Neural network approach to hoist deceleration control

This paper introduces a neural network approach to hoist deceleration control of industrial hoist mechanisms, with particular focus to crane applications. The necessity for investigation in this field arises from the increasing demands in terms of safety within in the industry. Should the industrial hoist feature too high deceleration this can lead to overstressing of the hoist mechanism and structure, further, damaging of the load due to large dynamical forces. Furthermore, too low deceleration can lead to incompliance with industrial standards and thus being a safety issue, due to potential loss of load in the worst case. Till this day various solutions and devices have been proposed to achieve controlled deceleration of the industrial hoist braking. However, there still lies a necessity for deeper study into this problem, to achieve quicker response towards the desired behavior of the hoist deceleration as well as improved adherence with the desired behavior. Thus, this paper analyses the potentials of hoist deceleration control by neural network architectures as such the linear, quadratic and cubic neural units with real-time recurrent learning and back-propagation through time approach when real measured data are used for experimental analysis.