Testing the quality of manufactured products based on their sound expression is becoming popular nowadays. To maintain low production costs, the testing is processed at the end of the assembly line. Such measurements are affected considerably by the factory noise even though they are performed in anechoic chambers. Before designing the quality control algorithm based on a convolutional neural network, we do not know the influence of the factory noise on the success rate of the algorithm that can potentially be obtained. Therefore, this contribution addresses this problem. The experiments were undertaken on a synthetic dataset of heat, ventilation, and air-conditioning devices. The results show that classification accuracy of the decision-making algorithm declines more rapidly at a high level of environmental noise.
This paper presents the current results of cooperation focused on automatic billet straightening machine development. First, an experimental study of three-point bending realized on small specimens is presented to explain the basic ideas of the straightening. Then, the main regimes of straightening and the algorithm itself are described together. Subsequent finite element simulations of operational experiments show the applicability of the developed theory. The significance of material parameters estimation is depicted in this work. At least four parameters have to be properly determined for a new material in the straightening process.
The article describes a unique method of processing scanned point cloud on a moving object with variable speed. Point clouds represent a set of trucks loaded by a scanning gate made up of four 2D laser scanners. Three scanners record the surface of the truck from three sides, the fourth scanner records the longitudinal change of the profile of one of a truck's side. From this profile, the method evaluates the motion movement function by which all other point clouds are recalculated. The result is a modified point clouds representing the surface of the standing set. The article also provides a way of point segmentation that separates the points representing the timber load from the truck.
Industrial companies focus on efficiency and cost reduction, which is very closely related to production process safety and secured environments enabling production with reduced risks and minimized cost on machines maintenance. Legacy systems are being replaced with new systems built into distributed production environments and equipped with machine learning algorithms that help to make this change more effective and efficient. A distributed control system consists of several subsystems distributed across areas and sites requiring application interfaces built across a control network. Data acquisition and data processing are challenging processes. This contribution aims to present an approach for the data collection based on features standardized in industry and for data classification processed with an applied machine learning algorithm for distinguishing exceptions in a dataset. Files with classified exceptions can be used to train prediction models to make forecasts in a large amount of data.
The paper deals with modeling and simulation of the event-driven model of the production line G. U. N. T. in real time. The event control was realized by the design of the simulation model which is composed in MATLAB & Simulink and with using blocks from Data Acquisition Toolbox. The simulation event-driven model was designed, realized and tested for supply, transport and sorting of material by the conveyor belt on the model of production line.
Every truck heading into the paper mill of company MONDI, Slovakia have to be checked if storage of its timber load meets all safety rules. Each truck is scanned by a group of 2D scanners. Then the evaluation of timber load is done by software with use of the data received from the scanners. Checking software is universal for every kind of storage of timber loads. This article deals with a problem of recognition a kind of storage of timber load on a semi-trailer. The image recognition is done with use of data recorded by a set of cameras situated with its view on the driveway. By detecting a type of storage, it is possible to evaluate the safety check for a specific type of storage with better result than the universal check.