Monitoring and validation of control for bio- and food processes
This project continues the work done in Multiparameter polynomial models for bioprocesses. It belongs to a larger Intelligent methods for management of dynamic systems (DYHA) -project coordinated by VTT Automation. DYHA is one of the projects of Adaptive and Intelligent Systems Applications 1994-1999 -program of TEKES.
In Wiener-NN modeling, described briefly in Multiparameter polynomial models, characteristics of dynamics are described with orthogonal polynomials. Kernel functions, generalized impulse responses, are approximated with orthogonal expansions. A priori structural information about the process can be contained in the model by selecting a suitable orthogonal basis and its rough parameters. Parameters in continuous formulation have clear physical meaning and are thus preferable, although discrete models are are needed for programs. In practical implementation, orthogonal expansions of input signals and also of output signals fed back (in Wiener-NN model with feedback) are actually calculated on-line.
In Wiener-NN models, instead of sliding data windows of signals, orthogonal representation of the signals are used. Sliding data windows are not economic nor robust way to represent long histories, but with orthogonal expansion long histories are. The same model can contain both long histories of certain signals, connected with with slow modes, and shortly histories of certain other variables responsible for fast dynamic modes. This property of describing robustly and economically different temporal scopes of signals is valuable not only in dynamic modeling but also in recognition of dynamic patterns for monitoring purposes. It can be applied in a natural way in fault diagnosis and validation of control particularly for batch type processes, where certain characteristic signal patterns should emerge during operation. Quality control can be enhanced with these kind of algorithms.
The Wiener-NN approach is utilized in fault diagnosis by implementing parallel estimators for key state variables, which base on information from different subsystems; cooling system, outgas measurements and pH-control. The process events can be seen in redundant way in all these sensor groups for these subsystems. The redundancy is utilized in fault detection and analysis
The goals of the project is to innovate and demonstrate generic methods for management, monitoring and validation of control of batch type or cyclical type continuos processes. The research is particularly aimed for biotechnical and food processing industry, where validation of processes is becoming really important. Roal Oy and Cultor Oy are participating in this project.
Tel.: +358-9-470 23306
Fax: +358-9-470 23308