Industrial IoT System for Automated AI-Defect ​Detection of Concrete Sleepers

Status

Complete

Partners

• Swinburne University of Technology
• Sunset Sleepers

Outputs

FINAL REPORT

Videos

 

Summary

• This project developed an automated quality control system to replace manual inspections in concrete precast manufacturing.
• The project demonstrated that the system works successfully in a real production environment at full speed and without interrupting workflows.
• With more accurate defect detection, manufacturers reduce rework, waste and associated costs while broader adoption would address cost challenges and skills shortages in the precast industry.

The industry problem

For Sunsets Sleeper and the broader precast manufacturing industry, manual inspection remains the primary method for detecting defects and verifying dimensions and reinforcement.

Subjective and prone to error, manual inspection can produce different results from one inspector to the next. This method also limits production speeds.

More precise inspections would reduce defects, returns and rework, delivering significant cost savings.

With a digitised system, manufacturers would also benefit from data that would help them improve processes such as determining where and why defects occur.

The solution

The project developed an automated quality control system that integrates industrial internet-of-things, AI, machine vision and industrial sensors.

Deployed at Sunset Sleepers at full production speed, it delivered reliable quality assurance and generated data for process improvement.

In one month of processing 377,164 frames, the system achieved:

• 99.75% pattern classification accuracy
• 94.23% colour classification accuracy
• 93% major defect detection
• 100% crack detection
• 0.09% rebar detection error rate

The system also produced stable and reliable height measurements, but produced variable width measurements that revealed a need to stabilise equipment.

Impact

Having operationalised this system, Sunset Sleepers stands to save costs both in the short and long term.

Offering far greater precision than manual inspection, the system’s deployment is already reducing the incidence of defects and product returns.

The system’s pace has been able to increase product output while the ongoing data capture offers potential for problem solving and process improvement.

Workers performing manual inspection have been reassigned to higher-value tasks such as troubleshooting, supervising and training.

Beyond these benefits to Sunset Sleepers, broader industry adoption would support the precast industry in addressing its challenges with skills shortages and rising costs.

Next steps

The system has demonstrated many of its technical capabilities in a production environment, with a small number of issues still being addressed.

The width measurement station needs to be stabilised to ensure the system can take accurate width measurements.

More work is also needed to increase the system’s dataset to enable it to detect rare forms of defects.

The operations team are establishing procedures for what operators should do when issues such as system disruptions or hardware failures arise.