Data science

Table of contents

1. 1. 1. F - 18 Machine learning for the optimization of FE simulations
      2. F - 17 Fibre optic measurements under the influence of high frequent loads
      3. F - 16 Numerical investigation of shear cracking in concrete beams
      4. F - 15 Crack propagation in cyclic loaded steel components
      5. F - 13 Detection of multidimensional states with fiber optics
      6. F - 12 Long-term effects on fiber optic sensors
      7. F - 10 Quality measures for distributed fiber optic sensor data
      8. F - 7 AI based design assistant für early stage bridge design
      9. F - 1 Damage assessment with ultrasonic measurements

F - 18 Machine learning for the optimization of FE simulations

The finite element method (FEM) is a key tool for calculating the load-bearing behavior of engineering structures. However, the calculation time can increase considerably, especially for complex structures and extensive load case combinations. The aim of this thesis is to investigate how artificial intelligence (AI) can be used to make FE analyses more efficient.

One focus is on the development and evaluation of surrogate models that are trained by machine learning to approximate FE simulations. Methods such as neural networks or Gaussian processes will be used to predict load-deformation relationships without relying on a full FE calculation.

The work includes:

• Literature research on existing AI approaches in numerical structural analysis,
• Creation of a simple, parameterized FE model,
• Creation of an AI model that is trained on the basis of the FE data,
• Comparison of the results with classical FE simulations in terms of accuracy and computing time.

This work is aimed at students with an interest in numerical methods and artificial intelligence. Previous knowledge of Python, machine learning and FE software is advantageous, but not essential.

Contact person:
Max Herbers, M.Sc.
0351 463 39620

F - 17 Fibre optic measurements under the influence of high frequent loads

Fibre optic sensors can be used to measure strain continuously over very long distances. The innovative technology of fibre optics, which is increasingly being used in the construction industry, offers great potential for the evaluation of existing bridge structures. By accurately detecting crack patterns or localised strain changes, such as those caused by tendon damage, critical changes in condition can be detected at an early stage. However, challenges arise when high frequent strain changes occur during a measurement, e.g. as a result of traffic impact. The aim of this student project is to investigate and define the limitations of fibre optic measurement using simple laboratory tests and existing data sets from real structures. The result will be a valuable information for fibre optic structural monitoring, which can be used to ensure the quality of measurement data even in the case of dynamic loads on bridges.

The specific task will be developed individually after a consultation.

Contact person:
Dipl.-Ing. Maria Walker
+49 351 463 39820

F - 16 Numerical investigation of shear cracking in concrete beams

The maintenance and assessment of bridge infrastructure is critical to the economy, as these structures are vital for transportation and commerce Many concrete bridges are aging and may not comply with modern technical standards due to outdated design methods and increased traffic loads. One of the failing mechanisms detected when recalculating existing bridges in accordance with the recalculation guidelines is insufficient shear load-bearing capacity, leading to the premature demolition of numerous bridges.

This project aims to model concrete beams with varying lengths and reinforcement ratios using ATENA software to simulate and analyze shear cracking behavior. The objective is to develop a reliable numerical model that accurately predicts shear cracking, providing a foundation for future experimental testing. Based on the key aspects, the following tasks should be solved:

• Literature review on shear cracking in concrete
• Literature review on ATENA software and shear modeling parameters
• Development of a numerical model with ATENA software
• Evaluation of shear cracking patterns
• Perform parametric studies
• Summarize all the results to draw conclusions and outlooks.

Contact person:
Kleo Lila, M.Sc.
0351 463 40471

F - 15 Crack propagation in cyclic loaded steel components

Full title: Measurement of crack propagation in cyclic loaded, repaired steel components

For the repair of fatigue cracks in steel components, adhesively bonded laminates made of carbon fiber-reinforced plastic (CFRP) offer particular potential due to their high stiffness and weight-specific strength. To strengthen trust in this innovative repair method, it is necessary to monitor crack propagation in the repaired component on the real structure and replicate it in the laboratory. However, the problem is that the steel component surface is covered by the CFRP laminate and therefore not accessible. This is where the work comes in.

As part of the work, strategies for tracking crack propagation under both laboratory conditions and application on a real structure will be developed and experimentally investigated. The accuracy, reliability, and consistency of the measurement and analysis methods will be highlighted through method validation. The objectives of the work are to derive a preferred methodology for the detection of crack propagation in repaired steel components, experimentally test the influencing factors for this variant, and create a procedural instruction for the optimal application of the measurement methodology.

Details of the task will be specified prior to commencement.

Contact person:
Max Herbers, M.Sc.
0351 463 39620

F - 13 Detection of multidimensional states with fiber optics

Full title: Measuring multi-dimensional tension and strain fields in concrete structures with distributed fiber optic sensors

Fiber optic sensors make quasi-continuous strain measurements possible. Therefore, this measurement technique has a great potential for monitoring the behavior of structures, such as global deformations. By mounting those linear sensors in a mesh-like manner, the three-dimensional deformation state of the structure can be detected. Hence, approaches to estimate two- and three-dimensional strain, stress and deformation fields from linear strain data are to be developed in this thesis. The algorithms are to be implemented into a Python software framework and validated by means of experimental data.

The work consists of the following tasks:

• Literature research regarding distributed fiber optic sensors and efficient mapping approaches
• Implementation of (geo-referenced) localization and orientation of the fiber optic sensor in three-dimensional space
• Linking the sensor to virtual models of the structure
• Reconstruction of the strain field of the structure by means of interpolation or numerical deformation calculations
• Calculation of the stress field and deformation state by means of analytical or numerical methods
• Implementation of the (analytical) algorithms into the Python software framework
• Validation of the developed methods using experimental data

Details of the task will be refined prior and while working on the project. Interest/experience in software development/programming is advantageous.

Contact person:
Dipl.-Ing. Bertram Richter
0351 463 32822

F - 12 Long-term effects on fiber optic sensors

Full title: Influence of long-term effects on distributed fiber optic sensors on concrete components

Fiber optic sensors make quasi-continuous strain measurements possible. Therefore, the interest in deployment of this measurement technique grows for structural health monitoring, especially for crack width monitoring. Concrete shows long-term load induced and load independend behavior. Real structures are subject to temperature fields. These influences alter measured strain signals and need to be compensated for a monitoring from the first hour.

Hence, approaches to compensate alduterating influences are to be developed in this thesis. The algorithms are to be implemented into a Python software framework and validated by means of experimental data. The work consists of the following tasks:

• Literature research regarding long-term and temperature behavior of distributed fiber optic sensors and concrete
• Development of a numerical model to investigate the temperature fields and long-term behavior of a concrete structure
• Development of compensation approaches
• Implementation of the algorithms into the Python software framework
• Validation of the developed methods using experimental data and the numerical model

Details of the task will be refined prior and while working on the project. Interest/experience in software development/programming is advantageous.

Contact person:
Max Herbers, M.Sc.
0351 463 39620

F - 10 Quality measures for distributed fiber optic sensor data

Fiber optic sensors make quasi-continuous strain measurements possible. Therefore, the interest in deployment of this measurement technique grows for structural health monitoring, especially for crack width monitoring. The quality of the resulting data depends on several faktors (sensor type, application method, measurement settings, etc.). Different sensor applications and their fitners for crack width monitoring were investigated. Due to missing quality measures, objective assessments are not yet possible.

Hence, such quality measures for fiber optic sensor data are to be developed in this thesis. The algorithms are to be implemented into a Python software framework and validated by means of experimental data. The work consists of the following tasks:

• Literature research regarding distributed fiber optic sensors and data quality measures
• Discussion of existing approaches for quality assessment and development of own approaches
• Development of quality measure for technical disturbances (noise, errors, missing data)
• Development of quality measure for strain peaks (detectability, stability)
• Implementation of the algorithms into the Python software framework
• Validation of the developed methods using experimental data
• Analysis of the experiments regarding data qualtity and deriving setup recommendations

Details of the task will be refined prior and while working on the project. Interest/experience in software development/programming is advantageous.

Contact person:
Dipl.-Ing. Bertram Richter
0351 463 32822

F - 7 AI based design assistant für early stage bridge design

Generally, bridge design is a creative process with multiple requirements and constraints. The design process is often manual and without (partially) automated support processes. Accordingly, it can be optimized and this is particularly important in times of high demand for new construction and growing sustainability requirements. Bridge design takes place within given, multiple boundary conditions. In the "preliminary design" phase, a preferred variant must be found, taking into account important boundary conditions such as topography and the road profile. Methods of artificial intelligence (AI) and machine learning (ML) enable nowadays design assistance for general design methods and also for bridge design. Data from built bridges can be taken into account and new design proposals can be derived from them. However, built bridges may not always be optimized or the reasons for the design determination may be manifold and therefore not always explainable.

The aim of this work is to develop a digital design assistant for the preliminary design of a frame bridge using AI methods that suggests possible bridge variants for given boundary conditions. The work can be processed by individually selecting between the following steps:

• Research and categorization of AI methods for the design process
• Development of a parametric FEM optimization model (Software: Grasshopper + Karamba + e.g. Galapogos)
• Development of a data set of synthetic bridge designs of superstructures based on existing data sets for real bridge designs
• Analyzing the synthetic database of optimized bridges for existing bridges (software: OrangeDataScience). For this purpose, a cluster analysis is to detect statical and constructive correlations. A correlation analysis should explain the correlations with the target variables for certain design parameters in more detail.
• Training of existing regression algorithms and evaluation of their reliability

The work is part of the research project mFUND-HyBridGen – Hybrid Bridge Generator: AI-based bridge generator with knowledge and experience data and early citizen participation. Support can be provided jointly with the planning and software company A+S Consult GmbH.

Contact person:
Jakob Grave, M.Sc.
030 220 777 74

F - 1 Damage assessment with ultrasonic measurements

Full title: Damage assessment of cyclically loaded concrete structures with ultrasonic measurements

Concrete structures under a given load do not fail because they abruptly change from a "normal" state to a fracture state, but because the degradation process progresses with increasing load until material failure occurs. When subjected to mechanical loads, stresses first concentrate around material defects or interfaces at the microscale, destroying bonds between individual molecules. With increasing mechanical load, the microcracks then grow and unite, leading to the formation of macrocracks. During this process, the lattice structure of the material, which serves as a propagation medium for the stress waves of an ultrasonic pulse, is progressively changed and in this way the damage can be detected.

The objective of this thesis is to relate ultrasonic measurements of degradation evolution from concrete specimens and beams subjected to cyclic loading to hypotheses of damage accumulation. From these correlations and using concepts of robustness and redundancy, safety factors will be determined and the remaining useful life will be evaluated.

Contact person:
Raúl Enrique Beltrán Gutiérrez, M.Sc.
0351 463 33675