Filtern
Dokumenttyp
Sprache
- Englisch (11)
Schlagworte
- Neural networks (11) (entfernen)
Organisationseinheit der BAM
Paper des Monats
- ja (1)
Eingeladener Vortrag
- nein (1)
Various applications for artificial intelligence in the context of spectroscopy will be presented. in particular, examples from bamline will be featured. After a short introduction to synchrotron radiation, artificial intelligence algorithms for the quantification of X-ray fluorescence measurement are discussed. In the second example, information retrieval by natural language processing is reviewed. As a last example the reconstruction of measurements with the X-ray color camera and coded apertures is presented.
This dataset accompanies the following publication, first published in Scientific Reports (www.nature.com/articles/s41598-021-84287-6):
B. Ruehle, J. Krumrey, V.-D. Hodoroaba, Scientific Reports, Workflow towards Automated Segmentation of Agglomerated, Non-Spherical Particles from Electron Microscopy Images using Artificial Neural Networks, DOI: 10.1038/s41598-021-84287-6
It contains electron microscopy micrographs of TiO2 particles, the corresponding segmentation masks, and their classifications into different categories depending on their visibility/occlusion. Please refer to the publication and its supporting information for more details on the acquisition and contents of the dataset, as well as the GitHub repository at https://github.com/BAMresearch/automatic-sem-image-segmentation
An approach to develop an arc sensor for gap width estimation during automated NG-GMAW with a weaving electrode motion is introduced by combining arc sensor readings with optical measurements of the groove shape to allow precise analyses of the process. The two test specimen welded for this study were designed to feature a variable groove geometry in order to maximize efficiency of the conducted experimental efforts, resulting in 1696 individual weaving cycle records with associated arc sensor measurements, process parameters and groove shape information. Gap width was varied from 18 mm to 25 mm and wire feed rates in the range of 9 m/min to 13 m/min were used in the course of this study. Artificial neural networks were applied as a modelling tool to derive an arc sensor for estimation of gap width suitable for online process control that can adapt to changes in process parameters as well as changes in the weaving motion of the electrode. Wire feed rate, weaving current, sidewall dwell currents and angles were defined as inputs to calculate the gap width. The evaluation of the proposed arc sensor model shows very good estimation capabilities for parameters sufficiently covered during the experiments.
The objective of this paper is to demonstrate a new simulation technique which allows the fast and automatic generation to temperature fields based on a combination of empirical and phenomenological modelling techniques. The automatic calibration of the phenomenological model is performed by a multi-variable global optimisation routine which yields the optimal fit between simulated and experimental weld charcteristics without the need for initial model parameters. For exemplary welding processes it is shown that linear 3D heat conduction models can approximate measured temperature fields with a high accuracy. The modelling approach presented comprises the automatic calibration against multiple experiments which permits simulating the temperature field for unknown process parameters. The validation of this composite simulation model is performed for exemplary welding processes and includes the prediction of the fusion line in the cross section and the corresponding thermal cycles.
Fast temperature field generation for welding simulation and reduction of experimental effort
(2011)
The quality of welding processes is governed by the occurring induced distortions yielding an increase in production costs due to necessary reworking. Especially for more complex specimens, it is difficult to evaluate the optimal configuration of welding sequences in order to minimize the distortion. Even experienced welding operators can solve this task only by trial and error which is time and cost consuming. In modern engineering the application of welding simulation is already known to be able to analyse the heat effects of welding virtually. However, the welding process is governed by complex physical interactions. Thus, recent weld thermal models are based on many simplifications. The state of the art is to apply numerical methods in order to solve the transient heat conduction equation. Therefore, it is not possible to use the real process parameters as input for the mathematical model. The model parameters which allow calculating a temperature field that is in best agreement with the experiments cannot be defined directly but inversely by multiple simulations runs. In case of numerical simulation software based on finite discretization schemes this approach is very time consuming and requires expert users. The weld thermal model contains an initial weakness which has to be adapted by finding an optimal set of model parameters. This process of calibration is often done against few experiments. The range of model validity is limited. An extension can be obtained by performing a calibration against multiple experiments. The focus of the paper is to show a combined modelling technique which provides an efficient solution of the inverse heat conduction problem mentioned above. On the one hand the inverse problem is solved by application of fast weld thermal models which are closed form solutions of the heat conduction equation. In addition, a global optimization algorithm allows an automated calibration of the weld thermal model. This technique is able to provide a temperature field automatically that fits the experimental one with high accuracy within minutes on ordinary office computers. This fast paradigm permits confirming the application of welding simulation in an industrial environment as automotive industry. On the other hand, the initial model weakness is compensated by calibrating the model against multiple experiments. The unknown relationship between model and process parameters is approximated by a neural network. The validity of the model is increased successively and enables to decrease experimental effort, For a test case, it is shown that this approach yields accurate temperature fields within very short amount of time for unknown process parameters as input data to the model contributing to the requirement to construct a substitute system of the real welding process.
Fast temperature field generation for welding simulation and reduction of experimental effort
(2009)
The quality of welding processes is governed by the occurring induced distortions yielding an
increase in production costs due to necessary reworking. Especially for more complex specimens
it is difficult to evaluate the optimal configuration of welding sequences in order to
minimise the distortion. Even experienced welding operators can solve this task only by trial
and error which is time and cost consuming.
In modern engineering the application of welding simulation is already known to be able to
analyse the heat effects of welding virtually. However, the welding process is governed by
complex physical interactions. Thus, recent weld thermal models are based on many simplifications.
The state of the art is to apply numerical methods in order to solve the transient heat
conduction equation. Therefore, it is not possible to use the real process parameters as input
for the mathematical model. The model parameters which allow calculating a temperature
field that is in best agreement with the experiments cannot be defined directly but inversely by
multiple simulations runs. In case of numerical simulation software based on finite discretisation
schemes this approach is very time consuming and requires expert users. The weld
thermal model contains an initial weakness which has to be adapted by finding an optimal set
of model parameters. This process of calibration is often done against few experiments. The
range of model validity is limited. An extension can be obtained by performing a calibration
against multiple experiments.
The focus of the paper is to show a combined modelling technique which provides an efficient
solution of the inverse heat conduction problem mentioned above. On the one hand the inverse
problem is solved by application of fast weld thermal models which are closed form
solutions of the heat conduction equation. In addition, a global optimisation algorithm allows
an automated calibration of the weld thermal model. This technique is able to provide a temperature
field automatically that fits the experimental one with high accuracy within minutes
on ordinary office computers. This fast paradigm permits confirming the application of welding
simulation in an industrial environment as automotive industry.
On the other hand, the initial model weakness is compensated by calibrating the model
against multiple experiments. The unknown relationship between model and process parameters
is approximated by a neural network. The validity of the model is increased successively
and enables to decrease experimental effort. For a test case it is shown, that this approach
yields accurate temperature fields within very short amount of time for unknown process parameters
as input data to the model contributing to the requirement to construct a substitute
system of the real welding process.
A long distance range over tens of kilometers is a prerequisite for a wide range of distributed fiber optic vibration sensing applications. We significantly extend the attenuation-limited distance range by making use of the multidimensionality of distributed Rayleigh backscatter data: Using the wavelength-scanning coherent optical time domain reflectometry (WS-COTDR) technique, backscatter data is measured along the distance and optical frequency dimensions. In this work, we develop, train, and test deep convolutional neural networks (CNNs) for fast denoising of these two-dimensional backscattering results. The very compact and efficient CNN denoiser “DnOTDR” outperforms state-of-the-art image denoising algorithms for this task and enables denoising data rates of 1.2 GB/s in real time. We demonstrate that, using the CNN denoiser, the quantitative strain measurement with nm/m resolution can be conducted with up to 100 km distance without the use of backscatter-enhanced fibers or distributed Raman or Brillouin amplification.
The objective of this paper is to demonstrate a new simulation technique which allows fast and automatic generation of temperature fields as input for subsequent thermomechanical welding simulation. The basic idea is to decompose the process model into an empirical part based on neural networks and a phenomenological part that describes the physical phenomena. The strength of this composite modelling approach is the automatic calibration of mathematical models against experimental data without the need for manual interference by an experienced user. As an example for typical applications in laser beam and GMA-laser hybrid welding, it is shown that even 3D heat conduction models of a low complexity can approximate measured temperature fields with a sufficient accuracy. In general, any derivation of model fitting parameters from the real process adds uncertainties to the simulation independent of the complexity of the underlying phenomenological model. The modelling technique presented hybridises empirical and phenomenological models. It reduces the model uncertainties by exploiting additional information which keeps normally hidden in the data measured when the model calibration is performed against few experimental data sets. In contrast, here the optimal model parameter set corresponding to a given process parameter is computed by means of an empirical submodel based on relatively large set of experimental data. The approach allows making a contribution to an efficient compensation of modelling inaccuracies and lack of knowledge about thermophysical material properties or boundary conditions. Two illustrating examples are provided.
We present a workflow for obtaining fully trained artificial neural networks that can perform automatic particle segmentations of agglomerated, non-spherical nanoparticles from scanning electron microscopy images “from scratch”, without the need for large training data sets of manually annotated images. The whole process only requires about 15 minutes of hands-on time by a user and can typically be finished within less than 12 hours when training on a single graphics card (GPU). After training, SEM image analysis can be carried out by the artificial neural network within seconds. This is achieved by using unsupervised learning for most of the training dataset generation, making heavy use of generative adversarial networks and especially unpaired image-to-image translation via cycle-consistent adversarial networks. We compare the segmentation masks obtained with our suggested workflow qualitatively and quantitatively to state-of-the-art methods using various metrics. Finally, we used the segmentation masks for automatically extracting particle size distributions from the SEM images of TiO2 particles, which were in excellent agreement with particle size distributions obtained manually but could be obtained in a fraction of the time.