Main competences in Physics, Material Science, and Informatics working inter-disciplinary! I am a senior researcher and lecturer at the University of Bremen, Germany. Research topics are distributed systems and distributed artificial intelligence, multi-agent systems and machine learning, sensor networks, materials informatics (computation within materials), compiler and programming design.
Manufacturing processes are increasingly adapted to multi-material part production to facilitate ... more Manufacturing processes are increasingly adapted to multi-material part production to facilitate lightweight design via improvement of structural performance. The difficulty lies in determining the optimum spatial distribution of the individual materials. Multi-Phase Topology Optimization (MPTO) achieves this aim via iterative, linear-elastic Finite Element (FE) simulations providing element- and part-level strain energy data under a given design load and using it to redistribute predefined material fractions to minimize total strain energy. The result us a part configuration offering maximum stiffness. The present study implements different material redistribution and optimization techniques and compares them in terms of optimization results and performance: Genetic algorithms are matched against simulated annealing, the latter with and without physics-based constraints. Both types employ partial randomization in generating new configurations to avoid settling into local rather tha...
Due to the increasing use of the different composite materials in lightweight applications, such ... more Due to the increasing use of the different composite materials in lightweight applications, such as in aerospace, it becomes crucial to understand the different damages occurring within them during life cycle and their possible inspection with different inspection techniques in different life cycle stages. A comprehensive classification of these damage patterns, measuring signals, and analysis methods using a taxonomical approach can help in this direction. In conjunction with the taxonomy, this work addresses damage diagnostics in hybrid and composite materials, such as fibre metal laminates (FMLs). A novel unified taxonomy atlas of damage patterns, measuring signals, and analysis methods is introduced. Analysis methods based on advanced supervised and unsupervised machine learning algorithms, such as autoencoders, self-organising maps, and convolutional neural networks, and a novel z-profiling method, are implemented. Besides formal aspects, an extended use case demonstrating dama...
This issue of Engineering Proceedings gathers the papers presented at the 8th International Elect... more This issue of Engineering Proceedings gathers the papers presented at the 8th International Electronic Conference on Sensors and Applications (ECSA-8), held online on 1–15 November 2021, through the sciforum.net platform developed by MDPI [...]
There is an emerging field of new materials, including, but not limited to, fibre-metal laminates... more There is an emerging field of new materials, including, but not limited to, fibre-metal laminates, foam materials, and materials processed by additive manufacturing, highly related to space applications. Typically, material properties such as yield strength or inelastic behaviour are determined from tensile tests. The main disadvantage of tensile testing is the irreversible modification of the device under test (only one experiment possible!). We develop and investigate the training of approximating predictor functions by Machine Learning (ML) and simple Artificial Neural Networks (ANN) for inelastic and fatigue prediction by history recorded data. The predictor functions should be able to predict irreversible effects like inelastic behaviour and material damage by data measured from simple tensile tests within the elastic range of the materials. We show some preliminary results from a broad range of materials and outline the challenges to derive such predictor functions by using re...
Sensorization aims at equipping technical structures with an analog of a nervous system by provid... more Sensorization aims at equipping technical structures with an analog of a nervous system by providing a network of sensors and communication facilities that link them. The objective is that, instead of having been designed to loads and tested to conditions, a structure can experience and report design constraint violations by means of real-time self-monitoring. Specialized electronic components and computational algorithms are needed to derive meaning from the combined signals. For this task, artificial intelligence approaches constantly gain importance; the more so as the trend of ever increasing sensor network size and density suggests that sensor and structure may soon become one, forming a sensorial material. Current simulation techniques capture many aspects of sensor networks and structures. For decision making and communication, the Intelligent Agent paradigm is an accepted approach, as is finite element analysis for structural behavior. To gain knowledge how sensorial structu...
Journal of Intelligent Material Systems and Structures, 2013
ABSTRACT The continuing decrease in size and energy demand of electronic sensor circuits allows e... more ABSTRACT The continuing decrease in size and energy demand of electronic sensor circuits allows endowing engineering structures and, to an increasing degree, materials with integrated sensing and data processing capabilities. Materials that adhere to this description are designated as Sensorial Materials. Their development is multidisciplinary and requires knowledge beyond materials science in fields like sensor science, computer science, energy harvesting, microsystems technology, low-power electronics, energy management, and communication. Development of such materials will benefit from systematic support for bridging research area boundaries. The present article introduces the backbone of an easy-to-use toolbox for layout of the energy supply of smart sensor nodes within a sensorial material. The fundamental approach is transferred from rapid control development, where a comparable MATLAB/Simulink tool chain is already in use. The main goal is to manage limited power resources without unacceptably compromising functionality in a given application scenario. The toolbox allows analysis of the modeled system in terms of energy and power and allows analyzing factors such as energy harvesting, use of predictive power estimation, power saving (e.g. sleep modes), model-based cognitive data reduction methods, and energy aware algorithm switching. It is linked to a simulation environment allowing analysis of energy demand and production in a specific application scenario. Its initial version presented here supports single selfpowered sensor nodes. A broad set of application cases is used to develop scenario-dependent solutions with minimum energy needs and thus demonstrate the use of the toolbox and the associated development process. The initial test case is a large-scale sensor network with optical fiber–based data and energy transmission, for which optimization of energy consumption is attempted. The toolbox can be used to improve the power-aware design of sensor nodes on digital hardware level using advanced high-level synthesis approaches and provides input for sensor node and sensor network level.
Recently emerging trends in engineering and micro-system applications such as the development of ... more Recently emerging trends in engineering and micro-system applications such as the development of sensorial materials show a growing demand for autonomous networks of miniaturized smart sensors and actuators embedded in technical structures [6]. With increasing miniaturization and sensor-actuator density, decentralized network and data processing architectures are preferred or required. A multi-agent system is used for a decentralized and self-organizing approach of data processing in a distributed system like a sensor network, enabling the mapping of distributed data sets to related information, for example, required for object manipulation with a robot manipulator. Traditionally, mobile agents are executed on generic computer architectures [7,8], which usually cannot easily be reduced to single-chip systems like they are required, e.g., in sensorial materials with high sensor node densities. We propose and compare two different data processing and communication architectures for th...
Common Structural Health Monitoring systems are used to detect past damages occurred in structure... more Common Structural Health Monitoring systems are used to detect past damages occurred in structures with sensor networks and external sensor data processing. The time of the damage creation event is commonly unknown. Numerical methods and Machine Learning are used to extract relevant damage information from sensor signals that is characterised by a high data volume and dimension. In this work, distributed multi-instance learning applied to raw time-series of sensor data is deployed to predict the event of the occurrence of a hidden damage in a mechanical structure using typical vibrations of the structure. The sensor processing and learning is performed locally on sensor node level with a global fusion of prediction results to estimate the damage location and the time of the damage creation. Recurrent neural networks with a long-short-term memory architecture are considered implementing a damage discriminator function. The sensor data required for the evaluation of the proposed appro...
Manufacturing processes are increasingly adapted to multi-material part production to facilitate ... more Manufacturing processes are increasingly adapted to multi-material part production to facilitate lightweight design via improvement of structural performance. The difficulty lies in determining the optimum spatial distribution of the individual materials. Multi-Phase Topology Optimization (MPTO) achieves this aim via iterative, linear-elastic Finite Element (FE) simulations providing element- and part-level strain energy data under a given design load and using it to redistribute predefined material fractions to minimize total strain energy. The result us a part configuration offering maximum stiffness. The present study implements different material redistribution and optimization techniques and compares them in terms of optimization results and performance: Genetic algorithms are matched against simulated annealing, the latter with and without physics-based constraints. Both types employ partial randomization in generating new configurations to avoid settling into local rather tha...
Due to the increasing use of the different composite materials in lightweight applications, such ... more Due to the increasing use of the different composite materials in lightweight applications, such as in aerospace, it becomes crucial to understand the different damages occurring within them during life cycle and their possible inspection with different inspection techniques in different life cycle stages. A comprehensive classification of these damage patterns, measuring signals, and analysis methods using a taxonomical approach can help in this direction. In conjunction with the taxonomy, this work addresses damage diagnostics in hybrid and composite materials, such as fibre metal laminates (FMLs). A novel unified taxonomy atlas of damage patterns, measuring signals, and analysis methods is introduced. Analysis methods based on advanced supervised and unsupervised machine learning algorithms, such as autoencoders, self-organising maps, and convolutional neural networks, and a novel z-profiling method, are implemented. Besides formal aspects, an extended use case demonstrating dama...
This issue of Engineering Proceedings gathers the papers presented at the 8th International Elect... more This issue of Engineering Proceedings gathers the papers presented at the 8th International Electronic Conference on Sensors and Applications (ECSA-8), held online on 1–15 November 2021, through the sciforum.net platform developed by MDPI [...]
There is an emerging field of new materials, including, but not limited to, fibre-metal laminates... more There is an emerging field of new materials, including, but not limited to, fibre-metal laminates, foam materials, and materials processed by additive manufacturing, highly related to space applications. Typically, material properties such as yield strength or inelastic behaviour are determined from tensile tests. The main disadvantage of tensile testing is the irreversible modification of the device under test (only one experiment possible!). We develop and investigate the training of approximating predictor functions by Machine Learning (ML) and simple Artificial Neural Networks (ANN) for inelastic and fatigue prediction by history recorded data. The predictor functions should be able to predict irreversible effects like inelastic behaviour and material damage by data measured from simple tensile tests within the elastic range of the materials. We show some preliminary results from a broad range of materials and outline the challenges to derive such predictor functions by using re...
Sensorization aims at equipping technical structures with an analog of a nervous system by provid... more Sensorization aims at equipping technical structures with an analog of a nervous system by providing a network of sensors and communication facilities that link them. The objective is that, instead of having been designed to loads and tested to conditions, a structure can experience and report design constraint violations by means of real-time self-monitoring. Specialized electronic components and computational algorithms are needed to derive meaning from the combined signals. For this task, artificial intelligence approaches constantly gain importance; the more so as the trend of ever increasing sensor network size and density suggests that sensor and structure may soon become one, forming a sensorial material. Current simulation techniques capture many aspects of sensor networks and structures. For decision making and communication, the Intelligent Agent paradigm is an accepted approach, as is finite element analysis for structural behavior. To gain knowledge how sensorial structu...
Journal of Intelligent Material Systems and Structures, 2013
ABSTRACT The continuing decrease in size and energy demand of electronic sensor circuits allows e... more ABSTRACT The continuing decrease in size and energy demand of electronic sensor circuits allows endowing engineering structures and, to an increasing degree, materials with integrated sensing and data processing capabilities. Materials that adhere to this description are designated as Sensorial Materials. Their development is multidisciplinary and requires knowledge beyond materials science in fields like sensor science, computer science, energy harvesting, microsystems technology, low-power electronics, energy management, and communication. Development of such materials will benefit from systematic support for bridging research area boundaries. The present article introduces the backbone of an easy-to-use toolbox for layout of the energy supply of smart sensor nodes within a sensorial material. The fundamental approach is transferred from rapid control development, where a comparable MATLAB/Simulink tool chain is already in use. The main goal is to manage limited power resources without unacceptably compromising functionality in a given application scenario. The toolbox allows analysis of the modeled system in terms of energy and power and allows analyzing factors such as energy harvesting, use of predictive power estimation, power saving (e.g. sleep modes), model-based cognitive data reduction methods, and energy aware algorithm switching. It is linked to a simulation environment allowing analysis of energy demand and production in a specific application scenario. Its initial version presented here supports single selfpowered sensor nodes. A broad set of application cases is used to develop scenario-dependent solutions with minimum energy needs and thus demonstrate the use of the toolbox and the associated development process. The initial test case is a large-scale sensor network with optical fiber–based data and energy transmission, for which optimization of energy consumption is attempted. The toolbox can be used to improve the power-aware design of sensor nodes on digital hardware level using advanced high-level synthesis approaches and provides input for sensor node and sensor network level.
Recently emerging trends in engineering and micro-system applications such as the development of ... more Recently emerging trends in engineering and micro-system applications such as the development of sensorial materials show a growing demand for autonomous networks of miniaturized smart sensors and actuators embedded in technical structures [6]. With increasing miniaturization and sensor-actuator density, decentralized network and data processing architectures are preferred or required. A multi-agent system is used for a decentralized and self-organizing approach of data processing in a distributed system like a sensor network, enabling the mapping of distributed data sets to related information, for example, required for object manipulation with a robot manipulator. Traditionally, mobile agents are executed on generic computer architectures [7,8], which usually cannot easily be reduced to single-chip systems like they are required, e.g., in sensorial materials with high sensor node densities. We propose and compare two different data processing and communication architectures for th...
Common Structural Health Monitoring systems are used to detect past damages occurred in structure... more Common Structural Health Monitoring systems are used to detect past damages occurred in structures with sensor networks and external sensor data processing. The time of the damage creation event is commonly unknown. Numerical methods and Machine Learning are used to extract relevant damage information from sensor signals that is characterised by a high data volume and dimension. In this work, distributed multi-instance learning applied to raw time-series of sensor data is deployed to predict the event of the occurrence of a hidden damage in a mechanical structure using typical vibrations of the structure. The sensor processing and learning is performed locally on sensor node level with a global fusion of prediction results to estimate the damage location and the time of the damage creation. Recurrent neural networks with a long-short-term memory architecture are considered implementing a damage discriminator function. The sensor data required for the evaluation of the proposed appro...
Uploads
Papers by Stefan Bosse