Model-Based Approach to Soft-Sensing and Diagnosis for Control of a Continuous Digester

The goal of this project is to develop and demonstrate computing based modeling and control methodologies that will facilitate integrated operations in a pulp mill.  In particular, the focus of this work is on the continuous pulp digester.  The technical work required to achieve these goals include development of efficient methods for soft-sensing using fundamental models, integration of fault diagnosis and control algorithms, and the development of computationally feasible formulations of model predictive control for profile management in the digester.  In collaboration with both the pulp and paper industry, as well as pulp and paper control vendors, the development platform for this work is an industrial distributed computer control system.  The developed operational methodology for digester grade transition control will be benchmarked against an industrial design in cooperation with collaborators at Weyerhaeuser and Westvaco.  The key outcomes of this project will be a set of prototype tools and concrete results demonstrated on a digester benchmark problem and a mill trial.  After the project, the software tools are expected to evolve into commercial products helping manufacturers to implement computer-integrated grade transitions for pulp digesters.

It is a three-year (2000-2003) project with the University of Delaware as the recipient and Honeywell and IETek as sub recipients.


The overall objectives of the project are as follows:

Develop a model-based approach to manage diagnostics and control of a pulp digester using an inferential approach in which fundamental knowledge is combined with available process measurements to develop estimates of key product variables. These key product variables include infrequently measured process variables (e.g., final kappa number), fault conditions, as well as internal process characteristics (e.g., lignin profile along digester) which will lead to improved operation of the unit. Develop effective control strategies for management of transitions through production rate and grade changes.

Demonstrate this approach using a virtual benchmark simulation interfaced to two actual industry DCS packages. Under the new “open-standards,” these results should generalize in a straightforward manner to other industry DCS vendors. The modeling effort will be coordinated to comply with standard industry interfaces to simulation and optimization packages, such as Microsoft OLE standards and the standards being proposed from CAPE OPEN. Ultimately, the approach will be demonstrated on an industrial continuous digester.

Deliver a series of short courses and workshops throughout the project to get industry feedback and to introduce the tools to potential users.

General Experimental Approach:

A key element of the project is reliance on a fundamental process model for the digester. The essential contribution of the new generation model is the introduction of physical and chemical (rate) properties as “states” of the system so that grade transitions can be tracked. It will manage dynamic chip size distributions as well as a dynamic compaction profile linking chip level dynamics with kappa number control dynamics. Stochastic variability in the feed properties and operating conditions will also be introduced. Owing to the complex phase behavior of the system with a mix of co- and counter-current flows and entrapped / free liquor conditions, it is impossible to model the dynamics of multiple grades of chips in a digester with previous models. The new approach in this project will make it possible to compute the appropriate material and energy balances and diffusion limitations as chips of different origin pass through the digester. However, this will come at the expense of significant increase in the number of ODEs that need to be solved. Compared to approximately 2,000 ODEs needed for present digester models, the new approach will require more than 65,000 ODEs. IETek has already made significant progress on numerical procedures to efficiently solve the model equations that represent the new generation fundamental digester model. 

In addition, the University of Delaware has already developed an approximate fundamental digester model, which will be used as the basis for simulation study in the early phases of the project. This model will be interfaced to two industrial PC NT-based DCS (digital control system) packages (Bailey Infi-90 and Honeywell PlantScape) as will the research results of this project (using C-code interface). In that manner, early virtual benchmark studies will evolve into actual mill testing in a seamless manner.

A complete and hierarchical control strategy will be considered, including the layers of coordinated production rate control, coordinated grade change control, kappa number control, digester level control, alkali-to-wood ratio control, extraction factor and dilution factor control, plug movement index, and blow line consistency control.

© IETek 1996-2002, all rights reserved.


Home ] Advanced Digester Model ] Advanced Digester Toolbox ] Batch Digester ] Paper Machine Monitoring & Control Workshop ] ACC 2000 Monitoring Workshop ] [ DOE Digester Control Project ] Digester Benchmark Model v1 ] Digester Benchmark Toolbox v1 ] Custom Short Courses ] UDel PCMC ] Publications ] Projects ]