Periodic Reporting for period 1 - DyVirt (Dynamic virtualisation: modelling performance of engineering structures)
Periodo di rendicontazione: 2018-02-01 al 2020-01-31
The aim of this innovative training network is to enable a new generation of early-stage researchers (ESRs) to face the urgent challenge of how to model the performance of engineering structures that operate in dynamic environments. Building trusted virtual models for structures subject to high dynamic loads is a process we call “dynamic virtualisation”. All of the ESRs who receive training through this network will (i) obtain a PhD from an internationally-recognised university, (ii) gain crucial skills in developing accurate models of dynamic structures much needed in industry, (iii) gain experience of applying their research skills in non-academic organisations, and (iv) receive training in transferable skills such as commercialisation and communication.Obtaining a valuable virtual model is no longer a question of computing power, but now rests in the more difficult problem of developing trust in the model through the process of Verification and Validation (V&V). Verification is concerned with the numerical accuracy of the virtual model. It addresses both the elimination of coding errors and estimation of the numerical errors that necessarily arise through discretisation of physical laws. Validation assesses the extent to which the virtual model accurately represents the system/structure being modelled, and thus the degree of trust that can be given to its predictions of real-world events. Successful development of dynamic virtualisation processes will be underpinned by the development and adoption of rigorous and appropriate V&V methodologies within industry. These challenges are perhaps most obvious in the renewable energy sector, where technology is developing at a very rapid pace, and more reliable models are required to cope with structures subjected to extreme loadings which lead to a high degree of nonlinearity and uncertainty. Applying our research to such problems will be accelerated by close interaction with the industrial partners in the network, with whom we intend to maintain and enhance an innovation-focussed relationship. All this will result in a training network where ESRs are able to be creative, entrepreneurial and innovative whilst receiving state-of-the-art training that will enable them to deal with future challenges in this important area of engineering.
Work performed & results achieved:
RO1: Development of a fundamental verification & validation methodology for nonlinear structures. Creation of new decision support capabilities for virtualization in structural dynamics applications.
RO2: Development of new methods for using multi-domain and multi-scale models that can be joined (or coupled) to model more complex phenomena, and progress on developing modal analysis for nonlinear structures.
RO3: Development of Bayesian tools for uncertainty quantification and propagation as well as optimal experimental design. Virtual sensing method for output-only modal analysis. Creation of a framework for uncertainty propagation and risk management for diagnostic and prognostic models.
RO4: Prediction of loads using Bayesian compressive sensing based evolutionary power spectrum estimation for incomplete process records. Novel theoretical methodologies for the determination of the operational envelope of large scale structures.
RO5: Creation real-time hybrid testing methods under uncertainty. Development of hybrid testing for virtualisation, via the development of nonlinear model order reduction methods.
RO1: Development of a fundamental verification & validation methodology for nonlinear structures. Creation of new decision support capabilities for virtualization in structural dynamics applications.
RO2: Development of new methods for using multi-domain and multi-scale models that can be joined (or coupled) to model more complex phenomena, and progress on developing modal analysis for nonlinear structures.
RO3: Development of Bayesian tools for uncertainty quantification and propagation as well as optimal experimental design. Virtual sensing method for output-only modal analysis. Creation of a framework for uncertainty propagation and risk management for diagnostic and prognostic models.
RO4: Prediction of loads using Bayesian compressive sensing based evolutionary power spectrum estimation for incomplete process records. Novel theoretical methodologies for the determination of the operational envelope of large scale structures.
RO5: Creation real-time hybrid testing methods under uncertainty. Development of hybrid testing for virtualisation, via the development of nonlinear model order reduction methods.
WP1
Progress beyond the state-of-the-art has been seen in subsystem-level validation approaches, the study of joints, ontologies, neuroevolution of augmenting topologies (NEAT) and graph neural networks. Results are expected to give a greater understanding of uncertainties in V & V processes. This will enable greater trust or confidence to be obtained from the V & V processes, and more accurate predictions of structural performance and longevity. Potential impacts, could be lower operation and maintenance costs for offshore wind, and/or more accurate plant life extension estimates
WP2
Progress beyond the state-of-the-art has been made by creating hybrid-coupling schemes. There has also been significant progress with modelling complex new material such as MXene/Polymer nanocomposites. We expect results that relate to assembling (or joining) complex models from sub-models to more accurately capture the dynamic behaviour of structures. There is the possibility of new modelling/software coupling techniques as well. New modelling/software coupling techniques can have potentially wide impact. We anticipate the impact to be based around more accurate predictions of dynamic behaviour of materials with multi-scale properties. Modelling complex new material such as MXene/Polymer which offer interesting future exploitation possibilities, as these materials have unique properties not exhibited by other materials.
WP3
Progress beyond the state-of-the-art has been made using the hierarchical Bayesian modelling framework, and the development of Bayesian optimal experimental design tools. There are also very promising developments in virtual sensing, and on developing an uncertainty quantification framework to estimate confidence bounds of wind turbine responses. The techniques being developed will lead to more sophisticated algorithms. Here the scope of application is more broadly on managing uncertainty in the engineering system of interest, and the results will therefore reflect that. The virtual sensing and uncertainty quantification framework results are expected to be more focused on specific industry application areas. Both the work on virtual sensing and the wind turbine responses are being developed in beneficiaries that are commercial organisations, and we anticipate there will be relatively rapid exploitation of these results in the near future.
WP4
Progress beyond the state-of-the-art has been made by considering Bayesian compressive sensing, identifying a generalized framework for determining non-stationary response of linear and nonlinear systems, and data reconstruction methods in the context of wind turbines. It is anticipated that this work will develop into a wider set of applications beyond just wind turbines in the next part of the project. In this work package, the results are expected to be more sophisticated methods for accurate estimation of loading and environmental effects for structures that have to operate in highly dynamic environments, such as wind turbines. Ultimately, having a greater understanding of load and environmental history will enable the useful life of a structure to be optimised and for safe operation to be guaranteed to a specified level.
WP5
Progress beyond the state-of-the-art has been made on applying uncertainty type analysis to the real-time hybrid test procedures. There has also been progress in the development of nonlinear model order reduction methods for integration into a dynamic substructuring scheme, the creation of a virtual nonlinear hybrid testing framework, including control considerations and integration schemes. In this WP the expected results will consist of a series of hybrid-testing experiments a that demonstrate the new concepts being developed. In particular, the incorporation of uncertainty propagation and nonlinear model order reduction. Hybrid-testing has potential impact for digital-twin technology.
Progress beyond the state-of-the-art has been seen in subsystem-level validation approaches, the study of joints, ontologies, neuroevolution of augmenting topologies (NEAT) and graph neural networks. Results are expected to give a greater understanding of uncertainties in V & V processes. This will enable greater trust or confidence to be obtained from the V & V processes, and more accurate predictions of structural performance and longevity. Potential impacts, could be lower operation and maintenance costs for offshore wind, and/or more accurate plant life extension estimates
WP2
Progress beyond the state-of-the-art has been made by creating hybrid-coupling schemes. There has also been significant progress with modelling complex new material such as MXene/Polymer nanocomposites. We expect results that relate to assembling (or joining) complex models from sub-models to more accurately capture the dynamic behaviour of structures. There is the possibility of new modelling/software coupling techniques as well. New modelling/software coupling techniques can have potentially wide impact. We anticipate the impact to be based around more accurate predictions of dynamic behaviour of materials with multi-scale properties. Modelling complex new material such as MXene/Polymer which offer interesting future exploitation possibilities, as these materials have unique properties not exhibited by other materials.
WP3
Progress beyond the state-of-the-art has been made using the hierarchical Bayesian modelling framework, and the development of Bayesian optimal experimental design tools. There are also very promising developments in virtual sensing, and on developing an uncertainty quantification framework to estimate confidence bounds of wind turbine responses. The techniques being developed will lead to more sophisticated algorithms. Here the scope of application is more broadly on managing uncertainty in the engineering system of interest, and the results will therefore reflect that. The virtual sensing and uncertainty quantification framework results are expected to be more focused on specific industry application areas. Both the work on virtual sensing and the wind turbine responses are being developed in beneficiaries that are commercial organisations, and we anticipate there will be relatively rapid exploitation of these results in the near future.
WP4
Progress beyond the state-of-the-art has been made by considering Bayesian compressive sensing, identifying a generalized framework for determining non-stationary response of linear and nonlinear systems, and data reconstruction methods in the context of wind turbines. It is anticipated that this work will develop into a wider set of applications beyond just wind turbines in the next part of the project. In this work package, the results are expected to be more sophisticated methods for accurate estimation of loading and environmental effects for structures that have to operate in highly dynamic environments, such as wind turbines. Ultimately, having a greater understanding of load and environmental history will enable the useful life of a structure to be optimised and for safe operation to be guaranteed to a specified level.
WP5
Progress beyond the state-of-the-art has been made on applying uncertainty type analysis to the real-time hybrid test procedures. There has also been progress in the development of nonlinear model order reduction methods for integration into a dynamic substructuring scheme, the creation of a virtual nonlinear hybrid testing framework, including control considerations and integration schemes. In this WP the expected results will consist of a series of hybrid-testing experiments a that demonstrate the new concepts being developed. In particular, the incorporation of uncertainty propagation and nonlinear model order reduction. Hybrid-testing has potential impact for digital-twin technology.