Hiroshima University Researchers Develop Non-Contact Vibration Monitoring for Industrial Robots

Industrial robots perform many critical tasks in manufacturing, including welding, assembly, painting, material handling, and inspection. Keeping these robots reliable depends on vibration monitoring, since abnormal vibration patterns often signal mechanical issues such as imbalance, loose components, bearing wear, or structural degradation.

Most vibration monitoring relies on physical sensors, like piezoelectric accelerometers. While accurate, these sensors add cost and weight to robot arms, can interfere with motion, and require cabling that is prone to flexing, wear, or failure over time. These challenges have driven interest in non-contact techniques that can support predictive maintenance without physically altering the robot.

Researchers at Hiroshima University in Japan have developed and tested a non-contact vibration monitoring system that combines Digital Image Correlation (DIC) with high-speed imaging from a Mikrotron camera. DIC is an optical method used to measure full-field surface deformation, displacement, and strain. The system captures vibration behavior across an industrial robot’s structure without the need for physical sensors.

To test the system, the researchers used a six-degree-of-freedom vertical articulated industrial robot. They applied a random speckle pattern to the robot's surface to enable optical tracking and attached a small vibration exciter to create controlled vibration.

Images were captured using a Mikrotron EoSens 2.0CXP2 high-speed camera with a 20 mm lens, mounted 1.2 meters from the robot. The camera's 2-megapixel CMOS global shutter sensor and four-lane CoaXPress 2.0 interface allowed high-resolution capture of 1920x1080 pixel images at 1,000 frames per second, with a spatial resolution of 0.6 mm per pixel.

Captured video was divided into individual frames and segmented into hundreds of smaller regions, creating a network of virtual vibration sensors across the robot. Rather than attaching physical sensors, the system tracked vibration at eight key points on the robot, defined in software using DIC. A GPU-accelerated algorithm developed in C++ processed the high-frame-rate image data and tracked changes in the speckle pattern over time to generate two-dimensional vibration maps.