Skip to content

ML usage on the dataset #7

@Anushka281195

Description

@Anushka281195
  1. Statistical Methods You Should Use (Specific to Your Dataset)
    A. Descriptive Statistics (For Each Sensor Column)

Apply to these columns only:

Air temperature (K)

Process temperature (K)

Rotational speed (rpm)

Torque (Nm)

Tool wear (min)
Include:

Mean

Median

Standard deviation

Minimum

Maximum

Range

25th, 50th, 75th percentiles

Purpose: Understand how machines normally operate before failure occurs.

B. Temperature Difference Analysis

Your dataset specifically allows:

Temperature Difference = Process Temperature – Air Temperature

This is an important engineered statistical feature.

Questions it answers:

Does a larger temperature difference indicate heat-related failures?

At what difference threshold do failures start to appear?

C. Correlation Analysis (Sensor-to-Failure Relationships)

Compute Pearson correlation for:

Tool wear vs Target

Torque vs Target

RPM vs Target

Temperature difference vs Target

This shows which measurements contribute most to failure.

D. Group-Based Statistical Comparison

Use "groupby statistics" for these:

Groups:

Failure vs No Failure (Target)

Failure Types (Heat, Tool Wear, Overstrain, etc.)

Machine Type (L, M, H)

Compare their:

Average torque

Average RPM

Temperature difference

Tool wear levels

These comparisons reveal patterns specific to your dataset.

E. Outlier Detection (Distribution-Based)

Check outliers for:

Torque

RPM

Tool wear

Process temperature

Your dataset typically shows:

Torque has extreme peaks

Tool wear increases linearly but failure cases cluster at high values

This is essential for understanding machine behavior.

F. Hypothesis Testing (Specifically for Your Data)

  1. T-test

Compare two groups:

Mean torque in Failure vs Non-Failure rows

Mean tool wear in Failure vs Non-Failure rows

  1. ANOVA

Compare:

Mean tool wear across different failure types

Mean torque across machine types L/M/H

  1. Chi-square Test

For categorical patterns:

Is failure type dependent on machine type?

These tests directly validate the relationships in your dataset.

  1. Statistical Questions You Should Answer (SPECIFIC to Your Dataset)

Below is the exact question set for your predictive_maintenance.csv file.

A. Sensor Behavior Questions

What is the average air temperature during machine operation?

What is the typical process temperature, and how much higher is it than air temperature?

What is the average rotational speed (rpm) of the machines?

How much does torque fluctuate across production cycles?

What is the distribution of tool wear values?

B. Failure-Specific Statistical Questions

Do failed machine cycles show higher torque than normal cycles?

How different is tool wear in failure rows compared to non-failure rows?

Do failures occur at higher RPM ranges?

Does the temperature difference increase significantly before heat-related failure?

C. Machine Type (L/M/H) Statistical Questions

Do machine types (L, M, H) operate at different average torque levels?

Do some machine types show higher average tool wear?

Are certain machine types more prone to failure?

D. Failure Type Statistical Questions

Using the "Failure Type" column:

What is the average torque for each failure type (Heat, Tool Wear, Overstrain)?

What are the temperature patterns associated with heat-related failures?

Does overstrain failure occur at high RPM levels?

How much tool wear occurs before tool-wear failure?

E. Outlier and Distribution Questions

Are torque outliers linked to mechanical failure?

Do extremely high process temperatures precede heat dissipation failures?

Is tool wear heavily right-skewed, indicating gradual degradation?

F. Hypothesis Testing Questions

Is the mean tool wear significantly higher in failure cases compared to normal cycles?

Do different failure types have statistically different torque levels?

Is failure occurrence independent of machine type, or related?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions