Uncertainty Aware Functional Behavior Prediction and Material Fatigue Assessment for Circular Factory

2026-06-06 · Source: cs.AI updates on arXiv.org · Field: Manufacturing & Industrial — Smart Manufacturing & Industry 4.0, Materials & Production Technology, Artificial Intelligence & Machine Learning · Depth: Expert, extended

Summary

This research introduces an uncertainty-aware framework for instance-specific assessment of returned angle grinders in circular factories. It combines a conditional sequence model for functional behavior prediction with a material-fatigue model for component integrity. The functional model, using an LSTM, achieved a mean 2%-tolerance accuracy of 0.9652 and R² of 0.8365 across nine outputs, with thermal variables predicted near-perfectly. Drive motor current and load speed were most challenging but showed high R² values (0.9750 and 0.9924). The material model, based on Paris-law crack propagation, predicted approximately 31 reuse cycles for the output shaft under nominal service loading, but this dropped to 3 cycles with a 1.6x amplification of the upper 10% of stress amplitudes. The integrated streaming algorithm coherently updates functional, material, and system reliability.

Key takeaway

For Machine Learning Engineers developing prognostic solutions for circular manufacturing, you should prioritize integrating both system-level functional prediction and component-level material fatigue assessment. Your models must incorporate uncertainty awareness and be sensitive to rare high-load events, as these significantly impact component reusability. Focus on robustly capturing dynamic output variables like motor current and load speed, which are critical for functional risk assessment.

Key insights

Jointly assessing functional behavior and material fatigue enables instance-specific redeployment decisions for returned products.

Principles

• Future functional behavior depends on current state and recent usage history.
• High-load events significantly reduce material reuse potential.
• Uncertainty-aware models improve reliability assessment.

Method

A convolutional encoder extracts loading patterns, an LSTM predicts nine functional variables as Gaussian mean/variance, and parallel finite-element/Paris-law analysis assesses material fatigue. A streaming replay consolidates reliability.

In practice

• Use force-torque usage windows for dynamic output prediction.
• Integrate FE stress reconstruction for component fatigue.
• Evaluate reliability under prescribed future usage scenarios.

Topics

• Circular Economy
• Prognostics and Health Management
• Material Fatigue Assessment
• LSTM Networks
• Uncertainty Quantification
• Angle Grinder
• Reliability Engineering

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Uncertainty Aware Functional Behavior Prediction and Material Fatigue Assessment for Circular Factory — The framework links functional prediction and material fatigue for instance-specific reliability in circular factories.
• FactoryBench: Evaluating Industrial Machine Understanding — FactoryBench evaluates time-series and LLM machine understanding using causal Q&A over industrial robotic telemetry.
• A Multi-Plant Machine Learning Framework for Emission Prediction, Forecasting, and Control in Cement Manufacturing — Data-driven ML models can significantly reduce cement plant NOx emissions and costs without hardware changes.
• News - NSWCPD Applies AI and Machine Learning to Predictive Machinery Health - DVIDS — AI/ML-driven vibration analysis can predict mechanical failures, enhancing condition-based maintenance for naval platforms.
• From Possible to Probable AI Models — AI reliability hinges on distinguishing between what models *can* do and what they *consistently* do.
• Flexible Functional Treatment Effect Estimation — A WMKRR method directly minimizes uniform balancing error for functional treatment effect estimation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.