Context
Industrial hydraulic equipment relies on accurate, real-time pressure readings for flow control, component protection, and fast response to load changes. The pressure that matters most is often located in a part of the circuit where mounting a dedicated sensor is impractical — too costly, mechanically awkward, or unreliable in service. The two conventional workarounds each have known limits: adding a physical sensor raises cost and adds another point of mechanical failure, while first-principles models struggle to hold accuracy across the full operating envelope as oil temperature, viscosity, and load conditions shift.
Approach
We developed an edge AI model that infers the target pressure from sensor signals already available on the equipment. The model is built on a class of lightweight machine-learning techniques (reservoir computing) well suited to time-series signals and edge deployment. It runs on the existing controller of the equipment, without additional hardware, an external runtime, or a cloud connection. Final tuning of the model to the customer's specific equipment and operating envelope is done in collaboration with the customer's engineering team.
Outcome
The model reproduces the measured pressure across the operating envelope, including transitions, with accuracy sufficient to use as the pressure input in the customer's control loop — all while running entirely on the existing edge controller.
This may apply to your equipment
If you build or operate machinery with a pressure sensor you'd rather avoid — for reasons of cost, reliability, or mechanical access — and the existing microcontroller has some spare cycles, a similar approach may be worth exploring. After a short technical conversation we can usually tell whether your case is a good fit. The technical details of the implementation are something we discuss case by case with our partners.