mills have to change grades frequently. Grade changes (GC) have a large impact on the machine production efficiency, so it is relevant to make all possible efforts to minimize the production losses they cause. Careful production planning does a lot. Some paper machines may even get rid of the losses with a production cycle where basis weight changes are so small that acceptance limits of the grades overlap. However, the minimum basis weight change between grades may be 20 g/m, there are simultaneous changes to fiber furnish type, filler content and type, color, and so on. This kind of issues keep the GC of paper and board machines as a very interesting and challenging research topic. In literature grade change has been a popular subject and many times dynamic simulation has had a major role in those studies. Already in 1988, Miyanishi et al. [1] demonstrated effects of machine chest volume, first pass retention, and dry broke ratio in GC. They also showed the effect of increasing filler loading in the initial period of the GC in which filler type is changed. Historically GC's have been accomplished manually by machine operators and many of the studies deal with the question how to automate the operations needed. The basic idea of automatic grade change (AGC) is simple: ramping of manipulated variables with preplanned targets and mutual coordination. However, it is challenging to tune the system to give good performance simultaneously for all quality variables, and many approaches have been presented to overcome this problem. Ihalainen and Ritala presented an idea to numerically optimize the actions in GC's by using dynamic simulation [2]. Murphy et al. proposed a dynamic coordinator to reduce the basis weight and moisture content upsets during GC's [3,4]. Thick stock flow was used as a coordinating actuator. Other than linear control curve pattern was used by Mori et al. too [5]. Additionally, Mori et al. presented a new iron plate drying model that enables fast calculation of steady state moisture and temperature profiles for the drying section. They told, that the model is tuned before each GC using on-line measurements, and after the tuning the model is able to accurately predict steam pressures for the new grade. For a successful GC it is very important that new grade's target values are reasonably good, and the most difficult to predict are steam pressures. Besides Mori et al., many other papers concern the different estimation methods [4,6,7]. The traditional AGC methods switch the machine's quality controls off when performing grade changes. Also approaches to do GC's under feedback control have been presented. Välisuo et al. suggested use of model predictive control with non-linear models [8]. Recently, Kuusisto et al. have discussed the same subject [9]. They emphasize the importance of moisture dynamics in the use of multivariable predictive control in AGC. The moisture model must handle the complicated dynamics due to the changing machine speed, paper basis weight and ash content, but still be easily commissioned. A totally different approach in speeding-up GC's has been presented in papers that introduce new papermaking concepts with considerably faster transient dynamics than before [10]. In this paper, we present our experiences in using dynamic simulation to improve GC's in a Finnish multi-grade, multi-ply board machine. Our approach has not been to develop a new algorithm or a control strategy, but to enhance the operation of the mill's present AGC system. The structure of the simulation model and the main calculation principles are described. Simulation results are shown and discussed. Finally, the way to use the simulator at the mill is presented before looking to the future work. PROBLEM FORMULATION The board machine in question produces liquid packaging boards having basis weight area of 170-350 g/m. The machine has three fourdrinier wires, one of which is equipped with a top wire unit. There are three press nips in the press section. Five two-tier dryer groups with conventional steam-heated cylinders are used to dry the base board to about 3% moisture prior to a size press. The following two steam groups before the on-line coating have still been included into the simulation model. The machine speed varies between 200 and 450 m/min. An average of one GC per day is done on the machine. For many years, a specific automatic grade change program has assisted operators at the mill. The control variables that are included in the AGC are: ! wire speed ! thick stock flows for 3 layers ! slice opening of the 3 headboxes ! jet-wire ratios for 3 headboxes ! steam pressures in steam groups 5 and 7 The AGC program calculates targets for the thick stock flows and the two steam pressures. Pressures in other steam groups are following the pressures in the 5th and 7th group with given ratios. The AGC also suggests typical grade operating values for the wire speed, slice openings and jet-wire ratios, and the operator may change them if necessary. After GC has been initiated, the AGC coordinates the mutual delays and handles ramping of the variables. This coordination is pre-planned by giving a start delay, maximum stepping rate and a stop delay for each variable in the GC. Additionally, there is a selection if to synchronize or not the ramping of each variable with the others. Operators select which of the variables are controlled automatically by the AGC. Usually they let the automatics handle all other variables but the slice openings, which they most often adjust manually. All operators use the AGC and consider it as a useful tool. On the other hand, the general feeling has been that the operation could be improved to give shorter GC times and to reduce moisture fluctuations during the GC's. The AGC's steam pressure prediction for the new grade has been at a satisfactory level, so no attention was paid on that part of the AGC in this study. Rather big unwanted excursions in the web moisture content have been one of the main concerns in GC's on the machine. Clearly the moisture bounds up or down when the machine speed starts to change and again just after reaching the target speed. The headbox dynamics during the speed change has been nominated as the dominant reason for this kind of moisture disturbance [3,4]. When browsing GC's in historical data it can be seen that this occurs in most of the GC's, but the amplitude of the disturbance varies from case to case. An intuitive reaction is that by better timing the phenomenon could be removed. However, it is very challenging to try to figure out the right actions needed to fix such a fluctuation using just a mind model. In a multi-ply machine like this the number of tuning parameters of the AGC is remarkable. Different ideas compete and conflict. This kind of "opinion engineering" is quite a fragile base to start experiments with the real machine. The simulator was seen as a possibility to play with different ideas before anything is done on the machine. MODELING The process model covers the board making process from pulp chests to the end of the base board drying. The model was built using the APROS platform [11]. It is a general-purpose modeling and dynamic simulation tool. The extension of the platform on pulp and paper mill applications is called APMS. Simulation models are built based on the P&ID's and equipment functional descriptions. High fidelity is achieved by using first principles of physics to describe process operation whenever possible. Conservation of mass, energy and momentum is used in solving pressures, flows, and temperatures in piping networks. Validated and tested model algorithms for process equipment, instrumentation and automation are integrated under the executive control of an advanced graphic user interface. The user builds graphically a model that looks analogous to the corresponding flowsheet in the P&ID's. The simulation model is configured while the flowsheet is being drawn and parameterized. The model structure can be changed any time and the model can be expanded without need to recompile or link the program. Most of the parameters needed were derived from P&ID's, like pipe diameters, tank volumes, nominal flows, and heads for pumps. Pipe lengths were calculated from piping drawings and some of them were ocular estimates. Equipment elevations from a certain reference level were obtained from layout drawings. Some of the parameters needed a look at construction drawings (like cylinder diameter and shell thickness, or number and dimensions of pipes inside a condensator) or were got from measurement data (like pressure drops in pressure screens). Also, general engineering knowledge is useful in cases when specific information of a piece of equipment is not easily available (like in our case pressure drops in the valves and pipes, and trim types of the control valves). Mass Preparation and Short Circulation Refining, mass proportioning and the short circulation of each layer form a large thermohydraulic pressure/flow network which is modeled by connecting model objects for pipes, pumps, valves and tanks etc. together. The incoming stocks to the chests before refining form the starting point for the model. Three fiber components, filler and water are carried along in the system. Ideal mixing was assumed in tanks, but wire pits were divided into several ideally mixed volumes. Refiners, pressure screens and centrifugal cleaners take also part in the pressure/flow network. Screens were defined to have constant separating ratios. Refining is not changing pulp properties in the model, though this feature would further broaden the scope to cases where refining is changed noticeably for the new grade. It would, however, require lot of experiments to capture the effects of refining quantitatively on e.g. water removing in the wire, wet press and drying sections, so we decided to take the first step without modeling any refining effects. Headbox is the last part of the pressure/flow network before the web model starts. Only the jet-wire ratio controls directly controlling the speed of each head box feed pump were needed of head box controls to get the dynamics correspond very well with measurements. Machine cross direction was neglected in this model. In the machine direction the web is described with the user defined number of elements moving with the machine speed. The web properties that are calculated are moisture content, temperature, basis weight, composition and thickness. Wire and Press Section In the wire section model, webs from three separate wires are interconnected into a single one and the compositions of the layers are averaged. We have used constant values for retention of various fiber components and still obtained very good correspondence to measured values, e.g. headbox consistencies. This is due to the relatively high grammage of all produced grades. Filler plays a minor role because its content is very low in all grades. So, no extra effort was taken to explain retention changes in this phase. Dewatering on the wire part was modeled in a straightforward way, as well. First approach was to use constant dewatering ratio in all wire model components. As an example, it means that the dry solids content of the web and total mass flow to the press section change when e.g. consistency in a headbox changes. Another approach was to use constant dry solids content of the outgoing web in the last part of the wire. Then a change in a head box consistency does not change the dry solids content but the total flow into the press section is still free to change. This approach seemed to work well. Maybe because the operators control the dry line position with slice opening thus keeping the dry solids content in a certain range. If the effect of the slice opening on the dry solids content is not known accurately, it may be better to keep the dry solids content as constant. It is worth mentioning here, that in this study the slice openings, before and after each simulated GC, could be taken from the corresponding historical data. The decreasing permeability model [12] was used to describe water removal in each of the three press nips: