Backstepping Observer for the Quasilinear Heat Equation with Linear Design Gains: Beyond Local Stability

Abstract: We consider the one-dimensional quasilinear heat equation with state-dependent heat capacity and thermal conductivity, and design a boundary-output observer based on the backstepping design for a linear heat equation with constant coefficients. Viewing the quasilinear system as a perturbation of the linear one, we establish exponential stability of the origin for the observation error dynamics in $H^1$, with an explicit region of attraction depending on the system parameters, observer gains, and the mismatch between the nonlinear diffusivity and the constant design diffusivity. Importantly, the observation error converges to zero rather than merely to a neighborhood scaling with this mismatch, even though, in contrast to backstepping-based stabilization of nonlinear PDEs, the mismatch need not decay along trajectories and may remain bounded away from zero, acting as a persistent state-dependent multiplicative perturbation. A technical challenge was to perform a sufficiently-fine Lyapunov analysis that does not yield overly conservative results such as mere boundedness of the observation error. Interestingly, while in the linear case the relationship between one of the backstepping observer gains and the convergence rate is monotonic, we show that in the nonlinear setting this is no longer the case: there may exist an optimal value of that gain, beyond which further increases deteriorate the system's performance. Such behavior cannot be predicted without our analysis: one might expect a priori the decay rate to be freely tunable at the expense of a region of attraction that shrinks to zero as the prescribed rate tends to infinity. However, our Lyapunov analysis (supported by numerical experiments) reveals that this intuition is incorrect.

• The paper introduces a backstepping observer that achieves exponential stability in the H¹ norm for quasilinear heat equations.
• It transforms the original PDE via an enthalpy change, enabling linear-design gains to ensure explicit performance guarantees.
• The study reveals that higher observer gains can deteriorate convergence due to kernel amplification, necessitating optimal gain selection.

Backstepping Observer Design for the Quasilinear Heat Equation: Beyond Local Stability

Introduction and Problem Context

This work addresses the observer design for a one-dimensional quasilinear heat equation with state-dependent heat capacity and thermal conductivity, a representative class of nonlinear parabolic PDEs found in various physical and engineering applications, including heat transfer in plasmas, phase-change materials, and porous media. The central challenge is to estimate the spatial temperature distribution solely from boundary measurements, under severe restrictions on sensor deployment. The equation under study generalizes the classical heat equation, where both the heat capacity $c(T)$ and thermal conductivity $\kappa(T)$ can be nonlinear functions of the temperature.

A key focus is the extension of backstepping observer techniques—well-established for linear parabolic PDEs—to the nonlinear, quasilinear setting. The methodology leverages the quasilinear equation's structure by treating the nonlinearity as a perturbation to a nominal linear model, thereby enabling the use of linear-design-based gains for the observer. This approach aims to establish exponential stability of the observation error in the $H^1$ norm, with explicit characterization of the region of attraction and performance dependencies on design parameters.

Technical Approach

Model Reformulation

To facilitate analysis and observer synthesis, the original equation is transformed by an enthalpy change of variables, allowing the quasilinear model to be recast into a form with a state-dependent ("quasilinear") diffusivity. The observer structure is inspired by the backstepping methodology for the linear heat equation, employing kernel-based correction terms with design parameters inherited from the linear setting.

Specifically, the observer is

$$\hat{v}_t = (\alpha(\hat{v})\hat{v}_x)_x + p_1(x) (y(t) - \hat{y}(t)),$$

with Neumann boundary conditions and gain functions $p_1(x)$, $p_{10}$ computed from the solution to a kernel PDE associated with the nominal linear equation. The observer error dynamics, after appropriate transformation (backstepping), are shown to consist of a target exponentially stable system perturbed by state-dependent terms proportional to the mismatch $\bar{\alpha}(v) = \alpha(v) - a$, where $a$ is the design diffusivity used for kernel synthesis.

Lyapunov Analysis and Region of Attraction

A main technical achievement is the construction of a Lyapunov functional—built on the $H^1$ norm—that allows a fine-grained analysis of the perturbed, nonlinear error system. Unlike the linear case, $L^2$-based functionals are insufficient due to the nonlinearity's effect on higher-order spatial norms via Agmon-type inequalities. The analysis yields explicit differential inequalities linking the convergence rate and region of attraction to the observer gains $\kappa(T)$0 and $\kappa(T)$1, as well as the magnitude of the diffusivity mismatch.

A series of lemmas establish that, under suitable conditions on the observer gains and mismatch (given by inequalities involving Lipschitz constants of the nonlinearities and kernel norms), the observation error converges exponentially to zero in $\kappa(T)$2 norm. This, in turn, guarantees max-norm convergence, and by the invertibility of the enthalpy transformation, the physical temperature profile can also be reconstructed with exponential accuracy.

Numerical and Theoretical Findings

A central and nontrivial theoretical result is the discovery that the interplay between the observer gain $\kappa(T)$3 and the system nonlinearity breaks the monotonicity present in the linear case: increasing the observer gain does not guarantee improved convergence rates. Instead, for sufficiently large $\kappa(T)$4, the induced kernel amplification of the persistent mismatch term causes the effective decay rate to deteriorate, and the region of attraction to shrink. Thus, an optimal value of $\kappa(T)$5 exists, contradicting intuition from the linear framework, where the convergence rate is typically monotonically and unboundedly tunable via $\kappa(T)$6 at the cost of robustness.

Additionally, while in the stabilization setting the mismatch between the system and design diffusivities can be made to vanish as the state approaches the equilibrium, in the pure observer design problem with independent system and observer initial conditions, the mismatch is persistent and independent of the observation error trajectory. This mandates a genuinely robust observer design with explicit performance guarantees tied to system nonlinearities.

The explicit estimates for the region of attraction, the dependence of the exponential rate, and the robustness margins are all computable from system parameters, kernel norms, and Lipschitz constants, providing practitioners with constructive tuning guidelines.

Implications and Future Directions

From a practical standpoint, the development of a backstepping observer with explicit robustness guarantees for quasilinear parabolic PDEs significantly enlarges the class of physically realistic systems amenable to advanced boundary state estimation. This holds direct relevance for thermal management systems, energy storage devices, and process engineering, where non-constant coefficients are intrinsic.

The results also sharpen the theoretical boundaries of the backstepping approach: the analysis exposes fundamental performance limitations arising from the mismatch between nonlinear plant dynamics and linear-design-based observers. This invites future research into the synthesis of truly nonlinear backstepping kernels, model adaptation techniques, or the development of more sophisticated Lyapunov functionals to reduce conservatism and enlarge the certified region of attraction.

Two prominent challenges for further work are (i) generalization to multi-dimensional domains, where the control of max norms involves higher ($\kappa(T)$7) regularity, and (ii) reduction of conservatism in the Lyapunov bounds, potentially by leveraging nonlinear functionals tailored to the structure of quasilinear PDEs. Extensions to more complex coupled PDE systems and output-feedback controller synthesis are suggested as promising applications.

Conclusion

This paper advances the theory and practice of observer design for nonlinear parabolic PDEs by presenting a rigorous, constructive methodology based on linear-design backstepping with explicit robustness analysis. By transcending local stability and systematically addressing the interaction between nonlinearity and observer gains, it provides a foundation for reliable state estimation in a broad range of real-world distributed parameter systems. The insights on the non-monotonicity of observer gain tuning and the limitations of linear-based design in nonlinear contexts have broad implications for the design of high-performance estimation schemes.