Please enable JavaScript.

Coggle requires JavaScript to display documents.

Paper - Coggle Diagram

- - - - Discussion
        
        Section 1
        
        1 paragraph
        
        1 paragraph
        
        1 paragraph
        
        Section 2
        
        1 paragraph
        
        1 paragraph
        
        1 paragraph
      - Subsection 2: Validation of the coupled ASM1–NN framework against coupled ASM1–CFD
        
        xfeed, xras and xe only between the 48 cfd coupled and NN coupled. Discuss where coupled errors are largest
        
        1 paragraph
        
        1 figure
        
        Optional: R2 for full field for the 48 coupled cases. Overall model accuracy on train/validation/test. Shows global R2, MSE, RMSE, and argues that the model generalizes to unseen test data. Discuss where coupled errors are largest
        
        1 paragraph
        
        1 figure
      - Subsection 4: Capacity
        
        Fixed-footprint internal-geometry optimization and baseline–optimized comparison
        
        The section tests whether internal geometry changes can expand the feasible region without changing tank footprint or volume.
        
        Define the optimization setup
        
        1 paragraph
        
        Define the objective and constraints
        
        Show design ranking and improvement
        
        1 paragraph
        
        1 figure
        
        Baseline vs optimized capacity envelope. discuss why they are better
        
        1 paragraph
        
        1 figure
        
        Optional: Selected CFD validation of optimized design
        
        Baseline plant-consistent capacity envelope
        
        Define the baseline geometry and loading grid
        
        1 paragraph
        
        Show the baseline capacity map
        
        Select the optimization loading condition or scenario set. Use the capacity envelope to choose the loading condition for Section 4.
        
        1 paragraph
        
        1 paragraph
        
        1 figure
      - Subsection 1: Predictive capability of the CFD-trained secondary clarifier neural surrogate (done) (The NN surrogate is first evaluated independently from ASM1.)
        
        Statiscs
        
        1000 cases R2 for full field. Overall model accuracy on train/validation/test. Shows global R2, MSE, RMSE, and argues that the model generalizes to unseen test data.
        
        1 paragaph
        
        posible xras and xe only
        
        1 figure
        
        best and worest cases
        
        1000 Case-by-case PM concentration evaluation. Separately evaluates c, gives high/medium/poor percentages, and explains why poor cases happen, especially extreme hydrograph parameter values and premature settling.
        
        1 paragraph
        
        1 figure
        
        optional: hyperparametertuning, I will put this in supp not in manuscript
        
        Overall runtime. AND speedup versus CFD in both standalone cfd and coupled cfd vs NN
        
        1 paragraph
        
        1 figure
      - Optinoal: Subsection 3: Sensitivity analysis to whole plant
        
        Motivation for sensitivity maps
        
        1 paragraph
        
        Direct NN sensitivity. Field-resolved sensitivity maps inside the clarifier
        
        1 paragraph
        
        1 figure
        
        Coupled biological-reactor sensitivity
        
        1 paragraph
        
        1 figure
        
        Define the baseline case and the sensitivity variables
        
        1 paragraph
        
        Optionaal: Interpretation, limitations, and connection to optimization
        
        1 paragraph
      - Conclusion
        
        1 paragraph
        
        1 paragraph
    - - State the overall framework. This study develops a multi-fidelity coupled surrogate framework in which the biological reactors are represented by ASM1/BSM1 and the secondary clarifier CFD model is replaced by a CFD-trained NN surrogate.
        
        1 paragaph
        
        1 figure
      - Describe the workflow from training to optimization. The workflow has four stages: generate CFD training cases, train/evaluate the NN surrogate, couple the NN to ASM1, then use the coupled framework for capacity mapping and geometry optimization.
        
        1 paragaph
      - Define main outputs and performance metrics
    - - CFD model summary
        
        1 paragaph
      - Clarifier parameterization
        
        1 paragaph
        
        1 figure
      - CFD sampling/design of experiments
        
        1 paragaph
    - - Define the surrogate mapping
        
        1 paragaph
      - Architecture
        
        1 paragaph
        
        1 figure
      - Training strategy and data split
        
        1 paragaph
    - - BSM1/ASM1 plant model summary
        
        1 paragaph
      - Mapping from ASM1 states to NN clarifier inputs and Mapping from NN outputs back to ASM1
        
        1 paragaph
        
        1 figure
      - Coupled steady-state solution algorithm
        
        1 paragaph
  - - - Our propose the solution: A NN surrogate can provide CFD higher-fidelity while giving predictions in real-time
      - 2.Historical and Current knowledge: All design optimization papers. What awesome things they did
      - There is a need for a multi-fidelity surrogate framework. and objectives
      - 1.Opening: importance:Why secondary clarifiers matter in activated sludge systems
      - Gap in knowledge: The main coupling problem: clarifier design cannot be optimized as an isolated unit and Plant-wide CFD simulation is too expensive for design optimization. I can mention that current plant-wide simulations are using 1d to save resources.