In the last few years, the reliability assessment acquired a fundamental role in many advanced technology applications. System downtime and unexpected failures massively affect the productivity of a system/product/plant. As a consequence, the Reliability, Availability, Maintainability, and Safety (RAMS) disciplines, together with diagnostics and prognostics tools are becoming more and more essential for several application fields, especially in case of complex industrial systems where environment, personnel, and equipment safety are mandatory features. Several works in recent literature deal with design for reliability methods that integrates one or more reliability tasks during the early phase of the design. However, all-around Reliability Life Cycle procedures that takes into account the complete system life cycle (from design and development to actual implementation) are rarely dealt with. Another fundamental aspect that is barely taken into account by recent literature is the importance of measurements and data within the context of a reliability life cycle approach. Usually, reliability parameters are estimated using probabilistic approaches, failure and degradation models, statistical analysis and failure data included in handbooks. However, instrumentation and measurements technologies could remarkably improve and optimize several different RAMS methodologies introducing suitable data analysis in spite of handbook data and probabilistic approaches. Trying to fill these gaps, the main aim of this work is to extend the classical idea of Design for Reliability introducing an innovative data-driven reliability life cycle procedure that integrates different RAMS techniques to optimize the reliability of complex industrial systems during both design and operational phases. However, it is not enough to simply provide a reliability procedure based on a set of different techniques without a thorough and structured study of the state-of-the art of each method. Therefore, the second aim of this project is the optimization of the techniques included in the proposed Reliability Life Cycle in order to overcome the major drawbacks highlighted by the literature review of each method. Firstly, the work deals with Failure Modes, Effects and Criticality Analysis (FMECA) providing a statistical comparison of the alternative approaches found in literature and applying all of them to the risk analysis of a real case study (Ventilation system for high-speed trains). Furthermore, the work shows how the FMECA could be integrated in the context of a data-driven approach. Then, an innovative method to easily and effectively estimate a risk threshold is presented and tested using the design of a control system for wind turbine as a case study. Reliability Allocation plays a central role in the proposed Reliability Life Cycle. In this point of view, this work presents an innovative method able to overcomes all the initial hypotheses required by the other approaches and test it on three complex systems (a numerical example, a sensor unit for railway systems and a lubrication system for gas turbines). The work also presents two test plans with the aim of characterize components and equipment by both system performance and system reliability point-of-views. The results of the experimental measurement campaigns provide significant information to improve the RAMS parameters and the electrical and metrological performances of the components under analysis (Inertial Measurement Units and DC-Dc converters for diagnostic devices). Furthermore, this research also proposes a new customized diagnostic-oriented decision-making diagram for maintenance management and apply it to maintenance planning of a wind turbine. Moreover, a new diagnostic method based on a data-driven Condition Monitoring tool is presented to efficiently monitor the health and detect damages in the wind turbine by means of measurements of critical parameters of the tested system. Finally, the work also deals with data-driven remaining useful life (RUL) estimation of Lithium-Ion batteries proposing a hybrid approach based on both condition monitoring and physic degradation model where a state-space estimation is used to generate a big dataset for the training of the proposed Recurrent neural Network. The application on a real battery dataset proves the superiority of the proposed degradation model and the effectiveness of the estimation with respect to the state of the art.

An innovative Data-Driven Reliability Life Cycle for complex systems / Gabriele Patrizi. - (2022).

An innovative Data-Driven Reliability Life Cycle for complex systems

Gabriele Patrizi
2022

Abstract

In the last few years, the reliability assessment acquired a fundamental role in many advanced technology applications. System downtime and unexpected failures massively affect the productivity of a system/product/plant. As a consequence, the Reliability, Availability, Maintainability, and Safety (RAMS) disciplines, together with diagnostics and prognostics tools are becoming more and more essential for several application fields, especially in case of complex industrial systems where environment, personnel, and equipment safety are mandatory features. Several works in recent literature deal with design for reliability methods that integrates one or more reliability tasks during the early phase of the design. However, all-around Reliability Life Cycle procedures that takes into account the complete system life cycle (from design and development to actual implementation) are rarely dealt with. Another fundamental aspect that is barely taken into account by recent literature is the importance of measurements and data within the context of a reliability life cycle approach. Usually, reliability parameters are estimated using probabilistic approaches, failure and degradation models, statistical analysis and failure data included in handbooks. However, instrumentation and measurements technologies could remarkably improve and optimize several different RAMS methodologies introducing suitable data analysis in spite of handbook data and probabilistic approaches. Trying to fill these gaps, the main aim of this work is to extend the classical idea of Design for Reliability introducing an innovative data-driven reliability life cycle procedure that integrates different RAMS techniques to optimize the reliability of complex industrial systems during both design and operational phases. However, it is not enough to simply provide a reliability procedure based on a set of different techniques without a thorough and structured study of the state-of-the art of each method. Therefore, the second aim of this project is the optimization of the techniques included in the proposed Reliability Life Cycle in order to overcome the major drawbacks highlighted by the literature review of each method. Firstly, the work deals with Failure Modes, Effects and Criticality Analysis (FMECA) providing a statistical comparison of the alternative approaches found in literature and applying all of them to the risk analysis of a real case study (Ventilation system for high-speed trains). Furthermore, the work shows how the FMECA could be integrated in the context of a data-driven approach. Then, an innovative method to easily and effectively estimate a risk threshold is presented and tested using the design of a control system for wind turbine as a case study. Reliability Allocation plays a central role in the proposed Reliability Life Cycle. In this point of view, this work presents an innovative method able to overcomes all the initial hypotheses required by the other approaches and test it on three complex systems (a numerical example, a sensor unit for railway systems and a lubrication system for gas turbines). The work also presents two test plans with the aim of characterize components and equipment by both system performance and system reliability point-of-views. The results of the experimental measurement campaigns provide significant information to improve the RAMS parameters and the electrical and metrological performances of the components under analysis (Inertial Measurement Units and DC-Dc converters for diagnostic devices). Furthermore, this research also proposes a new customized diagnostic-oriented decision-making diagram for maintenance management and apply it to maintenance planning of a wind turbine. Moreover, a new diagnostic method based on a data-driven Condition Monitoring tool is presented to efficiently monitor the health and detect damages in the wind turbine by means of measurements of critical parameters of the tested system. Finally, the work also deals with data-driven remaining useful life (RUL) estimation of Lithium-Ion batteries proposing a hybrid approach based on both condition monitoring and physic degradation model where a state-space estimation is used to generate a big dataset for the training of the proposed Recurrent neural Network. The application on a real battery dataset proves the superiority of the proposed degradation model and the effectiveness of the estimation with respect to the state of the art.
2022
Marcantonio Catelani
ITALIA
Gabriele Patrizi
File in questo prodotto:
File Dimensione Formato  
phd_thesis_Gabriele_Patrizi_CompressPdf.pdf

accesso aperto

Descrizione: Tesi di dottorato
Tipologia: Pdf editoriale (Version of record)
Licenza: Open Access
Dimensione 8.28 MB
Formato Adobe PDF
8.28 MB Adobe PDF

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1264675
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact