Researchers from the University of Derby have proposed a novel threshold-based governance framework designed to address this disconnect. The study, published in Discover Sustainability, quantifies the relationship between technical performance and sustainability metrics without requiring the retraining of complex machine learning models. By making these trade-offs visible and navigable, the framework equips maintenance practitioners with the ability to align their operations with strategic environmental, social, and governance goals. This approach represents a shift from purely algorithmic optimization to a more holistic management strategy that considers the broader impact of industrial decisions.

The core of the proposed framework relies on a utility function that incorporates strategic priority coefficients. These coefficients allow decision-makers to translate high-level organizational goals into specific operating points for their maintenance systems. For instance, a company prioritizing carbon reduction can adjust these coefficients to favor operating points that minimize environmental impact, even if it means accepting a slightly lower technical accuracy in failure detection. This flexibility is crucial for navigating the complex realities of industrial management, where sustainability targets often conflict with immediate reliability requirements.

To demonstrate the practical applicability of this framework, the researchers used a Random Forest classifier applied to the NASA C-MAPSS turbofan dataset as a proof-of-concept. This dataset is widely recognized in the field for its comprehensive representation of turbofan engine degradation. By varying thirty classification thresholds ranging from 0.3 to 0.9, the team generated a Pareto frontier of non-dominated solutions. This frontier maps out the optimal trade-offs between the F1-score, a measure of technical performance, and various sustainability metrics. The resulting map provides a clear visual and quantitative guide for where the system should operate based on specific strategic needs.

The analysis revealed a fascinating dual-regime trade-off structure that challenges conventional assumptions about performance optimization. In the initial regime, known as the win-win region, improving reliability was associated with a reduction in environmental impact. Specifically, as the F1-score increased from 0.458 to 0.614, the sustainability costs decreased. This suggests that there is a range of operation where enhancing the accuracy of failure detection actually leads to more efficient resource use, likely by preventing unnecessary maintenance interventions that waste energy and materials.

However, beyond the inflection point at an F1-score of approximately 0.614, the relationship shifts to a regime of diminishing returns. In this second phase, marginal gains in reliability correspond to increasingly rapid increases in sustainability costs. As the F1-score moves from 0.614 to 0.695, the environmental and financial penalties for pursuing higher accuracy become disproportionately large. This inflection point provides an evidence-based stopping rule for optimization, indicating that pushing for perfect technical accuracy is not always the most sustainable or cost-effective strategy for industrial operators.

To ensure the robustness of these findings, the researchers conducted extensive cross-validation, bootstrapping, and sensitivity analyses. These statistical methods were used to assess how the dual-regime structure holds up across different data splits and model choices. The results indicated that the qualitative trade-off structure was partially preserved, suggesting that the observed patterns are not merely artifacts of a specific dataset or model configuration. While the full details are provided in the supplementary materials, the consistency of the dual-regime behavior across various tests strengthens the credibility of the framework as a generalizable tool for industrial governance.

The implications of this work extend beyond the specific case study of turbofan engines. By closing a practical governance gap, the framework allows for a more nuanced approach to predictive maintenance. It acknowledges that sustainability and reliability are not always in direct conflict but can sometimes be aligned, depending on the operating point. This insight is particularly valuable for companies striving to meet stringent ESG targets without compromising the integrity of their industrial assets. The ability to quantify the carbon cost of improving failure detection by a specific percentage, such as five percent, provides a tangible metric for strategic planning.

Despite its contributions, the study acknowledges certain limitations that must be considered when applying the framework. The proof-of-concept was conducted using a single dataset and one model class, which limits the immediate generalizability to all industrial contexts. Furthermore, the impact values used in the analysis are illustrative and require domain-specific calibration to be applicable in real-world scenarios. Future research will likely focus on expanding the framework to include diverse datasets and model architectures, as well as integrating more detailed economic and environmental impact models to refine the utility function.

As industries continue to grapple with the dual mandates of reliability and sustainability, the need for practical governance tools becomes increasingly urgent. The threshold-based approach proposed by Massoud, Meziane, and AlZoubi offers a promising path forward by making the trade-offs explicit and manageable. By providing a structured way to navigate the complex relationship between technical performance and environmental impact, this framework empowers practitioners to make informed decisions that support both operational excellence and long-term sustainability goals. This shift in perspective is essential for the successful implementation of Industry 5.0 principles across various sectors.

Subject of Research: Threshold-based governance for balancing reliability and sustainability in predictive maintenance

Article Title: A threshold based governance approach to balancing reliability and sustainability in predictive maintenance

Article References: Massoud, A., Meziane, F., & AlZoubi, A. (2026). A threshold based governance approach to balancing reliability and sustainability in predictive maintenance. Discover Sustainability. https://doi.org/10.1007/s43621-026-04882-3

Image Credits: AI Generated

DOI: 10.1007/s43621-026-04882-3

Keywords: Predictive Maintenance, Industry 5.0, Sustainability, Reliability, Pareto Frontier, ESG Governance, Random Forest, Carbon Emissions, Operational Efficiency, Dual-Regime Trade-off, threshold, based