Fitting a circuit¶

Use Circuit.fit()

from fasteis import Circuit

circ = Circuit("R1-(R2,C2)")

res = circ.fit(f, Z)

You can also pass a 'Battery Data Format' style dataframe directly:

from fasteis import Circuit
import bdf

circ = Circuit("R1-(R2,C2)")

df = bdf.read("my/bdf/file.parquet")

res = circ.fit(df)

Machine learning guesses¶

fasteis has small convolutional neural networks trained on specific common circuits. It can use these models to guess good initial parameters to the fit, meaning faster, more robust fits, and no need to manually adjust input parameters.

This happens by default if initial parameters are not supplied to the circuit (as above).

To see the available models, use Circuit.ml_circuits().

A model will be used if it matches your circuit topology, even if elements are placed in a different way.

E.g. the "rc" circuit "R0-(R1,C1)" will match the following:

• "R1-(R3,C7)" - different labels
• "(C0,R1)-R2" - different order
• "R1-K1" - equivalent elements used

You can force or disable the machine learning guess with:

circ.fit(f, Z, guess_init=False)

Fixing parameters¶

Hold parameters constant during the fit with fixed, using the names from Circuit.param_names().

A dict holds each parameter at the given value:

circ = Circuit("sei_randles")

res = circ.fit(f, Z, fixed={"R0.r": 100.0})

Parameters start from the machine learning guess with R0.r replaced by 100.

You can also pass a list of parameter names for circuits with existing values:

circ = Circuit("R0-(R1,C1)").with_named_values({"R0.r": 10.0, "R1.r": 20.0, "C1.c": 1e-6})
res = circ.fit(f, Z, fixed=["R0.r"])

Fixed parameters appear in res.params with their held values, and as NaN in res.stderr.