Advanced Concrete and Construction Materials for Sustainable Structures

Mahdi Kioumarsi,Vagelis Plevris

DOI: https://doi.org/10.3390/su16041427

IF: 3.9

2024-02-09

Sustainability

Abstract:Innovation in structural engineering has sparked remarkable advancements in the building materials sector and the construction industry in general [...]

environmental sciences,environmental studies,green & sustainable science & technology

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper aims to address issues present in fault detection systems (such as Crimp Force Monitoring, CFM) during the wire harness manufacturing process. Specifically:

1. **Limitations of Traditional CFM Systems**:
- High Cost: Requires specialized equipment installation, significant investment for each production line, and high maintenance costs.
- Frequent Calibration: Needs recalibration every time the production process changes, relies on operator skills, and is prone to human error.
- Lack of Flexibility: As production volume and product variety increase, the system may struggle to maintain efficiency and accuracy.

2. **Challenges of Applying Artificial Intelligence in Fault Detection**:
- Data Variation: Changes in the manufacturing process can lead to data scale variations, causing existing AI models to fail.
- Insufficient Labeled Data: There is a lack of labeled defect data in actual manufacturing systems for training AI models, especially deep learning algorithms that require large amounts of data and computational resources.
- Limited Performance of Anomaly Detection Algorithms: Algorithms like Isolation Forest, which are trained only on normal data, may not ensure accurate detection of all potential faults.

To address these challenges, the paper proposes an AI-based fault detection system using Regional Selective Data Scaling (RSDS). This system enhances AI model training by generating synthetic anomaly data from normal data, achieving efficient and accurate fault detection. The specific steps are as follows:

1. **Data Collection**: Manually collect a small amount of normal production data as reference data.
2. **Data Augmentation**: Use RSDS to perform specific area up/down scaling on the reference data to generate synthetic anomaly data.
3. **Model Training**: Train the enhanced dataset using a Multilayer Perceptron (MLP) model.
4. **Fault Detection**: Use the trained model for real-time fault detection in subsequent production data.

This method addresses the issue of training models with only a small amount of normal data and significantly improves the accuracy and efficiency of fault detection.