图片详情
—
Skip to content

minimal_arrhenius.py ​

返回章节 · 配套代码说明

复制以下内容并保存为 minimal_arrhenius.py,使用方法见配套代码说明。

python
"""用五个速率常数拟合 Arrhenius 直线;仅依赖 Python 标准库。"""
import math

temperatures = [298.15, 308.15, 318.15, 328.15, 338.15]  # K
rate_constants = [3.46e-5, 1.35e-4, 4.98e-4, 1.50e-3, 4.87e-3]  # s^-1
x = [1000.0 / T for T in temperatures]
y = [math.log(k) for k in rate_constants]  # k 已按 s^-1 给出数值
m = len(x)

# 先把横坐标标准化,目标值仍使用 ln(k / s^-1)。
mean_x = sum(x) / m
std_x = math.sqrt(sum((xi - mean_x) ** 2 for xi in x) / m)
xs = [(xi - mean_x) / std_x for xi in x]

w_scaled, b_scaled = 0.0, 0.0
learning_rate = 0.1
for step in range(201):
    predictions = [w_scaled * xi + b_scaled for xi in xs]
    residuals = [yi - pi for yi, pi in zip(y, predictions)]
    loss = sum(ri ** 2 for ri in residuals) / (2 * m)
    if step in (0, 1, 2, 10, 50, 200):
        print(f"第 {step:3d} 步:J = {loss:.6f}")
    if step == 200:
        break
    # 两个梯度都用更新前的参数计算。
    grad_w = -sum(ri * xi for ri, xi in zip(residuals, xs)) / m
    grad_b = -sum(residuals) / m
    w_scaled -= learning_rate * grad_w
    b_scaled -= learning_rate * grad_b

# 还原为横坐标 1000/T 下的斜率、截距,再换算物理量。
w = w_scaled / std_x
b = b_scaled - w * mean_x
R = 8.314  # J/(mol K)
Ea_kJ_mol = -1000.0 * R * w / 1000.0  # 最后除以 1000 将 J 换为 kJ
A = math.exp(b)
print(f"斜率 = {w:.3f},截距 = {b:.3f}")
print(f"活化能 = {Ea_kJ_mol:.1f} kJ/mol,指前因子 = {A:.2e} s^-1")

《原子智能》· 纸质出版预备版 · PolyAI Team 著