Achieving 100% Inspection in Drilling EDM with Virtual Metrology

Introduction: What is EDM Virtual Metrology?

In today's highly competitive manufacturing industry, improving production efficiency and ensuring product quality are critical challenges. As technology advances, Electrical Discharge Machines (EDM) have become essential tools for machining difficult-to-process materials. In this field, virtual metrology technology has gradually become a key solution for achieving 100% inspection during machining.

EDM Virtual Metrology is a technology that uses data from the machining process to predict machining quality. Unlike traditional quality inspections that directly measure the workpiece, virtual metrology predicts quality indicators—such as hole diameter, roundness, surface roughness (Ra), and upper diameter—by collecting and analyzing critical data during the machining process. This method can significantly reduce the time required for quality inspection and decrease reliance on physical product inspection.

Virtual metrology relies on machine learning models that learn the relationship between the collected feature data during the machining process and the final product quality. Commonly used models include Convolutional Neural Networks (CNN), Multi-Input Convolutional Neural Networks (MICNN), and Multi-Input Multi-Layer Perceptrons (MIMLP). These models can accurately predict quality by inputting EDM feature values.

How Does EDM Virtual Metrology Work?

To define the inputs and outputs of a virtual metrology model, the following is explained:

Inputs (Feature Data):
The feature data input into the machine learning model includes the following variables:

1. Average Spark Frequency 
2. Average Discharge Current Pulse Duration 
3. Average Peak Current Discharge 
4. Average Discharge Energy 
5. Average Ignition Delay Time 
6. Average Gap Voltage 
7. Open Circuit Ratio 
8. Average of Laser Position
9. Standard Deviation of Laser Position

Outputs (Independent Variables):
The output of the machine learning model is quality measurement items, including:

1. Diameter
2. Roundness
3. Surface Roughness (Ra)
4. Upper Diameter

Supervised Learning and Prediction in Models

1. Data Collection and Preprocessing: Collect feature data and corresponding quality measurement data from the EDM process, remove outliers, and perform normalization, data filling, and smoothing.
2. Model Training: Train the machine learning model using labeled data (feature data and corresponding quality measurement data). The model learns the relationship between the feature data and quality measurement data.
3. Prediction and Evaluation: Use the trained model to make predictions and evaluate the model's performance by calculating errors such as MAE and MAPE.

Applications of Virtual Metrology in EDM

1. Enhancing Production Efficiency:
   The application of virtual metrology technology can significantly improve production efficiency. Traditional quality inspections require substantial time and resources, especially when each product needs detailed measurement. Virtual metrology can predict quality without disrupting the production process, reducing the dependency on physical inspections. Additionally, virtual metrology can provide real-time warnings when defective products are detected, allowing operators to adjust immediately and prevent the production of large quantities of non-conforming products.

2. Reducing Production Costs:
   Virtual metrology not only improves efficiency but also reduces production costs. By reducing reliance on physical quality inspections, companies can save on measurement equipment and labor costs.

3. Improving Product Quality:
   Virtual metrology technology helps manufacturers ensure product consistency and reliability by monitoring and predicting machining quality. This is especially important for industries with high-quality requirements, such as aerospace and medical devices. By using virtual metrology technology, manufacturers can better control the variables in the production process, ensuring that each batch of products meets the expected quality standards.

4. Promoting Smart Manufacturing:
   As Industry 4.0 advances, smart manufacturing has become the future direction of the manufacturing industry. Virtual metrology, as part of smart manufacturing, can achieve automation and digitization of the production process. By integrating IoT and big data analytics, virtual metrology can seamlessly connect with other manufacturing processes, building an intelligent production system.

Challenges in Implementing EDM Virtual Metrology

Although virtual metrology technology offers numerous advantages, there are still challenges in its implementation. Firstly, collecting and processing large amounts of high-frequency data requires efficient data processing capabilities and storage systems. Secondly, the accuracy of machine learning models depends on a large amount of training data. Currently, effective models are easier to establish under consistent processing programs, making them less suitable for diverse applications in contract manufacturing.

This article aims to introduce the application of virtual metrology technology in drilling EDM machines and how this technology can help improve production efficiency, reduce costs, and ensure product quality. If you are interested in the application of virtual metrology technology in the manufacturing industry, we welcome further discussions.


Article Source: Chinese Culture University, Denata Rizly Alimadji