← Selected work

Machine learning / Automotive

Using sensor data to improve welding

A summary of a project account published by Levintutu.

01 / The challenge

Start with the operating problem.

An automobile-parts manufacturer faced inconsistent welding quality despite having substantial sensor data from the production process.

02 / The approach

Connect the components.

The published approach cleaned missing values and outliers, then used welding parameters and quality information to train a model and recommend process changes.

03 / The practical lesson

Data preparation is part of the solution. Recommendations need to connect to the process parameters operators can change.

Discuss a similar challenge ↗