Michal Blaho (Humusoft)
MATLAB helps researchers, developers and engineers create new algorithms and devices. It is published twice a year with a lot of news in the basic module and individual add-ons. During the presentation, you'll see new options for modeling, simulating and sharing designs, as well as new tools to increase your productivity and create better code and models. News will focus on areas such as:
Marco Rossi (MathWorks)
MATLAB, Simulink, and RoadRunner help engineers to build automated driving systems with increasing levels of automation. In this session, you will discover new features and examples that will allow you to:
Jaroslav Jirkovský (Humusoft)
Virtual vehicle (virtual vehicle) is a term used for the virtualization of the development cycle of a vehicle, based on simulations of the vehicle's behavior at the system level. Simulations with a virtual vehicle allow you to quickly gain insight into the vehicle's behavior in the real world, perform virtual testing of various scenarios and verify the functionality of the embedded software (embedded sw). The use of a virtual vehicle helps to speed up the assessment of variants, safely study borderline cases and improve the overall quality of the developed system. Typical steps in creating a virtual vehicle include modeling the vehicle, integrating embedded software, defining test scenarios, simulating and analyzing the results.
In the field of electromobility, simulation models can be used to design and optimize systems typical of electric vehicles. It can be battery modeling and BMS (battery management system) development, fuel cell modeling, traction drive modeling and their control units, integration and testing of control algorithms, or the use of data-oriented approaches and artificial intelligence.
Michal Blaho (Humusoft)
Time series data is all around us. Whether it's sensors on automated vehicles and production lines, weather data or financial data from the stock market, they help us understand system behavior over time. However, real-world time series data can have many problems, such as missing data, outliers, or noise. Data must first be cleaned and prepared before it can be analyzed or used for model development. The process is iterative and can be very time consuming. MATLAB provides several tools for data preparation and preprocessing such as Live Editor Tasks, Data Cleaner and Signal Analyzer applications, which we will introduce.
Jaroslav Jirkovský (Humusoft)
Any application using artificial intelligence algorithms needs a sufficient amount of representative data, a good model for machine/deep learning, and suitable data manipulation functions that transform the raw data into a source of information suitable for automatic learning.
A data-oriented approach to artificial intelligence relies on expert knowledge of a specific domain, where appropriate data preparation leads to the improvement of the results provided by the AI algorithm and enables the deployment of artificial intelligence in real applications. We will show the current possibilities of data preparation to improve its quality, reduce variance and dimension, choose an optimized data representation using flags or determine an appropriate transformation.
In the resulting applications, artificial intelligence models are often used together with other algorithms. To optimize the behavior of the entire system, it is advisable to check not only the behavior of the AI models themselves, but also their interaction with other elements and external systems. The Model-Based Design method, used to develop algorithms based on simulation models, will enable the joint development and testing of artificial intelligence models with the rest of the system, optimize their interaction, detect potential integration problems in time and bring the design to the form of target deployment.
Martin Kožíšek (Humusoft)
In the lecture, the simulation tools COMSOL Multiphysics, COMSOL Server and COMSOL Compiler will be presented on a specific task. Emphasis will be placed on a demonstration of the simulation creation workflow from entering parameters to processing the results. How can software make the simulation engineer's job easier?
Jaroslav Jirkovský (Humusoft)
Predictive analytics, anomaly detection and visual inspection represent three areas of industrial deployment of modern computing methods based on systematic operational data collection, mathematical modeling and artificial intelligence.
Predictive analytics uses historical data to predict future events. Historical data is used to build a mathematical model that captures important trends. The model is then applied to the current data to predict what will happen next, or to suggest actions that need to be taken to achieve optimal results. Predictive analytics finds application in such diverse industries as the automotive industry, the aerospace industry, manufacturing, energy, finance, and others. A typical task for predictive analytics in industrial applications is predictive maintenance of systems and estimation of remaining time to failure, enabling efficient maintenance planning and timing of production shutdowns with minimal cost.
The second area of industrial deployment of data analysis methods is the monitoring and classification of the condition of technical equipment. Such a system may include anomaly detection tools. In principle, this is also a classification task, but it can also be used if we only have data about the system from „normal“ operation without previously known data from fault conditions.
A third industrial application is automated defect detection, which is essential for efficient quality control in manufacturing systems. Visual inspection is widely used in many industries to detect defects on various types of manufactured surfaces, such as metal components, semiconductor wafers, contact lenses, and so on. Recent developments in deep learning have brought new tools to automate visual inspection tasks with unprecedented accuracy and robustness. New methods make it possible to find arbitrary defects without the need to use „defect“ data during the learning of the detection algorithm.
Martina Mudrová (Humusoft), Martin Foltin (Humusoft)
A brief overview of the current possibilities that MATLAB and its complementary services offer for teaching and learning at universities, as well as for start-up companies.
Mauro Fusco (MathWorks)
As systems become more complex, engineers are being challenged to do more despite shrinking development timelines and budgets. Learn why engineers working in aerospace, automotive, robotics, medical devices, and industrial automation machinery and more rely on a suite of interconnected platforms from MathWorks to design, build, test, and deploy the systems of tomorrow. Model-Based System Engineering and Design incorporates verification and validation into the software development workflow. As a result engineers use intuitive programs to simulate and analyze architecture, algorithmically compare trade studies, and create a digital thread that spans the entire system.
By joining this talk, you will learn the best practices for improving software quality while reducing development time and costs.
Highlights
Tomáš Fridrich (Humusoft)
Real-Time testing has become an integral part of the development cycle. dSPACE offers a wide range of hardware and software solutions. The presentation will outline the basic procedures for creating applications for Real-time testing. We will show interesting areas of use of real-time platforms not only in automotive, but also from other industrial areas. Finally, I will tell you the news.
Anna Tocháčková (Humusoft)
You will learn about the various options for data visualization using graphs in the MATLAB environment. In MATLAB, you can draw „general purpose“ graphs, such as line, bar, pie, area, or volume charts. In addition, however, you can use charts specially prepared for specific purposes, such as charts for statistical visualization, oriented charts, map backgrounds and others. Some of them also provide additional computing equipment. In the samples, you will see several different types of graphs that you can use as inspiration for visualizations and presentations of your data.
Jakub Zábojník, Radek Papoušek (Continental Barum s.r.o.)
For the needs of our department and colleagues from abroad, we have created a series of tools in MATLAB AppDesigner that make our daily work easier. We will briefly introduce individual tools and their purpose and use. All applications are based on our specific needs. In some cases, they replaced outdated software, new requirements arose somewhere, and in other cases we replaced Excel. As non-programmers, we are still learning, and thanks to AppDesigner's ease of use and clarity, we are able to create our own applications easily and relatively quickly.
Filip Šroubek (ÚTIA AV ČR)
Deconvolution methods that remove blur in an image use the Fast Fourier Transform (FFT) for greater efficiency. Matlab methods such as „deconvwnr“, „deconvreg“ and „deconvlucy“ are no exception. The FFT assumes that the image is periodic, but this is not usually true in practice, and in that case disturbing artifacts are created at the edge of the image. We proposed a spectral adaptation (SPA) method that adjusts the blurred image to approximate the assumption of periodicity, and then deconvolution does not produce artifacts. The SPA method not only gives better results than the standard Matlab „edgetaper“ function, but is also more general. For example, it allows increasing the image resolution or compensates for problems associated with an imprecisely determined convolutional kernel.
Martin Šiler (Ústav přístrojové techniky AVČR, v.v.i.)
We can also use machine learning in MATLAB to identify (classify) dozens of types of bacteria in the blood that cause various diseases. For rapid analysis, we use light Raman spectroscopy, i.e. light scattering on bacterial cultures, during which the wavelength of light changes, thus obtaining a „unique fingerprint“ of the bacteria. However, the data entering the analysis suffers from a large number of ailments: noise, is mixed with another useless signal, few data, unevenly represented classes, etc. In the lecture, we will show how different pre-processing and feature selection procedures will affect the quality of the resulting classification.
Josef Souček (ČVUT FBMI a Hasselt University (Belgie))
A point defect based on a pair of nitrogen and a vacancy (Nitrogen-vacancy, NV) is among the intensively investigated spin systems in the crystal lattice of diamond. For the development of quantum chips for optimal excitation, a precise characterization of the quantum response of this center is needed. The work is focused on modeling the crystal environment of the diamond lattice and complex energy and charge transfer phenomena such as electron and photon emission or drift of charge carriers from the NV center during laser and microwave excitation.
Patrik Kováčik (Žilinská univerzita v Žiline)
The use of artificial intelligence in the engineering industry is a relatively new, rapidly developing issue. In this presentation, you will find out how to look at a classic engineering problem in a new way – we will go through the main functions of MATLAB that allow optimizing engineering structures, as well as the potential of artificial intelligence to improve the given process in the engineering world. Our project is aimed at optimizing drones intended for delivery of parcels so that they can carry the heaviest parcels for their „body proportions“.