Project Details
Projekt
Data- and Response-Surface-Driven Design Assistant for Controlled Flexible Multibody Systems
Subject Area
Mechanics
Term
since 2022
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 501840485
Flexible multibody systems appear in many applications, ranging from industrial robots in manufacturing to robotized support in surgery, usually including actuation and corresponding controllers. Thus, helping engineers to come up with improved designs of such systems has a large impact across many application fields. Today, the design of these systems is usually still based on an analysis-centric approach, where human engineers modify the mechanical and control design by hand based on manual inspection of simulation and experiment results, where the result is rather a function of individual expertise than being optimal with respect to formal criteria. Moreover, the mechanical design is often done first without paying attention to design criteria from control engineering. However, including control-engineering considerations already in the mechanical design phase could vastly improve the effectiveness of the controlled system but is often impossible in a purely expert-driven design workflow since engineers are rarely equally well-trained in several disciplines. The artificially intelligent design assistant developed in this project will bridge this gap, revolutionizing the design and analysis process of controlled flexible multibody systems by leveraging modern machine-learning techniques at several points in the design process. The assistant will jointly optimise the mechanical and the control parts of the system according to practical, often transient performance criteria. Large-scale statistical analyses of the system play a key role in the project to assess the system performance according to proper design criteria. However, many statistical analyses will only be made feasible by criteria-specific, machine-learned response surfaces, allowing to consider and come up with design criteria that previously were computationally inconceivable to be integrated in an automated optimisation-based design process. Moreover, in the assisted design procedure that will result from this project, machine learning will allow to treat flexible multibody systems where parts cannot conveniently be described by first-principles modelling, making the usage of data-driven learning methods unavoidable. Machine-learned surrogate models and model parts will ensure the real-time capability of model-based controllers designed jointly with the mechanical parts of the system. Software will be developed to formulate model-based controllers automatically using the system model and exporting them in a way directly ready for real-time execution. Techniques from artificial intelligence will better guide multicriteria optimisation algorithms into regions of the design space where better designs may be found than by traditional, heuristic schemes. All this methodological research will enable engineers to be automatically guided toward better designs.
DFG Programme
Priority Programmes
