Adaptive Safe Braking and Distance Prediction for Overhead Cranes With Multivariation Using MLP

The emergency braking and braking distance prediction of an overhead crane pose challenging issues in its safe operation. This paper employs a multilayer perceptron (MLP) to implement an adaptive safe distance prediction functionality for an overhead crane with multiple variations. First, a discrete model of an overhead crane is constructed, and a model predictive control (MPC) model with angle constraints is applied for safe braking. Second, we analysed and selected the input variations of the safe distance prediction model. Subsequently, we permuted the inputs to the MLP and analysed the effect of each input on the accuracy of the MLP in predicting safety distances separately. We constructed a training dataset, and a test dataset and we optimised the safe distance prediction model through the training dataset. Finally, we conducted a comparative analysis between the MLP and nlinfit algorithms, highlighting the superiority of MLP-based adaptive safety distance prediction for bridge cranes. Experiments confirm the method's ability to ensure minimal swing angle during the entire braking process to achieve safe braking. The results underscore the practical utility and novelty of the proposed algorithm.