Adaptive Manufacturing Based on Active Sampling for Multi-component Individual Assembly

The modular design of load-bearing structures out of precast concrete components offers considerable potential for more efficient and sustainable construction. A new concept of individual assembly can be used by compensating dimensional deviations of the individual modules through selective positioning of the components within the structure. The quality of the resulting structure depends on the combination of all component dimensions. In this paper, an adaptive process control for production is presented which aims at an optimal result over all components of the structure. Such a cross-component process control has not been considered in literature yet, since components of a product are only considered individually so far. The paper furthermore considers that the component dimensions depend in an unknown and stochastic way on certain setting parameters of the production process. By using active sampling to vary the setting parameters adaptively during production, knowledge about the transfer function can be gained efficiently. In particular, the heat treatment process for ultra-high performance concrete components is considered as use case. A cross-component process control is presented, which iteratively approximates the transfer function of the manufacturing process during operation by means of regression. Different strategies for the selection of parameter settings are developed, optimized and compared by simulation. In comparison with a benchmark strategy based on complete information, it can be shown that good results can be achieved by selected strategies depending on the overall number of components considered.