1. Introduction
In the foundry industry, the quality of castings is of utmost importance. One of the key factors affecting casting quality is the presence of defects in sand molds. Traditional methods of sand mold defect detection, such as manual inspection, are time – consuming, labor – intensive, and prone to human error. With the rapid development of technology, machine vision – based defect detection has emerged as a promising solution. This article delves into the application of machine vision in sand mold defect detection on hydrostatic automatic molding lines, exploring its principles, implementation, and benefits.
1.1 Background of the Foundry Industry
The foundry industry plays a crucial role in various sectors, including automotive, aerospace, and machinery manufacturing. Castings are used to produce a wide range of components, and their quality directly impacts the performance and reliability of the final products. Sand molds are commonly used in the casting process due to their cost – effectiveness and versatility. However, sand molds are prone to defects, such as ，These defects can lead to casting defects, resulting in increased production costs, reduced product quality, and potential safety hazards.
1.2 Limitations of Traditional Detection Methods
Manual inspection is the most common method for sand mold defect detection in the industry. Inspectors rely on their visual observation and experience to identify defects. However, this method has several limitations. Firstly, it is highly subjective, as different inspectors may have different standards for defect identification. Secondly, manual inspection is time – consuming, especially when dealing with a large number of sand molds. This can slow down the production process and increase labor costs. Thirdly, human fatigue can affect the accuracy of inspection, leading to missed defects. In addition, manual inspection is not suitable for real – time monitoring, which is essential for ensuring the quality of continuous production processes.
1.3 The Rise of Machine Vision Technology
Machine vision technology has made significant progress in recent years, driven by advancements in computer science, optics, and electronics. Machine vision systems can capture and analyze images of objects, enabling the detection of defects with high accuracy and speed. By using cameras, lighting systems, and image processing algorithms, machine vision can overcome the limitations of manual inspection. In the context of sand mold defect detection, machine vision offers the potential for automated, real – time, and objective defect detection, which can greatly improve the quality control process in the foundry industry.
2. Principles of Machine Vision – Based Sand Mold Defect Detection
2.1 Image Acquisition
The first step in machine vision – based sand mold defect detection is image acquisition. High – resolution cameras are used to capture images of sand molds. In the case of the sand mold defect detection system described in this article, two high – resolution cameras are combined, along with a ，to ensure clear and accurate image capture. The cameras are installed in a above a specific sand box position on the hydrostatic automatic molding line. The helps to eliminate external light interference and provides a stable lighting environment for image acquisition.
| Camera Parameter | Specification |
|---|---|
| Resolution | High – resolution (e.g., [X] megapixels) |
| Number of Cameras | 2 |
| Installation Location | Inside the above the sand box |
| Lighting Source |
2.2 Image Processing
Once the images are acquired, they need to be processed to extract relevant information. Image processing involves several steps, including image enhancement, segmentation, and feature extraction. In the sand mold defect detection system, the acquired images are first to obtain the detection area. Then, the images are pre – processed using techniques such as to reduce noise and improve image quality.
After pre – processing, the images are aligned. The alignment is achieved by calculating the between the left and right camera images. This ensures that the two images are in the same coordinate system, facilitating subsequent defect detection.
2.3 Defect Detection Algorithm
The core of the machine vision – based sand mold defect detection system is the defect detection algorithm. In this system, a combination of traditional non – deep learning methods and deep learning techniques is used.
First, the traditional method is used to compare the sand mold image with a . This comparison helps to identify the regions of interest that may contain defects. Once the regions of interest are determined, deep learning is employed for further verification.
Based on the balance between timeliness and accuracy required for sand mold defect detection, a lightweight convolutional neural network (MobileNet) is designed for the detection task. MobileNet is chosen because it can achieve relatively high accuracy while maintaining a low computational cost, making it suitable for real – time applications on the production line.
| Detection Algorithm Step | Method |
|---|---|
| Initial Defect Region Identification | Comparing with template image (traditional method) |
| Verification of Defect Region | Using MobileNet (deep learning) |
| Network Selection Consideration | Balance of timeliness and accuracy, low computational cost |
3. The Sand Mold Defect Detection System
3.1 System Architecture
The sand mold defect detection system consists of several components, including the image acquisition unit, the image processing unit, and the defect detection unit. The image acquisition unit is responsible for capturing images of sand molds, as described above. The image processing unit processes the acquired images to prepare them for defect detection. The defect detection unit uses the designed algorithm to identify and classify defects.
In addition, the system also includes a data storage and management unit. This unit is used to store the production data, defect images, and detection results. It enables the system to generate reports, perform statistical analysis, and provide data support for process improvement.
3.2 Function Interfaces
The sand mold defect detection system provides several user – friendly function interfaces, which facilitate the operation and management of the system.
- Production Data Visualization Interface: This interface displays real – time information about the sand mold production process, such as the current sand mold image, ， information, the number of detected sand molds, defects, and s. It also shows the detection date and time, as well as the history record. This interface allows operators to monitor the production process in real – time and quickly identify any potential issues.
- Suspicious Defect Processing Interface: When the system detects a potential defect in a sand mold, this interface pops up. It displays detailed information about the defect, including the defect image，detection date, and time, as well as the defect location. Operators can classify the defect as “normal sand mold”, , or and add it to the training set for the system to learn.
- Setting Detection Area Interface: This interface allows users to define the specific area of the sand mold that needs to be detected. By setting the detection area accurately, the system can focus on the critical parts of the sand mold, improving the efficiency and accuracy of defect detection.
- Sand Mold Defect Early Warning Interface: When the is changed, this interface shows the distribution of all defects on the current up to the last production in the form of a . It also allows users to view the number of defect occurrences at each position. By analyzing this information, operators can identify areas that are prone to defects and take preventive measures, such as maintenance and process optimization.
- Setting System Parameters Interface: This interface enables users to adjust various system parameters, such as camera settings, image processing parameters, and defect detection thresholds. By optimizing these parameters, the system can adapt to different production environments and improve the performance of defect detection.
- Exporting Detection Data Interface: This interface allows users to export the detection data, including production data, defect images, and detection results. The exported data can be used for further analysis, quality control reporting, and process improvement.
- Historical Waste Mold Display Interface: By double – clicking on the image in the production data visualization interface, users can access this interface. It provides detailed information about the ，such as the model, ，detection date, time, detection result, defect location, and the corresponding standard sand mold. Users can also zoom in and drag the sand mold defect image and the standard sand mold image for better viewing.
| Interface Name | Function Description |
|---|---|
| Production Data Visualization Interface | Display real – time production data and history |
| Suspicious Defect Processing Interface | Process detected defects and add to training set |
| Setting Detection Area Interface | Define the detection area of the sand mold |
| Sand Mold Defect Early Warning Interface | Provide defect distribution and early warning |
| Setting System Parameters Interface | Adjust system parameters |
| Exporting Detection Data Interface | Export detection data |
| Historical Waste Mold Display Interface | View detailed information |
4. Transformation of the Hydrostatic Automatic Molding Line
4.1 Data Interaction
To enable seamless communication between the sand mold defect detection system and the hydrostatic automatic molding line, a data interaction mechanism is established. In this case, the Snap7 开源通讯库 is used. Snap7 is an open – source library for Ethernet communication with Siemens S7 series PLCs. It supports communication with various PLC models, including S7 – 200, S7 – 200 Smart, S7 – 300, S7 – 400, S7 – 1200, S7 – 1500, and LOGO!(0BA 7/0BA8).
In the ，two DB data blocks are created. DB2006 is used for the sand mold defect detection system to read data, which includes current sand mold attributes, production data of the molding line, sand mold detection control signals, and equipment heartbeat signals. DB2007 is used for the sand mold defect detection system to write data, such as the state feedback of the detection system, detection results, and equipment heartbeat signals.
| Data Block | Function | Data Content |
|---|---|---|
| DB2006 | Read by the detection system | Sand mold attributes, production data, control signals, heartbeat signals |
| DB2007 | Written by the detection system | Detection system state, results, heartbeat signals |
4.2 Control Implementation
The control implementation of the sand mold defect detection system on the hydrostatic automatic molding line involves a specific control flow. When the sand mold defect detection function is enabled on the molding line, after each sand box is pushed in the molding section, the system queries the information of the specified sand box. It checks the state of the sand box. If the sand box is in a “good” state and is not empty, the system locks the molding section push and starts the sand mold defect detection. If the sand box is a or empty, the system skips the detection and does not lock the molding section push.
After the detection is completed, the system determines the next step based on the detection result. If the sand mold is identified as a ，it needs to be marked as a and then the molding section lock is released. If the sand mold is a good one, the molding section lock is directly released.
The sand box information query is achieved by the sand box attribute shift register data in DB91 of the molding line program. The relevant sand box information is then transferred to DB2006 and updated as the sand box is pushed, for the sand mold defect detection system to read.
The is also an important part of the control implementation. When the sand mold defect detection system returns a signal, the information needs to be matched with the corresponding sand mold based on the 、 information of the current sand mold and then written into the corresponding data bit in DB91.
| Control Step | Operation |
|---|---|
| Enable Detection | Check if the sand box is ready for detection |
| Detection Trigger | Lock molding section push and start detection for valid sand boxes |
| Result Handling | Mark s and release molding section lock based on results |
| Information Query | sand box information from DB91 and transfer to DB2006 |
| Waste Mold Identification | Match and write information in DB91 |
5. Benefits and Challenges of Machine Vision – Based Sand Mold Defect Detection
5.1 Benefits
- Improved Detection Accuracy: Machine vision – based sand mold defect detection can achieve higher accuracy compared to manual inspection. The combination of high – resolution cameras and advanced image processing algorithms, along with deep learning – based defect verification, can identify even small and subtle defects that may be missed by human inspectors. This helps to ensure that only high – quality sand molds are used in the casting process, reducing the occurrence of casting defects.
- Increased Productivity: The automated nature of the machine vision system allows for continuous and rapid defect detection. It can inspect a large number of sand molds in a short time, without being affected by human fatigue. This significantly increases the production efficiency of the foundry, as the production line can operate at a faster pace without sacrificing quality.
- Real – Time Monitoring: The system can provide real – time feedback on the quality of sand molds during the production process. This enables operators to take immediate corrective actions if defects are detected, such as adjusting the molding process or replacing faulty s. Real – time monitoring helps to prevent the production of a large number of defective castings, saving time and resources.
- Data – Driven Process Improvement: The sand mold defect detection system stores a large amount of data, including production data, defect images, and detection results. By analyzing this data, manufacturers can identify patterns and trends in defect occurrences. This data – driven approach can be used to optimize the molding process, improve 型板 design, and implement preventive maintenance measures, leading to continuous improvement in product quality and production efficiency.
| Benefit | Description |
|---|---|
| Improved Detection Accuracy | High – resolution cameras and advanced algorithms for accurate defect identification |
| Increased Productivity | Automated and fast detection, not affected by human fatigue |
| Real – Time Monitoring | Immediate feedback for corrective actions during production |
| Data – Driven Process Improvement | Analyze data to optimize processes and prevent defects |
5.2 Challenges
- Initial Investment: Implementing a machine vision – based sand mold defect detection system requires a significant initial investment. The cost includes the purchase of high – resolution cameras, lighting systems, image processing hardware, and software development. In addition, the cost of integrating the system with the existing hydrostatic automatic molding line, such as PLC programming and communication infrastructure, also needs to be considered. This initial investment may be a barrier for some small and medium – sized foundries.
- Complexity of System Integration: Integrating the machine vision system with the existing molding line can be a complex task. It requires a good understanding of both the machine vision technology and the operation of the molding line. The communication between different components, such as the cameras, image processing unit, and PLC, needs to be carefully configured to ensure seamless data transfer and system operation. Any errors or misconfigurations during integration can lead to system malfunctions.
- Model Adaptation: The performance of the machine vision – based defect detection system depends on the quality of the training data. As the production process may change over time, for example, due to changes in sand mold materials, molding processes, or designs, the system needs to be continuously updated and adapted. This requires regular collection and annotation of new data, as well as retraining of the deep – learning models, which can be time – consuming and resource – intensive.
| Challenge | Description |
|---|---|
| Initial Investment | High cost of equipment and system integration |
| Complexity of System Integration | Difficulties in integrating with existing molding line |
| Model Adaptation | Need for continuous model update due to production changes |
6. Future Trends in Sand Mold Defect Detection
6.1 Integration with Artificial Intelligence and Machine Learning Advancements
As artificial intelligence and machine learning continue to evolve, future sand mold defect detection systems are likely to incorporate more advanced techniques. For example, more powerful deep – learning architectures, such as convolutional neural networks with higher accuracy and faster processing speeds, may be developed specifically for sand mold defect detection. Additionally, techniques such as transfer learning can be used to leverage pre – trained models on large – scale image datasets, reducing the need for extensive data collection and training.
6.2 Multi – Sensor Fusion
To further improve the accuracy and reliability of defect detection, future systems may integrate multiple sensors. In addition to machine vision cameras, sensors such as laser scanners, ultrasonic sensors, or X – ray detectors can be used to obtain additional information about the sand mold. For example, laser scanners can provide 3D information about the surface of the sand mold, which can help to detect defects that are not easily visible in 2D images. Multi – sensor fusion can provide a more comprehensive view of the sand mold, enabling more accurate defect detection and classification.
6.3 Industry 4.0 and Smart Factory Integration
With the development of Industry 4.0, sand mold defect detection systems are expected to be more closely integrated into smart factory ecosystems. The data collected from the defect detection system can be shared with other production systems, such as enterprise resource planning (ERP) systems and manufacturing execution systems (MES). This integration enables real – time monitoring and control of the entire production process, from sand mold preparation to casting production. It also allows for better coordination between different departments, such as production, quality control, and maintenance, leading to more efficient and intelligent manufacturing operations.
| Future Trend | Description |
|---|---|
| Integration with AI and ML Advancements | Use of more advanced deep – learning architectures and transfer learning |
| Multi – Sensor Fusion | Combine multiple sensors for more comprehensive defect detection |
| Industry 4.0 and Smart Factory Integration | Share data with other production systems for intelligent manufacturing |
7. Conclusion
Machine vision – based sand mold defect detection technology represents a significant advancement in the foundry industry. By replacing traditional manual inspection methods with automated, accurate, and real – time defect detection systems, manufacturers can improve the quality of their castings, increase productivity, and reduce production costs. Although there are challenges associated with the implementation of these systems, such as initial investment and system integration complexity, the benefits far outweigh the costs in the long run.
As technology continues to evolve, the future of sand mold defect detection looks promising. With the integration of artificial intelligence, multi – sensor fusion, and Industry 4.0 concepts, we can expect even more advanced and intelligent defect detection systems in the coming years. These advancements will not only benefit the foundry industry but also contribute to the overall development of the manufacturing sector towards more efficient and sustainable production.
In conclusion, the adoption of machine vision – based sand mold defect detection technology is a crucial step towards achieving higher – quality casting production and driving the digital transformation of the foundry industry.
