Fakultät Maschinenbau
Refine
Document Type
Language
- English (4)
Has Fulltext
- no (4)
Is part of the Bibliography
- no (4)
Keywords
- optimization (4) (remove)
Institute
Begutachtungsstatus
- peer-reviewed (3)
Absorber-free laser transmission welding enables clean and precise joining of plastics without additives or adhesives. It is therefore well suited to produce optical and medical devices, which place high demands on cleanliness and accuracy.
However, the weld usually has an undesirably large vertical expansion, causing bulges and distortion. To improve this, the intensity distribution of the laser beam as well as the processing strategy must be adapted. Due to the complexity, this is aided by process simulation. However, simulation parameter calibration and verification are usually done considering the seam width and height, which is of limited significance. To overcome this, we propose a new method for image processing of microtome sections, determining the spatially resolved geometry of the weld. Thus, the deviation between experiment and simulation can be calculated pixel by pixel. This spatially resolved value is predestined for the calibration of the simulation parameters: For a parameter field with 18 different settings, the total deviation between experiment and simulation is less than 11 % after calibration.
Due to the complexity and the number of factors involved in factory layout planning, computers were identified as an efficient tool to support the process. However, so far no method for computer-aided layout planning has gained wide acceptance in practical application. One reason for this is that in present approaches either the user or the computer designs the layout, neglecting either the qualitative or the quantitative goals. To bridge this gap, this article introduces a concept for human-computer-integration based on evaluative feedback and inverse reinforcement learning. A key element of the concept is the interactive planning process in which user and computer alternately design and improve the layout until a satisfactory layout is found. The user evaluates the layouts according to qualitative criteria, adjusts them intuitively and specifies objectives and restrictions in an explorative way. The computer on the other hand – in form of a reinforcement algorithm – generates possible layouts and incorporates the user’s feedback into its policy. This synergy is expected to generate better results than an expert or an algorithm alone could. Furthermore, in the context of learning factories, it encourages critical thinking and allows students to develop a deeper understanding of the factors that contribute to efficient manufacturing processes. Both an architecture for the implementation is proposed and the requirements for the user interface are specified.
In recent years, the Tuned Mass Damper with inerter (TMDI) has received significant attention. The inerter is defined to exert a force that is in proportion to the relative acceleration of the two inerter terminals. Here, two TMDI topologies are investigated. The conventional topology is given by the inerter being in parallel to the spring and viscous damper of the TMDI. The other topology is the serial arrangement of spring, inerter and viscous damper being in parallel to the stiffness of the mass spring oscillator of the TMDI. While the first topology intends to increase the inertial force of the TMDI, the second topology aims at producing an additional degree of freedom. The considered TMDI concepts are simulated for harmonic and random excitations, with parameters set according to those described in the literature and with numerically optimized parameters which minimize the primary structure displacement response. The classical TMD is used as a benchmark. The findings are twofold. The conventional TMDI with typical inertance ratio of 1% and the very small value of 0.02% performs significantly worse than the classical TMD with the same mass ratio. In contrast, the TMDI with an additional degree of freedom can improve the mitigation of the primary structure if the inertance ratio is set very small and if the TMDI parameters are numerically optimized.
Thermo-mechanical simulation offers great opportunities to optimize welding processes of plastics. For realistic simulation, the temperature dependent mechanical properties need to be implemented from ambient temperature to temperatures above the flow temperature. Standard test methods are insufficient for characterization in the entire temperature range because close to the flow temperature the material is too soft for tensile tests and too stiff for rheometry. Therefore, an optimization strategy is developed, that determines unknown material parameters by testing in welding simulations. The unknown parameters are iteratively adjusted to minimize the mismatch between computed and measured set-paths. Thus, important process characteristics are calculated realistically, enabling the computer aided assessment of the weld quality.