betaFeedback

WorksPhysics and Astronomy

Irreversible Port-Hamiltonian Formulations for 1-Dimensional fluid systems

Abstract

The Irreversible Port-Hamiltonian Systems (IPHS) framework is extended to the modelling of non-isentropic fluids with viscous dissipation in the Eulerian description. Building on earlier IPHS formulations for diffusion-driven and non-convective distributed systems, it is shown that convective transport can be consistently encompassed by the framework by modifying the underlying differential operators. After revisiting the constitutive relations of non-isentropic fluids in both Eulerian and Lagrangian coordinates, it is demonstrate how these systems fit within an extended IPHS formulation. Furthermore, an extended parametrisation of the boundary port variables which ensures that the first and second laws of Thermodynamics are fulfilled allows to define a general class of boundary controlled IPHS.

Introduction

Irreversible port-Hamiltonian Systems (IPHS) formulations were first introduced in [11 b] and later developed as an alternative to pseudo port Hamiltonian formulations [3, 5] or metriplectic-GENERIC [4] formulations, to extend classical port-Hamiltonian theory to the modelling and control of irreversible thermodynamic systems whose dynamics depend strongly on temperature. Such systems include many cutting-edge engineering applications involving active materials, fluid flows, or energy conversion processes (e.g., chemical or nuclear reactions) [8, 2]. In contrast to classical port-Hamiltonian approaches, IPHS use the total energy as the potential function and incorporate the entropy balance directly into the system dynamics. This leads to a nonlinear structure that depends on the co-state variables and no longer satisfies the Jacobi identities. Nevertheless, the modulated skew-symmetric interconnection matrix still reflects energy conservation in accordance with the first law of thermodynamics, while the entropy balance enforces the second law. A nice property is that the modulation function explicitly depends on the topology and the irreversible thermodynamic forces of the irreversible processes involved in the systems. These features make IPHS a very promising framework, particularly for control design [13], which can be reinterpreted in terms of energy shaping and entropy assignment. IPHS formulations have recently been extended to distributed-parameter systems defined over one-dimensional spatial domains [9, 10]. This extension has proven particularly effective for representing diffusion-type systems, such as the heat equation, and for deriving boundary control laws that ensure their stabilization [6]. The same parametrization can be applied directly to non-convective hyperbolic systems, such as deformable mechanical structures, but it is not suitable for convective systems. In the latter case, the assumption of local thermodynamic equilibrium requires tracking the material particles, which naturally leads to formulations in the Lagrangian framework. The aim of this paper is to extend the IPHS formulations to convective systems formulated in the Eulerian description, focusing on the simple yet illustrative case of a non isentropic fluid with viscous damping.

The paper is organized as follows: § 1.1 provides the governing equations of 1D non isentropic compressible fluids in both Eulerian and Lagrangian coordinates. In § 2 are derived the irreversible port Hamiltonian formulations in these two different frames, leading to the general IPHS formulation in § 3. Some conclusions and perspectives are given in Section 5.

Non-isentropic compressible fluids

In this section, we focus on non-isentropic compressible fluids within a one-dimensional spatial domain. The behavior of fluid flows can be described from two different reference frames. In the Eulerian frame, or fixed coordinate system, the focus is on observing the fluid as it moves past a stationary point in space. In contrast, the Lagrangian frame, or moving coordinate system, follows the motion of individual fluid particles. Figure 1 vividly depicts the differences between the two coordinate frames. Now, we derive the governing equations of fluid flow using mass, momentum and energy (entropy) balance in both the reference frames.

Eulerian and Lagrangian reference frames

Figure 1. Eulerian and Lagrangian coordinates.

Governing equations : Eulerian coordinate

We denote by ρ=ρ(ζ,t)\rho=\rho(\zeta,t), v=v(ζ,t)v=v(\zeta,t), p=p(ζ,t)p=p(\zeta,t) the fluid mass density per unit length, velocity and pressure, respectively, with ζ∈[a,b]\zeta\in[a,b]. By applying Reynold’s transport theorem and divergence theorem, the conservation of mass can be represented as the continuity equation2

∂ρ∂t+∂ζ(ρv)=0.(1)\frac{\partial\rho}{\partial t}+\partial_{\zeta}(\rho v)=0. \tag*{(1)}

Next, using Newton’s second law for the material volume neglecting the gravitational forces and then applying Reynold’s transport theorem and divergence theorem, we obtain the Cauchy momentum equation

∂t(ρv)+∂ζ(ρv2)=∂ζσ,(2)\partial_t(\rho v)+\partial_{\zeta}(\rho v^2)=\partial_{\zeta}\sigma, \tag*{(2)}

where σ\sigma represents the stress tensor. Consider σ=−p−τ\sigma=-p-\tau, where pp is the thermodynamic pressure, τ\tau is the viscous stress. Now, considering a Newtonian fluid, we have the viscous stress τ=−μ∂ζv\tau=-\mu\partial_{\zeta}v, where μ\mu is the coefficient of dynamic viscosity. Using the continuity equation (1) we have the following momentum balance equation

ρ∂tv=−ρv∂ζv−∂ζp+∂ζ(μ∂ζv).(3)\rho\partial_t v=-\rho v\partial_{\zeta}v-\partial_{\zeta}p+\partial_{\zeta}(\mu\partial_{\zeta}v). \tag*{(3)}

Now, using Gibbs’ equation, i.e. du=−p dΦ+T dsdu=-p\,d\Phi+T\,ds where Φ=1ρ\Phi=\frac{1}{\rho} is the specific volume, we have

∂ζu=−p∂ζ(1ρ)+T∂ζs.(4)\partial_{\zeta}u=-p\partial_{\zeta}\left(\frac{1}{\rho}\right)+T\partial_{\zeta}s. \tag*{(4)}

By using the specific enthalpy h:=u+pρh:=u+\frac{p}{\rho} we have

1ρ∂ζp=∂ζh−T∂ζs.(5)\frac{1}{\rho}\partial_{\zeta}p=\partial_{\zeta}h-T\partial_{\zeta}s. \tag*{(5)}

From (3), we have

∂tv=−v∂ζv−∂ζh+T∂ζs+1ρ∂ζ(μ∂ζv).(6)\partial_t v=-v\partial_{\zeta}v-\partial_{\zeta}h+T\partial_{\zeta}s+\frac{1}{\rho}\partial_{\zeta}(\mu\partial_{\zeta}v). \tag*{(6)}

Finally, the rate of change of the internal energy per unit mass uu is equal to the sum of the heat conduction due to the heat flux qq and the mechanical work, and is given by

∂tu=−v∂ζu−1ρ∂ζq−pρ∂ζv−τρ∂ζv,(7)\partial_t u=-v\partial_{\zeta}u-\frac{1}{\rho}\partial_{\zeta}q-\frac{p}{\rho}\partial_{\zeta}v-\frac{\tau}{\rho}\partial_{\zeta}v, \tag*{(7)}

The heat flux follows Fourier’s law, given by q:=−k∂ζTq:=-k\partial_{\zeta}T where kk is the thermal conductivity. Using σ=−p+τ\sigma=-p+\tau, we have

∂tu=−v∂ζu+1ρ∂ζ(k∂ζT)−pρ∂ζv+μρ(∂ζv)2.(8)\partial_t u=-v\partial_{\zeta}u+\frac{1}{\rho}\partial_{\zeta}(k\partial_{\zeta}T)-\frac{p}{\rho}\partial_{\zeta}v+\frac{\mu}{\rho}(\partial_{\zeta}v)^2. \tag*{(8)}

Hence, (1), (6), and (8) constitute the governing differential equations of a compressible non-isentropic fluid in Eulerian coordinates flowing in one dimensional spatial domain. Again, using Gibbs’ equation we have3

Dts=1TDtu+pTDt(1ρ).(9)D_t s=\frac{1}{T}D_t u+\frac{p}{T}D_t\left(\frac{1}{\rho}\right). \tag*{(9)}

where Dt:=∂t+v∂ζD_t:=\partial_t+v\partial_{\zeta} is the material derivative. Then, expanding the material derivative, we have the following governing equation in terms of specific entropy ss instead of specific internal energy uu

∂ts=−v∂ζs+1ρT∂ζ(k∂ζT)+μρT(∂ζv)2.(10)\partial_t s=-v\partial_{\zeta}s+\frac{1}{\rho T}\partial_{\zeta}(k\partial_{\zeta}T)+\frac{\mu}{\rho T}(\partial_{\zeta}v)^2. \tag*{(10)}

To summarize, the governing equations of a non isentropic fluid subject to viscous damping and heat conduction in Eulerian coordinates are:

∂tρ=−∂ζ(ρv)(11)\partial_t\rho=-\partial_{\zeta}(\rho v) \tag*{(11)}
∂tv=−v∂ζv−∂ζh+T∂ζs+1ρ∂ζ(μ∂ζv)(12)\partial_t v=-v\partial_{\zeta}v-\partial_{\zeta}h+T\partial_{\zeta}s+\frac{1}{\rho}\partial_{\zeta}(\mu\partial_{\zeta}v) \tag*{(12)}
∂ts=−v∂ζs+1ρT∂ζ(k∂ζT)+μρT(∂ζv)2.(13)\partial_t s=-v\partial_{\zeta}s+\frac{1}{\rho T}\partial_{\zeta}(k\partial_{\zeta}T)+\frac{\mu}{\rho T}(\partial_{\zeta}v)^2. \tag*{(13)}

and the total energy is given by

H(ρ,v,s)=∫ab(12ρv2+ρu(ρ,s))dζ\mathcal{H}(\rho,v,s)=\int_a^b\left(\frac{1}{2}\rho v^2+\rho u(\rho,s)\right)d\zeta

Governing equations: Lagrangian coordinate

Consider now a small fluid element as in Figure 2.

Small fluid element in spatial and material coordinates

Figure 2. Small fluid element in spatial and material coordinates.

Mass conservation principle [7] implies

dm0=dmt(14)dm_0=dm_t \tag*{(14)}

leading to

ρ0(η) dη=ρ(ζ,t) dζ.(15)\rho_0(\eta)\,d\eta=\rho(\zeta,t)\,d\zeta. \tag*{(15)}

Thus,

∂η∂ζ=ρ(ζ,t)ρ0(η).(16)\frac{\partial\eta}{\partial\zeta}=\frac{\rho(\zeta,t)}{\rho_0(\eta)}. \tag*{(16)}

Let’s calculate the mass equation in Lagrangian coordinates. From (1), we have the mass balance equation in Eulerian coordinate as

∂tρ=−∂ζ(ρv)\partial_t \rho=-\partial_\zeta(\rho v)

Using the material derivative we have

Dtρ=∂tρ+v∂ζρ=−∂ζ(ρv)+v∂ζρ=−ρ∂ζv=−ρ(∂ζη)(∂ηv)=−ρ2ρ0∂ηv\begin{aligned} D_t\rho&=\partial_t\rho+v\partial_\zeta\rho\\ &=-\partial_\zeta(\rho v)+v\partial_\zeta\rho\\ &=-\rho\partial_\zeta v\\ &=-\rho(\partial_\zeta\eta)(\partial_\eta v)\\ &=-\frac{\rho^2}{\rho_0}\partial_\eta v \end{aligned}

leading to

Dt(1ρ)=1ρ0∂ηv.(17)D_t\left(\frac{1}{\rho}\right)=\frac{1}{\rho_0}\partial_\eta v. \tag*{(17)}

By performing similar calculations as outlined above and applying the definition of the material derivative, we can derive the velocity and entropy equations in Lagrangian coordinates as follows:

Dtv=−1ρ0∂ηp+μρ02∂η2v(18)D_t v=-\frac{1}{\rho_0}\partial_\eta p+\frac{\mu}{\rho_0^2}\partial_\eta^2v \tag*{(18)}
Dts=1ρ0T(k∂η2T+μ(∂ηv)2).(19)D_t s=\frac{1}{\rho_0 T}\left(k\partial_\eta^2T+\mu(\partial_\eta v)^2\right). \tag*{(19)}

Then, (17), (18), and (19) constitute the governing equations of a compressible non-isentropic fluid in Lagrangian coordinates flowing in one dimensional spatial domain via replacing the material derivative DtD_t with ∂t\partial_t.

IPHS FORMULATIONS OF NON ISENTROPIC COMPRESSIBLE FLUIDS

In this section we derive the IPHS formulation of non isentropic fluids in both Eulerian and Lagrangian coordinates.

IPHS formulation in Eulerian coordinates

In IPHS formulations [9], the transport and dissipative mechanisms are represented in a way that explicitly highlights energy flows and entropy production. This explicit representation is crucial for both physical interpretation and control design. With this goal in mind, we choose as state variables the energy variables x:=(xT s)T\mathbf{x}:=(x^{\mathsf{T}}\ s)^{\mathsf{T}}, where x:=(ρ v)T\mathbf{x}:=(\rho\ v)^{\mathsf{T}}. The total energy (Hamiltonian) of the fluid is defined as

H(x)=∫ab(12ρv2+ρu(ρ,s))dζ(20)\mathcal{H}(\mathbf{x})=\int_a^b\left(\frac{1}{2}\rho v^2+\rho u(\rho,s)\right)\mathrm{d}\zeta \tag*{(20)}
=∫ab(ρv22+ρh−p)dζ.(21)=\int_a^b\left(\frac{\rho v^2}{2}+\rho h-p\right)\mathrm{d}\zeta. \tag*{(21)}

The co-energy (effort) variables are defined as

δρH=12v2+h\delta_\rho\mathcal{H}=\frac{1}{2}v^2+h
δvH=ρv\delta_v\mathcal{H}=\rho v
δsH=δHδu∂u∂s=ρT(since du=−p d(1ρ)+T ds)\delta_s\mathcal{H}=\frac{\delta\mathcal{H}}{\delta u}\frac{\partial u}{\partial s}=\rho T\qquad\text{(since }du=-p\,\mathrm{d}\left(\frac{1}{\rho}\right)+T\,\mathrm{d}s\text{)}

Now, we can rewrite the governing equations (1), (6), and (10) in terms of energy and co-energy variables as follows

∂tρ=−∂ζδvH\partial_t\rho=-\partial_\zeta\delta_v\mathcal{H}
∂tv=−∂ζδρH+1ρ∂ζsδsH+μρ∂ζ(∂ζvδsHρT)\partial_t v=-\partial_\zeta\delta_\rho\mathcal{H}+\frac{1}{\rho}\partial_\zeta s\delta_s\mathcal{H}+\frac{\mu}{\rho}\partial_\zeta\left(\frac{\partial_\zeta v\delta_s\mathcal{H}}{\rho T}\right)
∂ts=−1ρ∂ζsδvH+μρT∂ζv∂ζ(δvHρ)+kρT∂ζ(∂ζ(1ρδsH))\partial_t s=-\frac{1}{\rho}\partial_\zeta s\delta_v\mathcal{H}+\frac{\mu}{\rho T}\partial_\zeta v\partial_\zeta\left(\frac{\delta_v\mathcal{H}}{\rho}\right)+\frac{k}{\rho T}\partial_\zeta\left(\partial_\zeta\left(\frac{1}{\rho}\delta_s\mathcal{H}\right)\right)

or equivalently as

(∂tρ∂tv∂ts)=(0−∂ζ0−∂ζ01ρ∂ζs+A0−1ρ∂ζs−BC)(δρHδvHδsH)(22)\begin{pmatrix} \partial_t\rho\\ \partial_t v\\ \partial_t s \end{pmatrix} = \begin{pmatrix} 0&-\partial_\zeta&0\\ -\partial_\zeta&0&\frac{1}{\rho}\partial_\zeta s+\mathcal{A}\\ 0&-\frac{1}{\rho}\partial_\zeta s-\mathcal{B}&\mathcal{C} \end{pmatrix} \begin{pmatrix} \delta_\rho\mathcal{H}\\ \delta_v\mathcal{H}\\ \delta_s\mathcal{H} \tag*{(22)} \end{pmatrix}

Where the operators A\mathcal{A} and B\mathcal{B} are defined from Hilbert space H1([a,b])H^1([a,b]) to H1([a,b])H^1([a,b]) as follows

A:f↦μρ∂ζ(∂ζvρTf)(23)\mathcal{A}:f\mapsto\frac{\mu}{\rho}\partial_\zeta\left(\frac{\partial_\zeta v}{\rho T}f\right) \tag*{(23)}
B:g↦−μρ∂ζvT∂ζ(gρ)(24)\mathcal{B}:g\mapsto-\frac{\mu}{\rho}\frac{\partial_\zeta v}{T}\partial_\zeta\left(\frac{g}{\rho}\right) \tag*{(24)}

and the operator C\mathcal{C} is given by

CδsH=kρT∂ζ(∂ζ(1ρδsH)).(25)\mathcal{C}\delta_s\mathcal{H}=\frac{k}{\rho T}\partial_\zeta\left(\partial_\zeta\left(\frac{1}{\rho}\delta_s\mathcal{H}\right)\right). \tag*{(25)}

For f∈dom⁡(C)f\in\operatorname{dom}(\mathcal{C}),

C:f↦1ρT∂ζ(k∂ζfρ)=−∂ζ(kρT∂ζfρ)−1ρ∂ζ(1T(k∂ζfρ)).(26)\mathcal{C}:f\mapsto\frac{1}{\rho T}\partial_\zeta\left(k\partial_\zeta\frac{f}{\rho}\right) =-\partial_\zeta\left(\frac{k}{\rho T}\partial_\zeta\frac{f}{\rho}\right)-\frac{1}{\rho}\partial_\zeta\left(\frac{1}{T}\left(k\partial_\zeta\frac{f}{\rho}\right)\right). \tag*{(26)}
1ρ∂ζ(1T(k∂ζfρ)).(27)\frac{1}{\rho}\partial_\zeta\left(\frac{1}{T}\left(k\partial_\zeta\frac{f}{\rho}\right)\right). \tag*{(27)}

Lemma 1. A\mathcal{A} is the formal adjoint of A\mathcal{A}, i.e. B=A∗\mathcal{B}=\mathcal{A}^*.

Proof. We have

⟨Af,g⟩L2([a,b])=∫abμρ∂ζv∂ζ(f∂ζvTρ) dζ=−∫abfμρT∂ζv∂ζ(gρ) dζ+[μρg∂ζvTρf]ab=⟨f,Bg⟩L2([a,b])+[μρg∂ζvTρf]ab.=⟨f,A∗g⟩L2([a,b])+[μρg∂ζvTρf]ab.\begin{aligned} \langle\mathcal{A}f,g\rangle_{L^2([a,b])}=\int_a^b\frac{\mu}{\rho}\partial_\zeta v\partial_\zeta\left(f\frac{\partial_\zeta v}{T\rho}\right)\,\mathrm{d}\zeta\\ &=-\int_a^b f\frac{\mu}{\rho T}\partial_\zeta v\partial_\zeta\left(\frac{g}{\rho}\right)\,\mathrm{d}\zeta+\left[\frac{\mu}{\rho}g\frac{\partial_\zeta v}{T\rho}f\right]_a^b \\ &=\langle f,\mathcal{B}g\rangle_{L^2([a,b])}+\left[\frac{\mu}{\rho}g\frac{\partial_\zeta v}{T\rho}f\right]_a^b. \\ &=\langle f,\mathcal{A}^*g\rangle_{L^2([a,b])}+\left[\frac{\mu}{\rho}g\frac{\partial_\zeta v}{T\rho}f\right]_a^b. \end{aligned}

If boundary conditions are considered equal to 0, then ⟨Af,g⟩L2([a,b])=⟨f,Bg⟩L2([a,b])\langle\mathcal{A}f,g\rangle_{L^2([a,b])}=\langle f,\mathcal{B}g\rangle_{L^2([a,b])}, i.e., B\mathcal{B} is the formal adjoint of A\mathcal{A}, i.e. B=A∗\mathcal{B}=\mathcal{A}^*.

Lemma 2. C\mathcal{C} is a formally skew-adjoint operator.

Proof. The proof is a direct application of Araque [1].

We then have

(∂tρ∂tv∂ts)=(0−∂ζ0−∂ζ01ρ∂ζs+A0−1ρ∂ζs−A∗C)(δρHδvHδsH)(28)\begin{pmatrix} \partial_t\rho\\ \partial_t v\\ \partial_t s \end{pmatrix} = \begin{pmatrix} 0&-\partial_\zeta&0\\ -\partial_\zeta&0&\frac{1}{\rho}\partial_\zeta s+\mathcal{A}\\ 0&-\frac{1}{\rho}\partial_\zeta s-\mathcal{A}^*&\mathcal{C} \end{pmatrix} \begin{pmatrix} \delta_\rho\mathcal{H}\\ \delta_v\mathcal{H}\\ \delta_s\mathcal{H} \tag*{(28)} \end{pmatrix}

The balance equation on the energy reads after integration by parts:

dHdt=∫abδHδx∂x∂t dζ=[−(12v2+h)ρv+μv∂ζv+k∂ζT]ab(29)\begin{aligned} \frac{\mathrm{d}\mathcal{H}}{\mathrm{d}t}=\int_a^b\frac{\delta\mathcal{H}}{\delta x}\frac{\partial x}{\partial t}\,\mathrm{d}\zeta\\ &=\left[-\left(\frac{1}{2}v^2+h\right)\rho v+\mu v\partial_\zeta v+k\partial_\zeta T\right]_a^b \tag*{(29)} \end{aligned}

encoding the first law of Thermodynamics, i.e. dHdt=0\frac{d\mathcal{H}}{dt}=0 if the boundary conditions are set to zero. The balance equation on the entropy reads:

dSdt=∫abδSδx∂x∂t dξ=∫ab(−∂ξ(ρv)−ρv∂ξs+μT(∂ξv)2+kT∂ξ(∂ξv)) dξ=∫ab(μT(∂ξv)2+kT2(∂ξT)2) dξ+[kT∂ξT−ρvs]ab(30)\frac{dS}{dt}=\int_a^b\frac{\delta S}{\delta x}\frac{\partial x}{\partial t}\,d\xi =\int_a^b\left(-\partial_\xi(\rho v)-\rho v\partial_\xi s+\frac{\mu}{T}(\partial_\xi v)^2+\frac{k}{T}\partial_\xi(\partial_\xi v)\right)\,d\xi =\int_a^b\left(\frac{\mu}{T}(\partial_\xi v)^2+\frac{k}{T^2}(\partial_\xi T)^2\right)\,d\xi+\left[\frac{k}{T}\partial_\xi T-\rho v s\right]_a^b \tag*{(30)}

encoding the second law of Thermodynamics, i.e. dSdt≥0\frac{dS}{dt}\geq0 if the boundary conditions are set to zero.

We will see in Section 4 that the system (28) with appropriate definition of the input/output map fits with the generalisation of the IPHS formulations proposed in [9] to systems with convective flows.

IPHS formulation in Lagrangian coordinates

Consider now the same compressible, non-isentropic fluid in Lagrangian coordinates whose dynamics is governed by the equations (17), (18), and (19). We choose as state variables z:=[zT s]T\mathbf{z}:=[z^{\mathsf{T}}\ s]^{\mathsf{T}}, where z:=[ϕ:=1/ρ v]Tz:=[\phi:=1/\rho\ v]^{\mathsf{T}}. The Hamiltonian is given by:

H(z)=∫ab(12ρv2+ρu) dξ=∫ab(12v2+u)ρ0 dη.\mathcal{H}(\mathbf{z})=\int_a^b\left(\frac{1}{2}\rho v^2+\rho u\right)\,d\xi=\int_a^b\left(\frac{1}{2}v^2+u\right)\rho_0\,d\eta.

Then the co-energy variables are:

δϕH=−ρ0p,δvH=ρ0v,δsH=ρ0T.\delta_\phi\mathcal{H}=-\rho_0p,\qquad\delta_v\mathcal{H}=\rho_0v,\qquad\delta_s\mathcal{H}=\rho_0T.

It can be easily verified that in Lagrangian coordinates, the viscous stress reads τ=−μϕρ0∂ηv\tau=-\frac{\mu}{\phi\rho_0}\partial_\eta v, and the heat flux q=−kϕρ0∂ηTq=-\frac{k}{\phi\rho_0}\partial_\eta T.

We can now rewrite the governing equations in terms of energy and co-energy variables in Lagrangian coordinates as

∂tϕ=1ρ0∂η(1ρ0∂vH),\partial_t\phi=\frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\partial_v\mathcal{H}\right),
∂tv=1ρ0∂η(1ρ0δϕH)+1ρ0∂η(μϕTρ02∂ηvδsH),\partial_tv=\frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\delta_\phi\mathcal{H}\right)+\frac{1}{\rho_0}\partial_\eta\left(\frac{\mu}{\phi T\rho_0^2}\partial_\eta v\delta_s\mathcal{H}\right),
∂ts=μTϕρ02∂ηv∂η(δvHρ0)+1ρ0T∂η(kϕρ0∂η(δsHρ0)).\partial_ts=\frac{\mu}{T\phi\rho_0^2}\partial_\eta v\partial_\eta\left(\frac{\delta_v\mathcal{H}}{\rho_0}\right)+\frac{1}{\rho_0T}\partial_\eta\left(\frac{k}{\phi\rho_0}\partial_\eta\left(\frac{\delta_s\mathcal{H}}{\rho_0}\right)\right).

Which is equivalent to:

(∂tϕ∂tv∂ts)=(01ρ0∂η(1ρ0⋅)01ρ0∂η(1ρ0⋅)0A10B1C1)(δϕHδvHδsH)(31)\begin{pmatrix}\partial_t\phi\\ \partial_tv\\ \partial_ts\end{pmatrix} = \begin{pmatrix} 0 & \frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right) & 0\\ \frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right) & 0 & \mathcal{A}_1\\ 0 & \mathcal{B}_1 & \mathcal{C}_1 \end{pmatrix} \begin{pmatrix}\delta_\phi\mathcal{H}\\ \delta_v\mathcal{H}\\ \delta_s\mathcal{H} \tag*{(31)}\end{pmatrix}

Where the operators A1\mathcal{A}_1 and B1\mathcal{B}_1 are defined from Hilbert space H1([a,b])H^1([a,b]) to H1([a,b])H^1([a,b]) as follows:

A1:f↦1ρ0∂η(μϕTρ02∂ηvf)(32)\mathcal{A}_1:f\mapsto\frac{1}{\rho_0}\partial_\eta\left(\frac{\mu}{\phi T\rho_0^2}\partial_\eta v f\right) \tag*{(32)}
B1:g↦μ∂ηvTϕρ02∂η(1ρ0g)(33)\mathcal{B}_1:g\mapsto\frac{\mu\partial_\eta v}{T\phi\rho_0^2}\partial_\eta\left(\frac{1}{\rho_0}g\right) \tag*{(33)}

and the operator C1\mathcal{C}_1 is given by

C1δsH=1ρ0T∂η(kϕρ0∂η(1ρ0δsH)).(34)\mathcal{C}_1\delta_s\mathcal{H}=\frac{1}{\rho_0T}\partial_\eta\left(\frac{k}{\phi\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\delta_s\mathcal{H}\right)\right). \tag*{(34)}

Lemma 3. The operator B1\mathcal{B}_1 is the formal adjoint of A1\mathcal{A}_1, i.e. B1=A1∗\mathcal{B}_1=\mathcal{A}_1^*. Moreover, C1\mathcal{C}_1 is formally skew-symmetric.

Proof. It is easy to show that μ∂ηvTϕρ02∂η(1ρ0⋅)\frac{\mu\partial_\eta v}{T\phi\rho_0^2}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right) is the formal adjoint of 1ρ0∂η(μTϕρ02∂ηv⋅)\frac{1}{\rho_0}\partial_\eta\left(\frac{\mu}{T\phi\rho_0^2}\partial_\eta v\cdot\right) by using integration by parts considering homogeneous boundary conditions. The skew adjointness of 1ρ0T∂η(kϕρ0∂η(1ρ0⋅))\frac{1}{\rho_0T}\partial_\eta\left(\frac{k}{\phi\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right)\right) can be proven in a similar way. □

Equation (31) is on the form

(∂tϕ∂tv∂ts)=(01ρ0∂η(1ρ0⋅)01ρ0∂η(1ρ0⋅)0A10A1∗C1)(δϕHδvHδsH)(35)\begin{pmatrix}\partial_t\phi\\ \partial_tv\\ \partial_ts\end{pmatrix} = \begin{pmatrix} 0 & \frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right) & 0\\ \frac{1}{\rho_0}\partial_\eta\left(\frac{1}{\rho_0}\cdot\right) & 0 & \mathcal{A}_1\\ 0 & \mathcal{A}_1^* & \mathcal{C}_1 \end{pmatrix} \begin{pmatrix}\delta_\phi\mathcal{H}\\ \delta_v\mathcal{H}\\ \delta_s\mathcal{H} \tag*{(35)}\end{pmatrix}

The balance equation on the energy reads after integration by parts:

dHdt=∫abδHδx∂x∂t dξ=[−ρv+μ∂ξv+f∂ξT]ab(36)\frac{d\mathcal{H}}{dt}=\int_a^b\frac{\delta\mathcal{H}}{\delta x}\frac{\partial x}{\partial t}\,d\xi=\left[-\rho v+\mu\partial_\xi v+f\partial_\xi T\right]_a^b \tag*{(36)}

encoding the first law of Thermodynamics, i.e. dHdt=0\frac{d\mathcal{H}}{dt}=0 if the boundary conditions are set to zero. The balance equation on the entropy reads:

dSdt=∫abδSδx∂x∂t dz=∫abμΦρ0T(∂ηv)2+kΦρ0T2(∂ηT)2+[kT∂ηT]ab(37)\frac{dS}{dt}=\int_a^b\frac{\delta S}{\delta x}\frac{\partial x}{\partial t}\,dz =\int_a^b\frac{\mu}{\Phi\rho_0T}(\partial_\eta v)^2+\frac{k}{\Phi\rho_0T^2}(\partial_\eta T)^2+\left[\frac{k}{T}\partial_\eta T\right]_a^b \tag*{(37)}

encoding the second law of Thermodynamics, i.e. dSdt≥0\frac{dS}{dt}\geq0 if the boundary conditions are set to zero.

In this case the system (35) with appropriate definition of the input/output boundary maps fits with the formulations given in [9] and their generalisation to convective systems proposed in Section 4.

Boundary controlled IPHS

Building on the previous examples, the IPHS definition for boundary-controlled systems in [9] can be extended to include irreversible phenomena and convective flows.

Definition 4.1. A general structure of an infinite-dimensional irreversible port-Hamiltonian system with potential convective flows undergoing mm irreversible processes is defined by

[∂tx∂ts]=[P0G1C(x)+G0R0−G1TC(x)−R0TG0T0][δxHδsH]+[P1∂ζf(⋅)G1∂ζf(R1⋅)−R1T∂ζf(G1T⋅)+rs∂ζf(gs⋅)+gs∂ζf(rs⋅)][δxHδsH](38)\begin{bmatrix}\partial_t x\\ \partial_t s\end{bmatrix} = \begin{bmatrix} P_0 & G_1 C(x)+G_0R_0\\ -G_1^{\mathsf T}C(x)-R_0^{\mathsf T}G_0^{\mathsf T} & 0 \end{bmatrix} \begin{bmatrix}\delta_x H\\ \delta_s H\end{bmatrix} + \begin{bmatrix} P_1\partial_\zeta^f(\cdot) & G_1\partial_\zeta^f(R_1\cdot)\\ -R_1^{\mathsf T}\partial_\zeta^f(G_1^{\mathsf T}\cdot)+r_s\partial_\zeta^f(g_s\cdot)+g_s\partial_\zeta^f(r_s\cdot) \end{bmatrix} \begin{bmatrix}\delta_x H\\ \delta_s H \tag*{(38)}\end{bmatrix}

where ∂ζf(g)=f(x)∂ζ(f(x)g)\partial_\zeta^f(g)=f(x)\partial_\zeta(f(x)g), R0,R1∈Rm×1\mathbf{R}_0,\mathbf{R}_1\in\mathbb{R}^{m\times1}, rs∈Rr_s\in\mathbb{R} are the modulating functions defined as follows

R0,i=γ0,i(x,δxH,δsH){S∣G0(:,i)∣H},R_{0,i}=\gamma_{0,i}(x,\delta_xH,\delta_sH)\{S\lvert G_0(:,i)\rvert\mathcal{H}\},
R1,i=γ1,i(x,δxH,δsH){S∣G1(:,i)∣∂ζfH},R_{1,i}=\gamma_{1,i}(x,\delta_xH,\delta_sH)\{S\lvert G_1(:,i)\rvert\partial_\zeta^f\mathcal{H}\},
rs=γs(x,δxH,δsH){S∣H},r_s=\gamma_s(x,\delta_xH,\delta_sH)\{S\lvert\mathcal{H}\},

γ0,i,γ1,i,γs\gamma_{0,i},\gamma_{1,i},\gamma_s are strictly positive-valued functions. G0,G1∈Rn−1×mG_0,G_1\in\mathbb{R}^{n-1\times m}, C(x)∈Rm×1C(x)\in\mathbb{R}^{m\times1}. The pseudo-Poisson brackets are defined as

{E∣G∣F}=[δxETδsET][0G−G∗0][δxFδsF],\{E\vert G\vert F\} = \begin{bmatrix}\delta_xE^{\mathsf T}&\delta_sE^{\mathsf T}\end{bmatrix} \begin{bmatrix}0&G\\-G^*&0\end{bmatrix} \begin{bmatrix}\delta_xF\\\delta_sF\end{bmatrix},
{E∣gs(x)∣F}=δsE∂ζ(gs(x)δsF)\{E\vert g_s(x)\vert F\}=\delta_sE\partial_\zeta(g_s(x)\delta_sF)

for some smooth function EE and FF, operator GG and its formal adjoint G∗G^* and a real continuous scalar function gs(x)g_s(x).

From this definition one can parametrize sets of boundary input and output in order to define the considered class of boundary controlled IPHS.

Assumption 1. The operator

[δxHδsH]T[P1∂ζf(⋅)−G1TC(x)]δxH\begin{bmatrix}\delta_xH\\\delta_sH\end{bmatrix}^{\mathsf T} \begin{bmatrix}P_1\partial_\zeta^f(\cdot)\\-G_1^{\mathsf T}C(x)\end{bmatrix}\delta_xH

is an exact differential form.

Definition 4.2. A Boundary Controlled IPHS (BC-IPHS) is an infinite dimensional IPHS according to Definition 4.1 augmented with the boundary port variables

v(t)=WB[e(t,b)e(t,a)],y(t)=WC[e(t,b)e(t,a)](39)v(t)=W_B\begin{bmatrix}e(t,b)\\e(t,a)\end{bmatrix},\qquad y(t)=W_C\begin{bmatrix}e(t,b)\\e(t,a)\end{bmatrix} \tag*{(39)}

as linear functions of the modified effort variable

e(t,z)=[f0(x)δHδx(t,ζ)f(x)R(x,δHδx)δsHδs(t,ζ)],(40)e(t,z)= \begin{bmatrix} f_0(x)\dfrac{\delta H}{\delta x}(t,\zeta)\\ f(x)\mathbf{R}\left(x,\dfrac{\delta H}{\delta x}\right)\dfrac{\delta_sH}{\delta s}(t,\zeta) \end{bmatrix}, \tag*{(40)}

with R(x,δHδx)=[1R1(x,δHδx)rs(x,δHδx)]\mathbf{R}\left(x,\dfrac{\delta H}{\delta x}\right)=\left[\begin{array}{cc}1&R_1\left(x,\dfrac{\delta H}{\delta x}\right)&r_s\left(x,\dfrac{\delta H}{\delta x}\right)\end{array}\right] and

WB=[12(Ξ2+Ξ1Pep)Mp12(Ξ2−Ξ1Pep)Mp],W_B= \left[ \dfrac{1}{\sqrt{2}}(\Xi_2+\Xi_1P_{ep})M_p\quad \dfrac{1}{\sqrt{2}}(\Xi_2-\Xi_1P_{ep})M_p \right],
WC=[12(Ξ1+Ξ2Pep)Mp12(Ξ1−Ξ2Pep)Mp],W_C= \left[ \dfrac{1}{\sqrt{2}}(\Xi_1+\Xi_2P_{ep})M_p\quad \dfrac{1}{\sqrt{2}}(\Xi_1-\Xi_2P_{ep})M_p \right],

where Mp=(MTM)−1MTM_p=(M^{\mathsf T}M)^{-1}M^{\mathsf T}, Pep=MTPeMP_{ep}=M^{\mathsf T}P_eM and M∈R(n+m+2)×kM\in\mathbb{R}^{(n+m+2)\times k} is spanning the columns of Pe∈Rn+m+2P_e\in\mathbb{R}^{n+m+2} of rank kk, defined by 4

Pe=[P10G10000gsG1T0000gs00](41)P_e= \begin{bmatrix} P_1&0&G_1&0\\ 0&0&0&g_s\\ G_1^{\mathsf T}&0&0&0\\ 0&g_s&0&0 \end{bmatrix} \tag*{(41)}

and where Ξ1\Xi_1 and Ξ2\Xi_2 in Rk×k\mathbb{R}^{k\times k} satisfy Ξ2TΞ1+Ξ1TΞ2=0\Xi_2^{\mathsf T}\Xi_1+\Xi_1^{\mathsf T}\Xi_2=0 and Ξ2TΞ2+Ξ1TΞ1=I\Xi_2^{\mathsf T}\Xi_2+\Xi_1^{\mathsf T}\Xi_1=I.

A direct consequence of the structure given in Definition 4.1 and 4.2 is the completion with the first and second law of Thermodynamics as stated in the following lemmas.

First law of Thermodynamics

Lemma 4 (First law of Thermodynamics). The total energy balance is

dH(t)dt=y(t)Tv(t)(42)\frac{dH(t)}{dt}=y(t)^{\mathsf T}v(t) \tag*{(42)}

which leads, when the input is set to zero, to ∂H=0\partial H=0 in accordance with the first law of Thermodynamics.

Second law of Thermodynamics

Lemma 5 (Second law of Thermodynamics). The total entropy balance is given by

dS(t)dt=∫abσt dξ+ys(t)Tvs(t)(43)\frac{dS(t)}{dt}=\int_a^b\sigma_t\,d\xi+y_s(t)^{\mathsf T}v_s(t) \tag*{(43)}

where ysy_s and vsv_s are the entropy conjugated input/output and σt\sigma_t is the total internal entropy production. This leads, when the input is set to zero, to dSdt=∫abσt dξ≥0\dfrac{dS}{dt}=\int_a^b\sigma_t\,d\xi\geq0 in accordance with the second law of Thermodynamics.

Proof. The proof follows the same development as in [9] with the additional use of Assumption 1 which accounts for the coupling between the material and thermal domain due to the convection.

The formulation of non-isentropic, non-reactive fluid defined in Eulerian and Lagrangian coordinates is given in the following propositions.

Proposition 4.3. A compressible, non-isentropic fluid in Eulerian coordinates can be described as an irreversible port-Hamiltonian system (38) considering the following boundary inputs and outputs

u=[−(12v2+h)(b)−1ρτ(b)−fs(b)(12v2+h)(a)+1ρτ(a)fs(a)],y=[ρv(b)T(b)ρv(a)T(a)](44)\mathbf{u}= \begin{bmatrix} -\left(\frac{1}{2}v^2+h\right)(b)-\frac{1}{\rho}\tau(b)\\ -f_s(b)\\ \left(\frac{1}{2}v^2+h\right)(a)+\frac{1}{\rho}\tau(a)\\ f_s(a) \end{bmatrix}, \qquad \mathbf{y}= \begin{bmatrix} \rho v(b)\\ T(b)\\ \rho v(a)\\ T(a) \end{bmatrix} \tag*{(44)}

respectively, then the total energy H\mathcal{H} satisfies

H˙=yTu,\dot{\mathcal{H}}=\mathbf{y}^{\mathsf T}\mathbf{u},

where fs:=kρ∂xTf_s:=\dfrac{k}{\rho}\partial_xT is the entropy flux, and the irreversible entropy production rate is

S˙=∫0Lσs dξ+ysTus,\dot{\mathcal{S}}=\int_0^L\sigma_s\,d\xi+\mathbf{y}_s^{\mathsf T}\mathbf{u}_s,

where S\mathcal{S} is the total entropy of the fluid, us\mathbf{u}_s and ys\mathbf{y}_s are the boundary entropy input and output respectively. Moreover, σs≥0\sigma_s\geq0 represents the total internal entropy production.

Proof. For a non-isentropic, non-reactive fluid defined in Eulerian coordinates, we have x=[ρ v]T\mathbf{x}=[\rho\ v]^{\mathsf T}, S=∫0bρs dξ\mathcal{S}=\int_0^b\rho s\,d\xi, δxS=[s 0 ρ]T\delta_x\mathcal{S}=[s\ 0\ \rho]^{\mathsf T}. The matrices in (38) are P0=02×2P_0=0_{2\times2}, P1=[0−1−10]P_1=\begin{bmatrix}0&-1\\-1&0\end{bmatrix}, G0=02×1G_0=0_{2\times1}, G1=[0 1]TG_1=[0\ 1]^{\mathsf T}, C(x)=∂ζsC(x)=\partial_\zeta s, gs=1g_s=1. We will have:

Pe=[0−1000−10010000000100000000],M=[−1200001000001120000010],P_e= \begin{bmatrix} 0&-1&0&0&0\\ -1&0&0&1&0\\ 0&0&0&0&0\\ 0&1&0&0&0\\ 0&0&0&0&0 \end{bmatrix}, \qquad M= \begin{bmatrix} -\frac{1}{2}&0&0&0\\ 0&1&0&0\\ 0&0&0&1\\ \frac{1}{2}&0&0&0\\ 0&0&1&0 \end{bmatrix},
Mp=[−10010010000000100100],Pep=[0100100000010010].M_p= \begin{bmatrix} -1&0&0&1&0\\ 0&1&0&0&0\\ 0&0&0&0&1\\ 0&0&1&0&0 \end{bmatrix}, \qquad P_{ep}= \begin{bmatrix} 0&1&0&0\\ 1&0&0&0\\ 0&0&0&1\\ 0&0&1&0 \end{bmatrix}.

We can compute R1=μ∂τvT\mathbf{R}_1=\frac{\mu\partial_{\tau}v}{T}, R0=0\mathbf{R}_0=0, rs=k∂τTT2r_s=\frac{k\partial_{\tau}T}{T^2}, and the irreversible entropy production rate is γ1=μT>0\gamma_1=\frac{\mu}{T}>0, γ0=0\gamma_0=0, γs=kT2>0\gamma_s=\frac{k}{T^2}>0. We have f0=1f_0=1 and f=1ρf=\frac{1}{\rho} leading to the extended set of effort variables

e(t)=[12v2+hρvTμ∂ξvkρT∂ξT],Mρe(t)=[−12v2−h+μρ∂ξvρvkρT∂ξTT]\mathbf{e}(t)=\begin{bmatrix}\frac{1}{2}v^2+h\\ \rho v\\ T\\ \mu\partial_{\xi}v\\ \frac{k}{\rho T}\partial_{\xi}T\end{bmatrix},\quad \mathbf{M}_{\rho}\mathbf{e}(t)=\begin{bmatrix}-\frac{1}{2}v^2-h+\frac{\mu}{\rho}\partial_{\xi}v\\ \rho v\\ \frac{k}{\rho T}\partial_{\xi}T\\ T\end{bmatrix}

leading by choosing Ξ1\Xi_1 and Ξ2\Xi_2 satisfying the condition of Definition 4.2 to the energy balance

H˙=∫ab(δρH∂τρ+δvH∂τv+δsH∂τs)dξ=∫ab(−δρH∂ξδvH−δvH∂ξδρH+1ρδvH∂ξδsH−δvHρ∂ξτ−1ρδsH∂ξδvH−τ∂ξ(δvHρ)−T∂ξ(qT)−qT∂ξT)dξ=−((12v2+h)ρv)∣ab−(τv)∣ab−(Tfs)∣ab=yTu,\begin{aligned} \dot{\mathcal{H}}=\int_a^b\left(\delta_{\rho}\mathcal{H}\partial_{\tau}\rho+\delta_v\mathcal{H}\partial_{\tau}v+\delta_s\mathcal{H}\partial_{\tau}s\right)d\xi\\ &=\int_a^b\left(-\delta_{\rho}\mathcal{H}\partial_{\xi}\delta_v\mathcal{H}-\delta_v\mathcal{H}\partial_{\xi}\delta_{\rho}\mathcal{H}+\frac{1}{\rho}\delta_v\mathcal{H}\partial_{\xi}\delta_s\mathcal{H}-\frac{\delta_v\mathcal{H}}{\rho}\partial_{\xi}\tau-\frac{1}{\rho}\delta_s\mathcal{H}\partial_{\xi}\delta_v\mathcal{H}-\tau\partial_{\xi}\left(\frac{\delta_v\mathcal{H}}{\rho}\right)-T\partial_{\xi}\left(\frac{q}{T}\right)-\frac{q}{T}\partial_{\xi}T\right)d\xi\\ &=-\left(\left(\frac{1}{2}v^2+h\right)\rho v\right)\Big|_a^b-(\tau v)\Big|_a^b-(Tf_s)\Big|_a^b \\ &=\mathbf{y}^{\mathsf{T}}\mathbf{u}, \end{aligned}

where the boundary inputs and outputs are as depicted in (44). Next, we define the total entropy of the system as

S=∫abρs dξ,δxS=[s 0 ρ]T.\mathcal{S}=\int_a^b \rho s\,d\xi,\quad\delta_x\mathcal{S}=[s\ 0\ \rho]^{\mathsf{T}}.

Then the entropy balance goes as follows

S˙=∫abδxST∂tx dξ=∫ab(−s∂ξ(ρv)−ρv∂ξs−τT∂ξv−1T∂ξT)dξ=∫ab[−s∂ξ(ρv)−ρv∂ξs−∂ξ(qT)]dξ+∫ab[μT(∂ξδvHρ)2+kT2(∂ξδsHρ)2]dξ=∫abσs dξ−(spv+qT)∣ab,\begin{aligned} \dot{\mathcal{S}}=\int_a^b\delta_x\mathcal{S}^{\mathsf{T}}\partial_t\mathbf{x}\,d\xi\\ &=\int_a^b\left(-s\partial_{\xi}(\rho v)-\rho v\partial_{\xi}s-\frac{\tau}{T}\partial_{\xi}v-\frac{1}{T}\partial_{\xi}T\right)d\xi\\ &=\int_a^b\left[-s\partial_{\xi}(\rho v)-\rho v\partial_{\xi}s-\partial_{\xi}\left(\frac{q}{T}\right)\right]d\xi\\ &+\int_a^b\left[\frac{\mu}{T}\left(\partial_{\xi}\frac{\delta_v\mathcal{H}}{\rho}\right)^2+\frac{k}{T^2}\left(\partial_{\xi}\frac{\delta_s\mathcal{H}}{\rho}\right)^2\right]d\xi\\ &=\int_a^b\sigma_s\,d\xi-\left(spv+\frac{q}{T}\right)\Big|_a^b, \end{aligned}

where σs:=μT(∂ξδvHρ)2+kT2(∂ξδsHρ)2≥0\sigma_s:=\frac{\mu}{T}\left(\partial_{\xi}\frac{\delta_v\mathcal{H}}{\rho}\right)^2+\frac{k}{T^2}\left(\partial_{\xi}\frac{\delta_s\mathcal{H}}{\rho}\right)^2\geq0 is the internal entropy production, ysTus:=(spv+qT)∣ab\mathbf{y}_s^{\mathsf{T}}\mathbf{u}_s:=\left(spv+\frac{q}{T}\right)\Big|_a^b. □\square

Proposition 4.4. A compressible, non-isentropic fluid in Lagrangian coordinates can be described as an irreversible port-Hamiltonian system (38) considering the following boundary inputs and outputs

u=[−p(b)−τ(b)−fs(b)p(a)+τ(a)fs(a)],y=[v(b)T(b)v(a)T(a)](45)\mathbf{u}=\begin{bmatrix}-p(b)-\tau(b)\\-f_s(b)\\p(a)+\tau(a)\\f_s(a)\end{bmatrix},\quad \mathbf{y}=\begin{bmatrix}v(b)\\T(b)\\v(a)\\T(a)\end{bmatrix} \tag*{(45)}

respectively, then the total energy H\mathcal{H} satisfies

H˙=yTu,\dot{\mathcal{H}}=\mathbf{y}^{\mathsf{T}}\mathbf{u},

and the irreversible entropy production rate is

S˙=∫0Lσs dη+ysTus,\dot{\mathcal{S}}=\int_0^L\sigma_s\,d\eta+\mathbf{y}_s^{\mathsf{T}}\mathbf{u}_s,

where us\mathbf{u}_s and ys\mathbf{y}_s are boundary entropy input and output respectively. Moreover, σs≥0\sigma_s\geq0 represents the total internal entropy production.

Proof. For a non-isentropic, non-reactive fluid defined in Lagrangian coordinates, we have x=[ϕ v s]T\mathbf{x}=[\phi\ v\ s]^{\mathsf{T}}, S=∫abρ0s dη\mathcal{S}=\int_a^b\rho_0s\,d\eta, δxS=[0 0 ρ0]T\delta_x\mathcal{S}=[0\ 0\ \rho_0]^{\mathsf{T}}. The underlying matrices are P0=02×2\mathbf{P}_0=\mathbf{0}_{2\times2}, P1=[0110]\mathbf{P}_1=\begin{bmatrix}0&1\\1&0\end{bmatrix}, G0=02×1\mathbf{G}_0=\mathbf{0}_{2\times1}, G1=[0 1]T\mathbf{G}_1=[0\ 1]^{\mathsf{T}}, C(x)=0\mathbf{C}(\mathbf{x})=0, f0=f=1ρ0f_0=f=\frac{1}{\rho_0}, R1=μ∂ηvρ0T\mathbf{R}_1=\frac{\mu\partial_{\eta}v}{\rho_0T}, R0=0\mathbf{R}_0=0, gs=1ρ0g_s=\frac{1}{\rho_0}, rs=k∂ηTρ0T2r_s=\frac{k\partial_{\eta}T}{\rho_0T^2}, γ1=μρ02T>0\gamma_1=\frac{\mu}{\rho_0^2T}>0, γ0=0\gamma_0=0, γs=kρ02T2>0\gamma_s=\frac{k}{\rho_0^2T^2}>0. Considering the state variables as x=[ϕ v s]T\mathbf{x}=[\phi\ v\ s]^{\mathsf{T}}, the energy balance in Lagrangian coordinates can be expressed as follows

H˙=∫ab(δϕH∂ηρ+δvH∂ηv+δsH∂ηs)dη=−(pv+τv)∣ab−(Tfs)∣ab=yTu,\begin{aligned} \dot{\mathcal{H}}=\int_a^b\left(\delta_{\phi}\mathcal{H}\partial_{\eta}\rho+\delta_v\mathcal{H}\partial_{\eta}v+\delta_s\mathcal{H}\partial_{\eta}s\right)d\eta\\ &=-\left(pv+\tau v\right)\Big|_a^b-\left(Tf_s\right)\Big|_a^b \\ &=\mathbf{y}^{\mathsf{T}}\mathbf{u}, \end{aligned}

where the boundary inputs and outputs are depicted in (45). Consider the total entropy of the system as

S=∫abρ0s dη,δxS=[0 0 ρ0]T.\mathcal{S}=\int_a^b\rho_0s\,d\eta,\quad\delta_x\mathcal{S}=[0\ 0\ \rho_0]^{\mathsf{T}}.

Then the entropy balance goes as follows

S˙=∫abδxST∂tx dη=∫ab(−1T∂ηv−1T∂ηT)dη=∫ab[−∂η(qT)]dη+∫ab[μϕρ0T(∂ηδvHρ)2+kϕρ0T2(∂ηδsHρ)2]dη=∫abσs dη−(qT)∣ab,\begin{aligned} \dot{\mathcal{S}}=\int_a^b\delta_x\mathcal{S}^{\mathsf{T}}\partial_{tx}\,d\eta\\ &=\int_a^b\left(-\frac{1}{T}\partial_{\eta}v-\frac{1}{T}\partial_{\eta}T\right)d\eta\\ &=\int_a^b\left[-\partial_{\eta}\left(\frac{q}{T}\right)\right]d\eta\\ &+\int_a^b\left[\frac{\mu}{\phi\rho_0T}\left(\partial_{\eta}\frac{\delta_v\mathcal{H}}{\rho}\right)^2+\frac{k}{\phi\rho_0T^2}\left(\partial_{\eta}\frac{\delta_s\mathcal{H}}{\rho}\right)^2\right]d\eta\\ &=\int_a^b\sigma_s\,d\eta-\left(\frac{q}{T}\right)\Big|_a^b, \end{aligned}

where σs:=μϕρ0T(∂ηδvHρ)2+kϕρ0T2(∂ηδsHρ)2≥0\sigma_s:=\frac{\mu}{\phi\rho_0T}\left(\partial_{\eta}\frac{\delta_v\mathcal{H}}{\rho}\right)^2+\frac{k}{\phi\rho_0T^2}\left(\partial_{\eta}\frac{\delta_s\mathcal{H}}{\rho}\right)^2\geq0 is the internal entropy production, ysTus:=(qT)∣ab\mathbf{y}_s^{\mathsf{T}}\mathbf{u}_s:=\left(\frac{q}{T}\right)\Big|_a^b. □\square

Conclusion

After recalling the constitutive relations of non-isentropic fluids in both Eulerian and Lagrangian coordinates, we show how these systems can be cast as Irreversible Port-Hamiltonian Systems by extending the definition in Ramirez et al. [9] to include convective transport. The key modification concerns the differential operators, whose structure introduces convective terms that must be treated appropriately. As in Ramirez et al. [9], we then derive a parametrisation of the boundary port variables that ensures consistency with the first and second laws of thermodynamics.

References

  1. [1]Araque, L.M. (2020). Port-Hamiltonian modeling of fluid-structure interactions in a longitudinal domain. Ph.D. thesis, Université Bourgogne Franche-Comté; Universidad técnica Federico Santa María . . . .
  2. [2]Cardoso-Ribeiro, F.L., Haine, G., Le Gorrec, Y., Matignon, D., and Ramirez, H. (2024). Port-Hamiltonian formulations for the modeling, simulation and control of fluids. Computers & Fluids, 283, 106407.
  3. [3]Favache, A., Dochain, D., and Maschke, B. (2010). An entropy-based formulation of irreversible processes based on contact structures. Chemical Engineering Science, 65, 5204–5216.DOI
  4. [4]Grmela, M. and Öttinger, H. (1997). Dynamics and thermodynamics of complex fluids. I. development of a general formalism. Physical Review E, 56(6), 6620–6632.DOI
  5. [5]Hoang, H., Couenne, F., Jallut, C., and Le Gorrec, Y. (2012). Lyapunov-based control of non isothermal continuous stirred tank reactors using irreversible thermodynamics. Journal of Process Control, 22(2), 412–422.DOI
  6. [6]L.A. Mora, Y.L.G. and Ramirez, H. (2024). Energy-shaping and entropy-assignment boundary control of the heat equation. Systems and Control Letters, 189, 105821.DOI
  7. [7]Malvern, L.E. (1969). Introduction to the Mechanics of a Continuous Medium. Monograph. Prentice Hall, Inc.
  8. [8]Mora, L.A., Le Gorrec, Y., Matignon, D., Ramirez, H., and Yuz, J.J. (2021). On port-Hamiltonian formulations of 3-dimensional compressible Newtonian fluids. Physics of Fluids, 33(11), 117117.DOI
  9. [9]Ramirez, H., Gorrec, Y.L., and Maschke, B. (2022). Boundary controlled irreversible port-hamiltonian systems. Chemical Engineering Science, 248, 117107.DOI
  10. [10]Ramirez, H. and Le Gorrec, Y. (2022). An overview on irreversible port-hamiltonian systems. Entropy, 10, 1478.DOI
  11. [11]Ramirez, H., Maschke, B., and Sbarbaro, D. (2013a). Irreversible port-Hamiltonian systems: A general formulation of irreversible processes with application to the CSTR. Chemical Engineering Science, 89(0), 223 – 234.DOI
  12. [13]Ramirez, H., Le Gorrec, Y., Maschke, B., and Couenne, F. (2016). On the passivity based control of irreversible processes: A port-Hamiltonian approach. Automatica, 64, 105 – 111.DOI

Paper details

vlnc:W_bxtkvnt9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents