Spaces:
Running on Zero
Running on Zero
File size: 15,694 Bytes
37dfff7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 | import numpy as np
import pandas as pd
from functools import partial
from scipy.optimize import minimize
from scipy.signal import savgol_filter
import jax
jax.config.update("jax_enable_x64", True)
import jax.numpy as jnp
from jax.lax import scan
import asyncio
def load_and_preprocess_cv_data(df, pot_col, cur_col, scan_rate_v_s, skip_factor):
if isinstance(df, str):
df = pd.read_csv(df, sep=None, engine='python')
if pot_col >= df.shape[1] or cur_col >= df.shape[1]:
raise ValueError(f"Selected column index (Potential: {pot_col}, Current: {cur_col}) exceeds total available columns ({df.shape[1]}).")
s_pot = pd.to_numeric(df.iloc[:, pot_col], errors='coerce').dropna()
s_cur = pd.to_numeric(df.iloc[:, cur_col], errors='coerce').dropna()
common_idx = s_pot.index.intersection(s_cur.index)
raw_potential = s_pot.loc[common_idx].values.astype(np.float64)
raw_current = s_cur.loc[common_idx].values.astype(np.float64)
if len(raw_potential) == 0:
raise ValueError(f"No valid numeric data found in Column {pot_col} (Potential) and Column {cur_col} (Current).")
voltage_steps = np.abs(np.diff(raw_potential, prepend=raw_potential[0]))
raw_time = np.cumsum(voltage_steps) / scan_rate_v_s
exp_potential = raw_potential[::skip_factor]
exp_current = raw_current[::skip_factor]
exp_time = raw_time[::skip_factor]
return exp_time, exp_potential, exp_current
def extract_physics_priors(potential, turn_idx, num_peaks, v_min, v_max):
global_params = [2.0, 1.0, 1.0, np.mean(potential), 0.0, 0.1, 1.0, 0.1, 1.0]
peaks_matrix = np.zeros((num_peaks, 3))
v_crits = np.linspace(v_min + 0.1, v_max - 0.1, num_peaks)
for i in range(num_peaks):
peaks_matrix[i] = [1.0, v_crits[i], 15.0]
return np.concatenate([global_params, peaks_matrix.flatten()])
def get_parameter_bounds(idx, val, num_globals, v_min, v_max):
var = np.abs(val) * 0.5
if idx == 0: return (max(1e-8, val - 5.0), val + 5.0)
if idx in (1, 2): return (val - 1.0, val + 1.0)
if idx == 3: return (val - 0.6, val + 0.6)
if idx == 4:
var = var if val != 0 else 10.0
return (val - var, val + var)
if idx in (5, 7): return (max(1e-8, val - (var + 0.1)), val + var + 0.1)
if idx in (6, 8): return (max(0.1, val - (var + 0.5)), val + var + 0.5)
offset = (idx - num_globals) % 3
if offset == 0: return (max(1e-4, val - (var + 1e-4)), val + 5.0)
if offset == 1: return (max(v_min, val - (var + 1e-4)), min(v_max, val + var + 1e-4))
return (max(0.1, val - (var + 1e-4)), val + 20.0)
def create_staged_bounds(target_params, active_indices, v_min, v_max, num_globals):
bounds = []
for i, val in enumerate(target_params):
if i not in active_indices:
bounds.append((val - 1e-9, val + 1e-9))
else:
bounds.append(get_parameter_bounds(i, val, num_globals, v_min, v_max))
return bounds
@partial(jax.jit, static_argnames=['num_terms'])
def run_fourier_simulation_with_data(time_array, potential_array, diffusivity, beta_left, beta_right, v_center, peaks_matrix, num_terms, thickness):
n_arr = jnp.arange(1.0, num_terms + 1.0, dtype=jnp.float64)
wavenumbers = (2.0 * n_arr - 1.0) * jnp.pi / (2.0 * thickness)
fourier_coeffs = 4.0 / ((2.0 * n_arr - 1.0) * jnp.pi)
sin_integrals = 1.0 / wavenumbers
dt_arr = jnp.diff(time_array)
dt_arr = jnp.where(dt_arr <= 0.0, 1e-6, dt_arr)
weights = peaks_matrix[:, 0, jnp.newaxis]
v_crits = peaks_matrix[:, 1, jnp.newaxis]
sharpnesses = peaks_matrix[:, 2, jnp.newaxis]
occ_matrix = weights / (1.0 + jnp.exp(-sharpnesses * (potential_array[jnp.newaxis, :] - v_crits)))
occ_eq_arr = jnp.sum(occ_matrix, axis=0)
occ_eq_old_arr = occ_eq_arr[:-1]
occ_eq_new_arr = occ_eq_arr[1:]
beta_arr = jnp.where(potential_array[1:] < v_center, beta_left, beta_right)
d_val_arr = diffusivity * jnp.exp(beta_arr * (potential_array[1:] - v_center)**2)
k_dt_matrix = jnp.outer(d_val_arr * dt_arr, wavenumbers**2)
decay_matrix = jnp.exp(-k_dt_matrix)
forcing_factor = jnp.where(
k_dt_matrix < 1e-8,
1.0 - k_dt_matrix / 2.0,
(1.0 - decay_matrix) / k_dt_matrix
)
base_forcing = jnp.outer(occ_eq_old_arr - occ_eq_new_arr, fourier_coeffs)
forcing_matrix = base_forcing * forcing_factor
def init_step(carry, xs):
cum_dec, acc_forc = carry
dec, forc = xs
return (cum_dec * dec, acc_forc * dec + forc), None
init_carry = (jnp.ones(num_terms, dtype=jnp.float64), jnp.zeros(num_terms, dtype=jnp.float64))
(final_cum_dec, final_acc_forc), _ = scan(init_step, init_carry, (decay_matrix, forcing_matrix))
T_m_0 = final_acc_forc / (1.0 - final_cum_dec + 1e-15)
def history_step(fourier_modes, xs):
dec, forc = xs
fourier_modes = fourier_modes * dec + forc
return fourier_modes, fourier_modes
_, fourier_history = scan(history_step, T_m_0, (decay_matrix, forcing_matrix))
sum_fourier = jnp.dot(fourier_history, sin_integrals)
total_ions_all = thickness * occ_eq_new_arr + sum_fourier
total_ions_old_init = thickness * occ_eq_arr[0] + jnp.sum(T_m_0 * sin_integrals)
total_ions_shifted = jnp.concatenate([jnp.array([total_ions_old_init]), total_ions_all])
simulated_currents = jnp.diff(total_ions_shifted) / dt_arr
return simulated_currents
def solve_cv(df, config, pot_col, cur_col, queue, loop):
USER_CONFIG = config
OPTIMIZER_CONFIG = {
"max_iter": int(config.get("max_iter", 100)),
"tol_ftol": float(config.get("tol_ftol", 1e-8)),
"tol_gtol": float(config.get("tol_gtol", 1e-7)),
"num_globals": 9,
"mult_diff": (USER_CONFIG["film_thickness"]**2) / 10.0,
"mult_beta": 1.0,
"mult_offset": 1e-4,
"mult_bg_a": 1e-4,
"mult_bg_k": 10
}
num_terms = int(config.get("num_terms", 50))
loss_weight_const = float(config.get("loss_weight_const", 1.0))
exp_time, exp_potential, exp_current = load_and_preprocess_cv_data(
df, pot_col, cur_col,
USER_CONFIG["scan_rate_v_s"], USER_CONFIG["skip_factor"]
)
exp_time_jax = jnp.array(exp_time)
exp_potential_jax = jnp.array(exp_potential)
global_target_current_jax = jnp.array(exp_current[1:].ravel())
if loop and queue:
loop.call_soon_threadsafe(
queue.put_nowait, {
"type": "init",
"exp_potential": exp_potential[1:].tolist(),
"exp_current": exp_current[1:].tolist()
}
)
turn_idx = np.argmax(np.abs(exp_potential - exp_potential[0]))
if turn_idx < len(exp_potential) * 0.1:
turn_idx = len(exp_potential) // 2
smoothed_current = savgol_filter(exp_current, window_length=51, polyorder=3)
d2I_raw = np.abs(np.diff(smoothed_current, n=2))
d2I = np.pad(d2I_raw, (1, 1), mode='edge')
loss_weights = (d2I / np.max(d2I)) + loss_weight_const
edge_threshold = (USER_CONFIG["v_max"] - USER_CONFIG["v_min"]) * 0.05
left_mask = exp_potential[1:] < (USER_CONFIG["v_min"] + edge_threshold)
right_mask = exp_potential[1:] > (USER_CONFIG["v_max"] - edge_threshold)
global_weights = loss_weights[1:].ravel().copy()
global_weights_masked = global_weights.copy()
global_weights_masked[left_mask] = 0.0
global_weights_masked[right_mask] = 0.0
dt_array = np.diff(exp_time)
dt_array[dt_array <= 0] = 1e-6
data_driven_initial_guess = extract_physics_priors(
exp_potential, turn_idx, USER_CONFIG["num_peaks"], USER_CONFIG["v_min"], USER_CONFIG["v_max"]
)
initial_peaks_matrix = data_driven_initial_guess[OPTIMIZER_CONFIG["num_globals"]:].reshape((-1, 3))
baseline_diffusivity = data_driven_initial_guess[0] * OPTIMIZER_CONFIG["mult_diff"]
calibration_sim = np.array(run_fourier_simulation_with_data(
exp_time_jax, exp_potential_jax,
baseline_diffusivity, 0.0, 0.0,
data_driven_initial_guess[3],
jnp.array(initial_peaks_matrix), num_terms, USER_CONFIG["film_thickness"]
))
pure_faradaic_target = exp_current[1:]
v_range = USER_CONFIG["v_max"] - USER_CONFIG["v_min"]
safe_min = USER_CONFIG["v_min"] + (v_range * 0.15)
safe_max = USER_CONFIG["v_max"] - (v_range * 0.15)
safe_mask = (exp_potential[1:] > safe_min) & (exp_potential[1:] < safe_max)
real_faradaic_ptp = np.ptp(pure_faradaic_target[safe_mask])
sim_ptp = np.ptp(calibration_sim[safe_mask])
if sim_ptp < 1e-12: sim_ptp = 1e-6
calibrated_scale = real_faradaic_ptp / sim_ptp
total_simulated_baseline = (calibration_sim * calibrated_scale)
real_mean = np.mean(exp_current[1:])
sim_mean = np.mean(total_simulated_baseline)
calibrated_offset = (real_mean - sim_mean) / OPTIMIZER_CONFIG["mult_offset"]
data_driven_initial_guess[4] = calibrated_offset
calibrated_scale_jax = jnp.array(calibrated_scale)
@jax.jit
def compute_forward(scaled_params, weights):
diffusivity = scaled_params[0] * OPTIMIZER_CONFIG["mult_diff"]
beta_left = scaled_params[1] * OPTIMIZER_CONFIG["mult_beta"]
beta_right = scaled_params[2] * OPTIMIZER_CONFIG["mult_beta"]
v_center = scaled_params[3]
baseline_offset = scaled_params[4] * OPTIMIZER_CONFIG["mult_offset"]
a_right = scaled_params[5] * OPTIMIZER_CONFIG["mult_bg_a"]
k_right = scaled_params[6] * OPTIMIZER_CONFIG["mult_bg_k"]
a_left = scaled_params[7] * OPTIMIZER_CONFIG["mult_bg_a"]
k_left = scaled_params[8] * OPTIMIZER_CONFIG["mult_bg_k"]
peaks_matrix = jnp.reshape(scaled_params[OPTIMIZER_CONFIG["num_globals"]:], (-1, 3))
simulated_currents = run_fourier_simulation_with_data(
exp_time_jax, exp_potential_jax,
diffusivity, beta_left, beta_right, v_center, peaks_matrix, num_terms, USER_CONFIG["film_thickness"]
)
bg_current = a_right * jnp.exp(k_right * (exp_potential_jax[1:] - USER_CONFIG["v_max"])) \
- a_left * jnp.exp(-k_left * (exp_potential_jax[1:] - USER_CONFIG["v_min"]))
final_sim = (simulated_currents * calibrated_scale_jax) + baseline_offset + bg_current
squared_errors = (final_sim - global_target_current_jax)**2
weighted_mse = jnp.average(squared_errors, weights=weights)
loss = jnp.sqrt(weighted_mse) * 1e6
return loss, final_sim
@jax.jit
def objective_function_jax(scaled_params, weights):
loss, _ = compute_forward(scaled_params, weights)
return loss
loss_and_grad_jax = jax.jit(jax.value_and_grad(objective_function_jax, argnums=0))
def scipy_objective(x, weights):
loss, grad = loss_and_grad_jax(jnp.array(x), jnp.array(weights))
return np.array(loss, dtype=np.float64), np.array(grad, dtype=np.float64)
class OptimizationTracker:
def __init__(self):
self.iter_count = 0
self.stage_label = ""
self.active_weights = None
def set_stage(self, label, weights):
self.stage_label = label
self.active_weights = weights
self.iter_count = 0
def __call__(self, xk):
if self.iter_count % 10 == 0:
loss, final_sim = compute_forward(jnp.array(xk), jnp.array(self.active_weights))
# Send update to queue via loop
if loop and queue:
loop.call_soon_threadsafe(
queue.put_nowait, {
"type": "update",
"stage": self.stage_label,
"iter": self.iter_count,
"loss": float(loss),
"sim_current": np.array(final_sim).tolist()
}
)
self.iter_count += 1
tracker = OptimizationTracker()
all_indices = list(range(len(data_driven_initial_guess)))
idx_baseline = [4]
idx_bg = [5, 6, 7, 8]
idx_diffusion_base = [0, 3]
idx_beta = [1, 2]
idx_peaks = list(range(OPTIMIZER_CONFIG["num_globals"], len(data_driven_initial_guess)))
optim_options = {
'maxiter': OPTIMIZER_CONFIG["max_iter"],
'ftol': OPTIMIZER_CONFIG["tol_ftol"],
'gtol': OPTIMIZER_CONFIG["tol_gtol"],
'disp': False
}
stages = [
("Stage 1: Pure Flat Baseline", idx_baseline, global_weights_masked),
("Stage 1.5: Background Tails", idx_bg, global_weights),
("Stage 2: Anchor Peaks (Constant D)", idx_baseline + idx_bg + idx_diffusion_base + idx_peaks, global_weights),
("Stage 3: Full Non-Linear Polish", all_indices, global_weights)
]
current_x = data_driven_initial_guess
# Send initial data arrays to client so it can setup the base plot
if loop and queue:
loop.call_soon_threadsafe(
queue.put_nowait, {
"type": "init",
"exp_potential": exp_potential[1:].tolist(),
"exp_current": exp_current[1:].tolist()
}
)
for label, active_idx, weights in stages:
tracker.set_stage(label, weights)
bounds = create_staged_bounds(current_x, active_idx, USER_CONFIG["v_min"], USER_CONFIG["v_max"], OPTIMIZER_CONFIG["num_globals"])
res = minimize(
scipy_objective,
current_x,
args=(weights,),
bounds=bounds,
jac=True,
method='L-BFGS-B',
callback=tracker,
options=optim_options
)
current_x = res.x
final_result = res
final_peaks = final_result.x[OPTIMIZER_CONFIG["num_globals"]:].reshape((-1, 3))
# Generate plot data
diffusivity = final_result.x[0] * OPTIMIZER_CONFIG["mult_diff"]
beta_left = final_result.x[1] * OPTIMIZER_CONFIG["mult_beta"]
beta_right = final_result.x[2] * OPTIMIZER_CONFIG["mult_beta"]
v_center = final_result.x[3]
baseline_offset = final_result.x[4] * OPTIMIZER_CONFIG["mult_offset"]
v_plot = np.linspace(USER_CONFIG["v_min"], USER_CONFIG["v_max"], 500)
beta_plot = np.where(v_plot < v_center, beta_left, beta_right)
d_of_v = diffusivity * np.exp(beta_plot * (v_plot - v_center)**2)
weights_p = final_peaks[:, 0, np.newaxis]
v_crits = final_peaks[:, 1, np.newaxis]
sharpnesses = final_peaks[:, 2, np.newaxis]
exp_terms = np.exp(-sharpnesses * (v_plot - v_crits))
dos_matrix = weights_p * sharpnesses * exp_terms / (1.0 + exp_terms)**2
dos_total = np.sum(dos_matrix, axis=0)
_, final_sim = compute_forward(jnp.array(final_result.x), jnp.array(global_weights))
result_data = {
"parameters": {
"diffusivity": float(diffusivity),
"beta_left": float(beta_left),
"beta_right": float(beta_right),
"baseline_offset": float(baseline_offset),
"v_center": float(v_center)
},
"plots": {
"v_plot": v_plot.tolist(),
"d_of_v": d_of_v.tolist(),
"dos_total": dos_total.tolist(),
"dos_matrix": dos_matrix.T.tolist(),
"exp_potential": exp_potential[1:].tolist(),
"exp_current": exp_current[1:].tolist(),
"sim_current": np.array(final_sim).tolist()
}
}
if loop and queue:
loop.call_soon_threadsafe(
queue.put_nowait, {
"type": "done",
"data": result_data
}
)
return result_data
|