Case Study: Doosan Heavy Industries & Construction achieves 98% AI model accuracy with Google Cloud Platform

Doosan Heavy Industries & Construction Identifying defective welds and extending manufacturing equipment lifespan with Google Cloud

Doosan Heavy Industries & Construction, a Korea-based engineering, procurement, and construction contractor, needed a faster and more accurate way to inspect welds in industrial facilities. Its existing AI-based non-destructive testing model struggled with long training cycles and limited computing power, and it turned to Google Cloud Platform with GPUs on Compute Engine to scale the workload.

Google Cloud Platform enabled Doosan Heavy Industries & Construction to retrain the model much faster, cutting training time from about 20 days to a few hours and improving weld-defect detection accuracy to 98%. The solution also supports continuous learning for different weld shapes and defect types, and Doosan plans to deliver it as a SaaS offering through Google Cloud Platform.

Doosan Heavy Industries & Construction

Jang Se-young

Digital Innovation Head