top of page

Developing Analysis Tools for Additive Manufacturing

 

Existing laser powder-bed fusion post print analysis tools are disjointed and ad-hoc. Our tool is looking to change that. 

Motivation

The revolutionary potential of additive manufacturing (AM) is currently limited by the lack of robust AM quality assurance capabilities. AM can only be implemented at an industrial-scale if businesses are able to ensure the structural integrity of the parts produced. In order to achieve this task, sensor networks can be installed on AM machines to collect data about a print job as it progresses. Once these sensor networks produce large quantities of data, both business executives and technical analysts will require an analysis platform to make sense of the data. It is imperative that this platform caters to the varying expertise of these user groups.

​

CCAM has outlined a need for a robust data analysis visualization tool to fulfil various quality assurance needs of researchers and partner corporations. By addressing this need with a set of Jupyter notebooks, it will be possible to integrate verbose documentation, explanation, and graphics alongside the underlying technical data analysis scripts. This approach will maximize the utility of the solution for both business and technical members of CCAM’s partner corporations. These partners include companies that utilize metal additive manufacturing in production and companies that produce metal additive manufacturing machines.

 

The team intends to create a tool that will synthesize and intuitively display large quantities of machine sensor data to enable users to make more informed decisions about their AM machines. Future applications of this project might include modifying the analysis pipeline to make recommendations about AM machine parameters, to identify build failure patterns, or to provide real-time failure alerts.

  1. To alpha test the proposed HDF5 sensor data organization structure and provide feedback

  2. To provide a mechanism to parse data from the HDF5 sensor data organization

  3. To provide a mechanism to group the parsed data by print layer

  4. To provide a reference implementation of a Jupyter notebook framework for utilizing all of the sensor groups in the sample data set, including:

    1. Infrared

    2. Acoustic Emissions

    3. High-Resolution Visible Light

    4. Photodiode

    5. Position

  5. To provide verbose documentation incorporated with the analysis tools

  6. To maintain a modular design strategy such that the solution is easily extensible to other sensor types and analysis needs beyond this specific application

  7. To iteratively seek and incorporate user feedback about the working prototypes produced

Project Bounds
The Importance of User Centered Design

User Center Design is critical to the development of successful analysis platform, It ensures that your product is addressing the root of the problem and enabling useful analysis. It also helps identify the systems boundaries and critical system components.

​

Understand root of problem

Identify 

useful analysis

Set

system

boundaries

Data Flow
Data Flow
Current Status

Completed:

  • Read the HDF5 file

  • Access sensor data points and metadata attributes

  • Partition data into layers using build plate position

  • Display IR images with slider

In progress:

  • Focusing on modularity and good software practices

  • Defining benign data

  • Filtering benign data

  • Exploring with existing MATLAB analysis

  • Interactive visualizations

bottom of page