AI predictive maintenance for rotating equipment

Requirement overview

The unit’s critical pumps and compressors already carry vibration and temperature sensors, but the data is only used for threshold alarms. The client wants a model that uses that same existing data to predict a failure at least two weeks in advance.

Background and rationale

Last year four unplanned shutdowns were caused by rotating-equipment failures, together costing more than 90 hours of lost production. Five years of sensor history and maintenance records will be made available to the contractor for model training.

Project objectives

• Predict failures with a horizon of at least 14 days
• Halve the number of unplanned shutdowns
• Identify the fault type alongside the time of occurrence
• Run the model on the client's own infrastructure

Scope of work

• Clean and consolidate the historical sensor data
• Develop and train the failure-prediction model
• Build the monitoring and alerting dashboard
• Transfer knowledge to the maintenance team

Expected deliverables

• Cleaned and documented dataset
• Trained model with an accuracy report
• Web dashboard for equipment health monitoring
• Technical documentation and an in-house training workshop

Applicant requirements

• Track record in machine learning on industrial data
• Familiarity with industrial data-acquisition protocols
• Commitment to keeping data on the client's infrastructure
• A full-time team of at least three people

Evaluation criteria

• Prediction accuracy and horizon40%
• Technical approach and solution architecture25%
• Price and model-ownership terms20%
• Knowledge-transfer plan15%

Call timeline

1. 1

   Call published

   2025-06-23

   The requirement document and the list of target equipment were published.

2. 2

   Sample data release

   2025-07-09

   An anonymised data sample is made available to applicants.

3. 3

   Proposal deadline

   2025-08-01

   Final deadline for uploading the technical and financial proposal.

4. 4

   Project kick-off

   2025-08-16

   The data-mining phase begins with the selected team.