Model

A suite of structures, their vibration responses to strong ground motion events, and analysis to investigate the relationships between ``ground truth'' damage state and estimated damage states obtained from inverse system identification.

Getting Started

1. get_data.py : finite element model analysis and data extraction.
   • Choose an analysis configuration via command-line flags
     • --structure frame or --structure bridge
     • --multisupport (bridge only)
     • --elastic (otherwise inelastic)
     • --field_only to save measured field data without running the FE model
   • Loads a suite of events
   • For each event:
     • saves measured field inputs (ground) and outputs (structure)
     • performs FEM analysis and saves:
       • pre- and post- earthquake natural frequencies from FEM eigenvalue analysis
       • displacement response histories at select output nodes
       • strain/stress (or force/deformation) response histories at select output elements
       • model inputs and outputs used for system identification
2. get_systems.py : system identification on the data produced by get_data.py.
   • Choose an analysis configuration via command-line flags
     • --structure frame or --structure bridge
     • --source field, --source elastic, or --source inelastic
     • --sid_method srim
     • --no_windowing to disable training data truncation
   • For each event, saves:
     • timestep (dt)
     • time array
     • inputs array
     • outputs array
     • system matrices (A,B,C,D)
3. get_prediction.py : simulates predictions from the system realizations produced by get_systems.py and compares them against measured output.
   • Choose an analysis configuration via command-line flags
     • --structure frame or --structure bridge
     • --source field, --source elastic, or --source inelastic
     • --output_quantity displacement or --output_quantity acceleration
     • --no_windowing, --no_signal_align
   • For each event, saves:
     • simulated (predicted) output
     • true vs. predicted timeseries plots (PNG + interactive HTML)
     • normalized L2 prediction error
   • After all events, saves cross-event error heatmaps
4. plot_inputs_outputs.py: plot the inputs and outputs used for system ID. Primarily used for debugging.
5. plot_series.py: plot timeseries.
   1. Prompts the user for:
      1. structure
      2. event
      3. quantity
   2. Adds onto an axis:
      1. source
      2. location
   3. Save the plot if desired.

Details on windowing and alignment throughout pipeline

1. get_systems: window inputs and outputs, create A,B,C,D with system ID. save inputs and outputs into System ID / structure / source / quantity / System ID Training Data.
2. get_prediction: take inputs from System ID Training Data. Use A,B,C,D to predict output from inputs. Window and align inputs, predicted outputs, and true outputs, save into System ID / structure / source / quantity / System ID Results / _processed. Use processed data for error computation and heatmap.

Running

Export and compare results

You can also run Build and the four comparisons from the VS Code Inspect sidebar. Install vscode-inspect/model-inspect-0.1.6.vsix using Extensions → … → Install from VSIX…, then select your Python interpreter and input folders. See the extension guide for setup and usage.

Export all three categories and the active Python environment with one command:

python inspect.py build naiqi

The supplied prefix determines the output location. To save in a particular run, use python inspect.py build runs/your_run/naiqi instead. Existing error CSV names are preserved. Each structure/quantity/source folder additionally contains training/{dt,time,ground,structure}/<event>.csv and systems/<event>/{A,B,C,D}.csv. Environment records go in <prefix>_environment/packages.txt and environment.json.

Compare one category per command, using the exact input directories you want:

python inspect.py compare errors chrystal_bridge runs/your_run/naiqi_bridge
python inspect.py compare realizations chrystal_bridge runs/your_run/naiqi_bridge
python inspect.py compare training chrystal_bridge runs/your_run/naiqi_bridge
python inspect.py compare errors chrystal_frame runs/your_run/naiqi_frame
python inspect.py compare realizations chrystal_frame runs/your_run/naiqi_frame
python inspect.py compare training chrystal_frame runs/your_run/naiqi_frame
python inspect.py compare environment reference_environment analysis_export_environment

Replace your_run with your actual directory name. No latest-run lookup is used. Each command prints its result and saves one report beside its second input folder: compare_bridge_errors.txt, compare_bridge_realizations.txt, compare_bridge_training.txt, and the corresponding three frame files. The standalone environment comparison writes compare_environment.txt beside the second environment folder. It compares recorded Python version, platform and package versions, ignoring paths and capture timestamps for match status. These snapshots describe the export environment, not necessarily the original computation environment; matching versions do not verify identical BLAS builds. Rerunning the same category in the same output location updates that report. Old compare_bridge.txt and compare_frame.txt files are left untouched.

realizations compares the existing systems/ CSV exports; their directory names do not change. Matrices are compared before prediction-time stabilization. Different state coordinates can yield different matrices with equivalent input/output behavior. Numeric comparisons check shapes, non-finite values, and symmetric relative/absolute tolerances. Missing data is reported; a category absent from both folders is not considered a successful match. Exit status is 0 for a match and 1 for differences or missing data.

Each report includes environment information when available. Package differences are informational and do not affect the numeric comparison exit status. Required source data must exist for all configured events during export. Files are not deleted, but reusing the same explicit export prefix updates same-name files.

build also collects heatmaps into <name>_heatmaps/; no separate command is needed.

The current run_full_comparison.sh runs data generation, system identification, and prediction, saving logs in its timestamped run directory. Export and compare with the commands above afterward. Modeling/ and System ID/ remain shared working directories. All inspection logic is in inspect.py.

Get data for all four configurations (frame/bridge × inelastic/elastic):

for s in frame bridge; do for e in "" "--elastic"; do python get_data.py --structure "$s" $e; done; done

Run system ID for all structures and sources (frame/bridge × field/elastic/inelastic):

for s in frame bridge; do for src in field elastic inelastic; do python get_systems.py --structure "$s" --source "$src"; done; done

Run prediction for all structures and sources (frame/bridge × field/elastic/inelastic):

for s in frame bridge; do for src in field elastic inelastic; do python get_prediction.py --structure "$s" --source "$src" --annotate_plots; done; done

Overall Directory Structure

tree.nathanfriend

.
├── Modeling/
│   ├── bridge/
│   │   ├── field/
│   │   │   ├── acceleration/
│   │   │   │   ├── ground/
│   │   │   │   │   ├── 1.csv
│   │   │   │   │   ├── 2.csv
│   │   │   │   │   └── ...
│   │   │   │   └── structure/
│   │   │   │       ├── 1.csv
│   │   │   │       └── ...
│   │   │   ├── displacement/
│   │   │   │   ├── ground/
│   │   │   │   │   ├── 1.txt
│   │   │   │   │   └── ...
│   │   │   │   └── ...
│   │   │   └── ...
│   │   ├── elastic/
│   │   │   ├── acceleration/
│   │   │   │   └── ground/
│   │   │   │       ├── 1.csv
│   │   │   │       └── ...
│   │   │   └── ...
│   │   └── inelastic/
│   │       └── ...
│   └── frame/
│       ├── field/
│       │   ├── acceleration/
│       │   │   ├── ground/
│       │   │   │   ├── 226.csv
│       │   │   │   └── ...
│       │   │   └── ...
│       │   └── ...
│       ├── elastic/
│       │   ├── acceleration/
│       │   │   └── ground/
│       │   │       ├── 226.csv
│       │   │       └── ...
│       │   └── ...
│       └── ...
└── System ID/
    ├── bridge/
    │   ├── acceleration/
    │   │   ├── field/
    │   │   │   ├── System ID Training Data/  
    │   │   │   │   ├── ground/
    │   │   │   │   │   └── 1.csv
    │   │   │   │   |   └── ...
    │   │   │   │   ├── structure/
    │   │   │   │   |   ├── 1.csv
    │   │   │   │   |   └── ...
    │   │   │   │   ├── dt/
    │   │   │   │   |   ├── 1.csv
    │   │   │   │   |   └── ...
    │   │   │   │   └── time/
    │   │   │   │       └── ...
    │   │   │   └── System ID Results/
    │   │   │       ├── system realization/
    │   │   │       │   ├── 1.pkl
    │   │   │       │   └── ...
    │   │   │       ├── prediction plots/
    │   │   │       │   ├── 1.pkl
    │   │   │       │   └── ...
    │   │   │       ├── inputs_processed
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── outputs_pred_processed
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── outputs_true_processed
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── time_processed
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── dt
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── errors
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── frequency ID (not yet implemented)/
    │   │   │       │   ├── 1.csv
    │   │   │       │   └── ...
    │   │   │       ├── heatmap.png
    │   │   │       └── heatmap_square.png
    │   │   ├── elastic/
    │   │   │   ├── System ID Training Data/
    │   │   │   └── System ID Results/
    │   │   └── inelastic/
    │   │       ├── System ID Training Data/
    │   │       └── System ID Results/
    │   │
    │   └── displacement/
    │       ├── field/
    │       │   ├── System ID Training Data/
    │       │   └── System ID Results/
    │       ├── elastic/
    │       │   ├── System ID Training Data/
    │       │   └── System ID Results/
    │       └── inelastic/
    │           ├── System ID Training Data/
    │           └── System ID Results/
    │
    └── frame/
        ├── acceleration/
        │   ├── field/
        │   │   ├── System ID Training Data/  
        │   │   │   ├── ground/
        │   │   │   │   └── 226.csv
        │   │   │   |   └── ...
        │   │   │   ├── structure/
        │   │   │   |   ├── 226.csv
        │   │   │   |   └── ...
        │   │   │   └── ...
        │   │   └── System ID Results/
        │   │       └── ...
        │   ├── elastic/
        │   │   ├── System ID Training Data/  
        │   │   └── System ID Results/
        │   └── inelastic/
        │       └── ...
        └── displacement/
            └── ...

Modeling Directory Structure

Level	Name	Quantities
1	Structure	frame, bridge
2	Source	field, elastic, inelastic
3	Quantity	time, dt, displacement, acceleration, stress, strain, frequency pre-eq, frequency post-eq
4	Location	ground (input), structure (output)
5	Event	1, 2, 3, ... or 226, 227, 228, ... etc.

See below for list of quantities and locations available in each Source's subdirectory.

Source	Quantities	Locations
field	time, dt, displacement, acceleration	ground (input), structure (output)
elastic	displacement, acceleration, stress, strain, frequency pre-eq, frequency post-eq	structure (output)
inelastic	displacement, acceleration, stress, strain, frequency pre-eq, frequency post-eq	structure (output)

System ID Directory Structure

Level	Name	Quantities
1	Structure	frame, bridge
2	Source	field, elastic, inelastic
3	Output Quantity	displacement, acceleration
4a*	System ID Training Data	ground acceleration (true input), structure response (true output), time, dt
4b	System ID Results	system realization (A,B,C,D), frequency ID, mode shapes, prediction, prediction error, heatmap (encompasses all events)
5	Event	1, 2, 3, ... or 226, 227, 228, ... etc.

*All time series are truncated and aligned according to true output.

Environment

Method 1

1. Install numba: conda install numba
2. Install requirements: pip install -r requirements.txt

Method 2

1. Set up a xara-friendly environment: https://xara.so/user/guides/compile.html
2. Install requirements: pip install -r requirements.txt