Master Thesis - Data Analysis and Machine Learning for optimization of testing process

Your role

Saltus, an Atlas Copco Group brand, designs and manufactures Geared Front Attachments (GFA) which are used to access difficult to reach tightening applications in various industries. The GFA are attached to Atlas Copco and Desoutter industrial tightening tools. 
To verify the quality, performance and durability, The GFA is subjected to extensive testing and run-in. These processes generate significant volumes of data. Today this data is used primarily for verification rather than product development. By applying modern data analysis and machine learning methods, there is potential to extract valuable insights that can improve both product development and manufacturing processes.

Purpose:

The purpose of this thesis is to explore how data from factory final testing and run-in can be analyzed using Exploratory Data Analysis (EDA) to identify parameters of interest to reduce run-in time and machine learning to identify patterns and predictive evaluation to determine how the GFA will behave during run-in. The focus is on understanding how different parameters such as temperature, torque and number of revolutions affect efficiency and performance, as well as comparing different GFA models to find correlation between models and run-in process.

 
Objectives:

• Collect and structure data from final testing and run-in rigs.
• Explore correlations between run-in and final testing.
• Analyze how various parameters like torque and temperature influence efficiency.
• Develop a model or variable to represent when a GFA has stable efficiency.
• Identify patterns and relationships between parameters for different GFA models.
• Evaluate and apply machine learning methods to draw conclusions and create predictive models.

Methodology: 

• Structuring data and preprocessing from internal systems.
• Exploratory Data Analysis (EDA) to identify relevant variables.
• Application of machine learning techniques such as regression analysis, classification, and clustering.
• Visualization of results and insights.
• Validation of models against real test results.

Expected outcome:

• A report describing an identified relationship between run-in parameters.
• Recommendations for more effective use of test data.
• A specification of when a GFA can be classified as having stable efficiency.

To succeed, you will need

• MSc in Data Engineering, Engineering Physics or similar.
• Good knowledge of Matlab and Python.
• Experience with machine learning techniques and model evaluation techniques.
• Experience with Python frameworks such as PyTorch.
• Experience from working with noisy datasets and performing data cleaning and preprocessing.
• General interest in mechatronics, data analysis, testing, and production processes.

Contact information

For more information, please contact mechatronics engineer Eric Marcus or manager Mattias Larsson. 
mattias.larsson@atlascopco.com, 0790 98 65 08 
eric.marcus@atlascopco.com, 0733 16 14 42