Volume 27 (2019): Issue 45 (September 2019)

The paper deals with a methodology of designing the machine parts subjected to elastic-plastic contact conditions. Elastic-plastic contact conditions are simulated by finite element (FE) simulation. As the result, the estimation of contact loading stresses in practice can be made using contact geometry, material datasheet or tensile test.

The paper presents the possibility of using machine vision in the industrial area. The case study is oriented to indirect image processing in a robotic cell using a Matlab tool. Theoretical part of the contribution is devoted to the comparative analysis of various methods of the object detection and recognition. Analysis of the functionality, speed, performance and reliability of selected methods in the object detection and recognition area is processed. In the practical part, a method of implementing an indirect machine vision is designed to control the handling of objects detected and recognized on the basis of an operator requirement. Based on the analysis of the sample robotic workplace and the identified limitations, possibility of using the indirect computer vision is suggested. In such a case, the image of the workspace scene is saved on the storage and then processed by an external element. The processing result is further distributed in a defined form by a selected channel to the control component of the production cell.

With incoming Industry 4.0 implementation, RFID technology is becoming one of the main part for (not only) automotive industry, as a source to identify any inconsistencies or failures in automated processes. Its implementation in the automated workplaces is conditioned by the real-time data collection depending on workflows and increasing demands for processing, and publishing of these data. As such, RFID technology enables the real-time data integration from objects as well as from people or inventory management, asset management and monitoring, alarm collection and evaluation, predictive maintenance, and overall critical digital communications.

The article is focused on testing selected communication protocols for the Internet of Things. The aim of the experiments was to find a suitable communication protocol to build a smart device of home automation. The recent developments in the field of communication protocols and data processing are providing a new form of constrained communication, while describing how the Things talk to each other in the Internet of Things. With growing number of connected devices in industry and commercial field, there is a need for the devices which can handle the new forms of communication, handle sensors and operate the battery power. The gathered data is either processed on the “edge” of the network or on the cloud platforms. Authors present a comparison of selected newest communication protocols, while reflecting on the results of their experimental testing.

The design and operation of modern industrial systems require modeling and analysis in order to select the optimal design alternative and operational policy. Discrete event system models are encountered in a variety of fields, for example computers, communication networks, manufacturing systems, sensors or actuators, faults diagnosis, robotics and traffic. The paper describes principles and methods of supervisory control of discrete event systems initiated by Ramadge and Wonham. Three supervisory control methods based on the Petri net models are introduced, and the key features of the Petri tool software application for the supervisory control of discrete event systems modeled by Petri nets are highlighted.

This paper is devoted to design of a non-linear model predictive controller (NMPC), which will swing-up and stabilize an inverse rotary pendulum known as the Furuta Pendulum. This paper presents a simulation validation of the NMPC strategy using a full-fidelity non-linear mathematical model of the Furuta pendulum obtained from the Euler-Lagrange motion equations. The NMPC strategy was implemented in MATLAB using the MATMPC toolbox.

The arrival of Cyber-physical systems provided space for a new emerging field of research oriented on this type of embedded systems. The aim of this paper is to provide a better understanding of this integrative research field focused on the concept, architecture and challenges in deployment of such systems within the concept of Industry 4.0. Cyber-physical systems represent an emerging area of research that attracts the interest of researchers around the world, because, in the field of design and development of future systems, they are expected to play a major role.

The article describes a possible way of implementing a neural network in recognizing the shape and position of the products in the production process. The neural network is designed as a multilayer perceptron (MLP), and the whole system is implemented in a form of attachment to robotic arm, where the primary task of neural network is to distinguish a position of product. The neural network is trained like a classifier and outputs are used to control the robot. The advantage of the solution is a high degree of reliability of product positioning under different lighting conditions.

In an optimal processes control, where the considered goals are in general observed as concurrently conflicted, a multi-objective approach fits the best. Commonly used scalarization techniques in multi-objective optimization need a transformation of the individual single-objective functions involved into a scalar multi-criteria objective function. There are many parameters which can influence the optimization results solutions, including an unreachable utopia point value. In this study, the authors compare the multi-objective problem solutions found via two ways of the individual objectives transformation with the respect to setting the utopia point. The methods are used in the area of production control in a case study for a batch production system. To find the solutions, The Weighted Sum Method with a priori articulated preferences under specific constraints as the scalar multi-objective optimization method is applied in simulation optimization.

This paper aims at the time-series data analysis. We propose the possibility of adding additional features to the existing time series data set, to improve the prediction performance of the prediction model. The main goal of our research was to find a proper method for building a prediction model for the time-series data, using also machine learning methods. In this phase of research, we aim at the data analysis and proposal of the ways to add additional features to our dataset. In this paper, we aim at adding derived parameters from one of the original features. We also propose incorporating LAG’s into the dataset as new features, to enhance the prediction performance on the time series based data.

In this paper, we present the impact of the data normalization on the classification model performance. In first part of this paper, we present the structure of our dataset, where we discuss the features of the data set and basic statistical analysis of the data. In this research, we worked with the medical data about the patients with the Parkinson disease. In second part of this paper, we present the process of data normalization and the impact of scaling data on the classification model performance. In this research, we used the XGBoost model as our classification model. The main classification task was to classify whether the patient is ill with Parkinson disease or not. Since the data set contains more numerical parameters of different scaling, the main aim of this paper was to investigate the impact of the data normalization (scaling) on the performance of the classification model.

The paper deals with the solution of the first–order passive filters (low–pass and high–pass) applying electrotechnical elements (resistor, capacitor - analogue filter) and digital Butterworth filter type IIR (Infinite Impulse Response). Procedure of the filters design and implementation is described, and the analogue and digital filter outputs with the same input signal are compared. The designed filters have already served for education purposes with the intention to bring an explanation of techniques for designing required functionality of the signal processing filters.

The contribution deals with measurement of acoustic absorption coefficient for different single or double-layer materials: cork, two layers of polyethylene, polyethylene and felt. The measurement was performed on an impedance tube of our own construction, using the two-microphone method transfer function (ISO 10534-2: 1998) and the PULSE measuring system. Values of the acoustic absorption coefficient for the frequencies from 100 Hz to 1600 Hz were determined experimentally. Subsequently, those values were processed graphically.

This paper deals with application of a modified Fréchet metric to self-organizing neural networks, called Kohenen maps. The methodology used allows us to put more emphasis on the selected parameters in the input data. It can simplify finding the minimal distance dF_j, since dF_j∈ 〈0,1〉

This paper aims at deeper exploration of the new field named auto-machine learning, as it shows promising results in specific machine learning tasks e.g. image classification. The following article is about to summarize the most successful approaches now available in the A.I. community. The automated machine learning method is very briefly described here, but the concept of automated task solving seems to be very promising, since it can significantly reduce expertise level of a person developing the machine learning model. We used Auto-Keras to find the best architecture on several datasets, and demonstrated several automated machine learning features, as well as discussed the issue deeper.

Deep learning is a kind of machine learning, and machine learning is a kind of artificial intelligence. Machine learning depicts groups of various technologies, and deep learning is one of them. The use of deep learning is an integral part of the current data classification practice in today’s world. This paper introduces the possibilities of classification using convolutional networks. Experiments focused on audio and video data show different approaches to data classification. Most experiments use the well-known pre-trained AlexNet network with various pre-processing types of input data. However, there are also comparisons of other neural network architectures, and we also show the results of training on small and larger datasets. The paper comprises description of eight different kinds of experiments. Several training sessions were conducted in each experiment with different aspects that were monitored. The focus was put on the effect of batch size on the accuracy of deep learning, including many other parameters that affect deep learning [1].