Final Report Summary - INSIST (Integrating Numerical Simulation and Geometric DesignTechnology)
The project ITN-INSIST has brought together academic and industrial partners from five European countries in an interdisciplinary effort which combines concepts from engineering, computer science and applied geometry. The scientific goal of the project has been to develop and study new methods related to the design and analysis of complex geometric models, which are of significant importance in industrial and manufacturing applications. We have also strived to train and promote the career development of the talented early stage and experienced researchers who have contributed to the achievement of these objectives.
The motivating idea behind the project is to narrow the gap between Computer-Aided Design (CAD) and Computer Aided Engineering (CAE) by using the same geometric description of a model in the design and analysis phases of product development. Traditionally, the geometric description in design is realized using smoother and higher-order B-Spline functions, while analysis is performed using lower-continuity and lower-degree polynomials. For example, a spherical object that is perfectly smooth in the design phase is often approximated by a collection of small polyhedrons (such as cubes and tetrahedrons) which introduces many corners and sharp edges into the model. This conversion process between the two representations is often time-consuming and can introduce errors in the analysis which are difficult to quantify, in particular if complex multi-physics simulations are performed.
A new paradigm that promises to overcome the divide between analysis and design is the “isogeometric analysis” (IGA) concept. In IGA, the same model representation is used for the design as well for performing numerical simulations, ensuring that no loss of accuracy is introduced by a conversion process. Nevertheless, often the design representation used by the CAD software can be incomplete (for example only the boundary representation is provided) or in a format that is not directly suitable for analysis. Some of these obstacles have been studied and at least partially overcome during the course of the ITN-INSIST project.
The main highlights of the work performed and the results obtained divided by the four areas of study are as follows:
CAD Processing: A declarative feature recognizer was developed which allows rapid identification of CAD features (such as notches, slots and through-holes). An important result is that these queries can be performed in linear time, while previous results were much slower, requiring quadratic or exponential time. This makes it feasible for this algorithm to be used in industrial applications with thousands or more components. There have also been some results obtained regarding feature simplification and error estimation, which aims to study the effect of adding and removing geometric features in a CAD model on the numerical simulation results. A related topic that has been considered is “isogeometric segmentation”, which aims to divide a complex model into simpler parts that can be more easily analyzed individually. Besides several journal publications by the fellows presenting these results, two open source packages have been developed to publicly provide reference implementations of the developed algorithms.
Pre-processing and Meshing: The work on this topic has been aimed to provide analysis-suitable isogeometric representations in the case these are not directly available from the CAD models. Algorithms have been developed that use a boundary representation of the domain as well as efficient methods for analysis of CT-scan based data. The later in particular has applications in medical fields leading for example to the potential development of better implants, or better methods for the diagnosis and treatment of osteoporosis from X-ray scans. Also within this work-package, algorithms for image registration have been developed, with the aim of computing a transformation between two images. The two images can be, for example, brain images of a patient taken at different times, and the process of registering can be helpful in the diagnosis and treatment of brain tumors or Alzheimer disease.
Numerical Analysis/CAE: This is the main scientific work package and thorough studies of the approximation properties of the basis functions used in IGA have been conducted by the fellows involved in this part of the project. Hierarchical methods, which use a tree-like structure to efficiently represent the domain of interest, have been developed and improved as part of this work. Also considered are so-called “meshless methods” which rely on point-cloud or particle-based representations to study complex models and simulate physical phenomena such as fracture. The work has resulted in more than 10 journal publications, several conference presentations, and additions to both open-source and proprietary software packages.
Voxel Processing: The lead beneficiary of this work package is the industrial partner Simpleware, who is developing parametrization and visualization software using voxel-based data. This work is also aimed at obtaining a better understanding and improved modeling capabilities from medical images. The research conducted in this project has resulted in algorithms for recovering sharp features from image-data and obtaining analysis-suitable models obtained by Boolean operations such as unions, intersections and subtractions of model elements.
In addition to the European universities and industrial partners who were directly involved in the project, associated partners from UK, France and the United States have also contributed to the supervision and training of the researchers. The associated partners consisted of both industrial and academic institutions, which have provided substantial support to the project. The fellows have participated in exchanges (secondments) with the associated and the full partners, which has greatly contributed to strengthening the collaboration between the project participants.
The summary of the recruitment, training and research output achievements is as follows:
• 13 Early-State Researcher and 2 Experienced Researcher positions have been filled
• The network has organized 6 global-training events (workshops) and has contributed to one international conference
• More than 20 journal publications and many conference talks have been co-authored/presented by the fellows
• Contributions to several open-source software have made, such as IGAFEM (https://sourceforge.net/projects/cmcodes/(se abrirá en una nueva ventana)) GIS+MO (http://gs.jku.at/gismo(se abrirá en una nueva ventana)) Efficient Feature Recognizer (https://bitbucket.org/ZBNIU/efficient-feature-recognizer(se abrirá en una nueva ventana)) SimErr (https://astarte.org/gitlab/xis10z/SimpErr(se abrirá en una nueva ventana)).
• The fellows have also contributed to closed-source or in-house software such as: Morfeo (www.cenaero.be/morfeo) DiffPack (http://www.diffpack.com/(se abrirá en una nueva ventana)) ScanIP (https://www.simpleware.com/software/scanip/(se abrirá en una nueva ventana))
The project has had a positive impact on the careers of the researchers involved. At the end of the project, the experienced researchers have found faculty positions at top-tier universities, and several early-stage researchers have defended or submitted their doctoral theses. Most of the doctoral students are in the final stages of their PhD programs and some are already employed or have received offers from academic and industrial institutions.
The results obtained during this project, which include journal publications, conference papers, open-source software and other multi-media documents are showcased on the project website: http://www.itn-insist.com/(se abrirá en una nueva ventana) They can also be accessed from online document archives such arxiv, orbiLU, ResearchGate, from open-source repositories, such as Sourceforge, and video websites such YouTube.
The motivating idea behind the project is to narrow the gap between Computer-Aided Design (CAD) and Computer Aided Engineering (CAE) by using the same geometric description of a model in the design and analysis phases of product development. Traditionally, the geometric description in design is realized using smoother and higher-order B-Spline functions, while analysis is performed using lower-continuity and lower-degree polynomials. For example, a spherical object that is perfectly smooth in the design phase is often approximated by a collection of small polyhedrons (such as cubes and tetrahedrons) which introduces many corners and sharp edges into the model. This conversion process between the two representations is often time-consuming and can introduce errors in the analysis which are difficult to quantify, in particular if complex multi-physics simulations are performed.
A new paradigm that promises to overcome the divide between analysis and design is the “isogeometric analysis” (IGA) concept. In IGA, the same model representation is used for the design as well for performing numerical simulations, ensuring that no loss of accuracy is introduced by a conversion process. Nevertheless, often the design representation used by the CAD software can be incomplete (for example only the boundary representation is provided) or in a format that is not directly suitable for analysis. Some of these obstacles have been studied and at least partially overcome during the course of the ITN-INSIST project.
The main highlights of the work performed and the results obtained divided by the four areas of study are as follows:
CAD Processing: A declarative feature recognizer was developed which allows rapid identification of CAD features (such as notches, slots and through-holes). An important result is that these queries can be performed in linear time, while previous results were much slower, requiring quadratic or exponential time. This makes it feasible for this algorithm to be used in industrial applications with thousands or more components. There have also been some results obtained regarding feature simplification and error estimation, which aims to study the effect of adding and removing geometric features in a CAD model on the numerical simulation results. A related topic that has been considered is “isogeometric segmentation”, which aims to divide a complex model into simpler parts that can be more easily analyzed individually. Besides several journal publications by the fellows presenting these results, two open source packages have been developed to publicly provide reference implementations of the developed algorithms.
Pre-processing and Meshing: The work on this topic has been aimed to provide analysis-suitable isogeometric representations in the case these are not directly available from the CAD models. Algorithms have been developed that use a boundary representation of the domain as well as efficient methods for analysis of CT-scan based data. The later in particular has applications in medical fields leading for example to the potential development of better implants, or better methods for the diagnosis and treatment of osteoporosis from X-ray scans. Also within this work-package, algorithms for image registration have been developed, with the aim of computing a transformation between two images. The two images can be, for example, brain images of a patient taken at different times, and the process of registering can be helpful in the diagnosis and treatment of brain tumors or Alzheimer disease.
Numerical Analysis/CAE: This is the main scientific work package and thorough studies of the approximation properties of the basis functions used in IGA have been conducted by the fellows involved in this part of the project. Hierarchical methods, which use a tree-like structure to efficiently represent the domain of interest, have been developed and improved as part of this work. Also considered are so-called “meshless methods” which rely on point-cloud or particle-based representations to study complex models and simulate physical phenomena such as fracture. The work has resulted in more than 10 journal publications, several conference presentations, and additions to both open-source and proprietary software packages.
Voxel Processing: The lead beneficiary of this work package is the industrial partner Simpleware, who is developing parametrization and visualization software using voxel-based data. This work is also aimed at obtaining a better understanding and improved modeling capabilities from medical images. The research conducted in this project has resulted in algorithms for recovering sharp features from image-data and obtaining analysis-suitable models obtained by Boolean operations such as unions, intersections and subtractions of model elements.
In addition to the European universities and industrial partners who were directly involved in the project, associated partners from UK, France and the United States have also contributed to the supervision and training of the researchers. The associated partners consisted of both industrial and academic institutions, which have provided substantial support to the project. The fellows have participated in exchanges (secondments) with the associated and the full partners, which has greatly contributed to strengthening the collaboration between the project participants.
The summary of the recruitment, training and research output achievements is as follows:
• 13 Early-State Researcher and 2 Experienced Researcher positions have been filled
• The network has organized 6 global-training events (workshops) and has contributed to one international conference
• More than 20 journal publications and many conference talks have been co-authored/presented by the fellows
• Contributions to several open-source software have made, such as IGAFEM (https://sourceforge.net/projects/cmcodes/(se abrirá en una nueva ventana)) GIS+MO (http://gs.jku.at/gismo(se abrirá en una nueva ventana)) Efficient Feature Recognizer (https://bitbucket.org/ZBNIU/efficient-feature-recognizer(se abrirá en una nueva ventana)) SimErr (https://astarte.org/gitlab/xis10z/SimpErr(se abrirá en una nueva ventana)).
• The fellows have also contributed to closed-source or in-house software such as: Morfeo (www.cenaero.be/morfeo) DiffPack (http://www.diffpack.com/(se abrirá en una nueva ventana)) ScanIP (https://www.simpleware.com/software/scanip/(se abrirá en una nueva ventana))
The project has had a positive impact on the careers of the researchers involved. At the end of the project, the experienced researchers have found faculty positions at top-tier universities, and several early-stage researchers have defended or submitted their doctoral theses. Most of the doctoral students are in the final stages of their PhD programs and some are already employed or have received offers from academic and industrial institutions.
The results obtained during this project, which include journal publications, conference papers, open-source software and other multi-media documents are showcased on the project website: http://www.itn-insist.com/(se abrirá en una nueva ventana) They can also be accessed from online document archives such arxiv, orbiLU, ResearchGate, from open-source repositories, such as Sourceforge, and video websites such YouTube.