# In[Plot]
plt.figure()
plt.plot(t, y_meas, 'k', label="$y$")
plt.plot(t, y_hat, 'b', label="$\hat y$")
plt.plot(t, y_meas - y_hat, 'r', label="$e$")
plt.grid(True)
plt.xlabel('Time (s)')
plt.ylabel('Voltage (V)')
plt.legend(loc='upper right')
plt.savefig('WH_fit.pdf')
# In[Inspect linear model]
n_imp = 128
G1_num, G1_den = G1.get_tfdata()
G1_sys = control.TransferFunction(G1_num, G1_den, ts)
plt.figure()
plt.title("$G_1$ impulse response")
_, y_imp = control.impulse_response(G1_sys, np.arange(n_imp) * ts)
# plt.plot(G1_num)
plt.plot(y_imp)
plt.savefig(os.path.join("models", model_name, "G1_imp.pdf"))
plt.figure()
mag_G1, phase_G1, omega_G1 = control.bode(G1_sys, omega_limits=[1e2, 1e5])
plt.suptitle("$G_1$ bode plot")
plt.savefig(os.path.join("models", model_name, "G1_bode.pdf"))
# G2_b = G2.G.weight.detach().numpy()[0, 0, ::-1]
G2_num, G2_den = G2.get_tfdata()
G2_sys = control.TransferFunction(G2_num, G2_den, ts)
in_nl = np.arange(y1_lin_min, y1_lin_max, (y1_lin_max - y1_lin_min) /
1000).astype(np.float32).reshape(-1, 1)
with torch.no_grad():
out_nl = F_nl(torch.as_tensor(in_nl))
plt.figure()
plt.plot(in_nl, out_nl, 'b')
plt.plot(in_nl, out_nl, 'b')
#plt.plot(y1_lin, y1_nl, 'b*')
plt.xlabel('Static non-linearity input (-)')
plt.ylabel('Static non-linearity input (-)')
plt.grid(True)
# In[Plot]
# Inspect process noise blocks
n_imp = 128
H_inv_num, H_inv_den = H_inv.get_tfdata()
H_inv_sys = 1 + control.TransferFunction(H_inv_num, H_inv_den, ts)
H_sys = 1 / H_inv_sys
plt.figure()
mag_H, phase_H, omega_H = control.bode(H_sys, omega_limits=[1e1, 1e5])
plt.suptitle("$H_inv$ bode plot")
# plt.savefig(os.path.join("models", model_name, "G1_bode.pdf"))
# In[]
control.bode(H_sys)
control.bode(Hud * 1000)