X-ray Detection of Porosity and Inclusions in Aluminum Die - Castings: Line - Level Scrap Judgment in <2 Seconds per Image

In the production of aluminum die-castings, internal defects such as porosity and inclusions seriously affect product quality. DaoAI by WeLinkirt uses advanced technology to achieve fast and accurate X-ray detection, bringing a new solution to the industry.

Industry Background and User Scenarios: Aluminum alloy die-casting is the mainstream process for chassis, steering, and new-energy structural components, playing a crucial role in industrial production. However, during the die-casting process, due to various factors such as the flow characteristics of molten metal, mold design, and process parameters, it is easy to form internal defects such as porosity, shrinkage, and inclusions in die-castings. These defects are completely invisible on the surface but directly weaken the strength and airtightness of parts, posing great potential risks to product performance and quality. For component manufacturers, ensuring product quality meets standards is of utmost importance. They need to conduct internal flaw detection on die-castings to find these hidden defects and prevent unqualified products from entering the market. The traditional detection method is to use X-ray for flaw detection and then manually read the X-ray images for judgment. However, with the expansion of production scale and the acceleration of production rhythm, this traditional method has gradually become difficult to meet the needs of enterprises.

Deep - Dive into Pain Points: Why It's Difficult

In terms of speed, the slow speed of manual image reading is a significant problem. Generally, it may take several minutes or even longer for a person to read a single X-ray image, which is completely out of sync with the high-rhythm production of the die-casting line. A die-casting line can usually produce several die-castings per minute, while the speed of manual image reading is far from keeping up with this pace, resulting in low production efficiency. For example, if a production line produces 5 die-castings per minute and it takes 3 minutes for each manual image reading, a large number of products will be积压, affecting the entire production process.

In terms of accuracy, there is a large subjective difference in manual image reading. Different inspectors may have different judgment criteria for defects, and even the same inspector is prone to fatigue after long-term image reading, leading to the omission of small pores. According to statistics, the omission rate of manual image reading may reach 10% -15%, which means that a considerable number of defective products may be misjudged as qualified and enter the market, bringing potential quality risks to the enterprise.

From a cost perspective, manual image reading requires a large amount of manpower input, and in order to ensure the accuracy of detection, professional training for inspectors is also needed, which increases the enterprise's labor cost and training cost. In addition, due to the low efficiency of manual image reading, additional detection time and equipment may be required, further increasing the enterprise's production cost.

The root cause of these problems lies in the fact that manual image reading is a method based on experience and subjective judgment. Human vision and cognitive abilities have certain limitations. When faced with a large number of X-ray images and small defects, it is easy to cause fatigue and misjudgment. Moreover, it is difficult to unify the standards of manual image reading, and different people may draw different conclusions due to differences in personal experience and judgment habits, which leads to the inconsistency and inaccuracy of detection results.

Technical Principle

WeLinkirt's DaoAI introduces deep-learning AI-AOI into the X-ray detection process. Its core algorithm is based on the convolutional neural network (CNN), which can automatically learn features and patterns from a large amount of data. First, the model is trained on a large number of annotated flaw-detection images, which contain various types of internal defects such as porosity, shrinkage, and inclusions. By learning the grayscale and morphological features of these defects, the model can accurately identify and locate them. In terms of imaging, the X-ray device can penetrate the die-casting and obtain an image of the internal structure. These images are transmitted to the DaoAI system, and the system uses the trained model to analyze and process the images.

Compared with the traditional manual image-reading method, DaoAI has obvious advantages. The traditional method relies on human experience and subjective judgment, while DaoAI is an objective analysis based on data and algorithms, which can avoid the influence of subjective differences and fatigue. Moreover, DaoAI has a very fast reasoning speed, and the reasoning time for a single image can be controlled within 2 seconds, which enables it to be directly embedded in the die-casting production line rhythm and achieve online full inspection of internal defects. In contrast, the traditional manual image reading can only perform offline sampling inspection, which cannot meet the needs of large-scale production.

Typical Application Scenarios

• Chassis Die - Casting Detection: The chassis is an important part of an automobile, and its quality is directly related to driving safety. In the detection of chassis die-castings, DaoAI can quickly and accurately detect internal defects such as porosity, shrinkage, and inclusions. The difficulty lies in the complex structure of the chassis die-casting, where defects may be hidden in various parts, and some small defects may have a significant impact on product performance. DaoAI can accurately locate and identify these defects in the complex structure through its high-precision algorithm and fast reasoning ability.
• Steering System Die - Casting Detection: The die-castings of the steering system need to have high-precision performance, and any internal defects may affect the accuracy and reliability of steering. DaoAI can conduct a comprehensive inspection of the steering system die-castings, automatically locate the defects, and give a scrap judgment based on the size, quantity, and location. The difficulty in detection is that the steering system die-castings have high precision requirements, and some small pores and inclusions may be difficult to detect. DaoAI uses the powerful feature extraction ability of its deep-learning model to detect defects at the micron level.
• New - Energy Structural Component Die - Casting Detection: The rapid development of new-energy vehicles has put forward higher requirements for the quality of structural components. Defects such as porosity and inclusions in the die-castings of new-energy structural components may affect the stability and safety of the battery system. DaoAI can detect these defects in real-time online to ensure product quality. The difficulty lies in the special materials and processes of new-energy structural components, and the morphology and characteristics of defects may be different from those of traditional die-castings. Through a large number of sample trainings, DaoAI can adapt to different materials and processes and accurately detect defects.
• Engine Component Die - Casting Detection: The engine is the core component of an automobile, and the quality of its components is crucial. DaoAI can detect defects such as porosity and inclusions in engine component die-castings to ensure the normal operation of the engine. The difficulty in detection is that the engine components operate in a harsh environment and have extremely high requirements for strength and airtightness. Any small defect may lead to serious consequences. DaoAI can effectively detect potential defects with its highly sensitive detection ability.

Implementation Case

A medium-sized component factory mainly produces aluminum die-castings for automobile chassis, steering systems, and new-energy structural components. The factory previously used the traditional manual image-reading method for X-ray detection. However, with the increase in orders and the acceleration of the production rhythm, the problems of inefficiency and inaccuracy of manual image reading became increasingly prominent. After introducing the DaoAI solution of WeLinkirt, the implementation process was relatively smooth. First, the WeLinkirt team connected and debugged the factory's X-ray equipment to ensure that the system could obtain images normally. Then, they used the existing flaw-detection image data of the factory to train the model specifically, making it better adapt to the product characteristics of the factory. After a period of testing and optimization, the DaoAI system was officially put into operation.

Before the implementation, the efficiency of manual image reading in the factory was low, with only about 200 die-castings being detected per day. The omission rate reached 12%, and the consistency of the reading results was poor. After the implementation, the DaoAI system achieved online full inspection of internal defects. The reasoning time for a single image was less than 2 seconds, and the number of die-castings that could be detected per day increased to more than 1000. The omission rate was reduced to 2%, and the consistency of the reading results was greatly improved.

DaoAI transforms the detection of aluminum die-castings from offline, experience-based manual image reading to online, quantitative automatic scrap judgment, improving production efficiency and product quality.

WeLinkirt's Solution and Product

WeLinkirt's DaoAI solution mainly includes a deep-learning AI-AOI model, a data-processing system, and a docking module with X-ray equipment. The deep-learning AI-AOI model is the core of the entire system. After being trained on a large number of annotated images, it can accurately identify internal defects such as porosity, shrinkage, and inclusions and give a scrap judgment based on the size, quantity, and location. The data-processing system is responsible for pre-processing, feature extraction, and analysis of X-ray images to ensure that the model can process data efficiently and accurately. The docking module with X-ray equipment realizes the smooth connection between the system and the X-ray equipment, enabling the images to be transmitted to the DaoAI system in real-time for processing.

This product has the following features: First, it has a fast reasoning speed, with the reasoning time for a single image being less than 2 seconds, which can match the high-rhythm of the die-casting production line. Second, it can accurately identify defects and automatically locate and classify internal defects such as porosity, shrinkage, and inclusions. Third, the scrap-judgment standard is unified, with quantitative judgment based on size, quantity, and location, eliminating subjective differences. Fourth, it can be deployed on the production line and can be connected to various X-ray equipment, getting rid of offline manual image reading.

Quantitative Results

By adopting the DaoAI solution of WeLinkirt, the enterprise has achieved significant quantitative results in multiple aspects. In terms of detection efficiency, the detection speed has increased by more than 400%, from detecting 200 die-castings per day to more than 1000, greatly shortening the detection cycle and improving production efficiency. In terms of accuracy, the omission rate has been reduced from 12% to 2%, and the false-alarm rate has been reduced by 30%, effectively improving the quality and reliability of products. In addition, the defect data also provides quantitative support for the adjustment of die-casting process parameters, helping the enterprise optimize the production process and further improve product quality and production efficiency.

This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.