Predictive maintenance is a useful industrial ML use case where AI-readiness often depends less on model selection and more on how the problem is structured.
This issue is for collecting general, non-confidential thoughts on what a lightweight AI-readiness checklist should capture for predictive-maintenance workflows.
Some questions:
- What signals are available?
- Are failure events clearly defined?
- Are labels reliable, sparse, delayed, or noisy?
- Are operating conditions stable or highly variable?
- What decisions can a model actually influence?
- What happens after a prediction is made?
- Is there a feedback loop from maintenance action back to the model or workflow?
- What are the biggest data-cleaning bottlenecks?
The goal is not to build a production predictive-maintenance system here, but to use this as an example of how physical / industrial ML problems can be scoped before model development.
Predictive maintenance is a useful industrial ML use case where AI-readiness often depends less on model selection and more on how the problem is structured.
This issue is for collecting general, non-confidential thoughts on what a lightweight AI-readiness checklist should capture for predictive-maintenance workflows.
Some questions:
The goal is not to build a production predictive-maintenance system here, but to use this as an example of how physical / industrial ML problems can be scoped before model development.