Scholarlink

<p>Accurate prediction of fuel consumption for driverless trucks is crucial for intelligent dispatching and cost reduction in open-pit mines. To address the challenges of limited sample data and complex, interrelated factors, this paper proposes a fuel consumption prediction model for mining haul trucks based on the Deep Siamese Transformer Network (DSTN). The model leverages the Siamese neural network structure to enhance learning ability with few-shot data and utilizes the multi-head self-attention mechanism of the Transformer model to analyze the complex interactions among influencing factors. Based on field data from open-pit mines, this study evaluates the proposed DSTN model and compares it with baseline models (SVM, Transformer, BP). The results indicate that the DSTN model significantly outperforms all baseline models, with maximum reductions in MSE, MAE, and RMSE of 31.92%, 18.14%, and 15.54%, respectively. Furthermore, ablation studies confirm the contribution of each architectural component. The DSTN model shows low prediction errors and high accuracy in fuel consumption prediction of driverless trucks in open-pit mines.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fuel consumption prediction of driverless trucks in Open-pit mines based on deep Siamese Transformer network model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月15日

查看全文
  • Jingchang Zhao,
  • Zihao Wang,
  • Peng Hou,
  • Shihao Ren

摘要

Accurate prediction of fuel consumption for driverless trucks is crucial for intelligent dispatching and cost reduction in open-pit mines. To address the challenges of limited sample data and complex, interrelated factors, this paper proposes a fuel consumption prediction model for mining haul trucks based on the Deep Siamese Transformer Network (DSTN). The model leverages the Siamese neural network structure to enhance learning ability with few-shot data and utilizes the multi-head self-attention mechanism of the Transformer model to analyze the complex interactions among influencing factors. Based on field data from open-pit mines, this study evaluates the proposed DSTN model and compares it with baseline models (SVM, Transformer, BP). The results indicate that the DSTN model significantly outperforms all baseline models, with maximum reductions in MSE, MAE, and RMSE of 31.92%, 18.14%, and 15.54%, respectively. Furthermore, ablation studies confirm the contribution of each architectural component. The DSTN model shows low prediction errors and high accuracy in fuel consumption prediction of driverless trucks in open-pit mines.