Evaluation method for crane luffing vibration based on threshold and BP neural network

Abstract

The issue of luffing vibration in truck cranes poses a significant threat to production safety. This paper presents the first vibration evaluation method for cranes based on thresholds and backpropagation neural network (BPNN). To address the unreliability of vibration judgment based on a single characteristic value, the method makes judgments based on four characteristic values derived from two types of signals (pressure and vibration acceleration). To tackle the issues of poor interpretability and low reliability in black-box model-based diagnostic methods, a threshold-based pre-diagnosis step is introduced. To resolve the contradiction in judgment results from different characteristic values, feature vectors and BPNN are introduced. The recognition effectiveness of the proposed method was validated using test data from luffing experiments under multiple working conditions of a crane, achieving an accuracy rate of 99.54%. The proposed method, based on test signals, avoids the complex modeling process of traditional methods. Both the threshold criteria and pre-diagnosis results can be intuitively understood by users, endowing the method with strong interpretability and significant potential for engineering applications.