The strength risks of aluminum die-castings hide where the eye can't reach—internal porosity and inclusions. A parts plant used DaoAI deep learning to auto-read films and fit X-ray judgment into line cadence.
Aluminum die-casting is the mainstream process for chassis, steering and new-energy structural parts, but the process readily forms internal porosity, shrinkage and inclusions. These defects are entirely invisible on the surface yet directly weaken a part's strength and air-tightness. This plant used X-ray for internal inspection of castings, but manual film reading was slow and subjective, and fatigue from long reading sessions made tiny pores easy to miss—hard to match the high cadence of a die-casting line.
Letting AI Read the X-ray Films
DaoAI brought deep-learning AI-AOI into the X-ray inspection step. Trained on large volumes of annotated inspection images, the model learns the grayscale and morphological signatures of internal defects such as porosity, shrinkage and inclusions, locating them automatically and issuing a reject decision by size, count and location. Per-image inference stays under 2 seconds, fitting directly into die-casting line cadence for full inline inspection rather than offline sampling.
- Under 2s per image: inference speed matches die-cast cadence, enabling full inline inspection
- Internal defect detection: porosity, shrinkage and inclusions auto-located and graded
- Unified reject criteria: quantified by size/count/location, removing subjective variance
- Line-deployable: integrates with X-ray equipment, ending offline manual film reading
Turn offline, experience-based, fatigue-prone manual film reading into online, quantified, tireless auto-judgment.
After deployment, internal defects in the plant's castings are screened fully inline, reading efficiency and consistency rose sharply, the risk of missed calls dropped notably, and the defect data gave quantified support for tuning die-casting parameters.