COMPONENT APPROACH OF SIGNAL TRANSMISSION AND PROCESSING OF RESULTS DURING ULTRASONIC CONTROL

Abstract

The paper addresses the challenge associated with the diversity of devices for ultrasonic testing and the necessity to unify the obtained data for analysis in a single format, which simplifies the control process and allows for the development of new analysis methods. The authors emphasize the importance of traceability to the original data source for a deeper understanding of the information received and its reliability. The primary goal of the research is described as the development of software and algorithms for the effective processing and analysis of data obtained from ultrasonic testing. The paper provides a review of various studies and developments in the field of ultrasonic testing and the use of machine learning for data analysis. Several approaches and technologies are described, including the use of a machine learning model to predict the strength of joints in 3D-printed concrete and methods for assessing the quality of coatings applied by additive methods. The authors delve deeply into the theoretical foundations of their approach to data analysis, using concepts of finite state automata (FSA) and constructive interaction models (CIM). Finite state automata are described as tuples that include a set of states, an alphabet, transition functions, an initial state, and a set of accepting states, allowing the modeling of system behavior through sequences of transitions between states. The concepts of services and their fragments are detailed in the context of software, where each component performs a specific function or processes messages. An important aspect is the modeling of system execution tracing using timed sequence diagrams, which allows for detailed interaction between components. The method of creating automated prefix trees (APT) for each service fragment is described, enabling detailed modeling of system behavior based on the executed program code. Significant attention is given to the analysis of timing annotations in state machines, with a particular focus on defining and interpreting changes in temporal behavior. Due to the specifics of working with time data, the research highlights challenges related to noise, non-normality of distributions, and small sample sizes, which are typical for analyzing the performance of real systems. The integration and functionality of the Mids+Time user interface are described in detail, combining timed automata, change detection heuristics, and visualizations in a single interface. The Mids+Time user interface provides interactive visualizations, including histograms, control charts, and time-shift graphs, which allow users to effectively analyze and compare the performance of different software components.