Scholarlink

This paper addresses the persistent and costly challenge of unplanned machine downtime in the candle manufacturing industry, highlighting the substantial operational disruptions and financial losses caused by unexpected equipment failures. Despite the widespread implementation of automation technologies—such as industrial robots, dispensers, and packaging lines—machine breakdowns remain a significant bottleneck in production. In 2024 alone, the analyzed production line experienced over 1,612 recorded failures, resulting in more than 1,065 h of cumulative downtime and estimated revenue losses of approximately PLN 330,000. Particular attention is given to the lid-applying machine, which emerged as the most failure-prone component of the line, exhibiting a mean time between failures (MTBF) of just 109.92 h. Its recurring malfunctions not only caused frequent halts in production but also led to the destruction of over 207,000 lids and additional material losses of PLN 19,580. To address this issue, the study investigates the application of artificial neural networks (ANNs) as a predictive maintenance tool aimed at anticipating failures before they occur. The predictive model incorporates a range of operational and environmental parameters, including machine operating times, vibration levels, temperature readings, and historical maintenance records. A feed-forward neural network architecture was implemented, utilizing ReLU and sigmoid activation functions to capture non-linear relationships between input signals and failure events. The trained model demonstrated strong predictive performance, achieving an accuracy of approximately 94%, sensitivity of 90%, and specificity exceeding 95%. These results underscore the feasibility and effectiveness of integrating AI-driven diagnostics into industrial operations. By enabling early detection of failure patterns, the proposed approach contributes to minimizing unplanned downtime, optimizing maintenance schedules, and improving overall production efficiency, in line with the goals of Industry 4.0 and smart manufacturing.

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

Predictive Maintenance of Machinery in Candle Manufacturing Using Artificial Neural Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kamil Musial,
  • Marlena Bentkowska,
  • Agnieszka Kotala

摘要

This paper addresses the persistent and costly challenge of unplanned machine downtime in the candle manufacturing industry, highlighting the substantial operational disruptions and financial losses caused by unexpected equipment failures. Despite the widespread implementation of automation technologies—such as industrial robots, dispensers, and packaging lines—machine breakdowns remain a significant bottleneck in production. In 2024 alone, the analyzed production line experienced over 1,612 recorded failures, resulting in more than 1,065 h of cumulative downtime and estimated revenue losses of approximately PLN 330,000. Particular attention is given to the lid-applying machine, which emerged as the most failure-prone component of the line, exhibiting a mean time between failures (MTBF) of just 109.92 h. Its recurring malfunctions not only caused frequent halts in production but also led to the destruction of over 207,000 lids and additional material losses of PLN 19,580. To address this issue, the study investigates the application of artificial neural networks (ANNs) as a predictive maintenance tool aimed at anticipating failures before they occur. The predictive model incorporates a range of operational and environmental parameters, including machine operating times, vibration levels, temperature readings, and historical maintenance records. A feed-forward neural network architecture was implemented, utilizing ReLU and sigmoid activation functions to capture non-linear relationships between input signals and failure events. The trained model demonstrated strong predictive performance, achieving an accuracy of approximately 94%, sensitivity of 90%, and specificity exceeding 95%. These results underscore the feasibility and effectiveness of integrating AI-driven diagnostics into industrial operations. By enabling early detection of failure patterns, the proposed approach contributes to minimizing unplanned downtime, optimizing maintenance schedules, and improving overall production efficiency, in line with the goals of Industry 4.0 and smart manufacturing.