Related Experiment Video
Updated: Sep 5, 2025
08:27
Author Spotlight: Efficient Image Recognition Using Directional Gradient Histogram Technique and Support Vector Machines
Published on: January 5, 2024
1.2K
Damage Progress Classification in AlSi10Mg SLM Specimens by Convolutional Neural Network and k-Fold Cross Validation
Claudia Barile1, Caterina Casavola1, Giovanni Pappalettera1
1Dipartimento di Meccanica, Matematica e Management, Politecnico di Bari, Via E. Orabona 4, 70125 Bari, Italy.
Materials (Basel, Switzerland)
|July 9, 2022
Summary
This study uses Acoustic Emission (AE) and Convolutional Neural Networks (CNNs) to identify damage stages in Selective Laser Melting (SLM) AlSi10Mg. The AI model accurately classifies signals from elastic, plastic, and fracture stages.
Area of Science:
- Materials Science and Engineering
- Additive Manufacturing
- Artificial Intelligence in Materials Testing
Background:
- Selective Laser Melting (SLM) is a key additive manufacturing technique for producing AlSi10Mg alloys.
- Understanding material damage evolution is crucial for ensuring the structural integrity and performance of SLM-processed components.
- Traditional methods for damage assessment can be time-consuming and may not capture real-time material behavior.
Purpose of the Study:
- To develop and validate an automated method for identifying damage evolution stages in AlSi10Mg specimens.
- To investigate the effectiveness of the Acoustic Emission (AE) technique combined with Convolutional Neural Networks (CNNs) for real-time material damage analysis.
- To classify AE signals corresponding to distinct damage phases: elastic, plastic, and fracture.
Main Methods:
- AE signals were recorded during mechanical testing of SLM-manufactured AlSi10Mg specimens.
- Continuous Wavelet Transform (CWT) spectrograms were employed to process and extract time-frequency features from AE signals.
- A modified SqueezeNet CNN architecture was trained using the processed AE signals from different damage stages.
- K-fold cross-validation was implemented to enhance the classification accuracy and robustness of the CNN model.
Main Results:
- The study successfully identified three distinct damage evolution stages (elastic, plastic, fracture) using AE signal analysis.
- The trained SqueezeNet CNN demonstrated high accuracy in classifying AE signals associated with each damage stage.
- The combination of AE technique and CNNs proved effective for real-time monitoring of material damage in SLM AlSi10Mg.
Conclusions:
- AE signal analysis, coupled with CNN-based classification, offers a powerful approach for non-destructive evaluation of material damage in SLM parts.
- The developed methodology can contribute to improved quality control and reliability assessment in additive manufacturing.
- Further research can explore the application of this technique to other materials and manufacturing processes.
