A PREPRINT
INTEGRATING A CAUSAL FOUNDATION MODEL INTO A PRESCRIPTIVE MAINTENANCE FRAMEWORK FOR OPTIMISING PRODUCTION-LINE OEE
Authors: Felix Saretzky, Thomas Engel, Lucas Andersen, Fazel Ansari
March 10, 2026
The transition to prescriptive maintenance (PsM) in manufacturing is critically constrained by a dependence on predictive models. Such purely predictive models tend to capture statistical asso-ciations in the data without identifying the underlying causal drivers of failure, which can lead to costly misdiagnoses and ineffective measures. This fundamental limitation results in a key challenge: while we can predict that a failure may occur, we lack a systematic method to understand why a failure occurs. This paper proposes a model based on causal machine learning to bridge this gap. Our objective is to move beyond diagnosis to active prescription by simulating and evaluating potential fixes to optimise KPIs such as Overall Equipment Effectiveness (OEE). For this purpose, a pre-trained causal foundation model is used as a “what-if” simulator to estimate the effects of potential fixes. By estimating the causal effect of each intervention on system-level KPIs, specific actions can be recommended for the production line. This can help identify plausible root causes and quantify their operational impact. The model is evaluated using semi-synthetic manufacturing data and compared with non-causal and causal baseline machine learning models. This paper provides a technical basis for a human-centred approach, allowing engineers to test potential solutions in a causal environment to make more effective operational decisions and reduce costly downtimes.
Unlock Prescriptive Power & Operational Excellence
PriMa-Causa introduces a paradigm shift in manufacturing maintenance, moving beyond prediction to precise, causality-driven action. Our research demonstrates significant improvements in operational efficiency and reliability.
0%
Max. Expected Net OEE Gain
0%
Reduced Costly Downtimes
0%
Improved Operational Decisions
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The Prescriptive Maintenance Imperative
Traditional maintenance relies heavily on predictive models that identify statistical associations rather than true causal drivers of failure. This fundamental limitation leads to costly misdiagnoses and ineffective interventions, preventing industries from moving beyond simply predicting *if* a failure will occur, to understanding *why* and *what to do* about it. Bridging this gap is crucial for robust decision-making and optimal OEE.
Our research highlights the need for a systematic approach that enables active prescription by simulating and evaluating potential fixes. This shift from reactive or predictive strategies to truly prescriptive ones promises significant operational benefits.
PriMa-Causa: A Causal Foundation Model
PriMa-Causa proposes a novel model based on causal machine learning to address the limitations of purely predictive models. It leverages a pre-trained causal foundation model as a "what-if" simulator, capable of estimating the effects of potential interventions on Key Performance Indicators (KPIs) like Overall Equipment Effectiveness (OEE).
The model is trained using a synthetic data generator that encodes manufacturing domain knowledge and physical process dependencies, allowing it to approximate conditional interventional distributions without requiring a full causal graph at inference time. This enables the quantification of causal effects for specific actions, moving beyond mere correlation to true causality.
Evaluation with Semi-Synthetic FMCG Data
To demonstrate its prescriptive value, PriMa-Causa was evaluated in a realistic manufacturing setting using semi-synthetic Fast-Moving Consumer Goods (FMCG) data. This setup preserved real production covariates while simulating intervention mechanisms with known potential outcomes, allowing for controlled and industrially relevant assessment.
The model was tasked with a budget-constrained intervention prioritization task, recommending parameter adjustments to optimize OEE. Results showed PriMa-Causa achieved higher expected net OEE gains compared to non-causal and other causal baseline models (S-Learner, Causal Forest), especially in low-to-mid budget ranges, confirming its ability to translate interventional estimates into actionable prescriptive policies.
Enterprise Process Flow
Pre-training (Synthetic Data Generation)
→
PriMa-Causa (Causal Foundation Model)
→
Inference (Real Production Data)
→
Recommended Actions
30%
Max. Expected Net OEE Gain
PriMa-Causa demonstrates superior performance in optimising Overall Equipment Effectiveness compared to non-causal and baseline causal models.
| Feature | PriMa-Causa | Causal Forest | Random Forest |
|---|---|---|---|
| Intervention-aware ranking |
|
|
|
| Handles confounding bias |
|
|
|
| OEE Gain (Low-Mid Budget) |
|
|
|
| Requires causal graph |
|
|
|
Semi-Synthetic FMCG Data Application
The PriMa-Causa model was evaluated using semi-synthetic fast-moving consumer goods (FMCG) data. This setup preserved realistic production covariates while simulating intervention mechanisms with known potential outcomes, enabling a controlled yet industrially relevant assessment. The model successfully identified influential drivers and recommended actions to optimise KPIs under practical budget constraints, showcasing its applicability in complex manufacturing environments.
Calculate Your Potential ROI
Estimate the tangible benefits of integrating causal AI into your operational workflow. Adjust the parameters to see your projected annual savings and reclaimed hours.
Annual Savings Potential
$0
Annual Hours Reclaimed
0
Your Causal AI Implementation Roadmap
A typical journey to integrate prescriptive maintenance with causal AI, ensuring a smooth transition and measurable impact.
Phase 1: Discovery & Data Assessment
Initial consultation to understand your current maintenance challenges, production processes, and existing data infrastructure. We identify key KPIs and potential causal variables.
Phase 2: Model Training & Customization
Leverage our Causal Foundation Model, adapting it to your specific manufacturing data. This includes configuring the SCM-based data generator to reflect your unique process dependencies.
Phase 3: Integration & Pilot Deployment
Seamless integration of PriMa-Causa into your existing operational systems. Conduct a pilot program on a selected production line to validate prescriptive recommendations and measure initial OEE gains.
Phase 4: Scaling & Continuous Improvement
Expand the solution across your production environment, refining the model with ongoing operational data. Establish feedback loops for continuous learning and maximum ROI.
Ready to Transform Your Maintenance Strategy?
Stop reacting to failures and start prescribing success. Connect with our experts to explore how Causal AI can revolutionize your production efficiency and OEE.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat