The Problem We Are Solving: Helping Machine Tending Cobots Detect Anomalies

Is the CNC door open?

Is the part properly loaded in the CNC?

Manufacturers aiming for higher efficiency in machine tending have increasingly adopted collaborative robots (cobots) to automate tasks. However, these robots are not immune to malfunctions or unexpected issues, which can cause significant disruptions. Issues such as incorrect setups, tools left in machines, incorrect or improperly loaded stock, doors failing to open, or cobots malfunctioning can lead to costly unplanned downtime. The repetitive and monotonous nature of machine tending makes it particularly ripe for automation, but existing solutions still leave gaps in anomaly detection.

Current Alternatives to Detecting Anomalies in Machine Tending:

While manufacturers have adopted various strategies to address the challenges of anomaly detection, these alternatives come with limitations:

• Do Nothing: Relying solely on cobots without proper anomaly detection means manufacturers simply accept the consequences of when errors occur. This can range from a robot hitting a protective stop (resulting in wasted time) to damage (such as a tool breaking or a robot’s gripper breaking). Not to mention the opportunity cost of having to design around every possible failure mode in advance.
• Proximity and Pressure Sensors: Sensors are designed for specific tasks, like detecting an object’s presence, but they lack flexibility in dynamic environments. They often miss out on other potential failure cases, making them insufficient for fully automated systems.  And they can become their own failure point. ‍
• Custom Computer Vision Solutions: Most solutions are expensive and require customized models whenever the environment changes. Additionally, these models are often unreliable without human oversight, leading to further costs and inefficiencies in maintaining their effectiveness.  

Groundlight’s Solution: Anomaly Detection Software

Groundlight AI’s Anomaly Detection software addresses these issues by providing manufacturers with a robust, adaptable solution that integrates easily with both Universal Robots and other cobot systems. The AI-based system begins working from day one, offering real-time anomaly detection and escalation.

Here’s an example where a Universal Robot cobot picks up parts from a chute and places them into a CNC machine, and Groundlight AI is being used to check the various steps along the way. A camera is pointed at the CNC machine, and Groundlight AI’s computer vision technology is analyzing the camera’s images to ensure that everything is running smoothly.

Here’s a look at the various “checks” that Groundlight is performing while the robot is operating:

Groundlight's AI software adapts to changing environments through human guidance. When analyzing images - like checking vise setup - if the system encounters an unfamiliar situation (such as a tool left in the machine), it flags the image for immediate human review. This human feedback helps the system learn and improve its accuracy over time.

As a result, Groundlight AI provides manufacturers with the confidence to operate fully automated, unattended shifts, reducing downtime and allowing employees to focus on higher-value tasks.

Key Features of the Solution:

• Instant Setup and Real-Time Monitoring: Groundlight’s solution is ready to work from day one, with instant notifications if an issue arises. Any uncertainty is escalated in real-time to a human for review, ensuring accuracy. Human reviewers can be either Groundlight’s staff or your own staff.

• High Accuracy: Groundlight AI’s system achieves high accuracy due to 24/7 human reviewers, meaning fewer disruptions and better overall performance.
• Flexible Integration: Groundlight integrates not only with Universal Robots but also with other robot systems, making it adaptable across different manufacturing environments.
• Scalability: The solution works seamlessly across multiple locations and machine types without requiring costly customizations or new models for each machine.
• Human-Guided AI: The solution becomes increasingly autonomous over time as it learns from human responses to anomalies it has never encountered before, eventually requiring minimal human oversight.

Impact:

Manufacturers using Groundlight AI’s Anomaly Detection software experience significant improvements in machine tending operations, including:

• Reduced Downtime: By catching anomalies early, manufacturers can prevent unplanned downtime and keep machines running smoothly, even during unattended shifts.
• Increased Efficiency: With real-time anomaly detection, employees are freed from monotonous oversight tasks and can focus on more innovative and higher-value work.
• Seamless Integration: Groundlight’s solution adapts to changing environments and machine setups, allowing manufacturers to use the same software across different machines and locations without additional customization costs.
• Rapid Deployment: Groundlight AI’s solution is operational from day one, offering immediate value and minimal disruption during implementation.

How Do I Get Started?

Groundlight AI’s Anomaly Detection solution works with Universal Robots as well as other collaborative robots. Schedule a Demo with us today and learn how you can run a lights-out operation.