A Multimodal Distribution Model of Stochastic Process Times in Resistance Spot Welding

Commonly used adaptive control systemsAdaptive control systems for resistance spot weldingResistance spot welding in car body construction lead to a multimodal distribution of weld times. This is due to the compensation of disturbances in the process with the help of adaption of parameters such as weld current and prolongation of weld time. In this work, a bimodal Gaussian Mixture Model for stochastic weld times in resistance spot weldingResistance spot welding is presented. On the basis of data from car body construction, expectation values of the model are specified by linear regression. An estimation method to determine the other parameters is introduced. The model is compared to a unimodal distribution model and actual weld times in balancing of a robotic assembly lineAssembly line for an automotive body-in-white backend. In contrast to the unimodal model, the developed model leads to more conservative balancing results than real data. This avoids over-optimistic balancing results, which might have lasting effects. Thus, the model enables methods for stochastic time planning during early planning phases without shop floor data being available.