How to Find Where a Hydraulic System Is Making Its Heat

Why this matters

You do not need a flow meter, a data logger or a shutdown to find a hydraulic system's heat source. Heat is made wherever pressure falls without an actuator moving, and a component doing that runs hotter on its outlet than on its inlet. That turns the whole question into a survey you can complete in one pass with a temperature instrument and a sheet of paper, on a machine that is running. What follows is that sheet, with what each field is for and what it costs you to leave blank.

This starts after the sibling card on why heat is a symptom has said the generation side moved. If the cooler has failed, none of this applies and you are fixing a cooler.

Before you fill in anything: two conditions and one hazard rule

Condition one: the machine has to be at its settled operating temperature. A machine still warming up is putting heat into metal rather than out of it, and every reading is a moving target. Give it a full duty cycle after the gauge stops climbing.

Condition two: you record one duty state and survey inside it. Holding a clamped load, cycling and idling at standby make heat in completely different places.

The hazard rule for the whole survey. Every reading is taken from outside the motion path, at arm's length, on a machine cycling under its own controls with guards in place. Line, manifold and cylinder surfaces at these temperatures burn skin on contact, so nothing is touched by hand and no reading is taken by feel. Never stand in line with a hose end or a fitting: a hose that lets go at working pressure delivers a stream that injects through skin, which is a surgical emergency covered in its own card. If a read point can only be reached by putting a hand or a head inside the machine's envelope, that point does not get read.

The survey sheet, field by field

Header fields, filled before any component row.

• Ambient air temperature, at the machine. Without it the reservoir number compares to nothing, including this machine last month.
• Reservoir temperature, and the rise above ambient. The sibling card on heat as a symptom gives the gate: a rise-above-ambient ratio within about 1.2 of the commissioning baseline is inside ordinary spread, and beyond that something moved.
• Cooler inlet, cooler outlet and the drop across it. This is the gate that says generation rather than rejection. A working cooler fed hotter oil shows a larger drop; a failed one shows a smaller one.
• Duty state. Holding, cycling or idling. One word, and every row below belongs to it. Surveying across a mixed cycle produces rows that cannot be compared.
• Instrument and surface treatment. See the next section. This is the field most often left blank and the one that invalidates the sheet.

Component rows, one per junction you can reach. Component name, its two read points (inlet and outlet, or the component and the line feeding it where only one port is reachable), inlet temperature, outlet temperature and raw rise. Then two more fields that do the actual work:

• Expected rise for that component. Not zero for everything. Some components are designed to dissipate.
• Net anomaly, which is raw rise minus expected rise. This is the column you read. A sheet without it invites you to chase the largest raw number, which is frequently a component doing its job.

The instrument field, and why it decides whether the sheet means anything

An infrared thermometer reads a surface, and how much infrared a surface emits depends on its finish. Bare polished steel and a chrome-plated fitting emit far less than paint or rubber at the same temperature, so a gun pointed at a shiny fitting reads badly low. That error is systematic and it belongs to the surface, which means it only cancels when both of your readings are on the same finish. Compare a painted manifold against a polished fitting and you have measured the paint.

The fix is cheap: read on painted surfaces, on hose cover, or on a strip of matte tape at each read point. Tape goes on with the machine shut down, the circuit relieved to zero and any accumulator discharged under the plant's energy control procedure, before the warm-up run, because the surfaces are cold then and nothing is moving.

The error figure, with its basis and its character, before it goes anywhere near the arithmetic. An infrared thermometer's specification is usually written as a percent of reading, or a fixed number of degrees, whichever is greater. Read yours, because the basis decides whether the error shrinks as the reading shrinks. The percent-of-reading part behaves as a fixed systematic offset for two readings taken minutes apart on the same finish, so it cancels in the difference and leaves only that percentage of the difference itself, which on a small rise is a fraction of a degree. What survives is the instrument's random spread on each reading, and independent spreads combine in quadrature, so two equal terms multiply by the square root of two. That residual is your threshold: a net anomaly smaller than it is not evidence.

Why a rise across a component is proof and a hot surface is not

A hot component tells you it is hot, not whether it made the heat or received it, and in a circuit where the whole charge circulates, most components are hot because the oil is. The difference across a component is the only reading that separates the two. So a component hot with an equally hot line on either side is a passenger, and one hot with a cooler line on each side of it is dissipating energy internally, which is the strongest single signal on the sheet.

The components that are almost always the answer

• A relief valve passing flow. Its tank line runs noticeably above reservoir temperature when oil is going over it, so a cold relief tank line kills the most-suspected cause in about a minute.
• A pressure-compensated pump on standby. Its case drain always runs above reservoir by design, so this row needs an expected value, not a zero. Same for a throttling or meter-out flow control, whose expected rise comes from the commissioning record.
• A directional valve leaking across its spool to tank. Warm tank port during a hold, when nothing should flow.
• A cylinder passing fluid across its piston. No external flow to read. The tell is the cylinder body running hot with cooler lines feeding it.

Worked example: the completed sheet on a machine 31 degrees F over

Header. Ambient 72 degrees F. Reservoir 148 degrees F. Rise = 148 - 72 = 76 degrees F. Commissioning baseline rise 45 degrees F. Ratio = 76 / 45 = 1.69, above the 1.2 band, so something moved. Cooler inlet 150, outlet 137, drop 13 degrees F against a 12 degree F baseline: the drop grew, so rejection is intact and this is a generation problem. Duty state: holding, which is about 60 percent of this machine's cycle. Instrument: one infrared gun, all read points on matte tape applied cold before the warm-up run.

The surface correction, printed. The first two readings were taken on the polished pump outlet fitting and read 18 degrees F below the taped point six inches away on the same line. Those two were discarded rather than corrected, and both points were retaken on tape. A correction factor was not applied because the emissivity of that plating is not known well enough to derive one.

Read the net anomaly column, not the raw rise column. The largest raw rise on the sheet is 22 degrees F at the meter-out flow control, and it is not the fault: that control is a designed restriction, its commissioning record shows a 20 degree F rise at this duty, and 22 minus 20 leaves a net anomaly of 2 degrees F.

Applying the residual threshold from the instrument section. With a percent-of-reading specification, the systematic part cancels on a same-finish difference and leaves that percentage of the difference, which on a 27 degree difference is well under a degree. The random spread on each of the two readings combines in quadrature, so the residual on any difference here is on the order of one to two degrees F. The 2 degree anomalies at the relief line and the flow control sit inside that residual and are not evidence. The 27 degree anomaly at the cylinder sits far outside it.

What the cold rows rule out. The relief tank line at 2 degrees F over reservoir says oil is not going over the relief during the hold, which kills the cause most techs start with, and the return line at its expected 4 degrees says no significant flow is leaving the cylinder circuit to tank either.

What the hot row means. The cylinder body is 27 degrees F above the line feeding it, with cooler lines on both sides. That is energy being dissipated inside the cylinder, which means fluid is crossing the piston. It is a localization, not yet a complete diagnosis: fluid crossing the piston of a cylinder with a rod on one side also needs somewhere to go, so there is a second path out of that cylinder circuit which this survey has not found. The sibling card on cylinder drift separates the two and says which one closes the complaint.

Confirming the survey without opening anything. The cylinder was taken to the hold position with the load on mechanical blocks rated for it and nobody underneath or reaching in, the controls left in hold, the rod marked and re-checked. It moved. A cylinder dissipating 27 degrees F internally that held perfectly would have contradicted the reading, and that contradiction sends the survey back rather than the cylinder to the bench.

The failure mode of skipping the expected-rise column. A sheet with only raw rises hands you a 22 at the flow control and a 27 at the cylinder, close enough that the wrong one gets picked about half the time. The flow control gets replaced and the machine returns at the same temperature with a healthy part on the invoice.

What would change the answer. Survey the same machine in the cycling duty state instead and the flow control's rise dominates every other row, correctly, because that is when it is doing its work. Nothing about that reading would be wrong and it would still not be the fault. That is why duty state is a header field rather than a note.

How to verify the finding before you spend the repair

• Re-run the two rows that decided it, at the same duty state and on the same tape, and confirm the net anomaly holds. One reading is a measurement; two agreeing readings are a finding.
• Predict what the fix should do to the reservoir rise, and write it down first. If you cannot say roughly how far the rise should fall, you have not established that this source is big enough to explain the whole 31 degrees F.
• Confirm the suspect has a second, independent symptom. Internal dissipation in a cylinder should come with drift, a relief passing flow with the pump loaded when nothing is moving, a leaking spool with a slow actuator. A thermal finding with no mechanical confirmation is one instrument's opinion.
• Re-survey after the repair, same sheet, same points, same duty state. Change the read points and you have lost the comparison rather than proved anything.

References

• Instrument manufacturer documentation for the infrared thermometer's accuracy specification and its basis, its distance-to-spot ratio and its emissivity setting
• Machine builder documentation for commissioning oil temperature rise, cooler capacity and the expected pressure drop across each throttling element
• 29 CFR 1910.147, for control of hazardous energy including stored hydraulic energy, before any read point is prepared or any component is opened in a general industry setting
• See related: Why Heat in a Hydraulic System Is a Symptom Rather Than a Fault; What Cylinder Drift Tells You About Which Seal Gave Up; What a Relief Valve Does in a Hydraulic Circuit and What It Costs