AI advances Robotics Automation in the Home Appliance Sector

The Solution Layout., copyright Mech-Mind Robotics

Case Background

A company, Turkey’s Lovemark and the leader of home appliances, engages in the production of durable goods with innovative, eco-friendly, superior design and technology. It has a wide product range, including white goods, built-in appliances, electronics, small appliances, water and home heaters, and air conditioners.

At A company’s factory, the halogen detection of air compressors, the core component of air conditioners, is a tricky task because it requires the ability to repeat the same movement over and over again for several hours with exactly the same precision.

A company wanted robots to undertake this monotonous task. The benefits are apparent. Robots can add efficiency and throughput to the production line because they can work 24/7, non-stop. However, A company found that robots were still limited when conducting halogen detection.

Project Info and Challenges

Welding points to be detected in each compressor are random in size, position, and orientation. Halogen leak detection requires extreme accuracy and the workspace is quite tight and occupied with tangled condenser pipes (See Figure 1), so robots need to exactly “see” all the welding points and “understand” their surroundings, then “plan” an optimum path to avoid collisions, and thus they can finally conduct a successful inspection.

This is where Mech-Mind AI + 3D vision solution comes to play. Mech-Mind AI + 3D vision solution comprising machine vision, motion planning, and deep learning technology empowers robots with greater accuracy and dexterity, even in the face of high-precision detection and inspection tasks.

The Mech-Mind Solution

Solution Components: Mech-Mind introduced a high-accuracy industrial 3D camera Mech-Eye PRO S and AI-powered software suites–machine vision software Mech-Vision, robot control software Mech-Viz, and deep learning software Mech-DLK, paired with a UR5e robot, to optimize halogen leak detection.

The Mech-Mind AI + 3D vision solution is easy to set up and integrate and gives A company agility to handle product variations and fast changeovers. Thanks to Mech-Mind AI + 3D solution, A company has upgraded the halogen detection process, increased detection consistency, reduced waste and costs, and improved productivity.

Technical Requirements

• The 3D vision system should accurately recognize all welding points varying in position and orientation in AC condensers.
• The 3D vision system can differentiate high- and low-pressure welding points. (There is one high-pressure welding point and three low-pressure welding points in each compressor.) The former does not need to be inspected, while the latter needs inspection.
• The vision-guided robot can conduct the inspection in a tight workspace with high agility.

Technical Challenges

• Difficulty in generating high-quality point clouds: Reflections from the background can cause unwanted noises and lead to missing data in point clouds. (See Figure 3)
• Difficulty in accurate recognition: The occupied workspace increases the possibility of misrecognition. (See Figure 1) Welding points come in various sizes, shapes, and positions; welding points without detection requirements should be excluded during leak detection. A 100% recognition accuracy is a challenge for the vision system.
• High requirement on model training speed: The deep learning software needs to train models rapidly to deliver stable and accurate recognition of various welding points.

The Advantages of Mech-Mind AI + 3D Vision Solution

• Accurate 3D point cloud data: Using structured light technology, Mech-Eye PRO S Industrial 3D Camera outputs accurate 3D point cloud data, even for randomly-located welding points and compressors with dark surfaces, ensuring the precision locating of each welding point.
• Reliable recognition: Each welding point in the object image can be specially and separately identified. Mech-Vision Machine Vision Software delivers accurate recognition results of each welding point and outputs its exact positions for robots.
• Advanced deep learning software for complex vision tasks: Product and welding point variations pose a difficult challenge for accurate recognition. Mech-DLK Deep Learning Software trains models rapidly to respond to product and object variations and fast changeovers.
• Collision detection and path optimization: Mech-Viz Robot Programming Software plans the optimum robot path for precision inspection without collisions.
• Faster inspection speed: Mech-Vision filters out the high-pressure welding point based on the information of the Z-axis and sets a threshold for outputting recognition results. So it can output the recognition results of the low-pressure welding points that we need, thus optimizing the detection cycle.
• Traceability and real-time data management: Mech-Mind AI-powered software gathers data in real-time to manage detection results and review issues, thus reducing unplanned downtime and improving efficiency.

Workflow

• Mech-Eye PRO S, fixing statically, takes images of welding points and maps the surroundings.
• Mech-Vision provides the position information of all welding points for the robot at one time and guides the robot to arrive at the designated position. It ensures that the robot can conduct the halogen detection of all welding points one after another with only one data capture, optimizing inspection efficiency.
• Mech-Vision excludes the high-pressure welding point based on the information of the Z-axis and limits the recognition result number to three.
• The vision system identifies the positions of welding points based on 3D point clouds. Then the vision system calculates the distances between the welding points and the compressor center to filter the high-pressure point. (See Figure 4)
• Mech-Viz plans the optimal path for robots to understand complex environments and avoid collisions.
• Robots carry out the halogen detection of welding points.
• Once the end effector detects leaks, the system will trigger the alert, and the degraded product will be sent for repair.
• Repeat the procedures above until all compressors are detected.

The Results

• 100% recognition accuracy: Machine learning and deep learning algorithms enable the accurate recognition of welding points of almost all sizes and positions.
• Vision time (image acquisition + vision processing for 4 welding points) < 4 s: Industrial robots can conduct the halogen detection of more welding points per minute.
• Optimized cycle time: The built-in intelligent path-planning and collision detection algorithms plan the optimal path for robotic operation and thus improve the cycle time.

The Outcomes

AI technologies, including Machine Learning, Motion Planning, and Deep Learning, give traditional robots human-like functions. AI enables robots to use machine vision, control motion, and detect collisions to deliver a better performance in halogen detection. The integration of AI, 3D vision, and robots brings numerous benefits to A company.

• Increased inspection efficiency: Integrating AI technologies into industrial automation optimizes halogen detection. With improved detection consistency and precision, A company has increased productivity and thus yielded more profits.
• Lowered costs: Thanks to advanced AI technologies, such as deep learning, A company has handled product variations and fast changeovers without re-programming robotic applications, reduced unplanned downtime, and thus saved costs.
• Improved reliability: In a complicated industrial environment, Mech-Mind AI + 3D solution helps A company reduce the risk of on-site accidents. The robot, paired with AI technologies, can sense obstructions along its pre-programmed path and movement. In A company’s factories, robots can avoid collisions with objects and surroundings while conducting halogen leak detection.

AI for Multiple Challenging Tasks

Empowered by advanced vision technology and AI-powered software, robots are finding their way into unchartered territory, including random bin picking, assembly, and inspection. Mech-Mind is doing its best to push AI technologies to the limits and create robotics that fit seamlessly into more demanding industrial applications.