Kornel Warwas

An application of a hybrid algorithm to identification of parameters of semi-empirical model describing a real process

In the paper the problem of identification is formulated and solved using a hybrid algorithm. The identification means calculation of parameters of a model describing a real process. The method proposed is applied to a semi-empirical model of the working cycle of an engine which enables calculations of the pressure courses in a cylinder. The identification procedure determines the values of the computational model parameters in order to ensure the minimal difference between the measured and calculated pressure courses in the cylinder. The identification is performed by solving dynamic optimization task using a hybrid approach. In the hybrid algorithm genetic algorithms are connected with the gradientless Nelder-Mead method. Results of the identification for different vectors of input to the model are presented as well.

Artificial neural networks to control braking moments on wheels of an articulated vehicle

The paper presents an application of artificial neural networks to control braking moments on wheels of an articulated vehicle in a critical situation. The trajectory of the articulated vehicle was calculated for given braking moments by means of the 3D computational model. Dynamic optimization was performed in order to use appropriate values of braking moments for each wheel of the vehicle. Solutions of the optimization tasks for different inputs formed a set of optimal values. In the next step the set of optimal values was used for training an artificial neural network. Results of the calculations with the Nelder-Mead method and for the MLP and RBF neural networks are presented.

On multivariate method of qualitative analysis of Hodgkin-Huxley model with decision tree induction

This paper presents a data mining technique for qualitative analysis of Hodgkin-Huxley model of cell excitability. Such problem cannot be solved analytically. Therefore we apply Monte-Carlo techniques for the generation of model parameters, and use data mining algorithm for classification of learning tuples obtained. As a result we attain a decision tree capable of classifying the excitability depending on the relationship between initial conditions and between rate constants. Details of software development and issues of computational complexity are presented. The software includes the package of Java-classes. An example for classification of excitability of three types is presented.

Dynamic optimisation of articulated vehicle motion for control of stability in critical situation

In the case of unforeseen obstacles, the nominal trajectory of articulated vehicle has to be modified, in order to avoid collisions. The paper presents comparison of dynamic optimisation methods used to calculate optimal braking torques applied to wheels of an articulated vehicle in lane following/changing manoeuvre in order to prevent vehicle before rollover. Gradient, gradientless and evolutionary methods have been employed in order to solve dynamic optimisation problem. The aim of the optimisation is to calculate appropriate braking torques for each wheel in order to restore stability of an articulated vehicle. Calculation of the objective function requires integration of the equations describing the motion of the vehicle. In the paper results obtained from different dynamic optimisation methods have been presented and compared.

Distributed Neural Network Used in Control of Brake Torque Distribution

The paper presents implementation application of a feed-forward multi-layer neural network for a distributed system. A modified distributed topology with a changeable number of blocks is proposed. This enables us not to carry out parallel calculations on single processors. The distributed network is used to solve an optimization task in order to find such a distribution of braking torques applied to car wheels which assure a stable motion in difficult situation on a road. The model of the car, optimization task and results of calculations are presented. The use of distributed neural network considerably shortens the calculation time.

An Application of Neural Networks to Control Stability of an Articulated Vehicle in Real Time

The paper presents an application of neural networks to control braking torques on wheels of an articulated vehicle in an untripped rollover manoeuvre. The numerical model of the articulated vehicle and dynamic optimisation were used to calculate appropriate braking torques for each wheel in order to restore stability. The optimisation problem requires the equations of motion to be integrated at each optimisation step and the models cannot be applied for controlling the motion in real time. Therefore, in the next step multilayer perceptron network has been proposed. Multiple dynamic optimisation tasks have been solved in order to calculate appropriate braking torques, which form a training set of the neural network. In the paper, results obtained from dynamic optimisation and neural network have been presented and compared.

The tupel model of basic components' set formation for cyberattacks

The unauthorized access to information systems resources, for instants cyberattacks, can impact on its surrounding area, causing the certain anomalies. Mostly this area is not formalizing and not clearly defined. To detect cyberattacks required specialized tools, which make it possible using a set of specific features. One of the approaches to solving this task is based on using corresponding models, methods and intrusion detection systems based on fuzzy sets focused on processing of semistructured data with a purpose to establish the facts of unauthorized access to information systems resources. In this work proposed, based on a tuple model, the formalization approach to build up the necessary dataset component, which are used to control the status of cyberspace parameters at a given time in a fuzzy conditions. This approach gives a possibility to identify the abnormal state generated by the action of the relevant cyberattacks class.

Mathematical modelling of infiltration effect on process of salts transfer in layered saturated-non-saturated soils

Mathematical modelling of the salt movement process under filtration salt transfer considering infiltration process in unsaturated-non-saturated layered soil has been carried out. The solution of this problem was received with a finite difference method. As a result of program implementation of the problem, numerical experiments have been performed and an analysis of the results has been done.

Analytic design of the optimum control system for five-degree-of-freedom stand simulator of the spacecraft motion

In this paper it is proposed the methodology of analytic design of the optimum structure for stochastic control system used on multi-degree-of-freedom stand simulator of spacecraft motion in the presence of deterministic and stochastic disturbances acting on it.

Generalized cost functions of avionics breakdown maintenance strategy

In this study, the generalized cost functions are proposed for choosing the optimal option of breakdown maintenance strategy of avionics systems. A mathematical model of avionics line replaceable unit (LRU) is developed. The model considers the main characteristics of the preflight checks and continuous testing of the LRU using built-in test equipment (BITE) in-flight. The equation for the mean time between unscheduled removals (MTBUR) is derived for an arbitrary and exponential distribution of time to failure. The cost functions are determined as the total operating costs separately for the warranty and post warranty period of operation. Different options of single-level, two-level and three-level maintenance are mathematically modeled and numerically analyzed for the warranty and post-warranty period of operation. The proposed analytical expressions take into account the trustworthiness of BITE, periodicity of preflight testing, cost of different maintenance operations, permanent and intermittent failure rate of LRUs and some other parameters. Numerical examples are included to illustrate the main features of the proposed mathematical models.