Scholarlink

<p>Turbochargers are an integral part of haul truck systems that enhance fuel efficiency and reduce carbon emissions. While crucial for heavy-duty mining equipment, a turbocharger failure can destroy the entire engine and even cause fires that could destroy the entire haul truck and harm the driver. Conventional approaches to maintenance, which are mainly scheduled inspections, often prove to be costly and inefficient when conditions change quickly. This research introduces a machine learning framework to monitor turbocharger health and proactively forecast failures using operational data from haul trucks at a Northern Ontario mine. This framework employs Random Forest, XGBoost, Support Vector Machines (SVM), Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM) approaches to model the relationship between key haul truck sensors and turbocharger turbine intake temperature. The proposed approach was tested on two different haul trucks, with SVM, LSTM and DNN outperforming the other ML approaches with their low mean absolute error, root mean squared error and a custom-defined asymmetric error metric that prioritizes safety. Moreover, the results identified anomalies correlated with a confirmed powertrain issue in one truck. The proposed framework offers early failure detection, contributing to enhanced safety and equipment longevity and reduced downtimes in mining operations.</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

Machine Learning Approach to Turbocharger Health Monitoring and Failure Forecasting in Mining Haul Trucks

  • 期刊论文

  • 发表日期：

    2026年4月15日

查看全文
  • Nuri Kalyoncuoglu,
  • Yuksel Asli Sari

摘要

Turbochargers are an integral part of haul truck systems that enhance fuel efficiency and reduce carbon emissions. While crucial for heavy-duty mining equipment, a turbocharger failure can destroy the entire engine and even cause fires that could destroy the entire haul truck and harm the driver. Conventional approaches to maintenance, which are mainly scheduled inspections, often prove to be costly and inefficient when conditions change quickly. This research introduces a machine learning framework to monitor turbocharger health and proactively forecast failures using operational data from haul trucks at a Northern Ontario mine. This framework employs Random Forest, XGBoost, Support Vector Machines (SVM), Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM) approaches to model the relationship between key haul truck sensors and turbocharger turbine intake temperature. The proposed approach was tested on two different haul trucks, with SVM, LSTM and DNN outperforming the other ML approaches with their low mean absolute error, root mean squared error and a custom-defined asymmetric error metric that prioritizes safety. Moreover, the results identified anomalies correlated with a confirmed powertrain issue in one truck. The proposed framework offers early failure detection, contributing to enhanced safety and equipment longevity and reduced downtimes in mining operations.