VARNISH: Inside Adaptive Dosing™: How SmartOil G3™ Can Close the Loop (Part 5)

3 days ago

9 min read

VARNISH: Inside Adaptive Dosing™: How SmartOil G3™ Can Close the Loop

How SmartOil G3™ Can Close the Loop on Hydraulic Varnish

Article by William Gillette (LogiLube, LLC)

Sense continuously. Sample when justified. Dose only when warranted. Verify the response.

A companion article, “Why Varnish Becomes an EBITDA Problem,” lays out the case for treating hydraulic-fluid condition as a production issue rather than a lubrication footnote on ultra-class mobile mining equipment: a fixed 1,000-hour drain interval is a weak proxy for oil condition, and a reservoir drain does not fully reset a machine such as the Liebherr R9800, whose hydraulic tank holds roughly 5,800 litres (1,530 gal) out of a 10,000-litre (2,640 gal) total system.

This article covers the technology built to close that gap: SmartOil G3™ Autonomous Fluid Intelligence™, multi-domain sensing, Exception Sampling™, laboratory confirmation and Adaptive Dosing™. A third companion piece, “Quantifying the EBITDA value,” puts a number on what this mechanism is worth across a large excavator fleet.

Autonomous Fluid Intelligence™: From Periodic Sampling to Continuous Evidence

Figure 1: SmartOil G3 Edge-AI Brain™ Autonomous Fluid Intelligence™.

Traditional oil analysis, machine telemetry, laboratory testing and additive treatment are usually independent activities: a sample is drawn periodically, sent out, and the result comes back days later, well after the condition that triggered it has moved on. SmartOil G3™'s premise is to combine those activities into one coordinated process, correlating real-time fluid measurements with hydraulic temperature, machine load, operating hours, reservoir level, oil additions, fluid losses and prior treatment history, continuously rather than at the next scheduled interval.

That does not mean automating an additive pump. It means being able to answer four questions on an ongoing basis: is treatment actually required, what condition caused the change, how much treatment is appropriate, and did the fluid, and the machine, respond. That is the distinction between simply automating a dosing pump and Autonomous Fluid Intelligence™.

Multi-Domain Sensing: Replacing the Hour Meter with Evidence

No single sensor should be expected to duplicate a laboratory oil analysis. SmartOil G3™'s approach is to let several independent sensor domains build a continuously updated fluid-health fingerprint, which the G3 Edge-AI Brain™ compares against the machine's own historical baseline, temperature-normalizes, and screens for abnormal rates of change.

Table 1: SmartOil G3™ Multi-Domain Sensing - Sensor Fusion.

G3-SEN1™ is particularly useful because it measures viscosity, density, dielectric constant and temperature from a single sensor, allowing several physical properties to be correlated rather than read as one number. G3-SEN2™ adds an independent electrochemical domain. G3-SEN3™ adds multi-frequency impedance and dissolved-water measurement, a third independent domain for anomaly detection. The OPCom II, G3-SEN4™, reports ISO 4406 cleanliness directly rather than inferring particle contamination from another fluid property.

Figure 2: SmartOil G3 Edge-AI Brain™ system architecture

Exception Sampling™: Sample the Event, Not the Calendar

Figure 3: How SmartOil G3™ converts fluid-condition evidence into a laboratory sample, and a laboratory sample into action.

The point of the sensor suite is not to eliminate laboratory oil analysis. It is to change when, and why, a sample gets taken. Instead of “1,000 hours reached, drain the oil,” the model becomes: continuously measure, identify abnormal multi-domain behavior, capture an Exception Sample™, obtain laboratory confirmation, then filter, treat, repair, continue operating or drain according to actual condition as shown in figure 3.

A practical control model sets an early-warning envelope ahead of the OEM action boundary. For example, a viscosity-deviation warning could trigger around 10% from the normalized new-oil baseline, leaving margin before the roughly 15% laboratory action level. Particle count could generate an escalating warning as the fluid trends toward ISO 21/19/16, with a sustained excursion automatically commanding an Exception Sample™. A rising water indication would similarly trigger sampling ahead of the 0.10% OEM boundary. These are commissioning parameters, not substitute OEM condemning limits, and would be validated against the specific hydraulic oil, operating environment and laboratory history.

Most importantly for varnish, an Exception Sample™ should also be triggered when the combined dielectric, viscosity, EIS/TAN Delta and temperature-normalized degradation signature departs materially from the established healthy-oil fingerprint, even while particle count, water and viscosity individually remain inside conventional limits. That sample then goes to the laboratory for a targeted confirmation slate: MPC varnish potential, RULER® antioxidant reserve, FTIR oxidation, TAN, kinematic viscosity at 40°C (104°F) and 100°C (212°F), Karl Fischer water, ISO 4406 particle count, PQ index and ICP elemental analysis including silicon and wear metals.

Traditional scheduled sampling, every 250, 500 or 1,000 operating hours, produces a valuable trend, but it cannot observe every abnormal thermal event, cooler problem or contamination episode that occurs between samples. The condition that matters most may occur hundreds of hours before the next scheduled draw. Exception Sampling™ ties the laboratory sample directly to the event that caused it.

Sensor Fusion, Not a Single Threshold

Figure 4: Why the G3 Edge-AI Brain™ should not trigger from one sensor crossing one number.

The G3 Edge-AI Brain™ should not command an Exception Sample™ or an adaptive dose because one sensor crosses one number. An isolated increase in dielectric constant, on its own, may not justify intervention. But increasing dielectric deviation together with increasing viscosity, worsening TAN Delta/EIS response, elevated hydraulic temperature and rising filter differential pressure is a materially stronger signal that the fluid is moving away from its established healthy baseline.

The system weighs absolute measurements, temperature-normalized measurements, rate of change, persistence, agreement across independent sensor domains, machine load and operating state, prior laboratory results, oil-addition history, prior dosing and prior post-dose response. That combination is what determines whether an observed condition looks like normal variation, contamination, lubricant degradation, an emerging mechanical fault, or developing varnish potential.

That last distinction matters because not every abnormal reading calls for chemistry. A rising MPC result with antioxidant depletion, in an otherwise normally operating machine, supports treatment. Elevated hydraulic temperature with normal varnish potential more likely points to a cooler, fan or mechanical problem. Rising particle count suggests dirt ingress or component wear. Rising water calls for contamination removal, not additive dosing. A sudden viscosity shift may indicate incorrect makeup oil. Dosing on top of any of these can obscure the real problem instead of fixing it, which is why the system's default response to uncertainty is not dose, it is Exception Sample™.

Adaptive Dosing™: the Closed Loop

Figure 5: The SmartOil G3™ Adaptive Dosing™ control loop, from continuous measurement to verified response.

LogiLube designed the G3 DOSE™ module around Fluitec's DECON™ AW, a hydraulic formulation of Fluitec's varnish and deposit control technology that also replenishes antiwear additives. Fluitec describes DECON as a lubricant-compatible varnish and deposit-control technology built on its patented SOLVANCER® chemistry. Rather than relying solely on filtration to remove insoluble contamination, it increases the oil's own ability to keep varnish-forming degradation products in solution, so existing deposits can progressively redissolve while the chemistry also reduces the tendency for new ones to form. Fluitec's recommended in-service treat rate for DECON AW is approximately 3 to 5% by volume.

“Varnish control works best as a maintained condition. DECON stays in the oil and keeps degradation products in solution throughout its service life, so deposits have little opportunity to dictate machine reliability. Most operators manage deposits with routine oil analysis and periodic top-ups, and automation gives fleets with constantly changing oil inventories another way to manage the process.”

— Greg Livingstone, Chief Innovation Officer, Fluitec

A manual DECON treatment (calculate the quantity, add it, check the oil later) becomes harder to manage on a mobile machine because the hydraulic-fluid inventory keeps changing: hose failures, cylinder repairs, leakage, top-ups, filter changes, partial drains and full reservoir drain-and-refill events all shift the concentration of DECON actually circulating in the system. Adding 500 litres (132 gal) of untreated oil dilutes it; draining 5,800 litres (1,530 gal) during a major service removes a portion of the treatment chemistry with the drained oil while leaving another portion behind in the residual, trapped volume. The useful question is therefore not “when was DECON last added?”, it is “what is the estimated DECON concentration in the hydraulic system right now?”

For the R9800's roughly 10,000-litre (2,640 gal) hydraulic system, a treatment concentration in the 3–5% range corresponds to a theoretical treatment inventory of roughly 300–500 litres (79–132 gal), distributed through the full circulating volume. Rather than one large manual addition, the G3 DOSE™ module maintains that target through small, controlled incremental additions. The G3 Edge-AI Brain™ will be running an estimated mass balance: current hydraulic-oil inventory, plus new-oil additions, minus known losses and oil removed during maintenance, against previous DECON additions minus DECON estimated lost with drained or leaked oil, to arrive at an estimated current concentration and the small top-up required to restore the approved target.

“DECON performs at its recommended treat rate, and the discipline is keeping it there. On mining equipment that constantly loses and receives oil, that depends on knowing the fluid inventory and its condition. The better that data, the easier it is to hold concentration where it belongs.”

— Dr. Cristian Soto, Chief Operations Officer, Fluitec

Measure → Interpret → Correlate → Exception Sample™ → Confirm → Calculate → Dose → Circulate → Verify → Repeat only when necessary

Adaptive Dosing™ is therefore a bounded, closed-loop control process. Treatment is not triggered because one sensor crosses one numerical threshold, or because a predetermined number of operating hours has elapsed. The system combines multi-domain evidence, determines whether laboratory confirmation is warranted, calculates the appropriate treatment quantity, records the intervention, and then monitors the post-dose response to verify that the hydraulic fluid is moving back toward its approved performance envelope. This functionality is the subject of a patent-pending application with the United States Patent and Trademark Office covering aspects of condition-responsive fluid treatment and Adaptive Dosing™ control.

“The next evolution in varnish mitigation is not simply detecting varnish or adding chemistry. It is knowing the condition of the oil, maintaining the correct treatment concentration and verifying that the machine and lubricant respond as intended.”

— William Gillette, Chief Executive Officer, LogiLube, LLC

Fluitec has already demonstrated the underlying principle in a published hydraulic-system case study, where approximately 3% DECON added to circulating oil was followed by declining varnish potential, contaminated components beginning to clean, and improved hydraulic performance, with the operator maintaining concentration as new oil was added. Applied to mobile mining, the same philosophy replaces “let degradation accumulate until a 1,000-hour drain dilutes it” with continuous condition monitoring plus a maintained DECON concentration, and a drain only when condition, not the hour meter, justifies it.

Fluitec provides the chemistry. SmartOil G3™ provides the intelligence, the fluid-inventory model, the physical sampling and the dosing control needed to apply that chemistry as part of a condition-based reliability strategy. Neither replaces engineering judgment. Both make it more evidence-based.

Recommended SmartOil G3™ Hardware Configuration for the R9800

Figure 6: The G3 Edge-AI Brain™, Power, HYD sampling and DOSE modules, with a G3-SEN sensor, as installed in the field.

Putting this on an R9800 takes more than a single sensor or an automated sampling valve. The recommended architecture combines multi-domain fluid sensing, autonomous physical sampling, Exception Sampling™, Adaptive Dosing™, machine operating data, power conditioning and edge computing into one integrated system, built around four functions: detect (continuously monitor fluid condition and machine state), confirm (capture a representative sample when conditions warrant laboratory analysis), act (deliver an approved quantity of DECON when justified), and verify (confirm the fluid is returning toward its approved performance envelope).

Table 2: SmartOil G3™ Adaptive Dosing System Architecture.

One Architecture, Four Closed Loops

Fluid Intelligence Loop

G3-SEN1™ + G3-SEN2™ + G3-SEN3™ + G4-SEN™+ particle count + water/EIS + temperature + ΔP → G3 Edge-AI Brain™ → anomaly probability and fluid-health trend.

Physical Verification Loop

Anomaly detected → Exception Sampling™ → G3-HYD™ → physical oil sample → laboratory confirmation.

Adaptive Treatment Loop

Confirmed varnish/degradation condition → Edge-AI mass-balance calculation → G3 DOSE™ → controlled Varnish Removal Additive addition → circulation → sensor and laboratory verification.

Economic Protection Loop

Improved fluid stability → fewer unnecessary oil drains → fewer maintenance interventions → greater R9800 availability → more productive loading hours → more tonnes moved → protected EBITDA.

The companion article “Quantifying the EBITDA Value of SmartOil G3™ Technology” puts numbers to this last loop across a large excavator fleet.

Sense Continuously. Sample When Justified. Dose Only When Warranted.

Sense continuously. Sample when the evidence warrants it. Dose only when justified. Verify the response. That is the practical foundation of SmartOil G3™ Autonomous Fluid Intelligence™: replacing an hour-meter proxy for oil condition with continuous, multi-domain evidence, and replacing a periodic manual additive treatment with a controlled, verifiable fluid-management process.

About LogiLube

LogiLube, LLC is a Denver, Colorado-based technology company focused on advancing real-time fluid condition monitoring and predictive maintenance through its patented SmartOil® platform. The company's SmartOil G3 Autonomous Fluid Intelligence™ layer combines third-party in situ sensors, Edge-AI processing, and Local Language Models (LoLM) to deliver continuous monitoring, anomaly detection, and remaining useful life (RUL) predictions for industrial assets. Serving industries such as data centers, mining, and energy, LogiLube enables operators, OEMs, and asset owners to reduce unplanned downtime, optimize maintenance strategies, and unlock the value of high-fidelity operational data across distributed fleets.

SmartOil G3™ technology is protected by U.S. Patent No. 10,466,152; 11,761,946; 12,681,003; Chilean Patent No. 71.366: Australian patent No. 2021202232; other International Patents, and other U.S. and International Patents Pending.