Paper Mill Modeling with Dymola

Dymola and Modelica

In the few years of research in modeling and simulation, the concept of object-oriented modeling has achieved a big relevance. Several works have demonstrated how objected oriented concepts can be successfully employed to support hierarchical structuring, reuse and evolution of large and complex models independent from the application domain and specialized graphical formalism.

To handle complex models, the reuse of standard model components is a key issue. But in order to exchange models between different packages an unified language is needed. Modelica is an object-oriented, general-purpose modeling language that is under development in an international effort to introduce an expressive standardized modeling language. Modelica supports object-oriented modeling using inheritance concepts taken from computer languages such as Simula and C++. It also supports non-causal modeling, meaning that model's terminals do not necessarily have to be assigned an input or output role. In fact, in the last few years it has been proved in several cases that non-causal simulation techniques perform much better than the ordinary object-oriented tools.

Dymola is a simulation tool for modeling of large systems, based Modelica language. There are already several Modelica libraries intend for use with Dymola for various applications domains, such as multibody systems, hydraulics, thermodynamical systems and chemical processes. Models are hierarchically decomposed into submodels. Reuse of modeling knowledge is supported by use of libraries containing model classes and by use of inheritance. Connections between submodels are conveniently described by defining cuts which model physical coupling. Special constructs are available for defining connection topology of composed models.

The Pulp and Paper Library

As an first effort in the project, a Modelica component library with process objects used in papermills was created. The initial focus was to model the wet-end part of the machine. This is basically a large hydrualic pipe network, consisting of open tanks, pipes, valves, pumps and specialized objects like pressure screens and pulp cyclones.

The wet end is of special interest from a system point of view because it contains dynamics of many different timescales, from orders of fractions of seconds (fast opening of valves) to hours, even days (slowly varying concentrations in the long circulation). It is generally agreed that the wet end system never reaches steady state. This causes obvius problems when shifting the plant between different operating points

Currently, the following component models have been implemented in the library:

1. Pipes, valves, pumps and tanks. These are basic components that make up the hydrualic flow networks.
2. Cyclones and screens(separators).
3. Headbox and wire sections.
4. Heat exchangers, heaters and coolers
5. DCS objects like pressure and flow indicators, PID controllers etc.
6. Dry end objects like press section and steam dryers are under development.

Modeling of the mill

The model and its structure are shown in the next page (See model).

In the real plant, the setpoints for the thick flow valve, the headbox pressure and the lip opening are calculated from the overall control system. In our simulations, these setpoints values were extracted from the Info database and fed to the model.

Despite the fact that only the main flows are taken into consideration, this model, all together, handles about 900 state variables (more than 10.000 general variables) and more than 1000 time delays.

The model was validate for different grade changes. (To see some of the validation click here). The simulation showed a good agreement with the measurements. The main causes of errors are unmodeled components and nonlinearities, and uncertanties in the concentrations.

Conclusions

At this stage of the project the results of the simulations were satisfactory. The main problem we encountered, was the settings of the initial conditions. In fact, the initial conditions have to be close to a set of stable values, otherwise the nonlinear solver may encounter problems in solving the differential equations.

The model can already be used for different aims: for example it can be employed to evaluate the effect of friction on the valves, to estimate time delays for an input change, qualitative testing of different control strategies or, in general, to get a better understanding of the wet end process.

Next, we are planning to complete the model, including the press and drying section, improving some of the present submodels to take into consideration some of the neglected nonlinearities and also chemical reactions.