O-Ring Gasket Swelling in Hydrocarbon Dielectric Fluid

Identify your fluid by SDS or product datasheet. Confirm 'synthetic hydrocarbon,' 'PAO,' 'GTL,' or 'mineral oil' basestock chemistry. If you cannot find paperwork, contact your fluid supplier (Engineered Fluids, BitCool, Castrol, ElectroCool) and request the material compatibility document. Print or save the compatibility page. Confirm flash point in the 170-250 C band and viscosity 4-10 cSt at 40 C as a cross-check.

Walk the tank perimeter with paper towels. Wipe every flange, gland, fitting, and bolt head. Mark each weep point with painter's tape and Sharpie, photograph for records, and do not touch anything yet. A wet sheen, oily residue, or weep mark is a leak path; documenting first prevents you from disturbing a marginal seal that wasn't actually leaking before you wiped it.

Pull one non-critical gasket sample (a spare from the install kit, or a non-leaking sensor O-ring you can replace easily). Drop it in a small jar of your dielectric at room temperature for 24 hours. Measure cross-section before and after with calipers. More than 5 percent swell confirms wrong elastomer family (EPDM or NBR in synthetic hydrocarbon) and means a full reseal is required, not a single-seal fix.

Source FKM (Viton-A or Viton-B) replacement parts in the right sizes for every elastomer in the wet path. McMaster-Carr ships FKM O-ring kits in 24 hours North America. ParkerStore stocks size-by-size online. For sheet gasket, order 1.6 mm FKM sheet (McMaster part class 9470K class). For cord stock, order 3 mm, 5 mm, and 8 mm diameter FKM. Total Tier-1 outlay typically 80-200 CAD for a home tank's elastomer inventory.

Power down all submerged miners at the breaker. Wait 30 minutes for fluid to fully de-energize and chip-side waste heat to dissipate. Confirm fluid temperature is below 40 C before disturbing seals. Hot hydrocarbon fluid is harder to clean off skin, slower to wipe off flanges, and more prone to flash off into vapor when the seal is broken.

Drain fluid through the drain valve into a clean dedicated dielectric container (HDPE jerry cans, never reused chemical containers, never PVC). Drain to below the lowest leaking seal: about 50 percent tank level for a lid gasket, just below the gland for a sensor seal, fully empty for a manifold or drain-valve seal. Label every drain container with date and source-tank ID; clean fluid is reusable, contaminated fluid is not.

Remove the leaking component. Lid bolts come out in a star pattern, opposite-corner sequence, quarter-turn at a time to release tension evenly. Don't pry; lift straight up. Sensor glands unscrew counterclockwise from the gland nut, leaving the gland body in the tank. Save every fastener. ASIC-tank fasteners are often metric and oddly sized (M5, M6, M8 hex socket cap, mixed) and replacements are slow to source if lost.

Pull the swollen gasket. Note its as-removed cross-section and length vs original spec for the swell-rate log. Wipe flange or gland mating surfaces with isopropyl alcohol on a lint-free wipe until residue stops transferring. A flange with bonded elastomer residue may need a plastic scraper, never metal. Metal scrapers score the substrate and create new leak paths along the score line; a substrate score is a permanent leak path even with a perfect FKM gasket.

Install the FKM replacement gasket. For O-rings: hand-place into the groove, no twist, no stretch. For sheet gasket: cut to flange shape with a sharp utility knife, transfer bolt-hole pattern from the flange itself, never reuse a cut gasket. For cord stock: cut to length, butt-joint or 30-degree bevel-joint the ends, bond with cyanoacrylate, let cure 5 minutes before installing. Do not stretch the gasket beyond 5 percent during install; stretched gasket relaxes inward over hours and creates initial leak.

Re-torque to fluid-supplier or flange spec. Typical: 8-14 N.m for sensor gland nuts, 12-25 N.m for tank-lid M8 bolts. Star pattern, quarter-turn passes, three full passes minimum. Final torque check after 24 hours of room-temperature settling, before refill. A flange torqued in one pass often crushes the gasket non-uniformly and leaks at the under-torqued side; the multi-pass star pattern is non-negotiable on a multi-bolt flange.

Schedule the reseal at the same time as annual fluid maintenance: drain fully, filter the fluid through a 5-micron pleated cartridge on a transfer pump, store filtered fluid in clean HDPE for refill. Replace every elastomer in the wet path on the same drain cycle: lid gasket, all sensor glands, all manifold flanges, drain-valve seal, sight-glass O-rings, fill-port cap, vent valves, pump-fitting elastomers, all quick-disconnects. Order all FKM parts in advance from one supplier to keep size consistency.

Replace cable jackets on submerged sensor wires if jackets have gone soft. Standard PVC and TPE jackets swell in hydrocarbons over months. Switch to PTFE-jacketed or FEP-jacketed sensor cable for permanent immersion service; these fluoropolymer jackets are essentially inert in hydrocarbons. Cable runs 3-8 CAD/m for FEP-jacketed Cat6 or RTD-extension cable. Not cheap, but the right answer for long-term immersion deployments.

Apply compatible thread sealant on threaded sensor glands and pipe fittings. PTFE tape (white plumber's tape) wrapped clockwise three to five turns is correct for tapered threads. Loctite 567 or Loctite 577 thread sealant is rated for petroleum service and acceptable. Never use silicone-based pipe dope or RTV form-a-gasket. These are silicone elastomers and dissolve into the dielectric over weeks, contaminating the fluid and leaving residue on every metal substrate they touched.

Pressure / vacuum test before refill. Cap or plug the drain valve and any open ports. Pull 0.1 bar (about 1.5 psi) vacuum on the empty tank with a hand vacuum pump and watch a manometer for 15 minutes. Pressure rise greater than 0.01 bar in 15 minutes indicates a leak path the gaskets are not covering, usually a missed gland or a substrate defect. Find it before refilling. A vacuum-tested empty tank is the cheapest insurance step in the entire build.

Refill with filtered fluid through a 5-micron filter on the transfer pump. Refill rate under 5 L/min to avoid foaming and air entrainment. Top up to the supplier's recommended level (typically 2.5-5 cm above the highest miner heatsink, never above the manifold or above the sensor gland mid-height). Re-energize miners one at a time after refill, in 5-minute intervals, monitoring each seal for new weep before bringing the next miner online.

Stop DIY when: fluid wicked up sensor wires onto a miner control board and the miner is now intermittent or dead, hashboards came out contaminated with sediment that will not wipe clean, a confirmed-FKM reseal leaks again within 30 days (substrate is the problem, not the elastomer), or you see scorching or corroded traces on a hashboard from a contaminated tank. Fluid-wicked boards are board-level repair territory, not field-resealable.

D-Central bench process for fluid-wicked or fluid-shorted hashboards: strip the affected board, ultrasonic clean in petroleum-compatible solvent, dry under controlled vacuum, reflow any compromised joints, full board-level test under load before return. Diagnosis distinguishes fluid-related failures from unrelated chip or PSU failures the tank was hiding. Consultation on dielectric-compatible firmware tuning available (immersion-cooled Antminer hardware can run higher frequency on DCENT_OS, Braiins OS+, LuxOS, or Vnish; non-Antminer follows native firmware).

Ship safely. Drain miners before shipping. Hashboards in anti-static bags, double-boxed with 5+ cm foam on every side. Note on the slip: dielectric brand and model, time in fluid, observed symptoms, any visible elastomer residue, whether the board ever ran dry. Fluid history matters: fluid-soaked boards have different failure signatures than dry boards, and the cleaning protocol depends on fluid chemistry. Saves diagnostic time and your repair cost.