基于 CSA-NARX 模型的电梯钢丝绳动态张力预测方法

为了预测曳引式电梯钢丝绳的动态张力,对带有外部输入的非线性自回归神经网络( NARX )进行研究,利用变色龙优化算法( CSA )对其关键参数进行优化,提出了 CSA-NARX 神经网络模型。该模型在计算速度以及预测精度方面皆优于 NARX 基础模型。最后,利用提出的神经网络模型对电梯上行过程中钢丝绳的动态张力进行预测,其预测精度达到了 97% 。以传统的非平稳时间序列分析模型 ARMA 和 LSTM 为对比,所提出模型的精度更高,验证了所提出模型的有效性。

Prediction Method of Elevator Wire Rope Dynamic Tension Based on CSA-NARX Model

In order to predict the dynamic tension of traction elevator wire rope , the nonlinear auto-regressive neural networkwith exogenous input( NARX ) is studied in this paper.The key parameters are optimized by chameleon swarm algorithm( CSA ), and the CSA-NARX neural network model is proposed. The proposed model is superior to the NARX model in both computation speed and prediction accuracy.Finally , the proposed neural network model is used to predict the dynamic tension of the wire rope in the process of elevator ascending , and the prediction accuracy reaches 97%.Compared with the traditional nonstationary time series analysis model ARMA and LSTM , the accuracy of the proposed model is higher , which verifies the effectiveness of the model proposed.