Results of fault diagnosis using TCN#

from ice.fault_diagnosis.datasets import FaultDiagnosisRiethTEP
from ice.fault_diagnosis.models import TCN
from sklearn.preprocessing import StandardScaler
import numpy as np
import pandas as pd

Download the dataset.

dataset = FaultDiagnosisRiethTEP()

Normalize the data.

scaler = StandardScaler()
dataset.df[dataset.train_mask] = scaler.fit_transform(dataset.df[dataset.train_mask])
dataset.df[dataset.test_mask] = scaler.transform(dataset.df[dataset.test_mask])

Create the TCN model.

model = TCN(
    window_size=60,
    batch_size=128,
    num_layers=1,
    kernel_size=3,
    hidden_dim=32,
    lr=1e-4,
    num_epochs=30,
    verbose=True,
    device='cpu',
    save_checkpoints=True,
    val_ratio=0.1,
)

Load the checkpoint.

model.load_checkpoint('tcn_fault_diagnosis_epoch_30.tar')

Evaluate the model on the test data.

metrics = model.evaluate(
    dataset.df[dataset.test_mask],
    dataset.target[dataset.test_mask]
)

idx = np.array([1, 2, 4, 5, 6, 7, 8, 10, 11, 12, 13, 14, 16, 17, 18, 19, 20]) - 1
pd.DataFrame({
    'Fault': idx,
    'TPR': np.array(metrics['true_positive_rate'])[idx],
    'FPR': np.array(metrics['false_positive_rate'])[idx],
}).round(4)

	Fault	TPR	FPR
0	0	0.9675	0.0000
1	1	0.9738	0.0000
2	3	0.9643	0.0000
3	4	0.9584	0.0000
4	5	0.9731	0.0000
5	6	0.9679	0.0000
6	7	0.9691	0.0000
7	9	0.9651	0.0000
8	10	0.9788	0.0000
9	11	0.9526	0.0000
10	12	0.9418	0.0001
11	13	0.9780	0.0000
12	15	0.9752	0.0000
13	16	0.9608	0.0000
14	17	0.9357	0.0000
15	18	0.9717	0.0000
16	19	0.9482	0.0000

print(f'Average TPR: {np.array(metrics["true_positive_rate"])[idx].mean():.2f}')

Average TPR: 0.96

for i in np.array(metrics["true_positive_rate"])[idx]*100:
    print(f'{i:.2f}')