Стохастические модели вибросигналов и их анализ для исследования состояния механических систем

Abstract

Описаны свойства корреляционно-спектральных характеристик периодически коррелированных случайных процессов и их обобщений – величин, которые целесообразно использовать для описания состояния вращающихся узлов механических систем и выявления их дефектов. Рассмотрены методы оценки вероятностных характеристик первого и второго порядков таких процессов по экспериментальным данным при известных и неизвестных периодах нестационарности.

Описано властивості кореляційно-спектральних характеристик періодично корельованих випадкових процесів та їх узагальнень – величин, які доцільно використовувати для опису стану обертових вузлів механічних систем та виявлення їх дефектів. Розглянуто методи оцінювання ймовірнісних характеристик першого та другого порядків таких процесів за експериментальними даними при відомих і невідомих періодах нестаціонарності.

Introduction. Faults detection and estimation of its stage is possible only on the base of the detail analysis of the vibration signal structure and its relation with kinematic and dynamic mechanisms. Description of such structure is possible only using the adequate mathematical models of vibration signals that contain necessary signals’ features needed for estimation of the mechanical system state. Recurrence and stochasticity are the features of the vibration oscillations. The faults appearance causes non-linearity of mechanical system and the interaction of these signal’s components. The character of such interaction contains those properties of vibrations that are necessary for the mechanism state estimation. Periodically and almost periodically non-stationary random processes are adequate models for the description and analysis of such interaction. Purpose. The purpose of the given paper is to propose a new approach for analysis of the vibration signal that allows one to detect the mechanical systems faults on the early stages of their initiation and to prevent accidents. Results. Properties of the correlation and spectral characteristics of periodically correlated random processes and their generalizations – quantities that are suitable for the description of mechanical systems' rotary units state and the detection of their faults, are described. The methods for estimation of probabilistic characteristics of the first and the second order on the base of experimental data at known and unknown period of non-stationarity are considered. Advantages and disadvantages of the estimators using the coherent, component methods and least squares methods are shown. Parameters that affect on the estimation accuracy are given. Definition of the aliasing effect is given and its influence on characteristics estimators is investigated. Methodology for vibration signals processing is proposed. Conclusions. On the base of the developed methods of spectral and correlation analysis of PCRP and their generalizations with uses of created information-measuring system the series of vibrodiagnostic investigations of mechanisms state on the factories in Ukraine are provided. Obtained results showed the high efficiency of the proposed approach for detection of the faults on the early stages of their initiation and its further growing.