Using machine learning to predict accurate load collectives for the rail/wheel system of stacker cranes

State-of-the art design methods for a stacker crane rail/wheel system use unrealistic load collectives. Therefore, the results of these design methods are often inaccurate. The BS EN 13001-3-3 introduces a more realistic load collective, called the contact force spectrum factor. However, in practice the contact force spectrum factor is difficult to compute, since all contact forces over the considered stacker crane's life cycle need to be known. One way to overcome this obstacle is to run a simulation. However, this approach is computationally expensive. Therefore, this paper introduces machine learning models for predicting the contact force spectrum factor. For data generation a simulation model is used. The presented machine learning models are compared to the standard approach for computing load collectives. In comparison to this standard approach the machine learning models' predictions are significantly more accurate. The presented results make it possible to predict realistic load collectives for the rail/wheel system of stacker cranes. Hence, with this paper's results the efficient application of the BS EN 13001-3-3 for stacker cranes becomes possible.