Free-surface flow past arbitrary topography and an inverse approach for wave-free solutions

Transcription

1 IMA Journal of Applied Mathematics (23) 78, doi:.93/imamat/hxt5 Advance Access publication on April 23, 23 Free-surface flow past arbitrary topography and an inverse approach for wave-free solutions Benjamin J. Binder School of Mathematical Sciences, University of Adelaide, Adelaide, Australia Corresponding author: M. G. Blyth School of Mathematical Sciences, University of East Anglia, Norwich, UK and Scott W. McCue School of Mathematical Sciences, Queensland University of Technology, Brisbane, Australia [Received on 8 December 22; revised on 9 March 23; accepted on 2 March 23] An efficient numerical method to compute nonlinear solutions for 2D steady free-surface flow over an arbitrary channel bottom topography is presented. The approach is based on a boundary integral equation technique which is similar to that of Vanden-Broeck s (996). The typical approach for this problem is to prescribe the shape of the channel bottom topography, with the free surface being provided as part of the solution. Here we take an inverse approach and prescribe the shape of the free surface aprioriand solve for the corresponding bottom topography. We show how this inverse approach is particularly useful when studying topographies that give rise to wave-free solutions, allowing us to easily classify basic flow types. Finally, the inverse approach is also adapted to calculate a distribution of pressure on the free surface, given the free surface shape itself. Keywords: free-surface flow; boundary integral equation method; potential flow; flow over bottom topography.. Introduction Two-dimensional free-surface flow past a finite-depth solid channel bottom topography has been studied extensively by many investigators in the past, including Havelock (97), Lamb (945), Forbes & Schwartz (982), King & Bloor (987, 99), Vanden-Broeck (987), Forbes (988), Dias & Vanden- Broeck (989, 22, 24), Binder & Vanden-Broeck (27), Chapman & Vanden-Broeck (26) and Lustriet al. (22). The channel bed topography consists of a series of connected straight-line segments (e.g. a triangular obstacle, Dias & Vanden-Broeck, 989; Binder et al., 25) or has a more arbitrary geometry (e.g. King & Bloor, 99; Dias & Vanden-Broeck, 24). In this paper, the latter case is considered with non-linear solutions to the problem obtained numerically by deriving and solving a boundary integral equation which is similar to the one used by Vanden-Broeck (996) amongst others for flow under a vertical sluice gate. The numerical procedure described herein is highly efficient as the computations involve only calculations on the free-surface streamline, even when the shape of the channel bottom streamline is unknown. c The authors 23. Published by Oxford University Press on behalf of the Institute of Mathematics and its Applications. All rights reserved.

2 686 B. J. BINDER ET AL. A y * =H+η * AIR B G V FLUID y * U H C y * =σ * D x * E Fig.. Sketch of flow over topography in the physical (x, y ) plane. Note: the plot is a non-linear solution where the topography is prescribed with F =.5, F u = 3, a = a 3 =., a 2 = a 4 =., b = b 2 = b 3 = b 4 =.5, p =., p 2 =., p 3 = 4., p 4 = 4., c =.5, d =.5 and q =.. We classify this flow as being Type. The standard approach with these problems is for the free surface to be unknown with the corresponding topography being prescribed (see the majority of references in this paper). We call this the forward problem. On the other hand, the inverse problem (e.g. Vanden-Broeck & Tuck, 985; Wu, 987; Chardard et al., 2) allows for the free surface to be prescribed and then solves for the bottom topography that gives rise to such a surface. One advantage of treating the inverse problem is that it provides a way of constructing solutions (directly) with a free surface (or portions of it) that is wavefree. An interesting physical implication of such an approach is that it provides flows that have zero wave-drag; solutions of this type have received particular attention in the literature (see Forbes, 982; Vanden-Broeck & Tuck, 985, for example). Non-linear solutions to both the forward and inverse problem are presented in Figs 7. We now classify these using basic flow types. For this purpose, we note that all flow types are characterized by two Froude numbers and the uniformity (or otherwise) of flow in the far field as x ±.Itis assumed that far downstream, as x, the flow approaches a uniform stream with constant velocity U and depth H. The downstream Froude number is defined by F = U. (.) (gh) /2 When the flow is uniform far upstream, as x, with constant velocity V and constant depth G, an additional upstream Froude number is defined by F u = V. () (gg) /2 A sketch of the flow configuration is shown in Fig.. The basic flow types are: () Supercritical flow with F = F u > (e.g. Vanden-Broeck (987) and Fig. 2(a)). (2) Supercritical flow with F > and F u >, with F = F u (e.g. Binder et al., 26 and Figs, 2(b) and 4(a)). (3) Subcritical flow with F < and waves as x (e.g. Binder et al., 25 and Fig. 2(c,d)). (4) Subcritical flow with F = F u < (e.g. Forbes, 982 and Figs 2(c), 4(d) and 7(c,d)).

3 FLOW PAST ARBITRARY TOPOGRAPHY 687 (a) (b) (c) (d) Fig. 2. The forward problem: flow with a prescribed topography. Non-linear solutions. (a) and (b) Supercritical flow, F =.5. (a) F u =.5, a =., b =.5, p =. (Type ). (b) F u = 3, c =.5, d =.5, q =. (Type 2). (c) and (d) Subcritical flow, F =.5. (c) a =., p =.. Solid curves, b =. (Type 3). Broken curves, F u =.5, b = (Type 4). (d) c =., q =.. Solid curves, d =.7 (Type 3). Broken curves, F u =.58, d =.32 (Type 5). (5) Subcritical flow with F < and F u <, with F = F u (e.g. Lustri et al., 22 and Figs 2(d) and 7(b)). (6) Flow with F = and waves as x (e.g. Binder et al., 26 and Fig. 3(a)). (7) Flow with F = F u = (e.g. Figs 4(c), 5(a) and 6(a,b)). (8) Flow with F = and F u > (e.g. Binder & Vanden-Broeck, 2 and Fig. 5(b)). (9) Flow with F = and F u < (e.g. Fig. 3(b)). () Generalized hydraulic fall with F > and waves as x (e.g. Dias & Vanden-Broeck, 22 and Fig. 3(c)). () Hydraulic fall with F > and F u < (e.g. Forbes, 988 and Figs 3(d) and 4(b)). It is well known that the radiation condition does not allow solutions with periodic waves in both limits x ±. To interpret the solution Types 3, 6 and, for which there are periodic waves as x, we must change the direction of flow so that fluid flows from right to left (in doing so, we

4 688 B. J. BINDER ET AL. (a) (b) (c) (d) Fig. 3. The forward problem: flow with a prescribed topography. Non-linear solutions. (a) and (b) Flow with F =, c =.2, d =.5, q =.. (a) (Type 6). (b) F u = 7, a =.2, b =.5, p = 4.92 (Type 9). (c) and (d) Generalized hydraulic fall and hydraulic fall with F =.5, a =., b =.5, p =., c =.7, d =.5, q =.. (c) (Type ). (d) F u = 9, a 2 =.4, b 2 =.5, p 2 =.97 (Type ). should interchange the words upstream with downstream ). For all other solutions types, the flow can be interpreted as being left-to-right, or right-to-left, as the mathematical problem for each case is the same. Although this study is primarily concerned with flow over a solid channel bottom topography, the more general idea of the inverse method is extended to model flow past a distribution of pressure (e.g. Akylas, 984; Vanden-Broeck & Tuck, 985; Vanden-Broeck, 22; Binder & Vanden-Broeck, 27) on the free surface. For a given shape of the free surface, the corresponding topography and distribution of pressure (both found inversely) are compared with each other, providing a quantitative way of relating the two different types of disturbances (Figs 6 and 7). The inverse method may potentially prove useful in the design of ship hulls (Vanden-Broeck, 98, 989; McCue & Forbes, 999; McCue & Stump, 2) that do not produce surface waves, reducing wave-drag (Vanden-Broeck & Tuck, 985; Farrow & Tuck, 995; Binder, 2; Ogilat et al., 2; Trinh et al., 2). The details of the formulation of the flow problem over a channel bottom topography are given in Section 2.

5 FLOW PAST ARBITRARY TOPOGRAPHY 689 (a) (c) (b) (d) Fig. 4. The inverse problem: flow with a prescribed free surface. Non-linear (solid curves) and weakly non-linear (broken curves) solutions. (a) Supercritical flow with F =.5, F u = 4 (non-linear), c =., d =. and q =. (Type 2). (b) Hydraulic fall with F =., F u =.9 (non-linear), c =., d =. and q =. (Type ). (c) Flow with F =., F u =., a =, b =.5 and p =. (Type 7). (d) Subcritical flow with F =.5, F u =.5, a =, b =.5 and p =. (Type 4). (a) (b) Fig. 5. The inverse problem: flow with a prescribed free surface. Non-linear (solid curves) and weakly non-linear (broken curves) solutions with F =., a = a 4 =., a 2 = a 3 =., b = b 2 = b 3 = b 4 =.5, p =., p 2 =., p 3 = 4., p 4 = 4.. (a) F u =. (Type 7). (b) F u =. (non-linear), c =.5, d =. and q =. (Type 8). 2. Formulation Consider the steady 2D irrotational flow of an incompressible inviscid fluid layer as shown in Fig.. The flow domain is bounded above by the free-surface AB and below by the bottom topography of

6 69 B. J. BINDER ET AL. (a) (b) Fig. 6. The inverse problem: flow with a prescribed free surface, F =. Non-linear solutions (solid curves) for a distribution of pressure on the free surface. The broken curves are for the distributions of pressure, and the support (on the free surface) is shown by the thick black curves (for P > 6 ). (a) The same values of the parameters as in Fig. 4(c) (Type 7). (b) The same values of the parameters as in Fig. 5(a) (Type 7). (a) (c) (b) (d) Fig. 7. The inverse problem: flow with a prescribed free surface, F =.5. Non-linear solutions. (a) and (c) Flow with a distribution of pressure on the free surface, the thick black curves indicate the support (for P > 6 ). (b) and (d) Flow past non-trivial channel bed topographies (solid curves). The broken curves are for the distributions of pressure in (a) and (c). (a) and (b) c =., d =. and q =. (Type 5). (c) and (d) c =., d =.5, q = 4., c 2 =., d 2 =. and q 2 = 4. (Type 4).

7 FLOW PAST ARBITRARY TOPOGRAPHY 69 the channel CDE. The origin of the Cartesian coordinate system (x, y ) is at D. The equations for the free-surface and topography are y = H + η (x ) and y = σ (x ), respectively. For uniform flow far downstream we have u U, v, η, σ as x, (2.) where u and v are the horizontal and vertical components of the velocity. Dimensionless quantities (x, y, η, σ, u, v) are defined by taking H as the reference length and U as the reference velocity. The dimensionless form of the dynamic boundary condition on the free surface AB is then given by 2 (u2 + v 2 ) + F 2 y = 2 + F 2 on y = + η. (2.2) This, together with the dimensionless equivalent of (2.) and the condition of no normal flow on the bottom, y = σ(x), completes the boundary conditions. Following Vanden-Broeck (996) and others a boundary integral equation is derived to solve the potential flow problem, which is formulated in terms of the complex potential function, f = φ + iψ, where φ is the velocity potential and ψ is the stream function, both defined in the usual way, and also in terms of the complex velocity, w = df /dz = u iv, where z = x + iy. Without loss of generality, we choose ψ = on the streamline AB, in which case it follows that ψ = on the channel bottom streamline CDE. The flow problem in the complex f -plane is mapped onto the lower half of the complex ζ -plane by the transformation ζ = α + iβ = e πf. (2.3) Note that the free streamline AB has been mapped to the positive α-axis, while the bottom streamline CDE is mapped to the negative α-axis. To proceed, we introduce the analytic function τ iθ defined by w = u iv = e τ iθ. (2.4) Next we apply Cauchy s integral formula to the function τ iθ at a point on the real axis in the ζ -plane using a contour consisting of the α-axis and a semicircle of infinitely large radius in the lower half plane. Taking the real part we obtain τ(α,) = π θ(α,) α α dα, <α<. (2.5) Note that the integral over the large semicircular part of the contour has made a vanishing contribution. We use a dash to indicate that the remaining integral in (2.5) is of Cauchy principal value type. The values of τ and θ on the free surface (α>) and channel bottom (α<) are denoted by τ f (α) and θ f (α) (free surface) and τ b (α) and θ b (α) (channel bottom). Applying (2.5) at a point on the free surface, we obtain the relationship τ f (α) = π θ b (α ) α α dα + π Equation (2.6) is re-written in terms of φ by applying the change of variables θ f (α ) α α dα, α>. (2.6) α = e πφ, α = e πφ (2.7)

8 692 B. J. BINDER ET AL. for the first integral, and α = e πφ, α = e πφ (2.8) for the second integral. Note that the first equation in each of (2.7) and (2.8) is required by (2.3), while the second is a substitution made for convenience. We find τ f (φ) = θ f (φ ) e π(φ φ ) dφ θ b (φ ) + e π(φ φ ) dφ, <φ<. (2.9) Using equation (2.4), the dynamic boundary condition (2.2) becomes Integrating the identity 2 e2τ f (φ) + F 2 y(φ) = 2 + F 2. (2.) dz df = u iv (2.) along AB from right to left provides the following parametric representation for the shape of the free surface, φ x(φ) = x( ) + e τ f (φ ) cos θ f (φ ) dφ, (2.2) φ y(φ) = + e τ f (φ ) sin θ f (φ ) dφ. (2.3) In computational practice we truncate the x-domain, so that x( ) marks the right-hand side end of the computational domain. For the forward problem the values of θ b (φ) are prescribed on the channel bottom and equations (2.9) (2.3) define an integro-differential system to be solved for the unknown function θ f (φ) on the free surface. For the inverse problem the values of θ f (φ) are prescribed instead, and the unknown function is θ b (φ) on the channel bottom. Both the forward and inverse problems are solved by using a numerical procedure described in Binder et al. (25) and others. After a converged numerical solution is obtained, the shape of the channel bottom topography is found by integrating (2.), using an expression similar to (2.9) for the values of τ b (φ) on the channel bottom. Note that even for the forward problem, with the present method the exact shape of the channel bottom topography in physical space is not strictly known in advance, but comes as part of the solution. The relationship between the topography prescribed by setting θ b as a function of φ and that obtained in physical space at the end of the calculation is discussed in detail in Section 3. We emphasize the improved efficiency of our method over those proposed in previous studies. Our method requires only the solution of a single integral equation, rather than the solution of a pair of integral equations (e.g. Dias & Vanden-Broeck, 22, 24 and others). 3. The forward problem Within the framework of this boundary integral formulation, the most obvious way to formulate the forward problem is to prescribe the channel bottom streamline θ b (φ) and to solve for the free surface

9 FLOW PAST ARBITRARY TOPOGRAPHY 693 θ f (φ). For definiteness we choose to study a linear combination of suitable functions given by [ ] d σ(φ) θ b (φ) = arctan, (3.) dφ where N M σ(φ)= a i exp [ b i (φ p i ) 2 ] + (c i tanh [d i (φ q i )] c i ), (3.2) i= where the a i, b i, c i, d i and p i and q i are chosen topographic parameters. In general, (3.2) is a combination of a Gaussian bump and a smoothed-out upwards or downwards step. Moreover, the functions have the desired properties θ b (φ), σ(φ) as φ. (3.3) Once the solution for θ f (φ) is obtained, the precise shape of the channel bottom topography y = σ(x) in the (x, y) plane can be found by integrating (2.) along the bottom streamline CDE. For flows that are small perturbations to the uniform stream, we find that φ x and σ(φ) σ(x). Non-linear solutions to the forward problem are shown in Figs 3, for non-zero values of the topographic parameters a i, b i, c i, d i, p i and q i in (3.2). The solutions are classified using the basic flow types listed in Section. We focus our attention on solutions with trains of periodic waves on the free surface and how to eliminate them. In general, for arbitrarily chosen values of the topographic parameters in (3.2), solutions with a subcritical stream in the far-field as x (see solid curves in Figs 2(c,d) and 3(a,c)) possess a train of periodic waves on the free-surface upstream. By carefully adjusting the topographic parameters in (3.2), wave-free profiles may be obtained (see the broken curves in Fig. 2(c,d)) or profiles with wave-free portions may be constructed (see Fig. 3(b,d)). This approach to eliminating the periodic waves has been successfully used in the past by Vanden-Broeck (22), Dias & Vanden-Broeck (24), Binder et al. (25), Binder et al. (28), Binder & Vanden-Broeck (27, 2) and others. 4. The inverse problem We handle the inverse problem by prescribing the free-surface shape θ f (φ) and solving for θ b (φ).inthis case, we choose members of the family [ ] d η(φ) θ f (φ) = arctan, (4.) dφ with similar to (3.2). We have i= i= N M η(φ) = a i exp [ b i (φ p i ) 2 ] + c i tanh [d i (φ q i )] c i, (4.2) i= θ f (φ), η(φ) as φ, (4.3) so that we restrict our attention to free-surface profiles that are wave-free in the far-field. In general, the free-surface profile prescribed via (4.2) and the physical free-surface profile are different, η(φ) = η(x), but for a small perturbation of a free stream they closely resemble one another, η(φ) η(x).

10 694 B. J. BINDER ET AL. Non-linear solutions to the inverse problem are shown in Figs 4 and 5 (solid curves) and are classified using the basic flow types listed in Section. Also shown in the figures are weakly non-linear solutions (broken curves) to the inverse problem obtained by solving a steady, forced Korteweg de Vries (KdV) equation (see Wu, 987; Shen, 995; Dias & Vanden-Broeck, 22, 24; Chardard et al., 2; Grimshaw, 2 and others). Following Binder et al. (25, 26) and Binder & Vanden-Broeck (27, 2), when written in terms of the dimensionless variables used in the non-linear computations, the forced KdV equation in the present case takes the form σ(x) = 2(F )η(x) d 2 η(x) 3 3 dx 2 2 η2 (x), (4.4) where the term on the left-hand side is the forcing due to the topography. The forced KdV equation (4.4) is valid for small disturbances and when F is close to (more precisely, if we set ɛ = max{ σ(x) }, then (4.4) holds for ɛ, η = O(ɛ /2 ) and F = O(ɛ /2 )). For the forward problem, σ(x) is an input and (4.4) acts as a differential equation that can be solved numerically for η(x). For the inverse problem, η(x) is the input and (4.4) reduces to a simple, explicit formula for the topography function σ(x). To prescribe the shape of the free surface in the weakly non-linear solutions, we express η using the right-hand side of (4.2) with φ replaced by x. As can be seen in Figs 4(c) and 5(a,b), there is excellent agreement between the fully non-linear solutions and the weakly non-linear solutions to the inverse problem when F =, even though the free surface is prescribed as a function of θ f (φ) in the non-linear computations and as η(x) in the weakly non-linear solutions. Next we consider the inverse problem for a flat channel bottom when there is an imposed distribution of pressure on the free surface. 4. Distribution of pressure Binder & Vanden-Broeck (27), and others, have shown that the forced KdV equation (4.4),with σ(x) on the left-hand side replaced by P(x), describes the situation when there is a distribution of pressure P(x) on the free surface and the channel bottom is flat. Thus, the weakly non-linear solutions shown in Figs 4 and 5 (broken curves) can also represent free-surface flow past a pressure distribution. The bottom broken curves in the figures are now for the distribution of pressure on the free surface (found inversely), instead of the channel bottom topography which is now flat (not shown). By adapting the method detailed in Section 2, non-linear solutions to the inverse problem for flow with a distribution of pressure on the free surface can be obtained. Following Binder & Vanden-Broeck (27), the dynamic boundary condition (2.) takes the form 2 e2τ f (φ) + F 2 (y(φ) + P(φ)) = 2 + F 2. (4.5) As the channel bottom is flat, we have θ b, and (2.9) becomes θ f (φ ) τ f (φ) = e dφ. (4.6) π(φ φ ) For the inverse problem the values of θ f (φ) are prescribed by (4.) (4.3), and then (2.2), (2.3), (4.5) and (4.6), define an integro-differential system for the unknown pressure P(φ), which is solved numerically.

11 FLOW PAST ARBITRARY TOPOGRAPHY 695 Non-linear solutions to the inverse problem, with a distribution of pressure on the free surface, are presented in Fig. 6(a,b), for F =. The free surfaces and flat channel bottom topographies are illustrated by the thinner solid curves and straight lines. The distributions of pressure are shown by the broken curves. The support (where P(φ) > 6 ) of the distributions on the free surface is indicated by the thick curves. The (non-linear) shape of the free surfaces in Fig. 6(a,b) is the same as those in Figs 4(c) and 5(a), respectively. We see that the non-linear pressure distributions (broken curves in Fig. 6(a,b)) are similar to both the non-linear topographies (lower solid curves in Figs 4(c) and 5(a)) and the weakly non-linear solutions (lower broken curves in Figs 4(c) and 5(a)). This is expected, as the weakly nonlinear analysis works extremely well for F =. As the value of the Froude number deviates from unity, so do the differences between the non-linear solutions (both pressure and topography) and the weakly non-linear solutions (e.g. Fig. 4(d)) increase. Furthermore, for the same shape of the free surface, the difference between the non-linear solutions for the distribution of pressure and channel bottom topography become more noticeable. This is illustrated by the lower broken (pressure distribution) and solid (topography) curves in Fig. 7(b,d). To a certain extent, the profiles in Fig. 7(a,c) can be interpreted as wave-free subcritical flows past the hull of a 2D ship. However, we must take some care with such a conclusion for two reasons. First, strictly speaking, the pressure does not vanish on the thin sections of the free surface in these figures, although it is certainly extremely small (P < 6 here). Secondly, treating the thick curves in Fig. 7(a,c) as ship hulls, we see the fluid meets the bow and leaves the stern with roughly zero slope at the height of the free stream. Genuine waveless free surface flows past 2D hulls would not necessarily have this property. Nevertheless, the solutions are an approximation for the much desired wave-free flows past the hulls of ships. 5. Concluding remarks We have considered both a forward and an inverse approach for determining the free-surface flow past an arbitrary channel bottom, and non-linear solutions for eleven basic flow types have been presented and discussed. The novelty of our method lies in the fact that the local tangent at a point on the free surface, or on the bottom, is prescribed as a function of the velocity potential, leading to a formulation requiring the solution of just a single integral equation. In particular, the inverse approach provides a useful way to construct wave-free solutions for flow past both an arbitrary channel bottom topography and a distribution of pressure on the free surface. With the goal of producing waveless solutions past ship hulls in mind, it would be interesting to study the free surface response to a pressure distribution with compact support working via an inverse approach. Such solutions should represent improved approximations to the free-surface flow past less idealized hull shapes than those discussed here. This is left as a topic for our future research. References Akylas, T. R. (984) On the excitation of long nonlinear water waves by a moving pressure distribution. J. Fluid Mech., 4, 455. Binder, B. J. (2) Free-surface flow at the stern of a ship. Phys. Fluids, 576, Binder, B. J., Dias, F. & Vanden-Broeck, J.-M. (25) Forced solitary waves and fronts past submerged obstacles. Chaos, 5, 376. Binder, B. J., Dias, F. & Vanden-Broeck, J.-M. (26) Steady free-surface flow past an uneven channel bottom. Theor. Comput. Fluid Dyn., 2, Binder, B. J. & Vanden-Broeck, J.-M. (27) The effect of disturbances on the free surface flow under a sluice gate. J. Fluid Mech., 576,

12 696 B. J. BINDER ET AL. Binder, B. J. & Vanden-Broeck, J.-M. (2) Hybrid free-surface flows in a two-dimensional channel. Phys. Rev. E, 84, 632. Binder, B. J., Vanden-Broeck, J.-M. & Dias, F. (28) Influence of rapid changes in a channel bottom on freesurface flows. IMA J. Appl. Math., 73, Chapman, J. & Vanden-Broeck, J.-M. (26) Exponential asymptotics and gravity waves. J. Fluid Mech., 567, Chardard, F., Dias, F., Nguyen, H. Y. & Vanden-Broeck, J.-M. (2) Stability of some stationary solutions to the forced KdV equation with one or two bumps. J. Eng. Math., 7, Dias, F. & Vanden-Broeck, J.-M. (989) Open channel flows with submerged obstructions. J. Fluid Mech., 26, Dias, F. & Vanden-Broeck, J.-M. (22) Generalized critical free-surface flows. J. Eng. Math., 42, Dias, F. & Vanden-Broeck, J.-M. (24) Trapped waves between submerged obstacles. J. Fluid Mech., 59, Farrow, D. E. & Tuck, E. O. (995) Further studies of stern wave-making. J. Aust. Math. Soc. B, 36, Forbes, L. K. (982) Non-linear, drag-free flow over a submerged semi-elliptical body. J. Eng. Math., 6, 7 8. Forbes, L. K. (988) Critical free-surface flow over a semi-circular obstruction. J. Eng. Math., 22, 3 3. Forbes, L. K. & Schwartz, L. W. (982) Free-surface flow over a semicircular obstruction. J. Fluid Mech., 4, Grimshaw, R. (2) Transcritical flow past an obstacle. ANZIAM J., 52,. Havelock, T. H. (97) Some cases of wave motion due to a submerged obstacle. Proc. R. Lond. A, 93, 52. King,A.C.&Bloor,M.I.G.(987) Free-surface flow over a step. J. Fluid Mech., 82, King,A.C.&Bloor,M.I.G.(99) Free surface flow of a stream obstructed by an arbitrary bed topography. Quart. J. Mech. Appl. Math., 43, Lamb, H. (945) Hydrodynamics (Chapter 9), 6th edn, Dover: Cambridge University Press, 4 pp. Lustri, C. J., McCue, S. W. & Binder, B. J. (22) European J. Appl. Math., 23, McCue, S. W. & Forbes, L. K. (999) Bow and stern flows with constant vorticity. J. Fluid Mech., 399, McCue, S. W. & Stump, D. M. (2) Linear stern waves in finite depth channels. Q. J. Mech. Appl. Math., 53, Ogilat, O., McCue, S. W., Turner, I. W., Belward, J. W. & Binder, B. J. (2) Minimising wave drag for free surface flow past a two-dimensional stern. Phys. Fluids, 23, 72. Shen, S. S.-P. (995) On the accuracy of the stationary forced Korteweg-de Vries equation as a model equation for flows over a bump. Q. Appl. Math., 53, Trinh, P. H., Chapman, S. J. & Vanden-Broeck, J.-M. (2) Do waveless ships exist? Results for singlecornered hulls. J. Fluid Mech., 685, Vanden-Broeck, J.-M. (98) Nonlinear stern waves. J. Fluid Mech., 96, 63. Vanden-Broeck, J.-M. (987) Free-surface flow over an obstruction in a channel. Phys. Fluids, 3, Vanden-Broeck, J.-M. (989) Bow flows in water of finite depth. Phys. Fluids A,, 328. Vanden-Broeck, J.-M. (996) Numerical calculations of the free-surface flow under a sluice gate. J. Fluid Mech., 33, Vanden-Broeck, J.-M. (22) Waves trapped under a moving pressure distribution. IUTAM Symposium on Diffraction and Scattering in Fluid Mechanics and Elasticity, vol. 4. Dordrecht: Kluwer Academic Publishers, pp Vanden-Broeck, J.-M. & Tuck, E. O. (985) Waveless free-surface pressure distributions. J. Ship. Res., 29, Wu, T. Y. (987) Generation of upstream advancing solitons by moving disturbances. J. Fluid Mech., 84,