This paper deals with different methods for minimum variance control of linear, timeinvariable and sampled single-input single-output systems. Furthermore the amplitude of the control signal will be constrained. The most natural method is to make use of a minimum variance strategy, which minimizes the variance of the output signal. But this only when no saturation is applied. Saturation of the inp

The master thesis intends to investigate the advantages and disadvantages with the simulation language SIMSCRIPT. The programming language was learned from manual and books about SIMSCRIPT (see references). Two models were made ato describe two different systems, one fictive and the other already existing. The first model simulates the work at an assembly line. The model illustrates the use of the

First in this master thesis some classical dynamic models for macroeconomic systems are presented. Then follow comparisons between keynesian and monetarist models. The greatest part of the thesis deals with different constructions of monetarist models for Sweden in the shape of linear difference equations with constant coefficients. The starting-point has been a well-known examination of the U.S.

This report includes: A definition of a language suitable for the description of nonlinear systems. Development of FORTRAN-routines that performs a compilation of the language to a pseudo-code and a routine to interpret and execute this code. The welding of the compiler and the execution-routine into a complete interactive simulation-package.

In this report a program for identification by the Maximum-Likelihood method is presented. The program is written in FORTRAN for a process computer, PDP-15. The application is limited to linear, single output, time-invariant dynamic systems with normal disturbances having rational spectra.

When regulating a process, the system being sampled, one can often regard the output as white noise. Often occuring is a wish to obtain an output as close to an optimal value as possible. At the same time, however, you demand this optimal value not to be exceeded too often. Thus, when you decide your set-point you have to consider the value of the standard deviation of the white noise. If the stan

This report deals with the problem of learning to classify observations from a mixture of random variables. The distributions of the random variables are of known functional forms but with unknown parameters. This problem is equivalent to that of determining separating hypersurfaces in the range of the random variables. Any rule of classification will be a function of random variables and therefo

I detta examensarbete redogörs för tre metoder att införa integratorer vid linjärkvadratisk optimering. För att jämföra metoderna har två system simulerats. Metoderna har i huvudsak jämförts med avseende på snabbhet, maximal amplitud och känslighet.