Abstract:

The article focuses on the potential of the accessible software environments used in distance learning of master degree in university engineering education. An example of such an environment is MATLAB, due to the fact that it combines a working environment set up for iterative analysis and design processes with a programming language, which is expressed in the mathematics of matrices and arrays. MATLAB applications allow learners to monitor the operation of various algorithms with the data entered by them until they obtain the desired results, after which the program is generated automatically to reproduce or automate the activity.

The authors of the article share their experience of the possibility of applying stochastic algorithms in order to identify the structure of linear systems, while using a genetic algorithm (GA) and optimization through the intelligence of the flock (Particle Swarm Optimization - PSO) to assess the order of extended autoregressive (ARX) model.

Topics such as teaching and self-training of intelligent systems are especially relevant and useful for modern engineering research. More and more managers from different companies justify their management decisions through an in-depth analysis of the accumulated large amounts of company data. Therefore, the authors believe that it is of a vital importance for students trained in master's degrees in disciplines such as "Hardware and Software Systems" and "Information Security" to understand and apply the structural identification of systems using artificial intelligence algorithms.

It is emphasized on the methods of structural identification by model parameters and the used criteria, which take into account the adequacy and order of the model: Information criterion Aikake (AIC), Bayes-Schwartz criterion (BSC) and Residual sum of the model error square (RSS).

The capabilities of the arxstruc functions in Matlab were used to identify the structure of the ARX (autoregressive external model). The report presents the syntax of the command, which is graphically illustrated.

The graphs of the position of the factors in the parameter space in the GA-tests and the position of the factors in the parameter space in the PSO tests are presented and analyzed. The laboratory model PT326 of the Feedback company is used during the experimental research. The input and output data are presented graphically. The results for the first 20 out of 90 tests obtained using GA are presented in a table. The results of the evaluation of the structure of a linear ARX model were obtained using PSO. A comparative analysis was made.

Through the conducted simulation and experimental researches, the master students have the opportunity to make the following conclusions:
- the structure of the ARX model is suitable for estimating the structure of real objects, using common or evolutionary algorithms;
- the size of the population of 30 individuals is sufficient for the procedure of estimating the parameters of the low-order ARX model, using an approach based on the genetic algorithm;
- In the task of identifying the structure of the ARX model, the PSO-based approach is not appropriate. In this case, GA works well and gives a relatively accurate identification of the structure.

Keywords:

Software environments, teaching, engineering education, structural identification, genetic algorithm, optimization, flock intelligence, ARX model.