Abstract: Hydrodynamic lubrication is concerned with a particular form of creeping flow between surfaces in relative motion where cavitation takes place. The determination of the free boundary of the cavitation area is then of fundamental importance for the computation of the characteristics of the mechanisms. Different conditions at the free boundary have been introduced. We study two of them and compare corresponding solutions with respect to film extent and pressure repartition. Introduction. The resolution of the so-called Reynolds elliptic equation using a variational inequation modeling is well known [1, 2, 3, 4], both from mathematical and numerical aspects. Unfortunately, since the cavitation area is restricted to rest in the divergent portion of the bearing, the solution obtained in this way may be unrealistic and does not always respect the mass flow conservation law in the cavitation area, especially when the supply line is not located at the maximum gap. Numerous models have been introduced in order to explain the various aspects of the cavitation phenomena [5, 6, 7, 8], If tensile strength and inertial effects are neglected, [5] gives a good basis for further developments by relating, in a one-dimensional cavitated convex slider bearing, the mass flow, the supply pressure and various breakdown conditions, the supply position being located at infinity. If the supply position and the supply pressure are given, as usually in a journal bearing, the differential problem studied in [5] becomes a two-point boundary value problem where not only the breakdown position but also the beginning of the oil film are unknown. It is common practice that a regular condition on the gradient of the pressure is taken at the film rupture whereas an eventual discontinuity is allowed for this gradient at film reformation (as in the articles by Elrod [6 p. 37, Floberg [6 p. 31,7 p. 138 and 9,10]). In the present paper, we study the mathematical aspects of this modeling (problem (P)) for an infinitely long journal bearing with zero supply and cavitation pressure whatever the eccentricity e and the supply position (f>. It will be noted that we must recall the variational inequation modeling (problem (PV)), especially the study of the film extent, before giving an existence and uniqueness theorem for the problem (P). It is of interest to notice that the solution of problem (P) is always less than that of problem (PV). The coincidence of both solutions is possible only under precise operating * Received June 12, 1980; revised version received March 13, 1981. 370 G. BAYADA AND M. CHAMBAT conditions; i.e., the input mass flow at the supply line must be the greatest compatible with the external boundary conditions. 1. Physical problem and preliminary results. The lubricating region Q = ]0, 2n[ of the bearing can be divided into two distinct zones. In the first zone i) +, where the fluid film is complete, the usual Reynolds equation (1) applies and the pressure p(x) is positive. In the second zone , which is cavitated, only an unknown fraction 0(x) of the film gap h(x) is assumed to be occupied by the fluid and the pressure is assumed to be zero (see Fig1). The unknown boundary (ct) between Q + and fi0 is the free boundary. Applying the continuity equation to the dimensionless mass flow [6], 9h — h3 (dp/dx), we obtain the following problem (P) with (0, p, a) as unknowns: Problem (P): on Q +: j(h3 = y, 0=1, p > 0; (1) ax \\ ax J ax on Q0: — (Oh) = 0, 0 < 0 < 1, p = 0; (2) dx on (a): h3 ̂ = (1 0)h; (3) dx on the supply line: p(0) = p(2n) = 0. (4) For a journal bearing, h(x) is defined by (5) where e and > are geometrical data: h{x) = 1 — e cos (x — 0); 0 < e < 1, 0 <
/ n. (5) Remarks. Due to the lack of side leakage, the input and output mass flow must be equal, but the supply line may be a discontinuity line for 6 if the film starts at x = 0. Let us note that S, = 0(0 —) = 0(271 — ); then we have ^(0) = 0(0 + )h(0) h3(0) d/(0 + ). (6) dx (3) implies that at the end (ct + ) of a non-cavitated area (i.e. dp/dx < 0) we have <0=1, dp/dx = 0 (7) whereas at the beginning (a —) of H + , 0 and dp/dx have a jump. 0+ TT °* no 2tt Fig. 1. Typical h(x), p(x), (Hx) aspects. FREE BOUNDARY PROBLEM IN PARTIAL LUBRICATION 371 Then the following results hold for the shape of the cavitated area and the existence of a lower bound for the input mass flow. Theorem 1. If ft + is not empty, ft + is a connected set and dh/dx > 0 on a +. Proof. Let [fcl5 a2] be a cavitated area between two non-cavitated zones Q, + and Q2 + ; we have, by integrating (1) in ft1+ and fi2 +: h\\x) ~ = h(x) hf, i = 1, 2 (8) where hf is the film thickness at the point where p(x) is maximum in each Q,+. From (2) and (7) we have: /if = h% = hibj = d(a2^)h(a2), and this is impossible for the given film gap (5). Corollary 1. For all £, with £h(Q) < hmin (hmin = 1 — e), problem (P) has the unique solution p = 0, 6(x) = £,h(0)/h(x). Proof. This is obvious from (2), (8) because existence of a point hmin. 2. The variational inequality modeling (Problem (PV)). We recall here the mathematical formulation [1] of problem (PV) and study the shape of the non-cavitated area S1R+. Let Hl0(Q) be the Sobolev space of square-integrable functions with square-integrable derivatives, which are zero for x = 0 and x = 2k. Let K be the closed convex set defined by: K = {> e //J(fi), > 0 a.e. in Q}. It is well known [1, 2, 3] that there is a unique function pRe K which is a solution of the variational inequality: h3 ̂ 2A dx > n dx dx dh (, > e K. n dx pR is continuously differentiable and satisfies: dh on CiR0 = {x e Cl/pR(x) 0}, pR = 0, — > 0, (9) d f dpR on QR+ = {x e Cl/pR(x) > 0}, — yh —J = dh/dx, dpR on aR (the Reynolds free boundary surface): pR = —— = 0. (10) dx Moreover, we have the following result: Theorem 2. Each connected set of QR+ has at most one free boundary, so 0^+ begins at x = 0 or ends at x = 2n. Proof. Integrating (1) on fis+ = ]a, ft[, where Eq. (10) holds, if pR is maximum for x = x*, we have dx dx dx and then h(a) = h(b) = h(x*), which is impossible from (5). So pR has one of the three patterns shown in Fig. 2. The usual one-hump pressure 372 G. BAYADA AND M. CHAMBAT