Python实现三台机器生产数据平滑折线图绘制求助

Got it, let's get those messy lines smoothed out! The problem you ran into with gaussian_filter1d is likely due to the None values in your filtered data (which become NaNs in numpy arrays) — the filter can't handle missing values properly by default. Here's a straightforward fix with step-by-step adjustments to your code:

Step 1: Handle Missing Values & Prepare Data for Smoothing #

First, we'll replace None with numpy's native np.nan, then filter out those NaN values to create clean datasets for smoothing. We'll also keep the corresponding Day-Shift labels so our x-axis stays aligned.

Step 2: Apply Gaussian Smoothing #

Use gaussian_filter1d from scipy.ndimage on the clean production values. Adjust the sigma parameter to control smoothness: higher values mean smoother lines, lower values keep more of the original data's shape.

Full Modified Code #

import numpy as np
import pandas as pd
from plotly.offline import iplot
import plotly.graph_objects as go
from scipy.ndimage import gaussian_filter1d  # Add this import

# Assume df1, df2, df3 are your preloaded DataFrames for each machine

# ----------------------
# Process Machine 1 Data
# ----------------------
# Filter Quality A production, replace non-A with NaN
b1 = np.where(df1['Quality'] == 'A', df1['Production'], np.nan)
# Get clean (non-NaN) x and y values
mask1 = ~np.isnan(b1)
x1_clean = df1['Day-Shift'][mask1]
y1_clean = b1[mask1]
# Apply Gaussian smoothing (sigma=1 is a good starting point)
y1_smoothed = gaussian_filter1d(y1_clean, sigma=1)

# ----------------------
# Process Machine 2 Data
# ----------------------
b2 = np.where(df2['Quality'] == 'A', df2['Production'], np.nan)
mask2 = ~np.isnan(b2)
x2_clean = df2['Day-Shift'][mask2]
y2_clean = b2[mask2]
y2_smoothed = gaussian_filter1d(y2_clean, sigma=1)

# ----------------------
# Process Machine 3 Data
# ----------------------
b3 = np.where(df3['Quality'] == 'A', df3['Production'], np.nan)
mask3 = ~np.isnan(b3)
x3_clean = df3['Day-Shift'][mask3]
y3_clean = b3[mask3]
y3_smoothed = gaussian_filter1d(y3_clean, sigma=1)

# ----------------------
# Create Plotly Traces
# ----------------------
traces = []
# Fix the name bug (your original code had all traces named 'Machine1')
traces.append({
    'x': x1_clean,
    'y': y1_smoothed,
    'name': 'Machine1',
    'line': {'color': 'red', 'dash': 'solid'}
})
traces.append({
    'x': x2_clean,
    'y': y2_smoothed,
    'name': 'Machine2',
    'line': {'color': 'blue', 'dash': 'solid'}
})
traces.append({
    'x': x3_clean,
    'y': y3_smoothed,
    'name': 'Machine3',
    'line': {'color': 'yellow', 'dash': 'solid'}
})

# ----------------------
# Plot the Smoothed Graph
# ----------------------
layout = go.Layout(
    title='Production meterage of Machine1/Machine2/Machine3 for Quality A',
    template='plotly_dark',
    xaxis=dict(autorange=True),
    yaxis=dict(autorange=True)
)
fig = go.Figure(data=traces, layout=layout)
iplot(fig)

Key Notes: #

• Adjust Smoothness: Tweak the sigma value in gaussian_filter1d (e.g., sigma=0.5 for less smoothing, sigma=2 for more).
• Fix Trace Names: Your original code had all three traces labeled "Machine1" — I corrected that to make the legend accurate.
• Handling Gaps: If you want to keep gaps in the line where there's no Quality A data (instead of dropping NaNs), you could use linear interpolation before smoothing, but dropping NaNs is simpler for clean, smooth lines where data exists.

内容的提问来源于stack exchange，提问作者user13419531