Main contents


Testing Manager

The goal of the Testing Manager project is to provide a flexible and comprehensible environment for simulation assisted testing and tuning of industrial automation systems. This means well-defined working practices, practical tools supporting the testing and commissioning phases, as well as open connectivity of the simulation and automation software.

In practice, a significant amount of industrial controllers behave unsatisfactorily. One common reason for this is inappropriate tuning of controllers. This may cause undesirable variations for end product quality, inefficient utilization of raw materials or increased emissions of pollutants. Traditionally, controller tuning was performed using single loop tuning methods, which is rather laborious and time-consuming. Furthermore, such approaches do not at all take into account interactions between control loops. Inevitably, this results in a non-optimal solution in terms of overall system performance.

Statistical multivariate regression methods are particularly suitable for modeling of multidimensional systems. These methods share the same fundamental idea of dimension reduction. It is assumed that relevant information on the original multidimensional measurement signals can be presented in a lower dimension with a set of appropriately selected latent variables. This reduction of data dimension works as a powerful tool against the numerical problems caused by noisy and collinear measurement signals.

Research concentrates on a novel iterative multivariate controller tuning method, Iterative Regression Tuning (IRT). The tuning technique is based on sequential simulations and model estimations of a static stochastic model. The model describes the dependency between control parameters and quality measures that are defined by the control objectives of the system. This information is used for iterative updating of parameters towards the optimal direction in the statistical sense. During the projects, tests were performed on nuclear and combustion power plant and pulp mill simulation models

Our research is financed by TEKES Intelligent automation systems. Our research partners are VTT, Tuotteet ja tuotanto, Systeemidynamiikka. Our industrial partners are Fortum Nuclear Services and Metso Automaatio.

  • Keywords: tuning
  • Duration: 2003-2004
  • Research area: Cybernetics
  • cURL error: name lookup timed out