Mamba-PINN Resilience Assessment Model: A Data–Physics Integrated Approach for Typhoon-Resistant Adaptation of Zhuhai Coastal Sewage Treatment Systems

We propose Mamba - Physics-Informed Neural Network(Mamba-PINN), a novel data–physics integrated resilience assessment model, to evaluate the anti-typhoon disturbance ability of coastal sewage treatment systems in Zhuhai. The increasing frequency of extreme weather events poses significant challenges to urban infrastructure; yet, existing methods often fail to capture the complex spatio-temporal dynamics of typhoon impacts. Our approach combines a Mamba neural network for high-frequency monitoring data processing with Physics-Informed Neural Networks (PINN) to quantify process recovery dynamics under typhoon conditions. The Mamba network extracts critical storm impact features, while the PINN embeds fluid inertia and microbial activity inhibition mechanisms to model system responses. Furthermore, we introduce a disturbance–recovery resilience index to provide a quantitative measure of system robustness, enabling targeted adaptive transformations for coastal sewage plants. The proposed method addresses the limitations of purely data-driven or physics-based models by integrating both paradigms, offering a more comprehensive understanding of resilience mechanisms. Experimental results demonstrate the model’s effectiveness in capturing nonlinear interactions between typhoon disturbances and treatment process recovery. This work contributes to the sustainable development of coastal cities by providing a scientifically grounded framework for infrastructure adaptation under climate change.