What We Did

• Developed an AI system to detect whether components had undergone the shot peening process

• Designed a GPU-based architecture with a TensorFlow-trained model to classify parts in real time

• Integrated existing camera equipment for automated image capture and inspection

• Built a user-friendly application with a React-based interface to display alerts and log outputs for quality assurance and IoT integration

The Challenge

Shot peening is a critical process in manufacturing to strengthen metal parts, but identifying whether a component has been properly peened can be difficult and time-consuming when relying solely on manual inspection. Manufacturers needed a solution that could leverage existing imaging equipment to accurately distinguish between peened and un-peened parts in real time.

Without such a system, quality control remained prone to inefficiencies, higher defect rates, and slower production.

Lifescale Analytics’ Solution

Lifescale Analytics created an AI-driven application to enhance quality assurance in the manufacturing process. Using images captured by existing cameras, a TensorFlow-trained model running on GPU hardware rapidly classified parts as peened or un-peened with 100% accuracy in testing.

​

The application included a React-based interface that provided operators with instant positive or negative alerts and logged results into a quality assurance database. Integration with IoT systems enabled further automation, including the ability to trigger downstream processes.

​

With detection times averaging between 200–300 milliseconds (and under 100 milliseconds with newer hardware), the system significantly improved the speed and consistency of inspections.

Impact

The AI-driven inspection system delivered immediate benefits by:

• Achieving 100% accuracy in identifying peened vs. un-peened parts during testing

• Reducing inspection times from seconds to milliseconds

• Improving overall manufacturing productivity and quality assurance

• Laying the foundation for future features such as component grading, predictive maintenance, and digital twin integration