Statistical Diagnostics of Electric Power Equipment

This chapter delves into the intricate field of statistical diagnostics for electric power equipment, emphasizing the integration of modern computational and statistical methods rooted in pattern recognition theory. It provides an extensive overview of the Ukrainian power industry, highlighting the distinct roles of thermal, hydro, and nuclear power plants and their associated reliability challenges. Key diagnostic methods—non-destructive testing, vibration, and acoustic emission signals—are discussed, focusing on their critical application in identifying faults in electric machines, particularly winding and bearing damages which contribute to the majority of failures. The chapter also examines the diagnostic processes, touching upon the historical evolution of measurement technologies, the development of diagnostic systems, and compliance with industry standards. Furthermore, it explores mathematical modeling and statistical spline functions for forecasting failures, demonstrating how these methods can enhance the prediction and reliability of industrial equipment by monitoring gradual changes in diagnostic parameters. The comprehensive analysis presented integrates theoretical foundations with practical applications, offering substantial insights into the methodologies for ensuring the optimal functioning and longevity of electrical machinery.

This comprehensive research delves into the mathematical modeling of diagnostic signals in electrical equipment diagnostics using linear random processes (LRPs) and their properties. Beginning with an exploration of stochastic processes with independent increments, the chapter establishes foundational concepts crucial for understanding LRPs. These processes, characterized by their reliance on past and present values, are further examined for their applicability in both continuous and discrete-time scenarios. Special attention is given to the properties of LRPs, including their closure under linear transformations and the simplification of probabilistic analysis through characteristic functions. The chapter extends the discussion to multidimensional random fields, showcasing the versatility of LRPs in modeling complex phenomena across various fields like radiophysics, radio engineering, and optics. Additionally, the exploration of stationarity and homogeneity within random fields elucidates the conditions under which these models can be simplified, further contributing to their practical utility in diagnostics. By providing a fundamental understanding of these processes and demonstrating their applicability through examples, this chapter aims to equip researchers and practitioners in the field of electrical equipment diagnostics with robust mathematical models for diagnostic signals.

This chapter presents a comprehensive mathematical framework for the analysis and diagnostic application of vibration and acoustic emission (AE) signals in electrical equipment (EE). Through the development of linear random processes and multi-resonant models, it offers a detailed representation of vibration and AE signals generated by rolling bearings and laminated magnetic circuits within electrical machines. The models are designed to decipher the complex condition of equipment components by analyzing vibrations caused by mechanical interactions and AE resulting from operational stresses. Key contributions include the establishment of a vibration model that simulates the multi-resonant system's response to generating impulses, reflecting the dynamics of rolling bearings, and a model for AE that integrates continuous and discrete signal components. This dual approach allows for a nuanced analysis of EE conditions, particularly focusing on bearing vibrations and laminated magnetic circuit responses to operational defects. The proposed models are centered on the understanding that diagnostic signals can be dissected into probabilistic characteristics, offering novel diagnostic signs through the statistical estimation of signal parameters. These models are crucial for identifying defects and monitoring the condition of EE, with significant implications for enhancing reliability and extending the lifespan of such equipment. The findings propose a methodological advancement in the diagnostic domain, providing a solid foundation for future research into the application of these models in real-world diagnostics and preventive maintenance strategies.

This comprehensive research explores the application of the linear periodic random processes (LPRP) and the linear random processes (LRP) models’ parameters in diagnosing and understanding the technical conditions and operational issues of diesel engine generators (DEG) and electric machines (EM). By integrating these models, the study aims to capture the cyclical and stochastic nature of the operational behaviors of EM and their components, particularly focusing on the vibrations and rotational dynamics of DEGs and EMs. Through the analysis of the uneven rotation of the crankshaft and the distribution of cylinder power, the research demonstrates how LPRP and LRP can be effectively used to diagnose potential malfunctions and optimize the performance of these systems. The methodology includes measuring the kinetic energy of the shaft, calculating acceleration deviations, and applying discrete Fourier transform to identify harmonics indicative of operational integrity or issues. The findings suggest that the absence or presence of specific harmonics can diagnose uneven cylinder power distribution, crucial for maintaining efficient and reliable operation. This chapter extends the application of LPRP and LRP in analyzing stochastically periodic behavior, offering significant insights for enhancing diagnostic techniques for DEGs and EMs, marking a pioneering step in applying periodic random processes to mechanical diagnostic fields.

This chapter presents a comprehensive exploration of statistical methods for the estimation and analysis of diagnostic signals in a variety of contexts, ranging from spectral analysis to parameter estimation and hypothesis testing. Beginning with numerical parameters of diagnostic signals, the discussion extends to spectral correlation analysis, focusing on the calculation of moments, the power spectral density, and the correlation function from signal samples. The utilization of the Goertzel algorithm for efficient spectral density estimation highlights the chapter's contribution to spectral analysis techniques. Further, the application of Pearson's system of curves for empirical data distribution approximation underscores the chapter’s innovative approach to statistical analysis. The extract also delves into stationarity testing methods (F-test, t-test, and Kolmogorov–Smirnov test) critical for diagnostic reliability and defect identification, illustrating the chapter's diagnostic focus. In addressing statistical hypothesis testing, the chapter introduces a method based on the maximum likelihood principle for distinguishing hypotheses about the parameters of a multidimensional Gaussian distribution, demonstrating the chapter's depth in statistical methodology. Lastly, it explores the use of the Neyman-Pearson criterion in planning diagnostic tests, offering a practical framework for hypothesis distinction regarding system states. This exploration underscores the chapter’s significant contribution to both the theoretical and practical aspects of diagnostic signal analysis, providing valuable insights and methods for researchers and practitioners in the field.

This chapter explores the advancement in diagnostic signal simulations for electrical equipment, emphasizing acoustic emission (AE) signals, vibration during shock diagnostics, and the detection of rod fastening defects using vibration analysis. The first section introduces a comprehensive approach for simulating diagnostic signals utilizing linear random processes (LRPs), providing a foundation for the efficient analysis and verification of diagnostic features without extensive physical experimentation. It elaborates on the development and application of a simulation model for AE signals in electrical equipment, showcasing how this model facilitates the understanding of signal propagation. Further, the chapter delves into the simulation of AE signals and vibration in electrical equipment under shock diagnostics, highlighting a developed program that generates AE signals mimicking those from a piezoelectric transducer. The program's ability to modify various parameters allows for a detailed analysis of their influence on AE and vibration signals, offering insights consistent with theoretical models and real measurements. Moreover, the study proposes the use of the parameter β, derived from the power spectrum, to detect defects in rod fastening through vibration analysis, marking a significant step towards predictive maintenance and fault diagnosis in mechanical systems. By comparing the β values of suspended and fixed rods, the chapter validates the effectiveness of numerical models in replicating real-world dynamics. The investigation extends to the simulation of failure modes in diesel-electric generator cylinders, suggesting a novel approach to diagnosing specific cylinder failures through amplitude-frequency and phase-frequency characteristics analysis. Collectively, this research presents a valuable contribution to the field of diagnostic signal simulation, offering novel methods and tools for the analysis, verification, and simulation of diagnostic signals in electrical and mechanical systems. The findings promise to enhance the precision of fault diagnostics, reduce dependency on physical experiments, and pave the way for more effective predictive maintenance strategies.

This chapter explores the implementation and effectiveness of statistical diagnostics systems for monitoring and diagnosing electrical equipment (EE) components using information-measuring systems (IMSs). The chapter delves into the limitations of traditional diagnostic methods and how IMSs, through advanced data analysis and automation, overcome these limitations. By proposing a generalized structure for IMSs ‒ incorporating sensors, processing units, and software components, the chapter showcases specific applications such as diagnosing rolling bearings and various other EE units including internal combustion engines and electrical machines. It emphasizes the role of statistical hypothesis testing, linear random process models, and the critical use of training sets in constructing accurate diagnostic decision rules. Experimental examples, including shock and vibration diagnostics, illustrate the system's capabilities in real-world applications. Through detailed analysis of errors including methodological, instrumental, discretization, and quantization errors, the chapter presents calculation relations for metrological control, thereby enhancing diagnostic accuracy. This comprehensive overview confirms the potential of IMSs to significantly improve the diagnostic processes of EE, highlighting the importance of statistical methods, advanced data processing, and the strategic selection of diagnostic signals in reducing errors and improving reliability.

This comprehensive research chapter delves into the application of innovative diagnostic methods for electrical equipment (EE), focusing on vibration and acoustic analyses to detect and diagnose faults in rolling bearings of electric motors (EM) and the technical condition of diesel-electric generators (DEGs). Through extensive experimental setups, the chapter validates probabilistic models for diagnostic signals by utilizing real-world EE units and bespoke test benches. It introduces a specialized installation for analyzing rolling bearing vibrations, demonstrating the significance of frequency characteristics and the impact hammer test in pinpointing natural frequencies crucial for diagnostics. Furthermore, the study explores the use of shock diagnostics and acoustic emission (AE) analysis for evaluating the mechanical loading of EE components, highlighting the effectiveness of vibration maxima, and AE signal coefficients as reliable diagnostic indicators. Employing statistical hypothesis testing, particularly the Neyman-Pearson procedure, the research identifies specific faults in DEG cylinder-piston groups, showcasing the method's potential in early fault detection and preventive maintenance to enhance engine reliability. The findings contribute to the advancement of diagnostic methodologies for EE, promising improved maintenance strategies and operational efficiency.

This chapter delves into the innovative development and assessment of environmental monitoring systems and atmospheric modeling concerning power engineering facilities. Emphasizing the pivotal role of unmanned aerial vehicles (UAVs) and unmanned aerial systems (UAS), alongside advanced mathematical and statistical models, this research aims to enhance the accuracy, efficiency, and reliability of environmental impact assessments on air quality, particularly concerning radioactive aerosol propagation and deposition. Authors contribute to a multidimensional analysis that focuses on: (1) developing a new class of environmental monitoring systems leveraging UAV technology for dynamic air quality data collection around power engineering facilities; (2) evaluating mathematical models for atmospheric transport of radionuclides, including vector random field models, Gaussian models, and integral models, to accurately predict the environmental and public health impacts of radioactive releases; (3) innovating “in-situ” experimentation methodology using UAS equipped with filter injection devices for direct and reverse problem-solving in radioactive aerosol concentration determination; (4) enhancing information support within environmental monitoring systems through comprehensive analysis of their structure, function, and life cycle stages, aimed at improving system reliability and effectiveness. The research spans theoretical, simulation, and experimental studies, intending to mitigate the environmental ramifications of power engineering facilities over an expected timeframe. Key environmental parameters monitored include radionuclide distribution in the atmosphere and their deposition, encompassing essential environmental dynamics such as wind speed variations, turbulent diffusion, and gravitational deposition. This chapter underscores the necessity for accurate models and systems in environmental monitoring, providing crucial insights for designing, implementing, and modernizing such systems to ensure minimal human health and environmental impact while addressing the challenges of traditional monitoring methods.

This chapter investigates the utility of Unmanned Aerial Measurement Systems (UAMS) for environmental monitoring around nuclear and thermal power plants. UAMS offer improved safety, flexibility, and cost-effectiveness over traditional methods. The systems can be categorized by various factors, including the type of measurement signals, data collection methods, and the number of UAVs deployed. UAMS are crucial in analyzing harmful emissions and radiation, ensuring environmental and personnel safety. The study discusses UAMS’s role and future development, emphasizing its potential to enhance international and state environmental protection programs. The research includes case studies, such as the Chornobyl and Fukushima Daiichi NPP incidents, demonstrating UAVs’ pivotal role in remote monitoring and emergency response.