Багатокласове розпізнавання стану складного просторового об’єкта нейромережевим класифікатором

Скорочена інформація про документ
dc.contributor.authorРупіч, Сергій Сергійович
dc.date.accessioned2019-04-16T11:05:38Z
dc.date.available2019-04-16T11:05:38Z
dc.date.issued2019
dc.description.abstractenThe thesis describes that complex spatial objects of aeronautical engineering, oil-andgas industry, special purpose objects are usually operated in difficult accessible places in zones with increased external influences and dynamic loads. It is described that multi-site damage can be aroused under such conditions of operation due to imperfection of the elements of the construction that are used under the action of complex loading. It is shown that in order to ensure the safe operation of complex spatial objects that are characterized by multi-class of possible technical states, it is necessary to carry out continuous monitoring of structural integrity and provide multi-class diagnostics. The basic principles and approaches of construction and organization a new type of intelligent multi-class monitoring systems have been considered. It is described approaches to building a network of sensitive elements. The basic methods of information processing in monitoring systems have been considered. The development tendencies and ways of improvement of newest diagnostic systems are described. It is shown diagnostic problems which are solved by intellectual monitoring systems that based on Structural Health Monitoring concept. It is shown that traditional methods of recognition are functionally limited to solving multi-class diagnostics. The basic principles of neural network approach and application features of neural networks in monitoring and diagnostic systems are considered. The general model of the neural network is given, the types of activation functions and the architectures of neural network construction are described. The basic models of training of artificial neural networks are also described. It shows that the tasks of multi-class diagnostics, estimation and future prediction of the technical condition of complex spatial objects with multi-site damage are not solved enough; the issue of methodological and algorithmic support for the development of multi-class diagnostic tools based on information technologies is not disclosed according to the literature. The thesis describes that the most rigid conditions for the preservation of integrity put forward to the welded tanks with ecologically dangerous substances. The basic general technical requirements for automated systems of early prediction of defects appearance, possible destruction and fuel leaks that based on implementing the concept of Structural Health Monitoring are described. It is presented the general structure of the functional diagnostics system for monitoring the technical condition of the tank. In the dissertation the subsystem of decision- making is substantiated and developed to improve the functional diagnostics system of the technical condition of the welded tank. It is substantiated that using Probabilistic Neural Network, which provides the best results of a multi-class recognition, for the development of the classifier. The general structure of the neural network classifier is developed. The functional building features of Probabilistic Neural Network are shown. Models of processes for forming sets of input vectors of diagnostic features for such diagnostic tasks as localization of single damage, localization of multi-site damage, monitoring of damage development and monitoring of structural degradation are developed. A generalized information model of the system of multi-class recognition, which combines the diagnostic tasks, is developed. The program implementation of the neural network classifier is carried out by the Matlab software environment. As a result, the probability of recognition from the network influence parameter, which shows the effectiveness of the neural network classifier for localization of single damage and localization of two defects, was established. Research has shown that it is possible to develop classifiers based on Probabilistic Neural Network with one set value of the spread parameter, which will achieve error-free multi-class recognition for the diagnostic tasks of localization. According to the results of the research of the efficiency of the classifier for the monitoring of a development of damage, it has been established that a error-free multi-class recognition of the object’s status is achieved. Range of values of the spread parameter established by an empirically way. The influence of the parameters of a neural network and characteristics of diagnostic vectors on the probability of multi-class recognition of the object’s state for monitoring of structural degradation is investigated. The research of the influence of number of training vectors on the recognition quality has been performed. It is shown that a decrease in the number of the training set impairs the recognition quality. The research of influence of different range of diagnostic features is carried out. It is shown that error-free recognition by the developed classifier is achieved provided that the values of the diagnostic features are within the same range. The research of the influence of number of diagnostic features in vectors on the recognition efficiency was performed. It was found that changes in a number of features from 5 to 7 lead to a significant complication for the construction of a neural network and an increase of the training set by 4 times. Also, the quality of recognition is reduced. The research of the possibility of the error-free recognition was conducted by the developed classifier based on the stress-strain state of the geometric model of the tank structural elements with multi-site damage, where sensors are located. The state recognition of the tank in cases for localization of damage of separate occurrence of three cracks, the simultaneous occurrence of several cracks and monitoring of the development of one crack were carried out. It is shown that the classifier provides an error-free recognition depending on established parameters of the neural network. It is identified that using a classifier based on Probabilistic Neural Network is effective for the recognition of cracks in welds of tanks. The physical model of the tank was made. Based on its statistical analysis of the results of the vibration measurements were researched the recognition of the technical state of its structure. The Hurst coefficients were determined from measured vibration signals. The two dominant frequencies of the oscillations were calculated and reduced to dimensionless magnitude using the sampling frequency. Both the Hurst coefficients and frequencies determine the functional dependence of changes in the current state of the tank on its filling with liquid from 0 % to 100 %. It is shown possibility of using the designed neural network classifier for recognizing real objects has been confirmed.en
dc.description.abstractruДиссертация посвящена проведению многоклассового распознавания состояния сложного пространственного объекта путем усовершенствования и внедрения подсистемы принятия решения в систему функциональной диагностики на основе разработки нейросетевого классификатора. Разработаны информационные модели процессов формирования учебных и тестовых множеств входных векторов диагностических признаков для многоклассового распознавания с целью локализации единичного и множественного повреждения, мониторинга развития повреждений и мониторинга деградации конструкции. Разработано программное обеспечение для предопределенных диагностических задач. Разработана обобщенная структура многоклассового распознавания. Исследована эффективность разработанного нейросетевого классификатора для обеспечения многоклассового распознавания, и проведено распознавание технического состояния компьютерной и физической моделей резервуара. Установлено влияние параметров нейронной сети и характеристик диагностических векторов на вероятность многоклассового распознавания состояния объекта для диагностических задач.ru
dc.description.abstractukДисертація присвячена проведенню багатокласового розпізнавання стану складного просторового об’єкта шляхом вдосконалення та впровадження підсистеми прийняття рішення в систему функціональної діагностики на основі розробки нейромережевого класифікатора. Розроблено інформаційні моделі процесів формування навчальних та тестових множин вхідних векторів діагностичних ознак для багатокласового розпізнавання з метою локалізації одиничного та багатоосередкового пошкодження, моніторингу розвитку пошкоджень і моніторингу деградації конструкції. Розроблено програмне забезпечення для визначених діагностичних завдань. Розроблено узагальнену структуру багатокласового розпізнавання. Досліджено ефективність розробленого нейромережевого класифікатора для забезпечення багатокласового розпізнавання, та проведено розпізнавання технічного стану комп’ютерної та фізичної моделей резервуару. Встановлено впливи параметрів нейронної мережі та характеристик діагностичних векторів на вірогідність багатокласового розпізнавання стану об’єкта для діагностичних завдань.uk
dc.format.page30 с.uk
dc.identifier.citationРупіч, С. С. Багатокласове розпізнавання стану складного просторового об’єкта нейромережевим класифікатором : автореф. дис. … канд. техн. наук : 05.11. 13 – Прилади і методи контролю та визначення складу речовин / Рупіч Сергій Сергійович. – Київ, 2019. – 30 с.uk
dc.identifier.urihttps://ela.kpi.ua/handle/123456789/27239
dc.language.isoukuk
dc.publisherКПІ ім. Ігоря Сікорськогоuk
dc.publisher.placeКиївuk
dc.subjectскладний просторовий об’єктuk
dc.subjectрезервуар зі зварними з’єднаннямиuk
dc.subjectбагатоосередкове пошкодженняuk
dc.subjectмоніторинг технічного стануuk
dc.subjectбагатокласове розпізнаванняuk
dc.subjectнейромережевий класифікатор стануuk
dc.subjectімовірнісна нейронна мережаuk
dc.subjectвектор діагностичних ознакuk
dc.subjectефективність класифікаціїuk
dc.subjectcomplex dimensional objecten
dc.subjecttank with welded jointsen
dc.subjectmulti-focal damageen
dc.subjectStructural Health Monitoringen
dc.subjectmulti-class recognitionen
dc.subjectneural network classifieren
dc.subjectProbabilistic Neural Networken
dc.subjectvector of diagnostic signsen
dc.subjectclassification efficiencyen
dc.subjectсложный пространственный объектru
dc.subjectрезервуар со сварными соединениямиru
dc.subjectмногоочаговое повреждениеru
dc.subjectмониторинг технического состоянияru
dc.subjectмногоклассовое распознаваниеru
dc.subjectнейросетевой классификатор состоянийru
dc.subjectвероятностная нейронная сетьru
dc.subjectвектор диагностических признаковru
dc.subjectэффективность классификацииru
dc.subject.udc629.735.083.2:620.179.1:004.032.26(043.3)uk
dc.titleБагатокласове розпізнавання стану складного просторового об’єкта нейромережевим класифікаторомuk
dc.typeThesisuk

Файли

Контейнер файлів
Зараз показуємо 1 - 1 з 1
Ескіз
Назва:
Rupich_aref.pdf
Розмір:
1.62 MB
Формат:
Adobe Portable Document Format
Опис:
Завантажити
Ліцензійна угода
Зараз показуємо 1 - 1 з 1
Ескіз недоступний
Назва:
license.txt
Розмір:
9.1 KB
Формат:
Item-specific license agreed upon to submission
Опис:
Завантажити

Зібрання

Автореферати (ПСОН)
Автореферати (вільний доступ)