Table of Contents
(Johan's speculation + Howard's notes:)
The thermal conductivity of the steel rotor shaft is roughly 200 times that of the PTFE seal lip. This is slightly offset by the lip's smaller thickness compared to the shaft wall, contributing to only a slightly larger lip thermal conductance. As a consequence, this model predicts that the vast majority (roughly 99%) of the friction heat is conducted through the shaft to the coolant inside the shaft and only a tiny fraction of the heat is conducted through the lip to the coolant on the wet side of the lip. As a consequence, presence of coolant at the wet side has little relevance, so both the primary and secondary sealing lip run at roughly the same temperature and wear at roughly the same rate. This means that when the primary lip has worn out, the secondary (backup) lip has probably worn out as well. In that sense, the secondary lip is not a backup for wear.
Advantage 1: After some wear, the chamber between primary and secondary lip can fill with coolant. The primary lip gap is still small, causing a large pressure drop accross that lip, so the pressure in that chamber will be relatively low. This may slow down the overall leakage. [Howard]: Visualize the chambers as a pressure step down ladder. Each chamber with a lower pressure than the previous one closer to the coolant. For both static and dynamic operation, except for the catastrophic case where all LDU chambers have been fully flooded. In prior conversation with seal manufacturer, they claimed they can even make quad lips. Didn't explain the reason at the time but probably yet another step in the pressure step down ladder.
Advantage 2: Although the primary lip can leak, it can still function as an excluder for larger contaminant particles in the coolant, protecting the secondary lip. This analysis corresponds to the LDU's specific design (hollow shaft with coolant), so it is not expected to be published in general sealing literature, unless this shaft design becomes a standard. Note that the addition of each contacting lip add roughly 100 to 200W of lost power at 10,000RPM, assuming a constant 4 MPa lip hoop stress after stress relaxation and a coefficient of friction of 0.25 [1]. So maximization of motor energy efficiency corresponds to minimizing the number of contacting lips. This seems to conflict with the design goal of leak minimization.
Steady state: Piecewise linear temperature profiles. (TODO: add analytical analysys and results).
Transient state: Non-linear temperature profiles. (requires numerical finite volumes modeling).
Note about thermal expansion: The coefficient of thermal expansion (CTE) of PTFE of 120με/K is an order of magniture larger than that of steel of 12με/K. This means that with increasing operating temperature, the lip expands much more than the shaft, reducing the radial load. The seal's metal housing restricts this expansion, but this is neglected because the lip's connection to the housing is somewhat flexible. Assuming a seal assembly temperature of 20°C and a lip operating temperature of 100°C, the lip expands roughly 1% more than the shaft. According to [1], this can reduce radial load by 50-70% reversibly or 10-40% irreversibly (so also after cooling down) in the long run. More data needed.
[1] Müller and Nau (1995), Fluid sealing technology
What is expected leak rate in static and dynamic operation?
In (at least) the year 1995, this was very difficult to answer by some in academia: "At high shaft speeds with dynamic run-out, plain PTFE sleeve-type lip seals leak unpredictably." (From: Müller and Nau, 1995, Fluid Sealing Technology, p95). So one factor may be the accuracy (related to run-out) of the production of the LDU at hand. Was the LDU machined with a properly calibrated CNC machine? Part of Electrified Garage's dismissive response (experienced by Dan K I believe) to "just a few drops of coolant on the speed sensor" could be explained by Tesla's egineers informing the service departments that slow leak rates are normal and expected. But only removing the end shield will really tell if it is possibly fatal.
Expected Thermal Profile on FKM Seal Lip and Coolant film
Multiple sources indicate FKM's heat resistance to water is poor and incompatible with coolant. SKF FKM spec indicates 100C for water temperature tolerance. Pg 9 of following link indicate FKM and FFKM both have compatibility issues with glycol
Pg 34-35 of this SKF doc shows lower temp resistance to water
0901d196807662c1-810-701_CRSeals_Handbook_Jan_2019_tcm_12-318140.pdf (skf.com)
From Müller and Nau (1995), Fluid sealing technology:
p35 Figure 10: The combination of FPM and water-glycol is left blank, so not sure if acceptable.
p33: "[...] notice that some highly fluorinated elastomers, such as FPM, are resistant to very aggressive chemicals over a wide temperature range but may be degraded by apparently innocuous polar fluids such as hot water (or steam), methanol, dilute acids, or alkalis. "
What would this be if can get one made by custom seal makers?
PTFE? Blend?
Triple lip? Quad lip? (seal maker can make) diameter of each lip?
concentric cuts behind 2-3mm shaft contact region to reduce pressure on shaft?
bidirectional pump aid on main lip? on all non excluder lips?
When the shaft is badly damaged, a SKF Speedy Sleeve may be the only affordable option.
Pros
covers surface defects
provides correct roughness
no risk of lead, e.g. caused by inproper grinding or sanding.
Cons
Added thermal conact resistance (use a thermal epoxy to reduce this effect!)
slightly larger diameter: risk of vacuum in case of using a rubber seal with excluder (dust lip) [Müller p90, p91]
sleeve edge may coincide with seal lip
risk of user installation error
future disassembly requires care
Note that adding a sleeve can decrease the shaft thermal conductance in radial direction by an order of magnitude, depending on the shaft damage and thus the tightness of the fit between shaft and sleeve. This can lead to seal lip overheating, especially with large lip contact areas like the Ceimin PTFE seals. This problem can be solved by carefully adding a layer of thermally conductive epoxy (or equivalent) between the shaft and sleeve before mounting the sleeve, just like when mounting a heatsink on a computer CPU.
Numerical example: The thermal contact conductance between two stainless steel surfaces ranges from 1500-4000W/(m²K) [1 2 3], which is a whopping factor 5 (optimistic) to 15 (pesimistic) smaller than the original shaft conductance of 50 [W/(mK)] / 2.25·10?³[m] = 22222W/(m²K). Alternatively, a gap of 10μm filled with Masterbond epoxy adhesive with a thermal conductivity of 2.6 W/(mK) has a conductance of 2.6 / 10·10E-6= 260,000 W/(m²K), which is an dramatic improvement.
❗❗❗ Lots of assumptions alert ❗❗❗ According to an anonymous ex-Tesla engineer that worked on the rotor shaft, the material is "hardened ten-something" steel (and "not stainless steel"). Given the severe corrosion of some shafts, indeed it is probably not stainless steel. A common steel for electric motor rotor shafts is SAE/AISI 1045 (sinotech | what-when-how | eng-tips.com), which matches the stated "ten-something" designation. Let's continue assuming it's 1045. This steel's production process can be either cold- or hot-rolled. It's assumed to be hot-rolled for less residual stresses (CTEMag) resulting in a higher precision as required for the seal. Hardening treatment options are Through-, Flame-, and Induction hardening (FushunSpecialSteel.com). For a 1045 steel, the maximum hardness after hardening is in the range of Rockwell C 54-60 (avg = 57) (DoubleEagleAlloys | azom). So it could be assumed that the hardness is in the upper fifties of Rockwell C. To express hardness on a different scale, use the hardness conversion table @ wikipedia. Note that Rockwell C is the USA standard scale for steel products.
Aegis supplied Tesla with shaft grounding rings, sometimes referred to as "Aegis ring" or "brush", located directly next to the inner rotor bearing at the gearbox side. Its function is to provide a path for "common mode" currents from rotor to housing, preventing those currents to travel through bearings and damaging them through micro-discharges (visualize small lighning strikes). The brush was most likely built into LDUs produced/rebuilt around the years 2011-2013, so not all LDUs have one. After that period, Tesla abandoned them, probably because they are not effective after contamination with excess bearing grease and gearbox oil. You can clean the brush, but only use lacquer thinner and nothing else as prescribed by Aegis. However, the brush likely gets contaminated quickly again. In 2022, Aegis presented a new brush design with thicker hairs that is supposed to be better than the version used by Tesla. However, that version has never been reported in any LDU. Also some claim that, as long as the rotor has hybrid bearings, the brush is not needed and instead driving style and climate play a larger role in causing bearing damage.
A nice summary is found at Prelon.de.
Huge discussion @ openinverter
According to Bauer2008 p17, the seal lip contact area is fully determined by the additives. PTFE is only used for its high temperature resistance, yet wears away quickly.
PTFE material blend/compound/additives info: Parker - Fluid Power Seal Design Guide, pp 3-7 | GFD | Eclipse Seal.