The problem
Mould changeovers without objective criteria
The forces and contact conditions that decide whether a mould is correctly set up are expensive, intrusive and slow to measure directly. Set-up quality today relies on operator experience or on instrumented fixtures that cost thousands per machine, so errors surface only once material is wasted.
The FORGESENSE approach
- Clamp-on accelerometers and ultrasonic microphones, fused with machine-side signals
- Reads the vibro-acoustic transients of clamping and settling, and the shift in the fixture’s modal response with preload
- A physics-informed estimator with learned residuals, not a black box
- Conformal prediction for verdicts with known error bounds
- Runs fully on a microcontroller, with no cloud dependency
- OPC UA and Modbus output, a real-time operator screen and a central audit log
Objectives
What FORGESENSE sets out to do
SO1
Sensing and acquisition
A clamp-on multimodal sensor cluster with synchronised time-stamping and on-edge feature extraction.
SO2
Augmented-sensing model
A hybrid estimator combining a reduced-order physical model of the fixture with learned corrections, mapped to seating-quality verdicts with tolerance limits.
SO3
Edge deployment
A quantised model on a low-cost microcontroller, delivering a verdict within seconds inside the set-up loop.
SO4
Integration and HMI
Industrial interfaces, a first-time-right indicator for the operator and centralised logging of every estimate.
SO5
Validation
Demonstration against measured baselines across representative mould changes on an industrial playground.
Proposed targets
Less waste, less downtime, first time right
≥80%
First-time-right mould adjustments in trial
≥15%
Less downtime per mould change
5–10%
Less energy per mould change
≥3 t/yr
Polyurethane waste avoided
<50 ms
On-device inference
≤2 s
End-to-end verdict
<€200
Hardware cost per machine
Targets are as proposed to the host project and will be reported against as the work progresses.
Work plan
Four stages over fourteen months
- Stage 1PlanMonth 1. Action plan, baseline protocol, data and IP set-up.
- Stage 2DevelopmentMonths 2 to 8. Sensing hardware, estimator, edge deployment and integration.
- Stage 3Test and validationMonths 9 to 10. Trials on the industrial playground and validation report.
- Stage 4Assessment and business planMonths 11 to 14. Solution assessment and route to market.
The host project
AID4SME
AID4SME helps SMEs develop combined AI and data solutions for large-scale industrial resource-optimisation challenges tied to the European Green Deal. It runs from December 2024 to November 2027.
Visit the AID4SME projectReal industrial challenges
Industry partners set the challenges and provide high-TRL playgrounds to integrate and demonstrate solutions, while scientific partners provide technology blocks and a Community of Practice that supports funded SMEs.
Funded by
the European Union
FORGESENSE receives funding from the AID4SME project, which is funded by the European Union’s Horizon Europe research and innovation programme under grant agreement No 101189562.
Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the granting authority. Neither the European Union nor the granting authority can be held responsible for them.