Hydraulic valve monitoring, answered plainly

It is a monitor fitted to an individual hydraulic valve that measures the valve's electrical and mechanical behaviour continuously and forecasts developing failures before they occur. It differs from plant-level condition monitoring, which infers a problem from process symptoms, and from a controller, which drives the valve. CANARY only observes and predicts.

No. The monitor passes supply and the full control signal straight through on a 12-pin input and output. It does not drive the valve, intercept commands or modify any signal, and the valve continues to operate normally even if the monitor loses power.

Hydraulic proportional and servo control valves with on-board electronics, on a 24 VDC supply drawing up to 5 A. Connectors are built to match the valve: 5-pin, 7-pin and 12-pin configurations are available and the pin assignment is specified from your valve data sheet at order. Send us that data sheet and we will confirm fit before you order.

No, and that is the most common misconception about the device. Four of the monitored pins are universal inputs covering 18 control-signal configurations: six voltage ranges from ±2 V to ±10 V including 0 to +10 V, and twelve current ranges from ±5 mA to ±100 mA including 4 to 20 mA. That covers proportional valves driven on a voltage command and servo valves driven directly into a torque motor on a current command. Each channel is configured independently, so the command and the actual-value feedback can use different signal types on the same valve.

Yes. That is the reason for the universal input design. A site running proportional valves on a ±10 V command alongside servo valves on a ±40 mA torque-motor drive uses the same monitor on both. What differs is the connector specified at order and the signal configuration set at commissioning — not a different product.

No. On every universal input the sense element stays permanently in series with the signal path, so no configuration setting and no monitor fault can open the connection between plant and valve. Current-mode measurement does not break the loop, and voltage-mode measurement does not load it. This is a fixed design constraint rather than a mode you select.

The design target is 8 to 10 days before a gradual failure reaches its limit, with a median lead of 9 to 18 days depending on the failure mode. These are simulation figures; field validation is in progress and the numbers will be restated against real data.

Simulation under a realistic plant-disturbance model gives 0.10 false warnings per valve-year. Under a deliberately harsh disturbance model that rises to roughly 1.6 per valve-year. The engine requires corroboration across multiple measurements before escalating, which is what holds the rate down.

External leakage or weeping, which needs a separate oil-presence sensor; a sudden mechanical break in the spring or pilot; a single large particle jamming the spool; a cut cable; and any failure with no measurable precursor. Internal leakage and edge erosion also require pressure transmitters on ports A and B — without them the monitor reports that mode as not monitored rather than showing a false OK.

No. Measurement, baseline learning and forecasting all happen on the device. Valve units report to a central unit over their own wireless mesh, and every unit serves its own web interface locally. There is no subscription and no cloud dependency for the core function.

Through a Modbus register map, an isolated RS-485 connection, and a hard-wired healthy contact for systems that want a dry contact. It is designed to fit alongside what a plant already runs rather than requiring the control layer to be rebuilt.

Not for the monitoring network. Valve units form a self-organising wireless mesh and relay for each other back to the central unit. You do need a 24 VDC supply at the valve, which is normally already there, and optional pressure transmitters if you want internal leakage detection.

The unit learns a baseline from your valve as installed over a commissioning period, and every later judgement is made against that valve's own history. You re-learn after a valve exchange, repair, oil flush or a change of setpoint regime, because the baseline has to describe the valve as it now is.

It is designed and tested against the relevant industrial standards for measurement equipment, including IEC 61010-1, IEC 61326-1 and IEC 60068-2. Formal certification submissions are in progress and no certification mark is claimed at this time. If your site requires a specific certification or a rated Safety Integrity Level before installation, talk to us about timing before ordering.

Yes. We ran a 75-day pilot programme from 1 August to 15 October 2026 on a ten-valve fleet — Bosch Rexroth, Eaton Vickers and Danfoss units in new, used and remanufactured condition, across 1,800 monitored hours each. Eighteen failures occurred. The engine warned on all seventeen that had a physical degradation path, with a minimum of 9.1 days’ lead and a median of 10.6 days. The eighteenth was a sudden stem-coupling fracture with no precursor, which no condition monitor can detect. Full results, including the calls we got wrong, are on the pilot results page.

A threshold alarm tells you a measurement has crossed a line, which means the damage is already done. CANARY runs a proprietary prognostic model built specifically for hydraulic proportional and servo valves. It fuses thirty engineering measurements, reads high-frequency electrical transients and multi-day degradation trend together, learns each valve’s own baseline rather than comparing against a generic specification, and forecasts when a limit will be reached. It also requires corroboration across several measurements before raising a warning, which is what keeps false alarms down. Inference runs on the device, so there is no cloud dependency and no subscription.

It means the engine is tuned to err early rather than late. In the pilot programme, all sixteen failures where a time-to-failure estimate was issued had the predicted date arrive before the actual failure — never after. The median gap was 5.8 hours against warnings issued nine to twenty-two days out, which is about 2.1 per cent of the lead time, and the worst case was 2.8 per cent. This direction matters: a system that predicts failure later than it happens tells a planner they have time they do not have. We would rather the maintenance date you schedule sits slightly before the valve actually needs it.

Closely, which is what we were testing for. The simulation programme projected 8 to 10 days of warning for gradual degradation with a median of 9 to 18 days depending on mode; the pilot produced a 9.1-day minimum and a 10.6-day median across seven modes. Where the two diverge is false-alarm rate — 75 days on ten valves is not long enough to characterise that properly, so we continue to quote the simulation figure of 0.10 per valve-year and label it as simulated.

The monitor is supplied with plant-side and valve-side connectors built to match your valve. Common interfaces include 6+PE and 11+PE per DIN EN 175201-804, the DIN 43563 6P+PE still quoted in much Bosch Rexroth literature, 7-pin and 12-pin MIL-C-5015 style connectors used by Atos and others, and the DIN EN 175301-803 Form A, B and C solenoid connectors formerly known as DIN 43650. The connectors page carries a per-valve reference table.

Only if the pin assignments match, and often they do not. A Bosch Rexroth 4WRPEH and a Moog D633 both use a 6+PE circular connector, but terminal C is the position-signal reference on one and the enable line on the other. The connector is interchangeable; the wiring is not. Tell us which valves are in the fleet and we will tell you which monitors are interchangeable across them.

Coverage reduces, and we say so rather than implying otherwise. Hysteresis, response threshold, null behaviour and silt lock all depend on comparing the command against the valve's reported spool position. On an interface that carries no feedback — a Vickers KADG5V, or a plain solenoid connector — the monitor still measures supply voltage and current, the solenoid-current monitor output where one exists, vibration, acoustic emission and oil temperature. That covers coil degradation, cavitation, and wiring and supply. The connectors page sets out what each interface class supports.

Retail is $3,995 per unit. Pre-order customers pay $2,946 all in, a saving of $1,049 per unit. That comes in two parts: a 20% pre-order discount brings the price to $3,196, and a further $250 per unit is deducted when the balance is invoiced. You pay a $799 deposit per unit up front and $2,147 per unit at invoicing. One monitor covers one valve.

We have opened pre-orders on the first 3,500 units. Of those, 1,800 are already accounted for in the initial production run, leaving 1,700 available. Allocation is strictly first come, first served by deposit date — not by order size or negotiation. When the 1,700 are committed, pre-ordering closes and remaining demand moves to the second production run at the then-current price.

The $799 per-unit deposit is paid up front, and it is what secures your allocation. Your pre-order agreement — setting out quantity, locked pricing, the delivery window, deposit terms and refund conditions — is issued once the deposit is received. If you would rather review the terms before paying, ask and we will send a sample agreement first.

First shipments are scheduled for March 2027. Production is allocated in order of deposit date — first come, first served — and we will keep pre-order customers informed of status against that schedule including if it moves.

The deposit terms, including refund conditions, are set out in the pre-order agreement, which is issued once your deposit is received. If you want to review those terms before paying, ask us and we will send you a sample agreement first — we would rather you were comfortable with the terms than surprised by them.

A typical unplanned hydraulic valve failure costs around $40,000 all-in, against roughly $800 to $1,400 for the same repair carried out in a planned window. At the $2,946 all-in pre-order price, one prevented failure covers about thirteen monitors. Your own numbers will differ, and the useful exercise is running them against your own downtime costs.